diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,56419 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.12601938285358827, + "eval_steps": 500, + "global_step": 47000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 5e-06, + "loss": 3.6187, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 1e-05, + "loss": 3.5952, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.5e-05, + "loss": 3.5556, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 2e-05, + "loss": 3.6457, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 2.5e-05, + "loss": 3.5799, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 3e-05, + "loss": 3.6072, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 3.5e-05, + "loss": 3.4903, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 4e-05, + "loss": 3.5139, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 4.5e-05, + "loss": 3.5147, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 5e-05, + "loss": 3.4908, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 5.500000000000001e-05, + "loss": 3.5049, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 6e-05, + "loss": 3.4844, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 6.500000000000001e-05, + "loss": 3.4641, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 7e-05, + "loss": 3.4053, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 7.500000000000001e-05, + "loss": 3.3768, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 8e-05, + "loss": 3.4016, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 8.5e-05, + "loss": 3.464, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 9e-05, + "loss": 3.2829, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 9.5e-05, + "loss": 3.371, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 0.0001, + "loss": 3.3552, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999995562965e-05, + "loss": 3.3274, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999982251862e-05, + "loss": 3.4382, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999960066688e-05, + "loss": 3.2754, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999929007444e-05, + "loss": 3.3766, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999889074132e-05, + "loss": 3.3042, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999984026675e-05, + "loss": 3.3027, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 9.9999997825853e-05, + "loss": 3.2065, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999716029778e-05, + "loss": 3.253, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999640600189e-05, + "loss": 3.2436, + "step": 145 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999556296531e-05, + "loss": 3.2938, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999463118803e-05, + "loss": 3.2245, + "step": 155 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999361067009e-05, + "loss": 3.2972, + "step": 160 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999250141145e-05, + "loss": 3.1364, + "step": 165 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999130341212e-05, + "loss": 3.1546, + "step": 170 + }, + { + "epoch": 0.0, + "learning_rate": 9.999999001667212e-05, + "loss": 3.2106, + "step": 175 + }, + { + "epoch": 0.0, + "learning_rate": 9.999998864119144e-05, + "loss": 3.2168, + "step": 180 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999871769701e-05, + "loss": 3.1428, + "step": 185 + }, + { + "epoch": 0.0, + "learning_rate": 9.999998562400806e-05, + "loss": 3.2215, + "step": 190 + }, + { + "epoch": 0.0, + "learning_rate": 9.999998398230536e-05, + "loss": 3.192, + "step": 195 + }, + { + "epoch": 0.0, + "learning_rate": 9.9999982251862e-05, + "loss": 3.1603, + "step": 200 + }, + { + "epoch": 0.0, + "learning_rate": 9.999998043267797e-05, + "loss": 3.2186, + "step": 205 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999785247533e-05, + "loss": 3.0613, + "step": 210 + }, + { + "epoch": 0.0, + "learning_rate": 9.999997652808794e-05, + "loss": 3.2031, + "step": 215 + }, + { + "epoch": 0.0, + "learning_rate": 9.999997444268194e-05, + "loss": 3.0355, + "step": 220 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999722685353e-05, + "loss": 3.2881, + "step": 225 + }, + { + "epoch": 0.0, + "learning_rate": 9.9999970005648e-05, + "loss": 3.1349, + "step": 230 + }, + { + "epoch": 0.0, + "learning_rate": 9.999996765402007e-05, + "loss": 3.1062, + "step": 235 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999652136515e-05, + "loss": 3.0767, + "step": 240 + }, + { + "epoch": 0.0, + "learning_rate": 9.999996268454229e-05, + "loss": 3.1813, + "step": 245 + }, + { + "epoch": 0.0, + "learning_rate": 9.999996006669245e-05, + "loss": 3.1079, + "step": 250 + }, + { + "epoch": 0.0, + "learning_rate": 9.999995736010199e-05, + "loss": 3.1125, + "step": 255 + }, + { + "epoch": 0.0, + "learning_rate": 9.999995456477091e-05, + "loss": 3.2153, + "step": 260 + }, + { + "epoch": 0.0, + "learning_rate": 9.999995168069923e-05, + "loss": 3.0924, + "step": 265 + }, + { + "epoch": 0.0, + "learning_rate": 9.999994870788691e-05, + "loss": 3.2386, + "step": 270 + }, + { + "epoch": 0.0, + "learning_rate": 9.9999945646334e-05, + "loss": 3.1615, + "step": 275 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999424960405e-05, + "loss": 3.0519, + "step": 280 + }, + { + "epoch": 0.0, + "learning_rate": 9.999993925700639e-05, + "loss": 3.1259, + "step": 285 + }, + { + "epoch": 0.0, + "learning_rate": 9.999993592923171e-05, + "loss": 3.0988, + "step": 290 + }, + { + "epoch": 0.0, + "learning_rate": 9.999993251271645e-05, + "loss": 3.0871, + "step": 295 + }, + { + "epoch": 0.0, + "learning_rate": 9.999992900746059e-05, + "loss": 3.1556, + "step": 300 + }, + { + "epoch": 0.0, + "learning_rate": 9.999992541346418e-05, + "loss": 3.2086, + "step": 305 + }, + { + "epoch": 0.0, + "learning_rate": 9.999992173072721e-05, + "loss": 3.1455, + "step": 310 + }, + { + "epoch": 0.0, + "learning_rate": 9.999991795924967e-05, + "loss": 3.1561, + "step": 315 + }, + { + "epoch": 0.0, + "learning_rate": 9.999991409903159e-05, + "loss": 3.222, + "step": 320 + }, + { + "epoch": 0.0, + "learning_rate": 9.999991015007296e-05, + "loss": 3.1018, + "step": 325 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999061123738e-05, + "loss": 3.0528, + "step": 330 + }, + { + "epoch": 0.0, + "learning_rate": 9.99999019859341e-05, + "loss": 3.1161, + "step": 335 + }, + { + "epoch": 0.0, + "learning_rate": 9.999989777075389e-05, + "loss": 3.0887, + "step": 340 + }, + { + "epoch": 0.0, + "learning_rate": 9.999989346683317e-05, + "loss": 2.9031, + "step": 345 + }, + { + "epoch": 0.0, + "learning_rate": 9.999988907417195e-05, + "loss": 3.0054, + "step": 350 + }, + { + "epoch": 0.0, + "learning_rate": 9.999988459277021e-05, + "loss": 3.0589, + "step": 355 + }, + { + "epoch": 0.0, + "learning_rate": 9.999988002262799e-05, + "loss": 3.0355, + "step": 360 + }, + { + "epoch": 0.0, + "learning_rate": 9.999987536374529e-05, + "loss": 3.0046, + "step": 365 + }, + { + "epoch": 0.0, + "learning_rate": 9.99998706161221e-05, + "loss": 3.0252, + "step": 370 + }, + { + "epoch": 0.0, + "learning_rate": 9.999986577975846e-05, + "loss": 2.9698, + "step": 375 + }, + { + "epoch": 0.0, + "learning_rate": 9.999986085465437e-05, + "loss": 3.047, + "step": 380 + }, + { + "epoch": 0.0, + "learning_rate": 9.999985584080982e-05, + "loss": 3.0807, + "step": 385 + }, + { + "epoch": 0.0, + "learning_rate": 9.999985073822484e-05, + "loss": 3.0396, + "step": 390 + }, + { + "epoch": 0.0, + "learning_rate": 9.999984554689941e-05, + "loss": 2.9717, + "step": 395 + }, + { + "epoch": 0.0, + "learning_rate": 9.999984026683357e-05, + "loss": 3.0606, + "step": 400 + }, + { + "epoch": 0.0, + "learning_rate": 9.999983489802733e-05, + "loss": 3.1319, + "step": 405 + }, + { + "epoch": 0.0, + "learning_rate": 9.999982944048067e-05, + "loss": 3.0232, + "step": 410 + }, + { + "epoch": 0.0, + "learning_rate": 9.999982389419363e-05, + "loss": 2.9903, + "step": 415 + }, + { + "epoch": 0.0, + "learning_rate": 9.99998182591662e-05, + "loss": 3.0394, + "step": 420 + }, + { + "epoch": 0.0, + "learning_rate": 9.999981253539841e-05, + "loss": 2.9271, + "step": 425 + }, + { + "epoch": 0.0, + "learning_rate": 9.999980672289025e-05, + "loss": 3.0966, + "step": 430 + }, + { + "epoch": 0.0, + "learning_rate": 9.999980082164174e-05, + "loss": 3.1027, + "step": 435 + }, + { + "epoch": 0.0, + "learning_rate": 9.999979483165288e-05, + "loss": 3.0962, + "step": 440 + }, + { + "epoch": 0.0, + "learning_rate": 9.999978875292368e-05, + "loss": 3.0701, + "step": 445 + }, + { + "epoch": 0.0, + "learning_rate": 9.999978258545417e-05, + "loss": 3.0215, + "step": 450 + }, + { + "epoch": 0.0, + "learning_rate": 9.999977632924435e-05, + "loss": 3.153, + "step": 455 + }, + { + "epoch": 0.0, + "learning_rate": 9.999976998429424e-05, + "loss": 2.9545, + "step": 460 + }, + { + "epoch": 0.0, + "learning_rate": 9.999976355060384e-05, + "loss": 3.0044, + "step": 465 + }, + { + "epoch": 0.0, + "learning_rate": 9.999975702817316e-05, + "loss": 2.917, + "step": 470 + }, + { + "epoch": 0.0, + "learning_rate": 9.999975041700221e-05, + "loss": 3.0225, + "step": 475 + }, + { + "epoch": 0.0, + "learning_rate": 9.999974371709102e-05, + "loss": 3.0239, + "step": 480 + }, + { + "epoch": 0.0, + "learning_rate": 9.999973692843958e-05, + "loss": 2.9507, + "step": 485 + }, + { + "epoch": 0.0, + "learning_rate": 9.999973005104792e-05, + "loss": 3.0276, + "step": 490 + }, + { + "epoch": 0.0, + "learning_rate": 9.999972308491604e-05, + "loss": 2.8835, + "step": 495 + }, + { + "epoch": 0.0, + "learning_rate": 9.999971603004396e-05, + "loss": 2.9019, + "step": 500 + }, + { + "epoch": 0.0, + "learning_rate": 9.999970888643168e-05, + "loss": 2.9554, + "step": 505 + }, + { + "epoch": 0.0, + "learning_rate": 9.999970165407924e-05, + "loss": 3.1321, + "step": 510 + }, + { + "epoch": 0.0, + "learning_rate": 9.999969433298662e-05, + "loss": 2.9113, + "step": 515 + }, + { + "epoch": 0.0, + "learning_rate": 9.999968692315385e-05, + "loss": 2.9836, + "step": 520 + }, + { + "epoch": 0.0, + "learning_rate": 9.999967942458093e-05, + "loss": 3.0215, + "step": 525 + }, + { + "epoch": 0.0, + "learning_rate": 9.999967183726788e-05, + "loss": 3.0717, + "step": 530 + }, + { + "epoch": 0.0, + "learning_rate": 9.999966416121473e-05, + "loss": 2.9074, + "step": 535 + }, + { + "epoch": 0.0, + "learning_rate": 9.999965639642149e-05, + "loss": 2.9473, + "step": 540 + }, + { + "epoch": 0.0, + "learning_rate": 9.999964854288815e-05, + "loss": 2.983, + "step": 545 + }, + { + "epoch": 0.0, + "learning_rate": 9.999964060061475e-05, + "loss": 3.0114, + "step": 550 + }, + { + "epoch": 0.0, + "learning_rate": 9.999963256960128e-05, + "loss": 2.9233, + "step": 555 + }, + { + "epoch": 0.0, + "learning_rate": 9.999962444984779e-05, + "loss": 2.9452, + "step": 560 + }, + { + "epoch": 0.0, + "learning_rate": 9.999961624135423e-05, + "loss": 2.9346, + "step": 565 + }, + { + "epoch": 0.0, + "learning_rate": 9.99996079441207e-05, + "loss": 2.9067, + "step": 570 + }, + { + "epoch": 0.0, + "learning_rate": 9.999959955814714e-05, + "loss": 2.8204, + "step": 575 + }, + { + "epoch": 0.0, + "learning_rate": 9.999959108343361e-05, + "loss": 3.0203, + "step": 580 + }, + { + "epoch": 0.0, + "learning_rate": 9.99995825199801e-05, + "loss": 3.0361, + "step": 585 + }, + { + "epoch": 0.0, + "learning_rate": 9.999957386778664e-05, + "loss": 3.0295, + "step": 590 + }, + { + "epoch": 0.0, + "learning_rate": 9.999956512685325e-05, + "loss": 2.9177, + "step": 595 + }, + { + "epoch": 0.0, + "learning_rate": 9.999955629717993e-05, + "loss": 2.9124, + "step": 600 + }, + { + "epoch": 0.0, + "learning_rate": 9.99995473787667e-05, + "loss": 2.8982, + "step": 605 + }, + { + "epoch": 0.0, + "learning_rate": 9.999953837161358e-05, + "loss": 2.8976, + "step": 610 + }, + { + "epoch": 0.0, + "learning_rate": 9.999952927572058e-05, + "loss": 2.847, + "step": 615 + }, + { + "epoch": 0.0, + "learning_rate": 9.999952009108773e-05, + "loss": 2.899, + "step": 620 + }, + { + "epoch": 0.0, + "learning_rate": 9.999951081771503e-05, + "loss": 2.8496, + "step": 625 + }, + { + "epoch": 0.0, + "learning_rate": 9.999950145560251e-05, + "loss": 3.0726, + "step": 630 + }, + { + "epoch": 0.0, + "learning_rate": 9.999949200475016e-05, + "loss": 3.054, + "step": 635 + }, + { + "epoch": 0.0, + "learning_rate": 9.999948246515805e-05, + "loss": 2.8893, + "step": 640 + }, + { + "epoch": 0.0, + "learning_rate": 9.999947283682613e-05, + "loss": 2.8911, + "step": 645 + }, + { + "epoch": 0.0, + "learning_rate": 9.999946311975446e-05, + "loss": 2.9062, + "step": 650 + }, + { + "epoch": 0.0, + "learning_rate": 9.999945331394305e-05, + "loss": 2.8179, + "step": 655 + }, + { + "epoch": 0.0, + "learning_rate": 9.999944341939192e-05, + "loss": 2.8872, + "step": 660 + }, + { + "epoch": 0.0, + "learning_rate": 9.999943343610107e-05, + "loss": 3.0942, + "step": 665 + }, + { + "epoch": 0.0, + "learning_rate": 9.999942336407056e-05, + "loss": 2.8936, + "step": 670 + }, + { + "epoch": 0.0, + "learning_rate": 9.999941320330035e-05, + "loss": 2.919, + "step": 675 + }, + { + "epoch": 0.0, + "learning_rate": 9.99994029537905e-05, + "loss": 2.9586, + "step": 680 + }, + { + "epoch": 0.0, + "learning_rate": 9.999939261554099e-05, + "loss": 2.9121, + "step": 685 + }, + { + "epoch": 0.0, + "learning_rate": 9.99993821885519e-05, + "loss": 3.001, + "step": 690 + }, + { + "epoch": 0.0, + "learning_rate": 9.999937167282318e-05, + "loss": 2.979, + "step": 695 + }, + { + "epoch": 0.0, + "learning_rate": 9.999936106835489e-05, + "loss": 2.8101, + "step": 700 + }, + { + "epoch": 0.0, + "learning_rate": 9.999935037514704e-05, + "loss": 2.9939, + "step": 705 + }, + { + "epoch": 0.0, + "learning_rate": 9.999933959319965e-05, + "loss": 2.9973, + "step": 710 + }, + { + "epoch": 0.0, + "learning_rate": 9.999932872251274e-05, + "loss": 2.9539, + "step": 715 + }, + { + "epoch": 0.0, + "learning_rate": 9.999931776308631e-05, + "loss": 2.9219, + "step": 720 + }, + { + "epoch": 0.0, + "learning_rate": 9.999930671492042e-05, + "loss": 2.9067, + "step": 725 + }, + { + "epoch": 0.0, + "learning_rate": 9.999929557801504e-05, + "loss": 2.7508, + "step": 730 + }, + { + "epoch": 0.0, + "learning_rate": 9.999928435237023e-05, + "loss": 2.9395, + "step": 735 + }, + { + "epoch": 0.0, + "learning_rate": 9.9999273037986e-05, + "loss": 2.9348, + "step": 740 + }, + { + "epoch": 0.0, + "learning_rate": 9.999926163486235e-05, + "loss": 2.9191, + "step": 745 + }, + { + "epoch": 0.0, + "learning_rate": 9.999925014299931e-05, + "loss": 2.8866, + "step": 750 + }, + { + "epoch": 0.0, + "learning_rate": 9.999923856239692e-05, + "loss": 2.9143, + "step": 755 + }, + { + "epoch": 0.0, + "learning_rate": 9.999922689305518e-05, + "loss": 2.9447, + "step": 760 + }, + { + "epoch": 0.0, + "learning_rate": 9.999921513497413e-05, + "loss": 2.8824, + "step": 765 + }, + { + "epoch": 0.0, + "learning_rate": 9.999920328815377e-05, + "loss": 2.8373, + "step": 770 + }, + { + "epoch": 0.0, + "learning_rate": 9.999919135259412e-05, + "loss": 2.9323, + "step": 775 + }, + { + "epoch": 0.0, + "learning_rate": 9.999917932829522e-05, + "loss": 2.9966, + "step": 780 + }, + { + "epoch": 0.0, + "learning_rate": 9.999916721525708e-05, + "loss": 2.9014, + "step": 785 + }, + { + "epoch": 0.0, + "learning_rate": 9.999915501347972e-05, + "loss": 2.821, + "step": 790 + }, + { + "epoch": 0.0, + "learning_rate": 9.999914272296317e-05, + "loss": 2.8028, + "step": 795 + }, + { + "epoch": 0.0, + "learning_rate": 9.999913034370744e-05, + "loss": 2.956, + "step": 800 + }, + { + "epoch": 0.0, + "learning_rate": 9.999911787571257e-05, + "loss": 2.945, + "step": 805 + }, + { + "epoch": 0.0, + "learning_rate": 9.999910531897856e-05, + "loss": 2.8033, + "step": 810 + }, + { + "epoch": 0.0, + "learning_rate": 9.999909267350543e-05, + "loss": 2.8858, + "step": 815 + }, + { + "epoch": 0.0, + "learning_rate": 9.999907993929322e-05, + "loss": 2.9663, + "step": 820 + }, + { + "epoch": 0.0, + "learning_rate": 9.999906711634197e-05, + "loss": 2.8677, + "step": 825 + }, + { + "epoch": 0.0, + "learning_rate": 9.999905420465166e-05, + "loss": 2.8702, + "step": 830 + }, + { + "epoch": 0.0, + "learning_rate": 9.999904120422234e-05, + "loss": 2.9527, + "step": 835 + }, + { + "epoch": 0.0, + "learning_rate": 9.999902811505403e-05, + "loss": 2.7591, + "step": 840 + }, + { + "epoch": 0.0, + "learning_rate": 9.999901493714674e-05, + "loss": 2.8844, + "step": 845 + }, + { + "epoch": 0.0, + "learning_rate": 9.999900167050052e-05, + "loss": 2.7509, + "step": 850 + }, + { + "epoch": 0.0, + "learning_rate": 9.999898831511536e-05, + "loss": 2.8329, + "step": 855 + }, + { + "epoch": 0.0, + "learning_rate": 9.99989748709913e-05, + "loss": 2.8901, + "step": 860 + }, + { + "epoch": 0.0, + "learning_rate": 9.999896133812837e-05, + "loss": 2.7646, + "step": 865 + }, + { + "epoch": 0.0, + "learning_rate": 9.99989477165266e-05, + "loss": 2.8065, + "step": 870 + }, + { + "epoch": 0.0, + "learning_rate": 9.999893400618598e-05, + "loss": 2.8348, + "step": 875 + }, + { + "epoch": 0.0, + "learning_rate": 9.999892020710657e-05, + "loss": 2.9766, + "step": 880 + }, + { + "epoch": 0.0, + "learning_rate": 9.999890631928838e-05, + "loss": 2.8015, + "step": 885 + }, + { + "epoch": 0.0, + "learning_rate": 9.999889234273143e-05, + "loss": 2.7151, + "step": 890 + }, + { + "epoch": 0.0, + "learning_rate": 9.999887827743575e-05, + "loss": 2.7859, + "step": 895 + }, + { + "epoch": 0.0, + "learning_rate": 9.999886412340139e-05, + "loss": 2.9127, + "step": 900 + }, + { + "epoch": 0.0, + "learning_rate": 9.99988498806283e-05, + "loss": 2.8461, + "step": 905 + }, + { + "epoch": 0.0, + "learning_rate": 9.99988355491166e-05, + "loss": 2.7719, + "step": 910 + }, + { + "epoch": 0.0, + "learning_rate": 9.999882112886626e-05, + "loss": 2.8639, + "step": 915 + }, + { + "epoch": 0.0, + "learning_rate": 9.999880661987731e-05, + "loss": 2.8807, + "step": 920 + }, + { + "epoch": 0.0, + "learning_rate": 9.999879202214978e-05, + "loss": 2.7215, + "step": 925 + }, + { + "epoch": 0.0, + "learning_rate": 9.999877733568372e-05, + "loss": 2.833, + "step": 930 + }, + { + "epoch": 0.0, + "learning_rate": 9.999876256047913e-05, + "loss": 2.8248, + "step": 935 + }, + { + "epoch": 0.0, + "learning_rate": 9.999874769653603e-05, + "loss": 2.7844, + "step": 940 + }, + { + "epoch": 0.0, + "learning_rate": 9.999873274385446e-05, + "loss": 2.7441, + "step": 945 + }, + { + "epoch": 0.0, + "learning_rate": 9.999871770243444e-05, + "loss": 2.7576, + "step": 950 + }, + { + "epoch": 0.0, + "learning_rate": 9.9998702572276e-05, + "loss": 2.8761, + "step": 955 + }, + { + "epoch": 0.0, + "learning_rate": 9.999868735337918e-05, + "loss": 2.8368, + "step": 960 + }, + { + "epoch": 0.0, + "learning_rate": 9.999867204574398e-05, + "loss": 2.8398, + "step": 965 + }, + { + "epoch": 0.0, + "learning_rate": 9.999865664937044e-05, + "loss": 2.8948, + "step": 970 + }, + { + "epoch": 0.0, + "learning_rate": 9.999864116425861e-05, + "loss": 2.8246, + "step": 975 + }, + { + "epoch": 0.0, + "learning_rate": 9.999862559040848e-05, + "loss": 2.9099, + "step": 980 + }, + { + "epoch": 0.0, + "learning_rate": 9.99986099278201e-05, + "loss": 2.8416, + "step": 985 + }, + { + "epoch": 0.0, + "learning_rate": 9.999859417649349e-05, + "loss": 2.9008, + "step": 990 + }, + { + "epoch": 0.0, + "learning_rate": 9.999857833642867e-05, + "loss": 2.8478, + "step": 995 + }, + { + "epoch": 0.0, + "learning_rate": 9.999856240762569e-05, + "loss": 2.9886, + "step": 1000 + }, + { + "epoch": 0.0, + "learning_rate": 9.999854639008457e-05, + "loss": 2.7783, + "step": 1005 + }, + { + "epoch": 0.0, + "learning_rate": 9.999853028380532e-05, + "loss": 2.6985, + "step": 1010 + }, + { + "epoch": 0.0, + "learning_rate": 9.999851408878799e-05, + "loss": 2.8538, + "step": 1015 + }, + { + "epoch": 0.0, + "learning_rate": 9.999849780503262e-05, + "loss": 2.8475, + "step": 1020 + }, + { + "epoch": 0.0, + "learning_rate": 9.99984814325392e-05, + "loss": 2.9108, + "step": 1025 + }, + { + "epoch": 0.0, + "learning_rate": 9.999846497130778e-05, + "loss": 2.8823, + "step": 1030 + }, + { + "epoch": 0.0, + "learning_rate": 9.999844842133839e-05, + "loss": 2.7578, + "step": 1035 + }, + { + "epoch": 0.0, + "learning_rate": 9.999843178263107e-05, + "loss": 2.8896, + "step": 1040 + }, + { + "epoch": 0.0, + "learning_rate": 9.999841505518582e-05, + "loss": 2.7881, + "step": 1045 + }, + { + "epoch": 0.0, + "learning_rate": 9.99983982390027e-05, + "loss": 2.6861, + "step": 1050 + }, + { + "epoch": 0.0, + "learning_rate": 9.999838133408173e-05, + "loss": 2.759, + "step": 1055 + }, + { + "epoch": 0.0, + "learning_rate": 9.999836434042292e-05, + "loss": 2.9172, + "step": 1060 + }, + { + "epoch": 0.0, + "learning_rate": 9.999834725802633e-05, + "loss": 2.8333, + "step": 1065 + }, + { + "epoch": 0.0, + "learning_rate": 9.999833008689198e-05, + "loss": 2.7991, + "step": 1070 + }, + { + "epoch": 0.0, + "learning_rate": 9.99983128270199e-05, + "loss": 2.8067, + "step": 1075 + }, + { + "epoch": 0.0, + "learning_rate": 9.99982954784101e-05, + "loss": 2.7929, + "step": 1080 + }, + { + "epoch": 0.0, + "learning_rate": 9.999827804106265e-05, + "loss": 2.8259, + "step": 1085 + }, + { + "epoch": 0.0, + "learning_rate": 9.999826051497756e-05, + "loss": 2.8768, + "step": 1090 + }, + { + "epoch": 0.0, + "learning_rate": 9.999824290015486e-05, + "loss": 2.7809, + "step": 1095 + }, + { + "epoch": 0.0, + "learning_rate": 9.999822519659458e-05, + "loss": 2.7952, + "step": 1100 + }, + { + "epoch": 0.0, + "learning_rate": 9.999820740429677e-05, + "loss": 2.7442, + "step": 1105 + }, + { + "epoch": 0.0, + "learning_rate": 9.999818952326144e-05, + "loss": 2.7215, + "step": 1110 + }, + { + "epoch": 0.0, + "learning_rate": 9.99981715534886e-05, + "loss": 2.8527, + "step": 1115 + }, + { + "epoch": 0.0, + "learning_rate": 9.999815349497834e-05, + "loss": 2.7883, + "step": 1120 + }, + { + "epoch": 0.0, + "learning_rate": 9.999813534773066e-05, + "loss": 2.8277, + "step": 1125 + }, + { + "epoch": 0.0, + "learning_rate": 9.999811711174558e-05, + "loss": 2.8714, + "step": 1130 + }, + { + "epoch": 0.0, + "learning_rate": 9.999809878702316e-05, + "loss": 2.6746, + "step": 1135 + }, + { + "epoch": 0.0, + "learning_rate": 9.999808037356341e-05, + "loss": 2.7569, + "step": 1140 + }, + { + "epoch": 0.0, + "learning_rate": 9.999806187136637e-05, + "loss": 2.8634, + "step": 1145 + }, + { + "epoch": 0.0, + "learning_rate": 9.999804328043207e-05, + "loss": 2.7646, + "step": 1150 + }, + { + "epoch": 0.0, + "learning_rate": 9.999802460076057e-05, + "loss": 2.8506, + "step": 1155 + }, + { + "epoch": 0.0, + "learning_rate": 9.999800583235185e-05, + "loss": 2.7151, + "step": 1160 + }, + { + "epoch": 0.0, + "learning_rate": 9.9997986975206e-05, + "loss": 2.7502, + "step": 1165 + }, + { + "epoch": 0.0, + "learning_rate": 9.999796802932301e-05, + "loss": 2.8083, + "step": 1170 + }, + { + "epoch": 0.0, + "learning_rate": 9.999794899470296e-05, + "loss": 2.758, + "step": 1175 + }, + { + "epoch": 0.0, + "learning_rate": 9.999792987134583e-05, + "loss": 2.8994, + "step": 1180 + }, + { + "epoch": 0.0, + "learning_rate": 9.999791065925169e-05, + "loss": 2.7969, + "step": 1185 + }, + { + "epoch": 0.0, + "learning_rate": 9.999789135842054e-05, + "loss": 2.847, + "step": 1190 + }, + { + "epoch": 0.0, + "learning_rate": 9.999787196885247e-05, + "loss": 2.8691, + "step": 1195 + }, + { + "epoch": 0.0, + "learning_rate": 9.999785249054746e-05, + "loss": 2.8196, + "step": 1200 + }, + { + "epoch": 0.0, + "learning_rate": 9.999783292350557e-05, + "loss": 2.6763, + "step": 1205 + }, + { + "epoch": 0.0, + "learning_rate": 9.999781326772684e-05, + "loss": 2.7302, + "step": 1210 + }, + { + "epoch": 0.0, + "learning_rate": 9.999779352321128e-05, + "loss": 2.8317, + "step": 1215 + }, + { + "epoch": 0.0, + "learning_rate": 9.999777368995895e-05, + "loss": 2.7518, + "step": 1220 + }, + { + "epoch": 0.0, + "learning_rate": 9.999775376796987e-05, + "loss": 2.7637, + "step": 1225 + }, + { + "epoch": 0.0, + "learning_rate": 9.999773375724409e-05, + "loss": 2.7313, + "step": 1230 + }, + { + "epoch": 0.0, + "learning_rate": 9.999771365778163e-05, + "loss": 2.7514, + "step": 1235 + }, + { + "epoch": 0.0, + "learning_rate": 9.999769346958254e-05, + "loss": 2.8422, + "step": 1240 + }, + { + "epoch": 0.0, + "learning_rate": 9.999767319264685e-05, + "loss": 2.8061, + "step": 1245 + }, + { + "epoch": 0.0, + "learning_rate": 9.999765282697459e-05, + "loss": 2.6965, + "step": 1250 + }, + { + "epoch": 0.0, + "learning_rate": 9.999763237256578e-05, + "loss": 2.7863, + "step": 1255 + }, + { + "epoch": 0.0, + "learning_rate": 9.999761182942051e-05, + "loss": 2.6574, + "step": 1260 + }, + { + "epoch": 0.0, + "learning_rate": 9.999759119753877e-05, + "loss": 2.7954, + "step": 1265 + }, + { + "epoch": 0.0, + "learning_rate": 9.999757047692061e-05, + "loss": 2.6935, + "step": 1270 + }, + { + "epoch": 0.0, + "learning_rate": 9.999754966756607e-05, + "loss": 2.7043, + "step": 1275 + }, + { + "epoch": 0.0, + "learning_rate": 9.999752876947517e-05, + "loss": 2.6389, + "step": 1280 + }, + { + "epoch": 0.0, + "learning_rate": 9.999750778264799e-05, + "loss": 2.7847, + "step": 1285 + }, + { + "epoch": 0.0, + "learning_rate": 9.99974867070845e-05, + "loss": 2.8634, + "step": 1290 + }, + { + "epoch": 0.0, + "learning_rate": 9.99974655427848e-05, + "loss": 2.7628, + "step": 1295 + }, + { + "epoch": 0.0, + "learning_rate": 9.999744428974888e-05, + "loss": 2.8584, + "step": 1300 + }, + { + "epoch": 0.0, + "learning_rate": 9.999742294797682e-05, + "loss": 2.7888, + "step": 1305 + }, + { + "epoch": 0.0, + "learning_rate": 9.999740151746864e-05, + "loss": 2.7657, + "step": 1310 + }, + { + "epoch": 0.0, + "learning_rate": 9.999737999822439e-05, + "loss": 2.801, + "step": 1315 + }, + { + "epoch": 0.0, + "learning_rate": 9.999735839024406e-05, + "loss": 2.7152, + "step": 1320 + }, + { + "epoch": 0.0, + "learning_rate": 9.999733669352775e-05, + "loss": 2.8158, + "step": 1325 + }, + { + "epoch": 0.0, + "learning_rate": 9.999731490807546e-05, + "loss": 2.7732, + "step": 1330 + }, + { + "epoch": 0.0, + "learning_rate": 9.999729303388725e-05, + "loss": 2.6808, + "step": 1335 + }, + { + "epoch": 0.0, + "learning_rate": 9.999727107096313e-05, + "loss": 2.8198, + "step": 1340 + }, + { + "epoch": 0.0, + "learning_rate": 9.999724901930317e-05, + "loss": 2.7975, + "step": 1345 + }, + { + "epoch": 0.0, + "learning_rate": 9.99972268789074e-05, + "loss": 2.7706, + "step": 1350 + }, + { + "epoch": 0.0, + "learning_rate": 9.999720464977586e-05, + "loss": 2.7736, + "step": 1355 + }, + { + "epoch": 0.0, + "learning_rate": 9.999718233190858e-05, + "loss": 2.7491, + "step": 1360 + }, + { + "epoch": 0.0, + "learning_rate": 9.99971599253056e-05, + "loss": 2.6852, + "step": 1365 + }, + { + "epoch": 0.0, + "learning_rate": 9.999713742996697e-05, + "loss": 2.7346, + "step": 1370 + }, + { + "epoch": 0.0, + "learning_rate": 9.999711484589271e-05, + "loss": 2.7463, + "step": 1375 + }, + { + "epoch": 0.0, + "learning_rate": 9.999709217308291e-05, + "loss": 2.7736, + "step": 1380 + }, + { + "epoch": 0.0, + "learning_rate": 9.999706941153755e-05, + "loss": 2.7937, + "step": 1385 + }, + { + "epoch": 0.0, + "learning_rate": 9.99970465612567e-05, + "loss": 2.809, + "step": 1390 + }, + { + "epoch": 0.0, + "learning_rate": 9.999702362224039e-05, + "loss": 2.7931, + "step": 1395 + }, + { + "epoch": 0.0, + "learning_rate": 9.999700059448869e-05, + "loss": 2.8154, + "step": 1400 + }, + { + "epoch": 0.0, + "learning_rate": 9.999697747800161e-05, + "loss": 2.671, + "step": 1405 + }, + { + "epoch": 0.0, + "learning_rate": 9.999695427277919e-05, + "loss": 2.773, + "step": 1410 + }, + { + "epoch": 0.0, + "learning_rate": 9.999693097882148e-05, + "loss": 2.7978, + "step": 1415 + }, + { + "epoch": 0.0, + "learning_rate": 9.999690759612852e-05, + "loss": 2.6553, + "step": 1420 + }, + { + "epoch": 0.0, + "learning_rate": 9.999688412470037e-05, + "loss": 2.7192, + "step": 1425 + }, + { + "epoch": 0.0, + "learning_rate": 9.999686056453703e-05, + "loss": 2.7625, + "step": 1430 + }, + { + "epoch": 0.0, + "learning_rate": 9.99968369156386e-05, + "loss": 2.7117, + "step": 1435 + }, + { + "epoch": 0.0, + "learning_rate": 9.999681317800506e-05, + "loss": 2.7583, + "step": 1440 + }, + { + "epoch": 0.0, + "learning_rate": 9.999678935163649e-05, + "loss": 2.7956, + "step": 1445 + }, + { + "epoch": 0.0, + "learning_rate": 9.999676543653292e-05, + "loss": 2.5792, + "step": 1450 + }, + { + "epoch": 0.0, + "learning_rate": 9.99967414326944e-05, + "loss": 2.645, + "step": 1455 + }, + { + "epoch": 0.0, + "learning_rate": 9.999671734012097e-05, + "loss": 2.6657, + "step": 1460 + }, + { + "epoch": 0.0, + "learning_rate": 9.999669315881267e-05, + "loss": 2.7049, + "step": 1465 + }, + { + "epoch": 0.0, + "learning_rate": 9.999666888876952e-05, + "loss": 2.7001, + "step": 1470 + }, + { + "epoch": 0.0, + "learning_rate": 9.999664452999163e-05, + "loss": 2.736, + "step": 1475 + }, + { + "epoch": 0.0, + "learning_rate": 9.999662008247896e-05, + "loss": 2.7313, + "step": 1480 + }, + { + "epoch": 0.0, + "learning_rate": 9.999659554623162e-05, + "loss": 2.6252, + "step": 1485 + }, + { + "epoch": 0.0, + "learning_rate": 9.999657092124962e-05, + "loss": 2.7718, + "step": 1490 + }, + { + "epoch": 0.0, + "learning_rate": 9.9996546207533e-05, + "loss": 2.7212, + "step": 1495 + }, + { + "epoch": 0.0, + "learning_rate": 9.999652140508183e-05, + "loss": 2.7717, + "step": 1500 + }, + { + "epoch": 0.0, + "learning_rate": 9.999649651389613e-05, + "loss": 2.7251, + "step": 1505 + }, + { + "epoch": 0.0, + "learning_rate": 9.999647153397595e-05, + "loss": 2.799, + "step": 1510 + }, + { + "epoch": 0.0, + "learning_rate": 9.999644646532135e-05, + "loss": 2.8541, + "step": 1515 + }, + { + "epoch": 0.0, + "learning_rate": 9.999642130793236e-05, + "loss": 2.7131, + "step": 1520 + }, + { + "epoch": 0.0, + "learning_rate": 9.999639606180902e-05, + "loss": 2.7205, + "step": 1525 + }, + { + "epoch": 0.0, + "learning_rate": 9.999637072695139e-05, + "loss": 2.6383, + "step": 1530 + }, + { + "epoch": 0.0, + "learning_rate": 9.999634530335948e-05, + "loss": 2.7404, + "step": 1535 + }, + { + "epoch": 0.0, + "learning_rate": 9.999631979103339e-05, + "loss": 2.8212, + "step": 1540 + }, + { + "epoch": 0.0, + "learning_rate": 9.999629418997312e-05, + "loss": 2.7046, + "step": 1545 + }, + { + "epoch": 0.0, + "learning_rate": 9.999626850017874e-05, + "loss": 2.7001, + "step": 1550 + }, + { + "epoch": 0.0, + "learning_rate": 9.999624272165027e-05, + "loss": 2.7109, + "step": 1555 + }, + { + "epoch": 0.0, + "learning_rate": 9.99962168543878e-05, + "loss": 2.5938, + "step": 1560 + }, + { + "epoch": 0.0, + "learning_rate": 9.999619089839133e-05, + "loss": 2.8318, + "step": 1565 + }, + { + "epoch": 0.0, + "learning_rate": 9.999616485366094e-05, + "loss": 2.7821, + "step": 1570 + }, + { + "epoch": 0.0, + "learning_rate": 9.999613872019663e-05, + "loss": 2.7239, + "step": 1575 + }, + { + "epoch": 0.0, + "learning_rate": 9.999611249799852e-05, + "loss": 2.6988, + "step": 1580 + }, + { + "epoch": 0.0, + "learning_rate": 9.99960861870666e-05, + "loss": 2.7907, + "step": 1585 + }, + { + "epoch": 0.0, + "learning_rate": 9.999605978740092e-05, + "loss": 2.6678, + "step": 1590 + }, + { + "epoch": 0.0, + "learning_rate": 9.999603329900153e-05, + "loss": 2.7428, + "step": 1595 + }, + { + "epoch": 0.0, + "learning_rate": 9.99960067218685e-05, + "loss": 2.7158, + "step": 1600 + }, + { + "epoch": 0.0, + "learning_rate": 9.999598005600187e-05, + "loss": 2.7083, + "step": 1605 + }, + { + "epoch": 0.0, + "learning_rate": 9.999595330140168e-05, + "loss": 2.7923, + "step": 1610 + }, + { + "epoch": 0.0, + "learning_rate": 9.999592645806797e-05, + "loss": 2.7433, + "step": 1615 + }, + { + "epoch": 0.0, + "learning_rate": 9.999589952600079e-05, + "loss": 2.6785, + "step": 1620 + }, + { + "epoch": 0.0, + "learning_rate": 9.999587250520018e-05, + "loss": 2.7438, + "step": 1625 + }, + { + "epoch": 0.0, + "learning_rate": 9.999584539566623e-05, + "loss": 2.701, + "step": 1630 + }, + { + "epoch": 0.0, + "learning_rate": 9.999581819739893e-05, + "loss": 2.75, + "step": 1635 + }, + { + "epoch": 0.0, + "learning_rate": 9.999579091039837e-05, + "loss": 2.7793, + "step": 1640 + }, + { + "epoch": 0.0, + "learning_rate": 9.999576353466459e-05, + "loss": 2.6655, + "step": 1645 + }, + { + "epoch": 0.0, + "learning_rate": 9.999573607019763e-05, + "loss": 2.6785, + "step": 1650 + }, + { + "epoch": 0.0, + "learning_rate": 9.999570851699753e-05, + "loss": 2.6342, + "step": 1655 + }, + { + "epoch": 0.0, + "learning_rate": 9.999568087506437e-05, + "loss": 2.6975, + "step": 1660 + }, + { + "epoch": 0.0, + "learning_rate": 9.999565314439817e-05, + "loss": 2.7826, + "step": 1665 + }, + { + "epoch": 0.0, + "learning_rate": 9.9995625324999e-05, + "loss": 2.6839, + "step": 1670 + }, + { + "epoch": 0.0, + "learning_rate": 9.999559741686688e-05, + "loss": 2.639, + "step": 1675 + }, + { + "epoch": 0.0, + "learning_rate": 9.99955694200019e-05, + "loss": 2.8456, + "step": 1680 + }, + { + "epoch": 0.0, + "learning_rate": 9.999554133440408e-05, + "loss": 2.7273, + "step": 1685 + }, + { + "epoch": 0.0, + "learning_rate": 9.999551316007347e-05, + "loss": 2.6203, + "step": 1690 + }, + { + "epoch": 0.0, + "learning_rate": 9.999548489701014e-05, + "loss": 2.7021, + "step": 1695 + }, + { + "epoch": 0.0, + "learning_rate": 9.999545654521412e-05, + "loss": 2.628, + "step": 1700 + }, + { + "epoch": 0.0, + "learning_rate": 9.999542810468547e-05, + "loss": 2.742, + "step": 1705 + }, + { + "epoch": 0.0, + "learning_rate": 9.999539957542423e-05, + "loss": 2.6653, + "step": 1710 + }, + { + "epoch": 0.0, + "learning_rate": 9.999537095743048e-05, + "loss": 2.6485, + "step": 1715 + }, + { + "epoch": 0.0, + "learning_rate": 9.999534225070424e-05, + "loss": 2.8301, + "step": 1720 + }, + { + "epoch": 0.0, + "learning_rate": 9.999531345524556e-05, + "loss": 2.79, + "step": 1725 + }, + { + "epoch": 0.0, + "learning_rate": 9.999528457105451e-05, + "loss": 2.7049, + "step": 1730 + }, + { + "epoch": 0.0, + "learning_rate": 9.999525559813114e-05, + "loss": 2.6837, + "step": 1735 + }, + { + "epoch": 0.0, + "learning_rate": 9.999522653647551e-05, + "loss": 2.7657, + "step": 1740 + }, + { + "epoch": 0.0, + "learning_rate": 9.999519738608763e-05, + "loss": 2.77, + "step": 1745 + }, + { + "epoch": 0.0, + "learning_rate": 9.999516814696759e-05, + "loss": 2.8128, + "step": 1750 + }, + { + "epoch": 0.0, + "learning_rate": 9.999513881911542e-05, + "loss": 2.6716, + "step": 1755 + }, + { + "epoch": 0.0, + "learning_rate": 9.999510940253119e-05, + "loss": 2.5943, + "step": 1760 + }, + { + "epoch": 0.0, + "learning_rate": 9.999507989721495e-05, + "loss": 2.6966, + "step": 1765 + }, + { + "epoch": 0.0, + "learning_rate": 9.999505030316674e-05, + "loss": 2.7577, + "step": 1770 + }, + { + "epoch": 0.0, + "learning_rate": 9.999502062038663e-05, + "loss": 2.6916, + "step": 1775 + }, + { + "epoch": 0.0, + "learning_rate": 9.999499084887465e-05, + "loss": 2.5991, + "step": 1780 + }, + { + "epoch": 0.0, + "learning_rate": 9.999496098863087e-05, + "loss": 2.7138, + "step": 1785 + }, + { + "epoch": 0.0, + "learning_rate": 9.999493103965534e-05, + "loss": 2.6465, + "step": 1790 + }, + { + "epoch": 0.0, + "learning_rate": 9.99949010019481e-05, + "loss": 2.6164, + "step": 1795 + }, + { + "epoch": 0.0, + "learning_rate": 9.999487087550923e-05, + "loss": 2.7684, + "step": 1800 + }, + { + "epoch": 0.0, + "learning_rate": 9.999484066033876e-05, + "loss": 2.6724, + "step": 1805 + }, + { + "epoch": 0.0, + "learning_rate": 9.999481035643676e-05, + "loss": 2.7304, + "step": 1810 + }, + { + "epoch": 0.0, + "learning_rate": 9.999477996380327e-05, + "loss": 2.7655, + "step": 1815 + }, + { + "epoch": 0.0, + "learning_rate": 9.999474948243834e-05, + "loss": 2.6767, + "step": 1820 + }, + { + "epoch": 0.0, + "learning_rate": 9.999471891234204e-05, + "loss": 2.6942, + "step": 1825 + }, + { + "epoch": 0.0, + "learning_rate": 9.999468825351441e-05, + "loss": 2.7191, + "step": 1830 + }, + { + "epoch": 0.0, + "learning_rate": 9.999465750595553e-05, + "loss": 2.7455, + "step": 1835 + }, + { + "epoch": 0.0, + "learning_rate": 9.999462666966542e-05, + "loss": 2.7208, + "step": 1840 + }, + { + "epoch": 0.0, + "learning_rate": 9.999459574464416e-05, + "loss": 2.7265, + "step": 1845 + }, + { + "epoch": 0.0, + "learning_rate": 9.999456473089179e-05, + "loss": 2.6334, + "step": 1850 + }, + { + "epoch": 0.0, + "learning_rate": 9.999453362840838e-05, + "loss": 2.7329, + "step": 1855 + }, + { + "epoch": 0.0, + "learning_rate": 9.999450243719399e-05, + "loss": 2.6806, + "step": 1860 + }, + { + "epoch": 0.01, + "learning_rate": 9.999447115724864e-05, + "loss": 2.6281, + "step": 1865 + }, + { + "epoch": 0.01, + "learning_rate": 9.999443978857241e-05, + "loss": 2.6649, + "step": 1870 + }, + { + "epoch": 0.01, + "learning_rate": 9.999440833116535e-05, + "loss": 2.7557, + "step": 1875 + }, + { + "epoch": 0.01, + "learning_rate": 9.999437678502753e-05, + "loss": 2.5978, + "step": 1880 + }, + { + "epoch": 0.01, + "learning_rate": 9.999434515015899e-05, + "loss": 2.7497, + "step": 1885 + }, + { + "epoch": 0.01, + "learning_rate": 9.999431342655978e-05, + "loss": 2.6419, + "step": 1890 + }, + { + "epoch": 0.01, + "learning_rate": 9.999428161422999e-05, + "loss": 2.7505, + "step": 1895 + }, + { + "epoch": 0.01, + "learning_rate": 9.999424971316964e-05, + "loss": 2.6116, + "step": 1900 + }, + { + "epoch": 0.01, + "learning_rate": 9.99942177233788e-05, + "loss": 2.7642, + "step": 1905 + }, + { + "epoch": 0.01, + "learning_rate": 9.999418564485752e-05, + "loss": 2.7384, + "step": 1910 + }, + { + "epoch": 0.01, + "learning_rate": 9.999415347760587e-05, + "loss": 2.7676, + "step": 1915 + }, + { + "epoch": 0.01, + "learning_rate": 9.999412122162392e-05, + "loss": 2.8212, + "step": 1920 + }, + { + "epoch": 0.01, + "learning_rate": 9.999408887691168e-05, + "loss": 2.6531, + "step": 1925 + }, + { + "epoch": 0.01, + "learning_rate": 9.999405644346925e-05, + "loss": 2.678, + "step": 1930 + }, + { + "epoch": 0.01, + "learning_rate": 9.999402392129667e-05, + "loss": 2.7302, + "step": 1935 + }, + { + "epoch": 0.01, + "learning_rate": 9.9993991310394e-05, + "loss": 2.6925, + "step": 1940 + }, + { + "epoch": 0.01, + "learning_rate": 9.99939586107613e-05, + "loss": 2.6337, + "step": 1945 + }, + { + "epoch": 0.01, + "learning_rate": 9.999392582239863e-05, + "loss": 2.6101, + "step": 1950 + }, + { + "epoch": 0.01, + "learning_rate": 9.999389294530605e-05, + "loss": 2.594, + "step": 1955 + }, + { + "epoch": 0.01, + "learning_rate": 9.999385997948361e-05, + "loss": 2.7358, + "step": 1960 + }, + { + "epoch": 0.01, + "learning_rate": 9.999382692493138e-05, + "loss": 2.6887, + "step": 1965 + }, + { + "epoch": 0.01, + "learning_rate": 9.99937937816494e-05, + "loss": 2.7229, + "step": 1970 + }, + { + "epoch": 0.01, + "learning_rate": 9.999376054963774e-05, + "loss": 2.7776, + "step": 1975 + }, + { + "epoch": 0.01, + "learning_rate": 9.999372722889646e-05, + "loss": 2.7553, + "step": 1980 + }, + { + "epoch": 0.01, + "learning_rate": 9.999369381942562e-05, + "loss": 2.6051, + "step": 1985 + }, + { + "epoch": 0.01, + "learning_rate": 9.999366032122527e-05, + "loss": 2.6612, + "step": 1990 + }, + { + "epoch": 0.01, + "learning_rate": 9.999362673429549e-05, + "loss": 2.6645, + "step": 1995 + }, + { + "epoch": 0.01, + "learning_rate": 9.999359305863631e-05, + "loss": 2.7081, + "step": 2000 + }, + { + "epoch": 0.01, + "learning_rate": 9.999355929424781e-05, + "loss": 2.7378, + "step": 2005 + }, + { + "epoch": 0.01, + "learning_rate": 9.999352544113005e-05, + "loss": 2.6468, + "step": 2010 + }, + { + "epoch": 0.01, + "learning_rate": 9.999349149928309e-05, + "loss": 2.655, + "step": 2015 + }, + { + "epoch": 0.01, + "learning_rate": 9.999345746870698e-05, + "loss": 2.5398, + "step": 2020 + }, + { + "epoch": 0.01, + "learning_rate": 9.999342334940179e-05, + "loss": 2.6937, + "step": 2025 + }, + { + "epoch": 0.01, + "learning_rate": 9.999338914136758e-05, + "loss": 2.657, + "step": 2030 + }, + { + "epoch": 0.01, + "learning_rate": 9.999335484460439e-05, + "loss": 2.7447, + "step": 2035 + }, + { + "epoch": 0.01, + "learning_rate": 9.999332045911232e-05, + "loss": 2.6687, + "step": 2040 + }, + { + "epoch": 0.01, + "learning_rate": 9.99932859848914e-05, + "loss": 2.6966, + "step": 2045 + }, + { + "epoch": 0.01, + "learning_rate": 9.999325142194169e-05, + "loss": 2.7051, + "step": 2050 + }, + { + "epoch": 0.01, + "learning_rate": 9.999321677026328e-05, + "loss": 2.7314, + "step": 2055 + }, + { + "epoch": 0.01, + "learning_rate": 9.99931820298562e-05, + "loss": 2.6274, + "step": 2060 + }, + { + "epoch": 0.01, + "learning_rate": 9.999314720072054e-05, + "loss": 2.6176, + "step": 2065 + }, + { + "epoch": 0.01, + "learning_rate": 9.999311228285634e-05, + "loss": 2.639, + "step": 2070 + }, + { + "epoch": 0.01, + "learning_rate": 9.999307727626367e-05, + "loss": 2.701, + "step": 2075 + }, + { + "epoch": 0.01, + "learning_rate": 9.99930421809426e-05, + "loss": 2.6415, + "step": 2080 + }, + { + "epoch": 0.01, + "learning_rate": 9.999300699689315e-05, + "loss": 2.6819, + "step": 2085 + }, + { + "epoch": 0.01, + "learning_rate": 9.999297172411545e-05, + "loss": 2.6842, + "step": 2090 + }, + { + "epoch": 0.01, + "learning_rate": 9.999293636260951e-05, + "loss": 2.7808, + "step": 2095 + }, + { + "epoch": 0.01, + "learning_rate": 9.999290091237544e-05, + "loss": 2.7607, + "step": 2100 + }, + { + "epoch": 0.01, + "learning_rate": 9.999286537341324e-05, + "loss": 2.6519, + "step": 2105 + }, + { + "epoch": 0.01, + "learning_rate": 9.999282974572303e-05, + "loss": 2.7588, + "step": 2110 + }, + { + "epoch": 0.01, + "learning_rate": 9.999279402930483e-05, + "loss": 2.5756, + "step": 2115 + }, + { + "epoch": 0.01, + "learning_rate": 9.999275822415874e-05, + "loss": 2.6153, + "step": 2120 + }, + { + "epoch": 0.01, + "learning_rate": 9.99927223302848e-05, + "loss": 2.6322, + "step": 2125 + }, + { + "epoch": 0.01, + "learning_rate": 9.99926863476831e-05, + "loss": 2.6346, + "step": 2130 + }, + { + "epoch": 0.01, + "learning_rate": 9.999265027635367e-05, + "loss": 2.6236, + "step": 2135 + }, + { + "epoch": 0.01, + "learning_rate": 9.99926141162966e-05, + "loss": 2.5553, + "step": 2140 + }, + { + "epoch": 0.01, + "learning_rate": 9.999257786751191e-05, + "loss": 2.6684, + "step": 2145 + }, + { + "epoch": 0.01, + "learning_rate": 9.999254152999973e-05, + "loss": 2.7854, + "step": 2150 + }, + { + "epoch": 0.01, + "learning_rate": 9.999250510376007e-05, + "loss": 2.6889, + "step": 2155 + }, + { + "epoch": 0.01, + "learning_rate": 9.999246858879303e-05, + "loss": 2.6561, + "step": 2160 + }, + { + "epoch": 0.01, + "learning_rate": 9.999243198509867e-05, + "loss": 2.6747, + "step": 2165 + }, + { + "epoch": 0.01, + "learning_rate": 9.999239529267704e-05, + "loss": 2.6599, + "step": 2170 + }, + { + "epoch": 0.01, + "learning_rate": 9.99923585115282e-05, + "loss": 2.5952, + "step": 2175 + }, + { + "epoch": 0.01, + "learning_rate": 9.999232164165224e-05, + "loss": 2.5862, + "step": 2180 + }, + { + "epoch": 0.01, + "learning_rate": 9.999228468304922e-05, + "loss": 2.7253, + "step": 2185 + }, + { + "epoch": 0.01, + "learning_rate": 9.999224763571917e-05, + "loss": 2.6569, + "step": 2190 + }, + { + "epoch": 0.01, + "learning_rate": 9.99922104996622e-05, + "loss": 2.7153, + "step": 2195 + }, + { + "epoch": 0.01, + "learning_rate": 9.999217327487837e-05, + "loss": 2.6923, + "step": 2200 + }, + { + "epoch": 0.01, + "learning_rate": 9.999213596136772e-05, + "loss": 2.6344, + "step": 2205 + }, + { + "epoch": 0.01, + "learning_rate": 9.999209855913034e-05, + "loss": 2.5422, + "step": 2210 + }, + { + "epoch": 0.01, + "learning_rate": 9.999206106816629e-05, + "loss": 2.6602, + "step": 2215 + }, + { + "epoch": 0.01, + "learning_rate": 9.999202348847562e-05, + "loss": 2.6296, + "step": 2220 + }, + { + "epoch": 0.01, + "learning_rate": 9.999198582005844e-05, + "loss": 2.7626, + "step": 2225 + }, + { + "epoch": 0.01, + "learning_rate": 9.999194806291477e-05, + "loss": 2.6839, + "step": 2230 + }, + { + "epoch": 0.01, + "learning_rate": 9.999191021704468e-05, + "loss": 2.6583, + "step": 2235 + }, + { + "epoch": 0.01, + "learning_rate": 9.999187228244828e-05, + "loss": 2.6298, + "step": 2240 + }, + { + "epoch": 0.01, + "learning_rate": 9.999183425912559e-05, + "loss": 2.4795, + "step": 2245 + }, + { + "epoch": 0.01, + "learning_rate": 9.999179614707672e-05, + "loss": 2.5798, + "step": 2250 + }, + { + "epoch": 0.01, + "learning_rate": 9.99917579463017e-05, + "loss": 2.7674, + "step": 2255 + }, + { + "epoch": 0.01, + "learning_rate": 9.999171965680061e-05, + "loss": 2.7022, + "step": 2260 + }, + { + "epoch": 0.01, + "learning_rate": 9.999168127857353e-05, + "loss": 2.7156, + "step": 2265 + }, + { + "epoch": 0.01, + "learning_rate": 9.99916428116205e-05, + "loss": 2.6714, + "step": 2270 + }, + { + "epoch": 0.01, + "learning_rate": 9.999160425594163e-05, + "loss": 2.643, + "step": 2275 + }, + { + "epoch": 0.01, + "learning_rate": 9.999156561153695e-05, + "loss": 2.6694, + "step": 2280 + }, + { + "epoch": 0.01, + "learning_rate": 9.999152687840656e-05, + "loss": 2.6299, + "step": 2285 + }, + { + "epoch": 0.01, + "learning_rate": 9.99914880565505e-05, + "loss": 2.6954, + "step": 2290 + }, + { + "epoch": 0.01, + "learning_rate": 9.999144914596886e-05, + "loss": 2.7215, + "step": 2295 + }, + { + "epoch": 0.01, + "learning_rate": 9.99914101466617e-05, + "loss": 2.691, + "step": 2300 + }, + { + "epoch": 0.01, + "learning_rate": 9.999137105862908e-05, + "loss": 2.6483, + "step": 2305 + }, + { + "epoch": 0.01, + "learning_rate": 9.999133188187109e-05, + "loss": 2.5671, + "step": 2310 + }, + { + "epoch": 0.01, + "learning_rate": 9.999129261638778e-05, + "loss": 2.635, + "step": 2315 + }, + { + "epoch": 0.01, + "learning_rate": 9.999125326217924e-05, + "loss": 2.6553, + "step": 2320 + }, + { + "epoch": 0.01, + "learning_rate": 9.999121381924552e-05, + "loss": 2.6948, + "step": 2325 + }, + { + "epoch": 0.01, + "learning_rate": 9.999117428758671e-05, + "loss": 2.7534, + "step": 2330 + }, + { + "epoch": 0.01, + "learning_rate": 9.999113466720284e-05, + "loss": 2.5862, + "step": 2335 + }, + { + "epoch": 0.01, + "learning_rate": 9.999109495809404e-05, + "loss": 2.6209, + "step": 2340 + }, + { + "epoch": 0.01, + "learning_rate": 9.999105516026034e-05, + "loss": 2.639, + "step": 2345 + }, + { + "epoch": 0.01, + "learning_rate": 9.999101527370183e-05, + "loss": 2.661, + "step": 2350 + }, + { + "epoch": 0.01, + "learning_rate": 9.999097529841855e-05, + "loss": 2.5943, + "step": 2355 + }, + { + "epoch": 0.01, + "learning_rate": 9.99909352344106e-05, + "loss": 2.5943, + "step": 2360 + }, + { + "epoch": 0.01, + "learning_rate": 9.999089508167804e-05, + "loss": 2.7074, + "step": 2365 + }, + { + "epoch": 0.01, + "learning_rate": 9.999085484022094e-05, + "loss": 2.7259, + "step": 2370 + }, + { + "epoch": 0.01, + "learning_rate": 9.99908145100394e-05, + "loss": 2.7394, + "step": 2375 + }, + { + "epoch": 0.01, + "learning_rate": 9.999077409113345e-05, + "loss": 2.6191, + "step": 2380 + }, + { + "epoch": 0.01, + "learning_rate": 9.999073358350318e-05, + "loss": 2.6671, + "step": 2385 + }, + { + "epoch": 0.01, + "learning_rate": 9.999069298714867e-05, + "loss": 2.5419, + "step": 2390 + }, + { + "epoch": 0.01, + "learning_rate": 9.999065230206997e-05, + "loss": 2.5665, + "step": 2395 + }, + { + "epoch": 0.01, + "learning_rate": 9.999061152826716e-05, + "loss": 2.7295, + "step": 2400 + }, + { + "epoch": 0.01, + "learning_rate": 9.999057066574034e-05, + "loss": 2.5256, + "step": 2405 + }, + { + "epoch": 0.01, + "learning_rate": 9.999052971448954e-05, + "loss": 2.7212, + "step": 2410 + }, + { + "epoch": 0.01, + "learning_rate": 9.999048867451486e-05, + "loss": 2.5893, + "step": 2415 + }, + { + "epoch": 0.01, + "learning_rate": 9.999044754581636e-05, + "loss": 2.5684, + "step": 2420 + }, + { + "epoch": 0.01, + "learning_rate": 9.999040632839413e-05, + "loss": 2.6042, + "step": 2425 + }, + { + "epoch": 0.01, + "learning_rate": 9.999036502224824e-05, + "loss": 2.6179, + "step": 2430 + }, + { + "epoch": 0.01, + "learning_rate": 9.999032362737873e-05, + "loss": 2.5587, + "step": 2435 + }, + { + "epoch": 0.01, + "learning_rate": 9.999028214378572e-05, + "loss": 2.7674, + "step": 2440 + }, + { + "epoch": 0.01, + "learning_rate": 9.999024057146924e-05, + "loss": 2.6018, + "step": 2445 + }, + { + "epoch": 0.01, + "learning_rate": 9.99901989104294e-05, + "loss": 2.6923, + "step": 2450 + }, + { + "epoch": 0.01, + "learning_rate": 9.999015716066626e-05, + "loss": 2.7107, + "step": 2455 + }, + { + "epoch": 0.01, + "learning_rate": 9.99901153221799e-05, + "loss": 2.6789, + "step": 2460 + }, + { + "epoch": 0.01, + "learning_rate": 9.999007339497037e-05, + "loss": 2.5812, + "step": 2465 + }, + { + "epoch": 0.01, + "learning_rate": 9.999003137903778e-05, + "loss": 2.4359, + "step": 2470 + }, + { + "epoch": 0.01, + "learning_rate": 9.998998927438217e-05, + "loss": 2.569, + "step": 2475 + }, + { + "epoch": 0.01, + "learning_rate": 9.998994708100365e-05, + "loss": 2.5509, + "step": 2480 + }, + { + "epoch": 0.01, + "learning_rate": 9.998990479890227e-05, + "loss": 2.523, + "step": 2485 + }, + { + "epoch": 0.01, + "learning_rate": 9.998986242807811e-05, + "loss": 2.731, + "step": 2490 + }, + { + "epoch": 0.01, + "learning_rate": 9.998981996853125e-05, + "loss": 2.626, + "step": 2495 + }, + { + "epoch": 0.01, + "learning_rate": 9.998977742026176e-05, + "loss": 2.5565, + "step": 2500 + }, + { + "epoch": 0.01, + "learning_rate": 9.998973478326972e-05, + "loss": 2.6879, + "step": 2505 + }, + { + "epoch": 0.01, + "learning_rate": 9.998969205755519e-05, + "loss": 2.6388, + "step": 2510 + }, + { + "epoch": 0.01, + "learning_rate": 9.998964924311827e-05, + "loss": 2.5264, + "step": 2515 + }, + { + "epoch": 0.01, + "learning_rate": 9.998960633995904e-05, + "loss": 2.7505, + "step": 2520 + }, + { + "epoch": 0.01, + "learning_rate": 9.998956334807755e-05, + "loss": 2.585, + "step": 2525 + }, + { + "epoch": 0.01, + "learning_rate": 9.998952026747389e-05, + "loss": 2.6648, + "step": 2530 + }, + { + "epoch": 0.01, + "learning_rate": 9.998947709814812e-05, + "loss": 2.6564, + "step": 2535 + }, + { + "epoch": 0.01, + "learning_rate": 9.998943384010033e-05, + "loss": 2.6261, + "step": 2540 + }, + { + "epoch": 0.01, + "learning_rate": 9.998939049333062e-05, + "loss": 2.7051, + "step": 2545 + }, + { + "epoch": 0.01, + "learning_rate": 9.998934705783902e-05, + "loss": 2.7724, + "step": 2550 + }, + { + "epoch": 0.01, + "learning_rate": 9.998930353362565e-05, + "loss": 2.6627, + "step": 2555 + }, + { + "epoch": 0.01, + "learning_rate": 9.998925992069057e-05, + "loss": 2.6003, + "step": 2560 + }, + { + "epoch": 0.01, + "learning_rate": 9.998921621903386e-05, + "loss": 2.5939, + "step": 2565 + }, + { + "epoch": 0.01, + "learning_rate": 9.998917242865558e-05, + "loss": 2.663, + "step": 2570 + }, + { + "epoch": 0.01, + "learning_rate": 9.998912854955583e-05, + "loss": 2.5347, + "step": 2575 + }, + { + "epoch": 0.01, + "learning_rate": 9.998908458173466e-05, + "loss": 2.6419, + "step": 2580 + }, + { + "epoch": 0.01, + "learning_rate": 9.998904052519219e-05, + "loss": 2.5139, + "step": 2585 + }, + { + "epoch": 0.01, + "learning_rate": 9.998899637992847e-05, + "loss": 2.6887, + "step": 2590 + }, + { + "epoch": 0.01, + "learning_rate": 9.998895214594358e-05, + "loss": 2.6063, + "step": 2595 + }, + { + "epoch": 0.01, + "learning_rate": 9.998890782323761e-05, + "loss": 2.5605, + "step": 2600 + }, + { + "epoch": 0.01, + "learning_rate": 9.998886341181061e-05, + "loss": 2.5954, + "step": 2605 + }, + { + "epoch": 0.01, + "learning_rate": 9.998881891166272e-05, + "loss": 2.7362, + "step": 2610 + }, + { + "epoch": 0.01, + "learning_rate": 9.998877432279395e-05, + "loss": 2.603, + "step": 2615 + }, + { + "epoch": 0.01, + "learning_rate": 9.998872964520441e-05, + "loss": 2.628, + "step": 2620 + }, + { + "epoch": 0.01, + "learning_rate": 9.998868487889419e-05, + "loss": 2.6347, + "step": 2625 + }, + { + "epoch": 0.01, + "learning_rate": 9.998864002386334e-05, + "loss": 2.6041, + "step": 2630 + }, + { + "epoch": 0.01, + "learning_rate": 9.998859508011197e-05, + "loss": 2.5921, + "step": 2635 + }, + { + "epoch": 0.01, + "learning_rate": 9.998855004764015e-05, + "loss": 2.7511, + "step": 2640 + }, + { + "epoch": 0.01, + "learning_rate": 9.998850492644793e-05, + "loss": 2.6989, + "step": 2645 + }, + { + "epoch": 0.01, + "learning_rate": 9.998845971653545e-05, + "loss": 2.6483, + "step": 2650 + }, + { + "epoch": 0.01, + "learning_rate": 9.998841441790274e-05, + "loss": 2.7287, + "step": 2655 + }, + { + "epoch": 0.01, + "learning_rate": 9.998836903054988e-05, + "loss": 2.5901, + "step": 2660 + }, + { + "epoch": 0.01, + "learning_rate": 9.9988323554477e-05, + "loss": 2.5738, + "step": 2665 + }, + { + "epoch": 0.01, + "learning_rate": 9.998827798968413e-05, + "loss": 2.603, + "step": 2670 + }, + { + "epoch": 0.01, + "learning_rate": 9.998823233617138e-05, + "loss": 2.5445, + "step": 2675 + }, + { + "epoch": 0.01, + "learning_rate": 9.99881865939388e-05, + "loss": 2.5991, + "step": 2680 + }, + { + "epoch": 0.01, + "learning_rate": 9.998814076298652e-05, + "loss": 2.6154, + "step": 2685 + }, + { + "epoch": 0.01, + "learning_rate": 9.998809484331458e-05, + "loss": 2.7085, + "step": 2690 + }, + { + "epoch": 0.01, + "learning_rate": 9.998804883492306e-05, + "loss": 2.6275, + "step": 2695 + }, + { + "epoch": 0.01, + "learning_rate": 9.998800273781207e-05, + "loss": 2.5713, + "step": 2700 + }, + { + "epoch": 0.01, + "learning_rate": 9.998795655198167e-05, + "loss": 2.6356, + "step": 2705 + }, + { + "epoch": 0.01, + "learning_rate": 9.998791027743197e-05, + "loss": 2.7083, + "step": 2710 + }, + { + "epoch": 0.01, + "learning_rate": 9.998786391416302e-05, + "loss": 2.6576, + "step": 2715 + }, + { + "epoch": 0.01, + "learning_rate": 9.99878174621749e-05, + "loss": 2.5845, + "step": 2720 + }, + { + "epoch": 0.01, + "learning_rate": 9.998777092146771e-05, + "loss": 2.5238, + "step": 2725 + }, + { + "epoch": 0.01, + "learning_rate": 9.998772429204153e-05, + "loss": 2.7617, + "step": 2730 + }, + { + "epoch": 0.01, + "learning_rate": 9.998767757389646e-05, + "loss": 2.6192, + "step": 2735 + }, + { + "epoch": 0.01, + "learning_rate": 9.998763076703255e-05, + "loss": 2.5263, + "step": 2740 + }, + { + "epoch": 0.01, + "learning_rate": 9.99875838714499e-05, + "loss": 2.5755, + "step": 2745 + }, + { + "epoch": 0.01, + "learning_rate": 9.998753688714859e-05, + "loss": 2.6263, + "step": 2750 + }, + { + "epoch": 0.01, + "learning_rate": 9.998748981412871e-05, + "loss": 2.6535, + "step": 2755 + }, + { + "epoch": 0.01, + "learning_rate": 9.998744265239033e-05, + "loss": 2.6977, + "step": 2760 + }, + { + "epoch": 0.01, + "learning_rate": 9.998739540193356e-05, + "loss": 2.5897, + "step": 2765 + }, + { + "epoch": 0.01, + "learning_rate": 9.998734806275844e-05, + "loss": 2.5941, + "step": 2770 + }, + { + "epoch": 0.01, + "learning_rate": 9.99873006348651e-05, + "loss": 2.5502, + "step": 2775 + }, + { + "epoch": 0.01, + "learning_rate": 9.998725311825359e-05, + "loss": 2.5913, + "step": 2780 + }, + { + "epoch": 0.01, + "learning_rate": 9.998720551292401e-05, + "loss": 2.6962, + "step": 2785 + }, + { + "epoch": 0.01, + "learning_rate": 9.998715781887645e-05, + "loss": 2.5815, + "step": 2790 + }, + { + "epoch": 0.01, + "learning_rate": 9.998711003611099e-05, + "loss": 2.6143, + "step": 2795 + }, + { + "epoch": 0.01, + "learning_rate": 9.99870621646277e-05, + "loss": 2.6835, + "step": 2800 + }, + { + "epoch": 0.01, + "learning_rate": 9.998701420442667e-05, + "loss": 2.6122, + "step": 2805 + }, + { + "epoch": 0.01, + "learning_rate": 9.998696615550802e-05, + "loss": 2.463, + "step": 2810 + }, + { + "epoch": 0.01, + "learning_rate": 9.998691801787178e-05, + "loss": 2.5916, + "step": 2815 + }, + { + "epoch": 0.01, + "learning_rate": 9.998686979151808e-05, + "loss": 2.7003, + "step": 2820 + }, + { + "epoch": 0.01, + "learning_rate": 9.998682147644698e-05, + "loss": 2.6212, + "step": 2825 + }, + { + "epoch": 0.01, + "learning_rate": 9.998677307265858e-05, + "loss": 2.6151, + "step": 2830 + }, + { + "epoch": 0.01, + "learning_rate": 9.998672458015295e-05, + "loss": 2.749, + "step": 2835 + }, + { + "epoch": 0.01, + "learning_rate": 9.998667599893019e-05, + "loss": 2.4858, + "step": 2840 + }, + { + "epoch": 0.01, + "learning_rate": 9.998662732899039e-05, + "loss": 2.4697, + "step": 2845 + }, + { + "epoch": 0.01, + "learning_rate": 9.998657857033363e-05, + "loss": 2.568, + "step": 2850 + }, + { + "epoch": 0.01, + "learning_rate": 9.998652972295998e-05, + "loss": 2.5946, + "step": 2855 + }, + { + "epoch": 0.01, + "learning_rate": 9.998648078686955e-05, + "loss": 2.6567, + "step": 2860 + }, + { + "epoch": 0.01, + "learning_rate": 9.998643176206241e-05, + "loss": 2.5577, + "step": 2865 + }, + { + "epoch": 0.01, + "learning_rate": 9.998638264853868e-05, + "loss": 2.6468, + "step": 2870 + }, + { + "epoch": 0.01, + "learning_rate": 9.99863334462984e-05, + "loss": 2.5586, + "step": 2875 + }, + { + "epoch": 0.01, + "learning_rate": 9.99862841553417e-05, + "loss": 2.5384, + "step": 2880 + }, + { + "epoch": 0.01, + "learning_rate": 9.998623477566865e-05, + "loss": 2.4253, + "step": 2885 + }, + { + "epoch": 0.01, + "learning_rate": 9.99861853072793e-05, + "loss": 2.5088, + "step": 2890 + }, + { + "epoch": 0.01, + "learning_rate": 9.99861357501738e-05, + "loss": 2.6544, + "step": 2895 + }, + { + "epoch": 0.01, + "learning_rate": 9.99860861043522e-05, + "loss": 2.7034, + "step": 2900 + }, + { + "epoch": 0.01, + "learning_rate": 9.998603636981462e-05, + "loss": 2.5884, + "step": 2905 + }, + { + "epoch": 0.01, + "learning_rate": 9.99859865465611e-05, + "loss": 2.6765, + "step": 2910 + }, + { + "epoch": 0.01, + "learning_rate": 9.998593663459178e-05, + "loss": 2.6262, + "step": 2915 + }, + { + "epoch": 0.01, + "learning_rate": 9.998588663390671e-05, + "loss": 2.6726, + "step": 2920 + }, + { + "epoch": 0.01, + "learning_rate": 9.9985836544506e-05, + "loss": 2.5509, + "step": 2925 + }, + { + "epoch": 0.01, + "learning_rate": 9.998578636638973e-05, + "loss": 2.6527, + "step": 2930 + }, + { + "epoch": 0.01, + "learning_rate": 9.998573609955799e-05, + "loss": 2.6621, + "step": 2935 + }, + { + "epoch": 0.01, + "learning_rate": 9.998568574401088e-05, + "loss": 2.6716, + "step": 2940 + }, + { + "epoch": 0.01, + "learning_rate": 9.998563529974848e-05, + "loss": 2.5777, + "step": 2945 + }, + { + "epoch": 0.01, + "learning_rate": 9.998558476677085e-05, + "loss": 2.6027, + "step": 2950 + }, + { + "epoch": 0.01, + "learning_rate": 9.998553414507814e-05, + "loss": 2.5961, + "step": 2955 + }, + { + "epoch": 0.01, + "learning_rate": 9.99854834346704e-05, + "loss": 2.6545, + "step": 2960 + }, + { + "epoch": 0.01, + "learning_rate": 9.998543263554772e-05, + "loss": 2.685, + "step": 2965 + }, + { + "epoch": 0.01, + "learning_rate": 9.998538174771021e-05, + "loss": 2.5673, + "step": 2970 + }, + { + "epoch": 0.01, + "learning_rate": 9.998533077115795e-05, + "loss": 2.6122, + "step": 2975 + }, + { + "epoch": 0.01, + "learning_rate": 9.998527970589102e-05, + "loss": 2.5136, + "step": 2980 + }, + { + "epoch": 0.01, + "learning_rate": 9.998522855190953e-05, + "loss": 2.551, + "step": 2985 + }, + { + "epoch": 0.01, + "learning_rate": 9.998517730921356e-05, + "loss": 2.5511, + "step": 2990 + }, + { + "epoch": 0.01, + "learning_rate": 9.998512597780321e-05, + "loss": 2.5867, + "step": 2995 + }, + { + "epoch": 0.01, + "learning_rate": 9.998507455767855e-05, + "loss": 2.7336, + "step": 3000 + }, + { + "epoch": 0.01, + "learning_rate": 9.998502304883969e-05, + "loss": 2.6387, + "step": 3005 + }, + { + "epoch": 0.01, + "learning_rate": 9.998497145128672e-05, + "loss": 2.6661, + "step": 3010 + }, + { + "epoch": 0.01, + "learning_rate": 9.998491976501972e-05, + "loss": 2.6729, + "step": 3015 + }, + { + "epoch": 0.01, + "learning_rate": 9.998486799003878e-05, + "loss": 2.5875, + "step": 3020 + }, + { + "epoch": 0.01, + "learning_rate": 9.998481612634403e-05, + "loss": 2.527, + "step": 3025 + }, + { + "epoch": 0.01, + "learning_rate": 9.998476417393551e-05, + "loss": 2.5996, + "step": 3030 + }, + { + "epoch": 0.01, + "learning_rate": 9.998471213281334e-05, + "loss": 2.5783, + "step": 3035 + }, + { + "epoch": 0.01, + "learning_rate": 9.998466000297761e-05, + "loss": 2.7528, + "step": 3040 + }, + { + "epoch": 0.01, + "learning_rate": 9.998460778442841e-05, + "loss": 2.6778, + "step": 3045 + }, + { + "epoch": 0.01, + "learning_rate": 9.998455547716583e-05, + "loss": 2.5096, + "step": 3050 + }, + { + "epoch": 0.01, + "learning_rate": 9.998450308118997e-05, + "loss": 2.7084, + "step": 3055 + }, + { + "epoch": 0.01, + "learning_rate": 9.998445059650092e-05, + "loss": 2.6099, + "step": 3060 + }, + { + "epoch": 0.01, + "learning_rate": 9.998439802309877e-05, + "loss": 2.5642, + "step": 3065 + }, + { + "epoch": 0.01, + "learning_rate": 9.998434536098362e-05, + "loss": 2.4893, + "step": 3070 + }, + { + "epoch": 0.01, + "learning_rate": 9.998429261015554e-05, + "loss": 2.5007, + "step": 3075 + }, + { + "epoch": 0.01, + "learning_rate": 9.998423977061467e-05, + "loss": 2.5669, + "step": 3080 + }, + { + "epoch": 0.01, + "learning_rate": 9.998418684236106e-05, + "loss": 2.6895, + "step": 3085 + }, + { + "epoch": 0.01, + "learning_rate": 9.998413382539483e-05, + "loss": 2.661, + "step": 3090 + }, + { + "epoch": 0.01, + "learning_rate": 9.998408071971605e-05, + "loss": 2.5864, + "step": 3095 + }, + { + "epoch": 0.01, + "learning_rate": 9.998402752532483e-05, + "loss": 2.5277, + "step": 3100 + }, + { + "epoch": 0.01, + "learning_rate": 9.998397424222127e-05, + "loss": 2.6087, + "step": 3105 + }, + { + "epoch": 0.01, + "learning_rate": 9.998392087040546e-05, + "loss": 2.5897, + "step": 3110 + }, + { + "epoch": 0.01, + "learning_rate": 9.998386740987748e-05, + "loss": 2.6086, + "step": 3115 + }, + { + "epoch": 0.01, + "learning_rate": 9.998381386063746e-05, + "loss": 2.6937, + "step": 3120 + }, + { + "epoch": 0.01, + "learning_rate": 9.998376022268546e-05, + "loss": 2.5386, + "step": 3125 + }, + { + "epoch": 0.01, + "learning_rate": 9.998370649602158e-05, + "loss": 2.4679, + "step": 3130 + }, + { + "epoch": 0.01, + "learning_rate": 9.998365268064593e-05, + "loss": 2.5384, + "step": 3135 + }, + { + "epoch": 0.01, + "learning_rate": 9.998359877655859e-05, + "loss": 2.5377, + "step": 3140 + }, + { + "epoch": 0.01, + "learning_rate": 9.998354478375967e-05, + "loss": 2.6301, + "step": 3145 + }, + { + "epoch": 0.01, + "learning_rate": 9.998349070224926e-05, + "loss": 2.6431, + "step": 3150 + }, + { + "epoch": 0.01, + "learning_rate": 9.998343653202747e-05, + "loss": 2.5421, + "step": 3155 + }, + { + "epoch": 0.01, + "learning_rate": 9.998338227309436e-05, + "loss": 2.622, + "step": 3160 + }, + { + "epoch": 0.01, + "learning_rate": 9.998332792545004e-05, + "loss": 2.5847, + "step": 3165 + }, + { + "epoch": 0.01, + "learning_rate": 9.998327348909465e-05, + "loss": 2.5546, + "step": 3170 + }, + { + "epoch": 0.01, + "learning_rate": 9.998321896402822e-05, + "loss": 2.6121, + "step": 3175 + }, + { + "epoch": 0.01, + "learning_rate": 9.998316435025089e-05, + "loss": 2.5354, + "step": 3180 + }, + { + "epoch": 0.01, + "learning_rate": 9.998310964776275e-05, + "loss": 2.5823, + "step": 3185 + }, + { + "epoch": 0.01, + "learning_rate": 9.998305485656388e-05, + "loss": 2.6302, + "step": 3190 + }, + { + "epoch": 0.01, + "learning_rate": 9.998299997665438e-05, + "loss": 2.573, + "step": 3195 + }, + { + "epoch": 0.01, + "learning_rate": 9.998294500803438e-05, + "loss": 2.6035, + "step": 3200 + }, + { + "epoch": 0.01, + "learning_rate": 9.998288995070396e-05, + "loss": 2.6334, + "step": 3205 + }, + { + "epoch": 0.01, + "learning_rate": 9.99828348046632e-05, + "loss": 2.7143, + "step": 3210 + }, + { + "epoch": 0.01, + "learning_rate": 9.99827795699122e-05, + "loss": 2.5722, + "step": 3215 + }, + { + "epoch": 0.01, + "learning_rate": 9.998272424645108e-05, + "loss": 2.6684, + "step": 3220 + }, + { + "epoch": 0.01, + "learning_rate": 9.998266883427992e-05, + "loss": 2.6468, + "step": 3225 + }, + { + "epoch": 0.01, + "learning_rate": 9.998261333339882e-05, + "loss": 2.5105, + "step": 3230 + }, + { + "epoch": 0.01, + "learning_rate": 9.998255774380789e-05, + "loss": 2.6549, + "step": 3235 + }, + { + "epoch": 0.01, + "learning_rate": 9.998250206550722e-05, + "loss": 2.7084, + "step": 3240 + }, + { + "epoch": 0.01, + "learning_rate": 9.998244629849692e-05, + "loss": 2.6089, + "step": 3245 + }, + { + "epoch": 0.01, + "learning_rate": 9.998239044277706e-05, + "loss": 2.5564, + "step": 3250 + }, + { + "epoch": 0.01, + "learning_rate": 9.998233449834777e-05, + "loss": 2.5973, + "step": 3255 + }, + { + "epoch": 0.01, + "learning_rate": 9.998227846520914e-05, + "loss": 2.7198, + "step": 3260 + }, + { + "epoch": 0.01, + "learning_rate": 9.998222234336126e-05, + "loss": 2.6149, + "step": 3265 + }, + { + "epoch": 0.01, + "learning_rate": 9.998216613280423e-05, + "loss": 2.6862, + "step": 3270 + }, + { + "epoch": 0.01, + "learning_rate": 9.998210983353817e-05, + "loss": 2.5603, + "step": 3275 + }, + { + "epoch": 0.01, + "learning_rate": 9.998205344556317e-05, + "loss": 2.6005, + "step": 3280 + }, + { + "epoch": 0.01, + "learning_rate": 9.998199696887931e-05, + "loss": 2.4267, + "step": 3285 + }, + { + "epoch": 0.01, + "learning_rate": 9.998194040348672e-05, + "loss": 2.5448, + "step": 3290 + }, + { + "epoch": 0.01, + "learning_rate": 9.998188374938548e-05, + "loss": 2.6667, + "step": 3295 + }, + { + "epoch": 0.01, + "learning_rate": 9.99818270065757e-05, + "loss": 2.7352, + "step": 3300 + }, + { + "epoch": 0.01, + "learning_rate": 9.998177017505748e-05, + "loss": 2.6618, + "step": 3305 + }, + { + "epoch": 0.01, + "learning_rate": 9.998171325483092e-05, + "loss": 2.5272, + "step": 3310 + }, + { + "epoch": 0.01, + "learning_rate": 9.998165624589613e-05, + "loss": 2.4815, + "step": 3315 + }, + { + "epoch": 0.01, + "learning_rate": 9.998159914825318e-05, + "loss": 2.574, + "step": 3320 + }, + { + "epoch": 0.01, + "learning_rate": 9.99815419619022e-05, + "loss": 2.6554, + "step": 3325 + }, + { + "epoch": 0.01, + "learning_rate": 9.998148468684329e-05, + "loss": 2.5903, + "step": 3330 + }, + { + "epoch": 0.01, + "learning_rate": 9.998142732307655e-05, + "loss": 2.6075, + "step": 3335 + }, + { + "epoch": 0.01, + "learning_rate": 9.998136987060207e-05, + "loss": 2.6008, + "step": 3340 + }, + { + "epoch": 0.01, + "learning_rate": 9.998131232941997e-05, + "loss": 2.6902, + "step": 3345 + }, + { + "epoch": 0.01, + "learning_rate": 9.998125469953032e-05, + "loss": 2.5969, + "step": 3350 + }, + { + "epoch": 0.01, + "learning_rate": 9.998119698093326e-05, + "loss": 2.5934, + "step": 3355 + }, + { + "epoch": 0.01, + "learning_rate": 9.998113917362888e-05, + "loss": 2.5464, + "step": 3360 + }, + { + "epoch": 0.01, + "learning_rate": 9.998108127761728e-05, + "loss": 2.6413, + "step": 3365 + }, + { + "epoch": 0.01, + "learning_rate": 9.998102329289854e-05, + "loss": 2.5005, + "step": 3370 + }, + { + "epoch": 0.01, + "learning_rate": 9.99809652194728e-05, + "loss": 2.5799, + "step": 3375 + }, + { + "epoch": 0.01, + "learning_rate": 9.998090705734016e-05, + "loss": 2.55, + "step": 3380 + }, + { + "epoch": 0.01, + "learning_rate": 9.998084880650069e-05, + "loss": 2.509, + "step": 3385 + }, + { + "epoch": 0.01, + "learning_rate": 9.998079046695452e-05, + "loss": 2.63, + "step": 3390 + }, + { + "epoch": 0.01, + "learning_rate": 9.998073203870175e-05, + "loss": 2.6122, + "step": 3395 + }, + { + "epoch": 0.01, + "learning_rate": 9.998067352174249e-05, + "loss": 2.5986, + "step": 3400 + }, + { + "epoch": 0.01, + "learning_rate": 9.998061491607683e-05, + "loss": 2.6486, + "step": 3405 + }, + { + "epoch": 0.01, + "learning_rate": 9.998055622170489e-05, + "loss": 2.4694, + "step": 3410 + }, + { + "epoch": 0.01, + "learning_rate": 9.998049743862674e-05, + "loss": 2.6475, + "step": 3415 + }, + { + "epoch": 0.01, + "learning_rate": 9.998043856684251e-05, + "loss": 2.418, + "step": 3420 + }, + { + "epoch": 0.01, + "learning_rate": 9.998037960635231e-05, + "loss": 2.4598, + "step": 3425 + }, + { + "epoch": 0.01, + "learning_rate": 9.998032055715625e-05, + "loss": 2.5016, + "step": 3430 + }, + { + "epoch": 0.01, + "learning_rate": 9.998026141925442e-05, + "loss": 2.6466, + "step": 3435 + }, + { + "epoch": 0.01, + "learning_rate": 9.998020219264692e-05, + "loss": 2.7314, + "step": 3440 + }, + { + "epoch": 0.01, + "learning_rate": 9.998014287733385e-05, + "loss": 2.5649, + "step": 3445 + }, + { + "epoch": 0.01, + "learning_rate": 9.998008347331535e-05, + "loss": 2.6389, + "step": 3450 + }, + { + "epoch": 0.01, + "learning_rate": 9.99800239805915e-05, + "loss": 2.5555, + "step": 3455 + }, + { + "epoch": 0.01, + "learning_rate": 9.99799643991624e-05, + "loss": 2.6308, + "step": 3460 + }, + { + "epoch": 0.01, + "learning_rate": 9.997990472902816e-05, + "loss": 2.6593, + "step": 3465 + }, + { + "epoch": 0.01, + "learning_rate": 9.997984497018888e-05, + "loss": 2.4938, + "step": 3470 + }, + { + "epoch": 0.01, + "learning_rate": 9.99797851226447e-05, + "loss": 2.6341, + "step": 3475 + }, + { + "epoch": 0.01, + "learning_rate": 9.997972518639569e-05, + "loss": 2.4717, + "step": 3480 + }, + { + "epoch": 0.01, + "learning_rate": 9.997966516144197e-05, + "loss": 2.584, + "step": 3485 + }, + { + "epoch": 0.01, + "learning_rate": 9.997960504778366e-05, + "loss": 2.6754, + "step": 3490 + }, + { + "epoch": 0.01, + "learning_rate": 9.997954484542084e-05, + "loss": 2.5988, + "step": 3495 + }, + { + "epoch": 0.01, + "learning_rate": 9.997948455435362e-05, + "loss": 2.5453, + "step": 3500 + }, + { + "epoch": 0.01, + "learning_rate": 9.997942417458212e-05, + "loss": 2.6509, + "step": 3505 + }, + { + "epoch": 0.01, + "learning_rate": 9.997936370610645e-05, + "loss": 2.4562, + "step": 3510 + }, + { + "epoch": 0.01, + "learning_rate": 9.997930314892672e-05, + "loss": 2.6292, + "step": 3515 + }, + { + "epoch": 0.01, + "learning_rate": 9.997924250304301e-05, + "loss": 2.5629, + "step": 3520 + }, + { + "epoch": 0.01, + "learning_rate": 9.997918176845546e-05, + "loss": 2.597, + "step": 3525 + }, + { + "epoch": 0.01, + "learning_rate": 9.997912094516415e-05, + "loss": 2.533, + "step": 3530 + }, + { + "epoch": 0.01, + "learning_rate": 9.99790600331692e-05, + "loss": 2.6229, + "step": 3535 + }, + { + "epoch": 0.01, + "learning_rate": 9.997899903247072e-05, + "loss": 2.4839, + "step": 3540 + }, + { + "epoch": 0.01, + "learning_rate": 9.997893794306883e-05, + "loss": 2.4875, + "step": 3545 + }, + { + "epoch": 0.01, + "learning_rate": 9.997887676496362e-05, + "loss": 2.5201, + "step": 3550 + }, + { + "epoch": 0.01, + "learning_rate": 9.99788154981552e-05, + "loss": 2.4486, + "step": 3555 + }, + { + "epoch": 0.01, + "learning_rate": 9.997875414264369e-05, + "loss": 2.5383, + "step": 3560 + }, + { + "epoch": 0.01, + "learning_rate": 9.997869269842919e-05, + "loss": 2.4835, + "step": 3565 + }, + { + "epoch": 0.01, + "learning_rate": 9.99786311655118e-05, + "loss": 2.5129, + "step": 3570 + }, + { + "epoch": 0.01, + "learning_rate": 9.997856954389166e-05, + "loss": 2.5588, + "step": 3575 + }, + { + "epoch": 0.01, + "learning_rate": 9.997850783356884e-05, + "loss": 2.5936, + "step": 3580 + }, + { + "epoch": 0.01, + "learning_rate": 9.99784460345435e-05, + "loss": 2.5665, + "step": 3585 + }, + { + "epoch": 0.01, + "learning_rate": 9.997838414681569e-05, + "loss": 2.6724, + "step": 3590 + }, + { + "epoch": 0.01, + "learning_rate": 9.997832217038555e-05, + "loss": 2.5934, + "step": 3595 + }, + { + "epoch": 0.01, + "learning_rate": 9.99782601052532e-05, + "loss": 2.6039, + "step": 3600 + }, + { + "epoch": 0.01, + "learning_rate": 9.997819795141874e-05, + "loss": 2.6738, + "step": 3605 + }, + { + "epoch": 0.01, + "learning_rate": 9.997813570888225e-05, + "loss": 2.6536, + "step": 3610 + }, + { + "epoch": 0.01, + "learning_rate": 9.997807337764391e-05, + "loss": 2.5388, + "step": 3615 + }, + { + "epoch": 0.01, + "learning_rate": 9.997801095770378e-05, + "loss": 2.5633, + "step": 3620 + }, + { + "epoch": 0.01, + "learning_rate": 9.997794844906196e-05, + "loss": 2.5103, + "step": 3625 + }, + { + "epoch": 0.01, + "learning_rate": 9.997788585171861e-05, + "loss": 2.4752, + "step": 3630 + }, + { + "epoch": 0.01, + "learning_rate": 9.99778231656738e-05, + "loss": 2.5149, + "step": 3635 + }, + { + "epoch": 0.01, + "learning_rate": 9.997776039092765e-05, + "loss": 2.5571, + "step": 3640 + }, + { + "epoch": 0.01, + "learning_rate": 9.99776975274803e-05, + "loss": 2.6113, + "step": 3645 + }, + { + "epoch": 0.01, + "learning_rate": 9.99776345753318e-05, + "loss": 2.5928, + "step": 3650 + }, + { + "epoch": 0.01, + "learning_rate": 9.997757153448233e-05, + "loss": 2.7143, + "step": 3655 + }, + { + "epoch": 0.01, + "learning_rate": 9.997750840493196e-05, + "loss": 2.5958, + "step": 3660 + }, + { + "epoch": 0.01, + "learning_rate": 9.997744518668081e-05, + "loss": 2.5555, + "step": 3665 + }, + { + "epoch": 0.01, + "learning_rate": 9.9977381879729e-05, + "loss": 2.5865, + "step": 3670 + }, + { + "epoch": 0.01, + "learning_rate": 9.997731848407663e-05, + "loss": 2.4613, + "step": 3675 + }, + { + "epoch": 0.01, + "learning_rate": 9.997725499972383e-05, + "loss": 2.6457, + "step": 3680 + }, + { + "epoch": 0.01, + "learning_rate": 9.99771914266707e-05, + "loss": 2.4966, + "step": 3685 + }, + { + "epoch": 0.01, + "learning_rate": 9.997712776491735e-05, + "loss": 2.4952, + "step": 3690 + }, + { + "epoch": 0.01, + "learning_rate": 9.99770640144639e-05, + "loss": 2.565, + "step": 3695 + }, + { + "epoch": 0.01, + "learning_rate": 9.997700017531046e-05, + "loss": 2.5669, + "step": 3700 + }, + { + "epoch": 0.01, + "learning_rate": 9.997693624745716e-05, + "loss": 2.6552, + "step": 3705 + }, + { + "epoch": 0.01, + "learning_rate": 9.997687223090409e-05, + "loss": 2.6181, + "step": 3710 + }, + { + "epoch": 0.01, + "learning_rate": 9.997680812565138e-05, + "loss": 2.5041, + "step": 3715 + }, + { + "epoch": 0.01, + "learning_rate": 9.997674393169911e-05, + "loss": 2.5534, + "step": 3720 + }, + { + "epoch": 0.01, + "learning_rate": 9.997667964904744e-05, + "loss": 2.4774, + "step": 3725 + }, + { + "epoch": 0.01, + "learning_rate": 9.997661527769646e-05, + "loss": 2.5612, + "step": 3730 + }, + { + "epoch": 0.01, + "learning_rate": 9.99765508176463e-05, + "loss": 2.5632, + "step": 3735 + }, + { + "epoch": 0.01, + "learning_rate": 9.997648626889704e-05, + "loss": 2.4512, + "step": 3740 + }, + { + "epoch": 0.01, + "learning_rate": 9.997642163144884e-05, + "loss": 2.5344, + "step": 3745 + }, + { + "epoch": 0.01, + "learning_rate": 9.997635690530179e-05, + "loss": 2.4266, + "step": 3750 + }, + { + "epoch": 0.01, + "learning_rate": 9.997629209045598e-05, + "loss": 2.596, + "step": 3755 + }, + { + "epoch": 0.01, + "learning_rate": 9.997622718691157e-05, + "loss": 2.6278, + "step": 3760 + }, + { + "epoch": 0.01, + "learning_rate": 9.997616219466866e-05, + "loss": 2.4567, + "step": 3765 + }, + { + "epoch": 0.01, + "learning_rate": 9.997609711372736e-05, + "loss": 2.558, + "step": 3770 + }, + { + "epoch": 0.01, + "learning_rate": 9.997603194408779e-05, + "loss": 2.5814, + "step": 3775 + }, + { + "epoch": 0.01, + "learning_rate": 9.997596668575006e-05, + "loss": 2.6373, + "step": 3780 + }, + { + "epoch": 0.01, + "learning_rate": 9.997590133871428e-05, + "loss": 2.663, + "step": 3785 + }, + { + "epoch": 0.01, + "learning_rate": 9.99758359029806e-05, + "loss": 2.5524, + "step": 3790 + }, + { + "epoch": 0.01, + "learning_rate": 9.99757703785491e-05, + "loss": 2.5404, + "step": 3795 + }, + { + "epoch": 0.01, + "learning_rate": 9.99757047654199e-05, + "loss": 2.6005, + "step": 3800 + }, + { + "epoch": 0.01, + "learning_rate": 9.997563906359313e-05, + "loss": 2.5176, + "step": 3805 + }, + { + "epoch": 0.01, + "learning_rate": 9.99755732730689e-05, + "loss": 2.5875, + "step": 3810 + }, + { + "epoch": 0.01, + "learning_rate": 9.997550739384733e-05, + "loss": 2.5158, + "step": 3815 + }, + { + "epoch": 0.01, + "learning_rate": 9.997544142592853e-05, + "loss": 2.6811, + "step": 3820 + }, + { + "epoch": 0.01, + "learning_rate": 9.997537536931262e-05, + "loss": 2.7012, + "step": 3825 + }, + { + "epoch": 0.01, + "learning_rate": 9.997530922399972e-05, + "loss": 2.6801, + "step": 3830 + }, + { + "epoch": 0.01, + "learning_rate": 9.997524298998995e-05, + "loss": 2.6548, + "step": 3835 + }, + { + "epoch": 0.01, + "learning_rate": 9.997517666728343e-05, + "loss": 2.6168, + "step": 3840 + }, + { + "epoch": 0.01, + "learning_rate": 9.997511025588026e-05, + "loss": 2.6905, + "step": 3845 + }, + { + "epoch": 0.01, + "learning_rate": 9.997504375578058e-05, + "loss": 2.495, + "step": 3850 + }, + { + "epoch": 0.01, + "learning_rate": 9.99749771669845e-05, + "loss": 2.5876, + "step": 3855 + }, + { + "epoch": 0.01, + "learning_rate": 9.997491048949213e-05, + "loss": 2.504, + "step": 3860 + }, + { + "epoch": 0.01, + "learning_rate": 9.997484372330357e-05, + "loss": 2.6118, + "step": 3865 + }, + { + "epoch": 0.01, + "learning_rate": 9.997477686841899e-05, + "loss": 2.6117, + "step": 3870 + }, + { + "epoch": 0.01, + "learning_rate": 9.997470992483848e-05, + "loss": 2.542, + "step": 3875 + }, + { + "epoch": 0.01, + "learning_rate": 9.997464289256216e-05, + "loss": 2.6754, + "step": 3880 + }, + { + "epoch": 0.01, + "learning_rate": 9.997457577159013e-05, + "loss": 2.6353, + "step": 3885 + }, + { + "epoch": 0.01, + "learning_rate": 9.997450856192254e-05, + "loss": 2.6287, + "step": 3890 + }, + { + "epoch": 0.01, + "learning_rate": 9.997444126355951e-05, + "loss": 2.5373, + "step": 3895 + }, + { + "epoch": 0.01, + "learning_rate": 9.997437387650113e-05, + "loss": 2.5818, + "step": 3900 + }, + { + "epoch": 0.01, + "learning_rate": 9.997430640074754e-05, + "loss": 2.7585, + "step": 3905 + }, + { + "epoch": 0.01, + "learning_rate": 9.997423883629886e-05, + "loss": 2.5746, + "step": 3910 + }, + { + "epoch": 0.01, + "learning_rate": 9.997417118315519e-05, + "loss": 2.5737, + "step": 3915 + }, + { + "epoch": 0.01, + "learning_rate": 9.99741034413167e-05, + "loss": 2.6422, + "step": 3920 + }, + { + "epoch": 0.01, + "learning_rate": 9.997403561078345e-05, + "loss": 2.5596, + "step": 3925 + }, + { + "epoch": 0.01, + "learning_rate": 9.997396769155559e-05, + "loss": 2.6658, + "step": 3930 + }, + { + "epoch": 0.01, + "learning_rate": 9.997389968363324e-05, + "loss": 2.4064, + "step": 3935 + }, + { + "epoch": 0.01, + "learning_rate": 9.997383158701652e-05, + "loss": 2.5619, + "step": 3940 + }, + { + "epoch": 0.01, + "learning_rate": 9.997376340170554e-05, + "loss": 2.5176, + "step": 3945 + }, + { + "epoch": 0.01, + "learning_rate": 9.997369512770045e-05, + "loss": 2.6683, + "step": 3950 + }, + { + "epoch": 0.01, + "learning_rate": 9.997362676500132e-05, + "loss": 2.5646, + "step": 3955 + }, + { + "epoch": 0.01, + "learning_rate": 9.997355831360833e-05, + "loss": 2.4585, + "step": 3960 + }, + { + "epoch": 0.01, + "learning_rate": 9.997348977352157e-05, + "loss": 2.6046, + "step": 3965 + }, + { + "epoch": 0.01, + "learning_rate": 9.997342114474115e-05, + "loss": 2.5631, + "step": 3970 + }, + { + "epoch": 0.01, + "learning_rate": 9.997335242726722e-05, + "loss": 2.702, + "step": 3975 + }, + { + "epoch": 0.01, + "learning_rate": 9.99732836210999e-05, + "loss": 2.5669, + "step": 3980 + }, + { + "epoch": 0.01, + "learning_rate": 9.997321472623928e-05, + "loss": 2.5259, + "step": 3985 + }, + { + "epoch": 0.01, + "learning_rate": 9.99731457426855e-05, + "loss": 2.4996, + "step": 3990 + }, + { + "epoch": 0.01, + "learning_rate": 9.997307667043872e-05, + "loss": 2.5803, + "step": 3995 + }, + { + "epoch": 0.01, + "learning_rate": 9.997300750949899e-05, + "loss": 2.6835, + "step": 4000 + }, + { + "epoch": 0.01, + "learning_rate": 9.997293825986649e-05, + "loss": 2.7114, + "step": 4005 + }, + { + "epoch": 0.01, + "learning_rate": 9.997286892154133e-05, + "loss": 2.5416, + "step": 4010 + }, + { + "epoch": 0.01, + "learning_rate": 9.99727994945236e-05, + "loss": 2.5914, + "step": 4015 + }, + { + "epoch": 0.01, + "learning_rate": 9.997272997881348e-05, + "loss": 2.6193, + "step": 4020 + }, + { + "epoch": 0.01, + "learning_rate": 9.997266037441105e-05, + "loss": 2.5593, + "step": 4025 + }, + { + "epoch": 0.01, + "learning_rate": 9.997259068131645e-05, + "loss": 2.6373, + "step": 4030 + }, + { + "epoch": 0.01, + "learning_rate": 9.99725208995298e-05, + "loss": 2.4203, + "step": 4035 + }, + { + "epoch": 0.01, + "learning_rate": 9.997245102905124e-05, + "loss": 2.4702, + "step": 4040 + }, + { + "epoch": 0.01, + "learning_rate": 9.997238106988086e-05, + "loss": 2.4993, + "step": 4045 + }, + { + "epoch": 0.01, + "learning_rate": 9.99723110220188e-05, + "loss": 2.4883, + "step": 4050 + }, + { + "epoch": 0.01, + "learning_rate": 9.99722408854652e-05, + "loss": 2.5822, + "step": 4055 + }, + { + "epoch": 0.01, + "learning_rate": 9.997217066022018e-05, + "loss": 2.5487, + "step": 4060 + }, + { + "epoch": 0.01, + "learning_rate": 9.997210034628384e-05, + "loss": 2.4815, + "step": 4065 + }, + { + "epoch": 0.01, + "learning_rate": 9.997202994365633e-05, + "loss": 2.5326, + "step": 4070 + }, + { + "epoch": 0.01, + "learning_rate": 9.997195945233776e-05, + "loss": 2.6387, + "step": 4075 + }, + { + "epoch": 0.01, + "learning_rate": 9.997188887232826e-05, + "loss": 2.5863, + "step": 4080 + }, + { + "epoch": 0.01, + "learning_rate": 9.997181820362797e-05, + "loss": 2.6602, + "step": 4085 + }, + { + "epoch": 0.01, + "learning_rate": 9.997174744623699e-05, + "loss": 2.5008, + "step": 4090 + }, + { + "epoch": 0.01, + "learning_rate": 9.997167660015547e-05, + "loss": 2.6069, + "step": 4095 + }, + { + "epoch": 0.01, + "learning_rate": 9.997160566538351e-05, + "loss": 2.6835, + "step": 4100 + }, + { + "epoch": 0.01, + "learning_rate": 9.997153464192127e-05, + "loss": 2.4615, + "step": 4105 + }, + { + "epoch": 0.01, + "learning_rate": 9.997146352976883e-05, + "loss": 2.5113, + "step": 4110 + }, + { + "epoch": 0.01, + "learning_rate": 9.997139232892635e-05, + "loss": 2.5662, + "step": 4115 + }, + { + "epoch": 0.01, + "learning_rate": 9.997132103939394e-05, + "loss": 2.6735, + "step": 4120 + }, + { + "epoch": 0.01, + "learning_rate": 9.997124966117175e-05, + "loss": 2.5504, + "step": 4125 + }, + { + "epoch": 0.01, + "learning_rate": 9.99711781942599e-05, + "loss": 2.5237, + "step": 4130 + }, + { + "epoch": 0.01, + "learning_rate": 9.997110663865847e-05, + "loss": 2.5644, + "step": 4135 + }, + { + "epoch": 0.01, + "learning_rate": 9.997103499436767e-05, + "loss": 2.5537, + "step": 4140 + }, + { + "epoch": 0.01, + "learning_rate": 9.997096326138754e-05, + "loss": 2.6656, + "step": 4145 + }, + { + "epoch": 0.01, + "learning_rate": 9.997089143971829e-05, + "loss": 2.534, + "step": 4150 + }, + { + "epoch": 0.01, + "learning_rate": 9.997081952935998e-05, + "loss": 2.4803, + "step": 4155 + }, + { + "epoch": 0.01, + "learning_rate": 9.997074753031276e-05, + "loss": 2.5295, + "step": 4160 + }, + { + "epoch": 0.01, + "learning_rate": 9.99706754425768e-05, + "loss": 2.4997, + "step": 4165 + }, + { + "epoch": 0.01, + "learning_rate": 9.997060326615215e-05, + "loss": 2.5858, + "step": 4170 + }, + { + "epoch": 0.01, + "learning_rate": 9.997053100103899e-05, + "loss": 2.5436, + "step": 4175 + }, + { + "epoch": 0.01, + "learning_rate": 9.997045864723744e-05, + "loss": 2.6045, + "step": 4180 + }, + { + "epoch": 0.01, + "learning_rate": 9.997038620474763e-05, + "loss": 2.5479, + "step": 4185 + }, + { + "epoch": 0.01, + "learning_rate": 9.997031367356968e-05, + "loss": 2.42, + "step": 4190 + }, + { + "epoch": 0.01, + "learning_rate": 9.997024105370371e-05, + "loss": 2.5555, + "step": 4195 + }, + { + "epoch": 0.01, + "learning_rate": 9.997016834514988e-05, + "loss": 2.5618, + "step": 4200 + }, + { + "epoch": 0.01, + "learning_rate": 9.997009554790831e-05, + "loss": 2.569, + "step": 4205 + }, + { + "epoch": 0.01, + "learning_rate": 9.997002266197908e-05, + "loss": 2.5937, + "step": 4210 + }, + { + "epoch": 0.01, + "learning_rate": 9.996994968736239e-05, + "loss": 2.6287, + "step": 4215 + }, + { + "epoch": 0.01, + "learning_rate": 9.996987662405833e-05, + "loss": 2.5191, + "step": 4220 + }, + { + "epoch": 0.01, + "learning_rate": 9.996980347206704e-05, + "loss": 2.5944, + "step": 4225 + }, + { + "epoch": 0.01, + "learning_rate": 9.996973023138865e-05, + "loss": 2.5008, + "step": 4230 + }, + { + "epoch": 0.01, + "learning_rate": 9.996965690202329e-05, + "loss": 2.4822, + "step": 4235 + }, + { + "epoch": 0.01, + "learning_rate": 9.996958348397108e-05, + "loss": 2.5088, + "step": 4240 + }, + { + "epoch": 0.01, + "learning_rate": 9.996950997723216e-05, + "loss": 2.5299, + "step": 4245 + }, + { + "epoch": 0.01, + "learning_rate": 9.996943638180666e-05, + "loss": 2.6813, + "step": 4250 + }, + { + "epoch": 0.01, + "learning_rate": 9.99693626976947e-05, + "loss": 2.6152, + "step": 4255 + }, + { + "epoch": 0.01, + "learning_rate": 9.996928892489644e-05, + "loss": 2.5546, + "step": 4260 + }, + { + "epoch": 0.01, + "learning_rate": 9.996921506341198e-05, + "loss": 2.6099, + "step": 4265 + }, + { + "epoch": 0.01, + "learning_rate": 9.996914111324146e-05, + "loss": 2.5888, + "step": 4270 + }, + { + "epoch": 0.01, + "learning_rate": 9.996906707438502e-05, + "loss": 2.452, + "step": 4275 + }, + { + "epoch": 0.01, + "learning_rate": 9.996899294684278e-05, + "loss": 2.4869, + "step": 4280 + }, + { + "epoch": 0.01, + "learning_rate": 9.996891873061486e-05, + "loss": 2.4223, + "step": 4285 + }, + { + "epoch": 0.01, + "learning_rate": 9.996884442570144e-05, + "loss": 2.5151, + "step": 4290 + }, + { + "epoch": 0.01, + "learning_rate": 9.99687700321026e-05, + "loss": 2.521, + "step": 4295 + }, + { + "epoch": 0.01, + "learning_rate": 9.996869554981851e-05, + "loss": 2.4921, + "step": 4300 + }, + { + "epoch": 0.01, + "learning_rate": 9.996862097884928e-05, + "loss": 2.6987, + "step": 4305 + }, + { + "epoch": 0.01, + "learning_rate": 9.996854631919503e-05, + "loss": 2.4546, + "step": 4310 + }, + { + "epoch": 0.01, + "learning_rate": 9.996847157085592e-05, + "loss": 2.5139, + "step": 4315 + }, + { + "epoch": 0.01, + "learning_rate": 9.996839673383207e-05, + "loss": 2.5773, + "step": 4320 + }, + { + "epoch": 0.01, + "learning_rate": 9.996832180812361e-05, + "loss": 2.6551, + "step": 4325 + }, + { + "epoch": 0.01, + "learning_rate": 9.996824679373069e-05, + "loss": 2.581, + "step": 4330 + }, + { + "epoch": 0.01, + "learning_rate": 9.996817169065342e-05, + "loss": 2.4511, + "step": 4335 + }, + { + "epoch": 0.01, + "learning_rate": 9.996809649889196e-05, + "loss": 2.6355, + "step": 4340 + }, + { + "epoch": 0.01, + "learning_rate": 9.99680212184464e-05, + "loss": 2.5136, + "step": 4345 + }, + { + "epoch": 0.01, + "learning_rate": 9.996794584931692e-05, + "loss": 2.4373, + "step": 4350 + }, + { + "epoch": 0.01, + "learning_rate": 9.996787039150364e-05, + "loss": 2.5403, + "step": 4355 + }, + { + "epoch": 0.01, + "learning_rate": 9.996779484500667e-05, + "loss": 2.5738, + "step": 4360 + }, + { + "epoch": 0.01, + "learning_rate": 9.996771920982618e-05, + "loss": 2.3989, + "step": 4365 + }, + { + "epoch": 0.01, + "learning_rate": 9.996764348596228e-05, + "loss": 2.5097, + "step": 4370 + }, + { + "epoch": 0.01, + "learning_rate": 9.996756767341511e-05, + "loss": 2.6498, + "step": 4375 + }, + { + "epoch": 0.01, + "learning_rate": 9.996749177218481e-05, + "loss": 2.5166, + "step": 4380 + }, + { + "epoch": 0.01, + "learning_rate": 9.99674157822715e-05, + "loss": 2.6611, + "step": 4385 + }, + { + "epoch": 0.01, + "learning_rate": 9.996733970367533e-05, + "loss": 2.6209, + "step": 4390 + }, + { + "epoch": 0.01, + "learning_rate": 9.996726353639644e-05, + "loss": 2.552, + "step": 4395 + }, + { + "epoch": 0.01, + "learning_rate": 9.996718728043495e-05, + "loss": 2.5537, + "step": 4400 + }, + { + "epoch": 0.01, + "learning_rate": 9.9967110935791e-05, + "loss": 2.6233, + "step": 4405 + }, + { + "epoch": 0.01, + "learning_rate": 9.996703450246474e-05, + "loss": 2.5815, + "step": 4410 + }, + { + "epoch": 0.01, + "learning_rate": 9.996695798045627e-05, + "loss": 2.5098, + "step": 4415 + }, + { + "epoch": 0.01, + "learning_rate": 9.996688136976577e-05, + "loss": 2.5335, + "step": 4420 + }, + { + "epoch": 0.01, + "learning_rate": 9.996680467039334e-05, + "loss": 2.6411, + "step": 4425 + }, + { + "epoch": 0.01, + "learning_rate": 9.996672788233913e-05, + "loss": 2.4623, + "step": 4430 + }, + { + "epoch": 0.01, + "learning_rate": 9.996665100560328e-05, + "loss": 2.4191, + "step": 4435 + }, + { + "epoch": 0.01, + "learning_rate": 9.996657404018592e-05, + "loss": 2.4118, + "step": 4440 + }, + { + "epoch": 0.01, + "learning_rate": 9.99664969860872e-05, + "loss": 2.5339, + "step": 4445 + }, + { + "epoch": 0.01, + "learning_rate": 9.996641984330724e-05, + "loss": 2.5186, + "step": 4450 + }, + { + "epoch": 0.01, + "learning_rate": 9.996634261184619e-05, + "loss": 2.5556, + "step": 4455 + }, + { + "epoch": 0.01, + "learning_rate": 9.996626529170417e-05, + "loss": 2.5094, + "step": 4460 + }, + { + "epoch": 0.01, + "learning_rate": 9.996618788288133e-05, + "loss": 2.6038, + "step": 4465 + }, + { + "epoch": 0.01, + "learning_rate": 9.99661103853778e-05, + "loss": 2.5905, + "step": 4470 + }, + { + "epoch": 0.01, + "learning_rate": 9.996603279919372e-05, + "loss": 2.5427, + "step": 4475 + }, + { + "epoch": 0.01, + "learning_rate": 9.996595512432925e-05, + "loss": 2.5896, + "step": 4480 + }, + { + "epoch": 0.01, + "learning_rate": 9.996587736078448e-05, + "loss": 2.3125, + "step": 4485 + }, + { + "epoch": 0.01, + "learning_rate": 9.99657995085596e-05, + "loss": 2.494, + "step": 4490 + }, + { + "epoch": 0.01, + "learning_rate": 9.996572156765472e-05, + "loss": 2.4965, + "step": 4495 + }, + { + "epoch": 0.01, + "learning_rate": 9.996564353806997e-05, + "loss": 2.7065, + "step": 4500 + }, + { + "epoch": 0.01, + "learning_rate": 9.996556541980552e-05, + "loss": 2.5263, + "step": 4505 + }, + { + "epoch": 0.01, + "learning_rate": 9.996548721286148e-05, + "loss": 2.5831, + "step": 4510 + }, + { + "epoch": 0.01, + "learning_rate": 9.9965408917238e-05, + "loss": 2.5962, + "step": 4515 + }, + { + "epoch": 0.01, + "learning_rate": 9.996533053293522e-05, + "loss": 2.5593, + "step": 4520 + }, + { + "epoch": 0.01, + "learning_rate": 9.996525205995327e-05, + "loss": 2.5366, + "step": 4525 + }, + { + "epoch": 0.01, + "learning_rate": 9.996517349829228e-05, + "loss": 2.5562, + "step": 4530 + }, + { + "epoch": 0.01, + "learning_rate": 9.996509484795245e-05, + "loss": 2.4178, + "step": 4535 + }, + { + "epoch": 0.01, + "learning_rate": 9.996501610893383e-05, + "loss": 2.3881, + "step": 4540 + }, + { + "epoch": 0.01, + "learning_rate": 9.996493728123664e-05, + "loss": 2.5981, + "step": 4545 + }, + { + "epoch": 0.01, + "learning_rate": 9.996485836486095e-05, + "loss": 2.4508, + "step": 4550 + }, + { + "epoch": 0.01, + "learning_rate": 9.996477935980696e-05, + "loss": 2.4686, + "step": 4555 + }, + { + "epoch": 0.01, + "learning_rate": 9.996470026607479e-05, + "loss": 2.4925, + "step": 4560 + }, + { + "epoch": 0.01, + "learning_rate": 9.996462108366455e-05, + "loss": 2.5791, + "step": 4565 + }, + { + "epoch": 0.01, + "learning_rate": 9.996454181257642e-05, + "loss": 2.6235, + "step": 4570 + }, + { + "epoch": 0.01, + "learning_rate": 9.996446245281052e-05, + "loss": 2.5015, + "step": 4575 + }, + { + "epoch": 0.01, + "learning_rate": 9.9964383004367e-05, + "loss": 2.4028, + "step": 4580 + }, + { + "epoch": 0.01, + "learning_rate": 9.996430346724601e-05, + "loss": 2.4236, + "step": 4585 + }, + { + "epoch": 0.01, + "learning_rate": 9.996422384144766e-05, + "loss": 2.4065, + "step": 4590 + }, + { + "epoch": 0.01, + "learning_rate": 9.996414412697212e-05, + "loss": 2.5082, + "step": 4595 + }, + { + "epoch": 0.01, + "learning_rate": 9.996406432381953e-05, + "loss": 2.446, + "step": 4600 + }, + { + "epoch": 0.01, + "learning_rate": 9.996398443199002e-05, + "loss": 2.4575, + "step": 4605 + }, + { + "epoch": 0.01, + "learning_rate": 9.996390445148372e-05, + "loss": 2.6574, + "step": 4610 + }, + { + "epoch": 0.01, + "learning_rate": 9.99638243823008e-05, + "loss": 2.4611, + "step": 4615 + }, + { + "epoch": 0.01, + "learning_rate": 9.99637442244414e-05, + "loss": 2.6219, + "step": 4620 + }, + { + "epoch": 0.01, + "learning_rate": 9.996366397790565e-05, + "loss": 2.6078, + "step": 4625 + }, + { + "epoch": 0.01, + "learning_rate": 9.996358364269367e-05, + "loss": 2.4855, + "step": 4630 + }, + { + "epoch": 0.01, + "learning_rate": 9.996350321880566e-05, + "loss": 2.5521, + "step": 4635 + }, + { + "epoch": 0.01, + "learning_rate": 9.99634227062417e-05, + "loss": 2.5444, + "step": 4640 + }, + { + "epoch": 0.01, + "learning_rate": 9.996334210500198e-05, + "loss": 2.5534, + "step": 4645 + }, + { + "epoch": 0.01, + "learning_rate": 9.996326141508662e-05, + "loss": 2.6037, + "step": 4650 + }, + { + "epoch": 0.01, + "learning_rate": 9.996318063649577e-05, + "loss": 2.5392, + "step": 4655 + }, + { + "epoch": 0.01, + "learning_rate": 9.996309976922957e-05, + "loss": 2.5852, + "step": 4660 + }, + { + "epoch": 0.01, + "learning_rate": 9.996301881328817e-05, + "loss": 2.6119, + "step": 4665 + }, + { + "epoch": 0.01, + "learning_rate": 9.99629377686717e-05, + "loss": 2.3955, + "step": 4670 + }, + { + "epoch": 0.01, + "learning_rate": 9.996285663538032e-05, + "loss": 2.4615, + "step": 4675 + }, + { + "epoch": 0.01, + "learning_rate": 9.996277541341417e-05, + "loss": 2.4958, + "step": 4680 + }, + { + "epoch": 0.01, + "learning_rate": 9.996269410277339e-05, + "loss": 2.4504, + "step": 4685 + }, + { + "epoch": 0.01, + "learning_rate": 9.996261270345812e-05, + "loss": 2.464, + "step": 4690 + }, + { + "epoch": 0.01, + "learning_rate": 9.996253121546852e-05, + "loss": 2.5381, + "step": 4695 + }, + { + "epoch": 0.01, + "learning_rate": 9.996244963880472e-05, + "loss": 2.4668, + "step": 4700 + }, + { + "epoch": 0.01, + "learning_rate": 9.996236797346687e-05, + "loss": 2.5809, + "step": 4705 + }, + { + "epoch": 0.01, + "learning_rate": 9.996228621945511e-05, + "loss": 2.4865, + "step": 4710 + }, + { + "epoch": 0.01, + "learning_rate": 9.99622043767696e-05, + "loss": 2.5496, + "step": 4715 + }, + { + "epoch": 0.01, + "learning_rate": 9.996212244541048e-05, + "loss": 2.626, + "step": 4720 + }, + { + "epoch": 0.01, + "learning_rate": 9.996204042537786e-05, + "loss": 2.5498, + "step": 4725 + }, + { + "epoch": 0.01, + "learning_rate": 9.996195831667195e-05, + "loss": 2.5715, + "step": 4730 + }, + { + "epoch": 0.01, + "learning_rate": 9.996187611929284e-05, + "loss": 2.4865, + "step": 4735 + }, + { + "epoch": 0.01, + "learning_rate": 9.996179383324072e-05, + "loss": 2.4541, + "step": 4740 + }, + { + "epoch": 0.01, + "learning_rate": 9.996171145851569e-05, + "loss": 2.4882, + "step": 4745 + }, + { + "epoch": 0.01, + "learning_rate": 9.996162899511793e-05, + "loss": 2.5475, + "step": 4750 + }, + { + "epoch": 0.01, + "learning_rate": 9.996154644304758e-05, + "loss": 2.5209, + "step": 4755 + }, + { + "epoch": 0.01, + "learning_rate": 9.996146380230478e-05, + "loss": 2.5372, + "step": 4760 + }, + { + "epoch": 0.01, + "learning_rate": 9.996138107288966e-05, + "loss": 2.6316, + "step": 4765 + }, + { + "epoch": 0.01, + "learning_rate": 9.996129825480241e-05, + "loss": 2.6269, + "step": 4770 + }, + { + "epoch": 0.01, + "learning_rate": 9.996121534804315e-05, + "loss": 2.4314, + "step": 4775 + }, + { + "epoch": 0.01, + "learning_rate": 9.996113235261202e-05, + "loss": 2.5533, + "step": 4780 + }, + { + "epoch": 0.01, + "learning_rate": 9.996104926850918e-05, + "loss": 2.4352, + "step": 4785 + }, + { + "epoch": 0.01, + "learning_rate": 9.996096609573478e-05, + "loss": 2.5772, + "step": 4790 + }, + { + "epoch": 0.01, + "learning_rate": 9.996088283428898e-05, + "loss": 2.4497, + "step": 4795 + }, + { + "epoch": 0.01, + "learning_rate": 9.996079948417187e-05, + "loss": 2.468, + "step": 4800 + }, + { + "epoch": 0.01, + "learning_rate": 9.996071604538366e-05, + "loss": 2.6215, + "step": 4805 + }, + { + "epoch": 0.01, + "learning_rate": 9.99606325179245e-05, + "loss": 2.5124, + "step": 4810 + }, + { + "epoch": 0.01, + "learning_rate": 9.996054890179447e-05, + "loss": 2.5624, + "step": 4815 + }, + { + "epoch": 0.01, + "learning_rate": 9.996046519699379e-05, + "loss": 2.5595, + "step": 4820 + }, + { + "epoch": 0.01, + "learning_rate": 9.996038140352257e-05, + "loss": 2.5637, + "step": 4825 + }, + { + "epoch": 0.01, + "learning_rate": 9.996029752138098e-05, + "loss": 2.3108, + "step": 4830 + }, + { + "epoch": 0.01, + "learning_rate": 9.996021355056915e-05, + "loss": 2.4952, + "step": 4835 + }, + { + "epoch": 0.01, + "learning_rate": 9.996012949108726e-05, + "loss": 2.4181, + "step": 4840 + }, + { + "epoch": 0.01, + "learning_rate": 9.996004534293542e-05, + "loss": 2.4989, + "step": 4845 + }, + { + "epoch": 0.01, + "learning_rate": 9.995996110611379e-05, + "loss": 2.5226, + "step": 4850 + }, + { + "epoch": 0.01, + "learning_rate": 9.995987678062253e-05, + "loss": 2.4691, + "step": 4855 + }, + { + "epoch": 0.01, + "learning_rate": 9.99597923664618e-05, + "loss": 2.5101, + "step": 4860 + }, + { + "epoch": 0.01, + "learning_rate": 9.995970786363172e-05, + "loss": 2.5981, + "step": 4865 + }, + { + "epoch": 0.01, + "learning_rate": 9.995962327213247e-05, + "loss": 2.3827, + "step": 4870 + }, + { + "epoch": 0.01, + "learning_rate": 9.995953859196417e-05, + "loss": 2.5007, + "step": 4875 + }, + { + "epoch": 0.01, + "learning_rate": 9.9959453823127e-05, + "loss": 2.6254, + "step": 4880 + }, + { + "epoch": 0.01, + "learning_rate": 9.995936896562109e-05, + "loss": 2.5605, + "step": 4885 + }, + { + "epoch": 0.01, + "learning_rate": 9.99592840194466e-05, + "loss": 2.6333, + "step": 4890 + }, + { + "epoch": 0.01, + "learning_rate": 9.995919898460368e-05, + "loss": 2.4521, + "step": 4895 + }, + { + "epoch": 0.01, + "learning_rate": 9.995911386109247e-05, + "loss": 2.608, + "step": 4900 + }, + { + "epoch": 0.01, + "learning_rate": 9.995902864891314e-05, + "loss": 2.4649, + "step": 4905 + }, + { + "epoch": 0.01, + "learning_rate": 9.995894334806581e-05, + "loss": 2.5698, + "step": 4910 + }, + { + "epoch": 0.01, + "learning_rate": 9.995885795855069e-05, + "loss": 2.4729, + "step": 4915 + }, + { + "epoch": 0.01, + "learning_rate": 9.995877248036787e-05, + "loss": 2.5913, + "step": 4920 + }, + { + "epoch": 0.01, + "learning_rate": 9.995868691351753e-05, + "loss": 2.5248, + "step": 4925 + }, + { + "epoch": 0.01, + "learning_rate": 9.99586012579998e-05, + "loss": 2.5741, + "step": 4930 + }, + { + "epoch": 0.01, + "learning_rate": 9.995851551381487e-05, + "loss": 2.5241, + "step": 4935 + }, + { + "epoch": 0.01, + "learning_rate": 9.995842968096287e-05, + "loss": 2.5985, + "step": 4940 + }, + { + "epoch": 0.01, + "learning_rate": 9.995834375944396e-05, + "loss": 2.4096, + "step": 4945 + }, + { + "epoch": 0.01, + "learning_rate": 9.995825774925829e-05, + "loss": 2.4206, + "step": 4950 + }, + { + "epoch": 0.01, + "learning_rate": 9.9958171650406e-05, + "loss": 2.5033, + "step": 4955 + }, + { + "epoch": 0.01, + "learning_rate": 9.995808546288725e-05, + "loss": 2.4103, + "step": 4960 + }, + { + "epoch": 0.01, + "learning_rate": 9.995799918670219e-05, + "loss": 2.4963, + "step": 4965 + }, + { + "epoch": 0.01, + "learning_rate": 9.995791282185099e-05, + "loss": 2.5737, + "step": 4970 + }, + { + "epoch": 0.01, + "learning_rate": 9.995782636833378e-05, + "loss": 2.6329, + "step": 4975 + }, + { + "epoch": 0.01, + "learning_rate": 9.995773982615073e-05, + "loss": 2.5031, + "step": 4980 + }, + { + "epoch": 0.01, + "learning_rate": 9.9957653195302e-05, + "loss": 2.5558, + "step": 4985 + }, + { + "epoch": 0.01, + "learning_rate": 9.995756647578772e-05, + "loss": 2.5539, + "step": 4990 + }, + { + "epoch": 0.01, + "learning_rate": 9.995747966760806e-05, + "loss": 2.5113, + "step": 4995 + }, + { + "epoch": 0.01, + "learning_rate": 9.995739277076317e-05, + "loss": 2.5191, + "step": 5000 + }, + { + "epoch": 0.01, + "learning_rate": 9.99573057852532e-05, + "loss": 2.5636, + "step": 5005 + }, + { + "epoch": 0.01, + "learning_rate": 9.995721871107832e-05, + "loss": 2.5681, + "step": 5010 + }, + { + "epoch": 0.01, + "learning_rate": 9.995713154823867e-05, + "loss": 2.4807, + "step": 5015 + }, + { + "epoch": 0.01, + "learning_rate": 9.99570442967344e-05, + "loss": 2.6448, + "step": 5020 + }, + { + "epoch": 0.01, + "learning_rate": 9.995695695656568e-05, + "loss": 2.5366, + "step": 5025 + }, + { + "epoch": 0.01, + "learning_rate": 9.995686952773267e-05, + "loss": 2.4624, + "step": 5030 + }, + { + "epoch": 0.01, + "learning_rate": 9.995678201023549e-05, + "loss": 2.4419, + "step": 5035 + }, + { + "epoch": 0.01, + "learning_rate": 9.995669440407434e-05, + "loss": 2.5459, + "step": 5040 + }, + { + "epoch": 0.01, + "learning_rate": 9.995660670924933e-05, + "loss": 2.6548, + "step": 5045 + }, + { + "epoch": 0.01, + "learning_rate": 9.995651892576066e-05, + "loss": 2.5432, + "step": 5050 + }, + { + "epoch": 0.01, + "learning_rate": 9.995643105360846e-05, + "loss": 2.5735, + "step": 5055 + }, + { + "epoch": 0.01, + "learning_rate": 9.99563430927929e-05, + "loss": 2.5381, + "step": 5060 + }, + { + "epoch": 0.01, + "learning_rate": 9.99562550433141e-05, + "loss": 2.3988, + "step": 5065 + }, + { + "epoch": 0.01, + "learning_rate": 9.995616690517226e-05, + "loss": 2.6054, + "step": 5070 + }, + { + "epoch": 0.01, + "learning_rate": 9.995607867836752e-05, + "loss": 2.5637, + "step": 5075 + }, + { + "epoch": 0.01, + "learning_rate": 9.995599036290005e-05, + "loss": 2.4666, + "step": 5080 + }, + { + "epoch": 0.01, + "learning_rate": 9.995590195876999e-05, + "loss": 2.5511, + "step": 5085 + }, + { + "epoch": 0.01, + "learning_rate": 9.995581346597748e-05, + "loss": 2.3998, + "step": 5090 + }, + { + "epoch": 0.01, + "learning_rate": 9.995572488452272e-05, + "loss": 2.4357, + "step": 5095 + }, + { + "epoch": 0.01, + "learning_rate": 9.995563621440583e-05, + "loss": 2.5704, + "step": 5100 + }, + { + "epoch": 0.01, + "learning_rate": 9.995554745562699e-05, + "loss": 2.607, + "step": 5105 + }, + { + "epoch": 0.01, + "learning_rate": 9.995545860818634e-05, + "loss": 2.513, + "step": 5110 + }, + { + "epoch": 0.01, + "learning_rate": 9.995536967208406e-05, + "loss": 2.595, + "step": 5115 + }, + { + "epoch": 0.01, + "learning_rate": 9.99552806473203e-05, + "loss": 2.5767, + "step": 5120 + }, + { + "epoch": 0.01, + "learning_rate": 9.995519153389519e-05, + "loss": 2.5215, + "step": 5125 + }, + { + "epoch": 0.01, + "learning_rate": 9.995510233180892e-05, + "loss": 2.5745, + "step": 5130 + }, + { + "epoch": 0.01, + "learning_rate": 9.995501304106165e-05, + "loss": 2.5306, + "step": 5135 + }, + { + "epoch": 0.01, + "learning_rate": 9.995492366165352e-05, + "loss": 2.4468, + "step": 5140 + }, + { + "epoch": 0.01, + "learning_rate": 9.99548341935847e-05, + "loss": 2.5536, + "step": 5145 + }, + { + "epoch": 0.01, + "learning_rate": 9.995474463685535e-05, + "loss": 2.4986, + "step": 5150 + }, + { + "epoch": 0.01, + "learning_rate": 9.995465499146562e-05, + "loss": 2.6726, + "step": 5155 + }, + { + "epoch": 0.01, + "learning_rate": 9.995456525741566e-05, + "loss": 2.577, + "step": 5160 + }, + { + "epoch": 0.01, + "learning_rate": 9.995447543470567e-05, + "loss": 2.565, + "step": 5165 + }, + { + "epoch": 0.01, + "learning_rate": 9.995438552333577e-05, + "loss": 2.5887, + "step": 5170 + }, + { + "epoch": 0.01, + "learning_rate": 9.995429552330613e-05, + "loss": 2.4109, + "step": 5175 + }, + { + "epoch": 0.01, + "learning_rate": 9.995420543461691e-05, + "loss": 2.5185, + "step": 5180 + }, + { + "epoch": 0.01, + "learning_rate": 9.995411525726827e-05, + "loss": 2.4117, + "step": 5185 + }, + { + "epoch": 0.01, + "learning_rate": 9.995402499126039e-05, + "loss": 2.5225, + "step": 5190 + }, + { + "epoch": 0.01, + "learning_rate": 9.995393463659339e-05, + "loss": 2.5672, + "step": 5195 + }, + { + "epoch": 0.01, + "learning_rate": 9.995384419326745e-05, + "loss": 2.437, + "step": 5200 + }, + { + "epoch": 0.01, + "learning_rate": 9.995375366128275e-05, + "loss": 2.3797, + "step": 5205 + }, + { + "epoch": 0.01, + "learning_rate": 9.995366304063942e-05, + "loss": 2.4658, + "step": 5210 + }, + { + "epoch": 0.01, + "learning_rate": 9.995357233133765e-05, + "loss": 2.5608, + "step": 5215 + }, + { + "epoch": 0.01, + "learning_rate": 9.995348153337758e-05, + "loss": 2.5834, + "step": 5220 + }, + { + "epoch": 0.01, + "learning_rate": 9.995339064675938e-05, + "loss": 2.5438, + "step": 5225 + }, + { + "epoch": 0.01, + "learning_rate": 9.99532996714832e-05, + "loss": 2.5492, + "step": 5230 + }, + { + "epoch": 0.01, + "learning_rate": 9.99532086075492e-05, + "loss": 2.4893, + "step": 5235 + }, + { + "epoch": 0.01, + "learning_rate": 9.995311745495755e-05, + "loss": 2.6605, + "step": 5240 + }, + { + "epoch": 0.01, + "learning_rate": 9.995302621370843e-05, + "loss": 2.6313, + "step": 5245 + }, + { + "epoch": 0.01, + "learning_rate": 9.995293488380197e-05, + "loss": 2.5724, + "step": 5250 + }, + { + "epoch": 0.01, + "learning_rate": 9.995284346523835e-05, + "loss": 2.4324, + "step": 5255 + }, + { + "epoch": 0.01, + "learning_rate": 9.995275195801774e-05, + "loss": 2.6707, + "step": 5260 + }, + { + "epoch": 0.01, + "learning_rate": 9.995266036214028e-05, + "loss": 2.5461, + "step": 5265 + }, + { + "epoch": 0.01, + "learning_rate": 9.995256867760614e-05, + "loss": 2.585, + "step": 5270 + }, + { + "epoch": 0.01, + "learning_rate": 9.99524769044155e-05, + "loss": 2.543, + "step": 5275 + }, + { + "epoch": 0.01, + "learning_rate": 9.99523850425685e-05, + "loss": 2.4448, + "step": 5280 + }, + { + "epoch": 0.01, + "learning_rate": 9.995229309206531e-05, + "loss": 2.4265, + "step": 5285 + }, + { + "epoch": 0.01, + "learning_rate": 9.99522010529061e-05, + "loss": 2.5415, + "step": 5290 + }, + { + "epoch": 0.01, + "learning_rate": 9.995210892509104e-05, + "loss": 2.5168, + "step": 5295 + }, + { + "epoch": 0.01, + "learning_rate": 9.995201670862027e-05, + "loss": 2.5215, + "step": 5300 + }, + { + "epoch": 0.01, + "learning_rate": 9.995192440349397e-05, + "loss": 2.6094, + "step": 5305 + }, + { + "epoch": 0.01, + "learning_rate": 9.99518320097123e-05, + "loss": 2.5555, + "step": 5310 + }, + { + "epoch": 0.01, + "learning_rate": 9.995173952727543e-05, + "loss": 2.3882, + "step": 5315 + }, + { + "epoch": 0.01, + "learning_rate": 9.995164695618351e-05, + "loss": 2.6239, + "step": 5320 + }, + { + "epoch": 0.01, + "learning_rate": 9.99515542964367e-05, + "loss": 2.403, + "step": 5325 + }, + { + "epoch": 0.01, + "learning_rate": 9.99514615480352e-05, + "loss": 2.4849, + "step": 5330 + }, + { + "epoch": 0.01, + "learning_rate": 9.995136871097914e-05, + "loss": 2.6149, + "step": 5335 + }, + { + "epoch": 0.01, + "learning_rate": 9.995127578526869e-05, + "loss": 2.5914, + "step": 5340 + }, + { + "epoch": 0.01, + "learning_rate": 9.995118277090402e-05, + "loss": 2.4116, + "step": 5345 + }, + { + "epoch": 0.01, + "learning_rate": 9.99510896678853e-05, + "loss": 2.5205, + "step": 5350 + }, + { + "epoch": 0.01, + "learning_rate": 9.99509964762127e-05, + "loss": 2.5336, + "step": 5355 + }, + { + "epoch": 0.01, + "learning_rate": 9.995090319588638e-05, + "loss": 2.6312, + "step": 5360 + }, + { + "epoch": 0.01, + "learning_rate": 9.99508098269065e-05, + "loss": 2.559, + "step": 5365 + }, + { + "epoch": 0.01, + "learning_rate": 9.995071636927322e-05, + "loss": 2.5321, + "step": 5370 + }, + { + "epoch": 0.01, + "learning_rate": 9.995062282298672e-05, + "loss": 2.4178, + "step": 5375 + }, + { + "epoch": 0.01, + "learning_rate": 9.995052918804717e-05, + "loss": 2.624, + "step": 5380 + }, + { + "epoch": 0.01, + "learning_rate": 9.99504354644547e-05, + "loss": 2.5034, + "step": 5385 + }, + { + "epoch": 0.01, + "learning_rate": 9.995034165220954e-05, + "loss": 2.4599, + "step": 5390 + }, + { + "epoch": 0.01, + "learning_rate": 9.995024775131179e-05, + "loss": 2.5376, + "step": 5395 + }, + { + "epoch": 0.01, + "learning_rate": 9.995015376176164e-05, + "loss": 2.4196, + "step": 5400 + }, + { + "epoch": 0.01, + "learning_rate": 9.995005968355929e-05, + "loss": 2.5241, + "step": 5405 + }, + { + "epoch": 0.01, + "learning_rate": 9.994996551670485e-05, + "loss": 2.4466, + "step": 5410 + }, + { + "epoch": 0.01, + "learning_rate": 9.994987126119852e-05, + "loss": 2.5971, + "step": 5415 + }, + { + "epoch": 0.01, + "learning_rate": 9.994977691704048e-05, + "loss": 2.4972, + "step": 5420 + }, + { + "epoch": 0.01, + "learning_rate": 9.994968248423089e-05, + "loss": 2.5765, + "step": 5425 + }, + { + "epoch": 0.01, + "learning_rate": 9.994958796276989e-05, + "loss": 2.4853, + "step": 5430 + }, + { + "epoch": 0.01, + "learning_rate": 9.994949335265766e-05, + "loss": 2.6405, + "step": 5435 + }, + { + "epoch": 0.01, + "learning_rate": 9.994939865389439e-05, + "loss": 2.5002, + "step": 5440 + }, + { + "epoch": 0.01, + "learning_rate": 9.994930386648022e-05, + "loss": 2.6242, + "step": 5445 + }, + { + "epoch": 0.01, + "learning_rate": 9.994920899041536e-05, + "loss": 2.5121, + "step": 5450 + }, + { + "epoch": 0.01, + "learning_rate": 9.994911402569992e-05, + "loss": 2.4733, + "step": 5455 + }, + { + "epoch": 0.01, + "learning_rate": 9.994901897233411e-05, + "loss": 2.4444, + "step": 5460 + }, + { + "epoch": 0.01, + "learning_rate": 9.994892383031808e-05, + "loss": 2.589, + "step": 5465 + }, + { + "epoch": 0.01, + "learning_rate": 9.994882859965201e-05, + "loss": 2.4131, + "step": 5470 + }, + { + "epoch": 0.01, + "learning_rate": 9.994873328033607e-05, + "loss": 2.5094, + "step": 5475 + }, + { + "epoch": 0.01, + "learning_rate": 9.994863787237041e-05, + "loss": 2.4842, + "step": 5480 + }, + { + "epoch": 0.01, + "learning_rate": 9.994854237575521e-05, + "loss": 2.6548, + "step": 5485 + }, + { + "epoch": 0.01, + "learning_rate": 9.994844679049067e-05, + "loss": 2.5653, + "step": 5490 + }, + { + "epoch": 0.01, + "learning_rate": 9.99483511165769e-05, + "loss": 2.433, + "step": 5495 + }, + { + "epoch": 0.01, + "learning_rate": 9.994825535401414e-05, + "loss": 2.561, + "step": 5500 + }, + { + "epoch": 0.01, + "learning_rate": 9.994815950280249e-05, + "loss": 2.4968, + "step": 5505 + }, + { + "epoch": 0.01, + "learning_rate": 9.994806356294215e-05, + "loss": 2.4869, + "step": 5510 + }, + { + "epoch": 0.01, + "learning_rate": 9.99479675344333e-05, + "loss": 2.468, + "step": 5515 + }, + { + "epoch": 0.01, + "learning_rate": 9.99478714172761e-05, + "loss": 2.5346, + "step": 5520 + }, + { + "epoch": 0.01, + "learning_rate": 9.994777521147074e-05, + "loss": 2.4849, + "step": 5525 + }, + { + "epoch": 0.01, + "learning_rate": 9.994767891701736e-05, + "loss": 2.6244, + "step": 5530 + }, + { + "epoch": 0.01, + "learning_rate": 9.994758253391615e-05, + "loss": 2.5334, + "step": 5535 + }, + { + "epoch": 0.01, + "learning_rate": 9.994748606216728e-05, + "loss": 2.504, + "step": 5540 + }, + { + "epoch": 0.01, + "learning_rate": 9.994738950177091e-05, + "loss": 2.4409, + "step": 5545 + }, + { + "epoch": 0.01, + "learning_rate": 9.994729285272721e-05, + "loss": 2.4927, + "step": 5550 + }, + { + "epoch": 0.01, + "learning_rate": 9.994719611503638e-05, + "loss": 2.5659, + "step": 5555 + }, + { + "epoch": 0.01, + "learning_rate": 9.994709928869856e-05, + "loss": 2.5208, + "step": 5560 + }, + { + "epoch": 0.01, + "learning_rate": 9.994700237371392e-05, + "loss": 2.441, + "step": 5565 + }, + { + "epoch": 0.01, + "learning_rate": 9.994690537008266e-05, + "loss": 2.5524, + "step": 5570 + }, + { + "epoch": 0.01, + "learning_rate": 9.994680827780495e-05, + "loss": 2.4663, + "step": 5575 + }, + { + "epoch": 0.01, + "learning_rate": 9.994671109688094e-05, + "loss": 2.5656, + "step": 5580 + }, + { + "epoch": 0.01, + "learning_rate": 9.99466138273108e-05, + "loss": 2.5732, + "step": 5585 + }, + { + "epoch": 0.01, + "learning_rate": 9.994651646909474e-05, + "loss": 2.4707, + "step": 5590 + }, + { + "epoch": 0.02, + "learning_rate": 9.99464190222329e-05, + "loss": 2.4111, + "step": 5595 + }, + { + "epoch": 0.02, + "learning_rate": 9.994632148672545e-05, + "loss": 2.5853, + "step": 5600 + }, + { + "epoch": 0.02, + "learning_rate": 9.994622386257257e-05, + "loss": 2.3985, + "step": 5605 + }, + { + "epoch": 0.02, + "learning_rate": 9.994612614977446e-05, + "loss": 2.5352, + "step": 5610 + }, + { + "epoch": 0.02, + "learning_rate": 9.994602834833125e-05, + "loss": 2.5764, + "step": 5615 + }, + { + "epoch": 0.02, + "learning_rate": 9.994593045824315e-05, + "loss": 2.5349, + "step": 5620 + }, + { + "epoch": 0.02, + "learning_rate": 9.99458324795103e-05, + "loss": 2.4178, + "step": 5625 + }, + { + "epoch": 0.02, + "learning_rate": 9.994573441213291e-05, + "loss": 2.5873, + "step": 5630 + }, + { + "epoch": 0.02, + "learning_rate": 9.994563625611113e-05, + "loss": 2.4967, + "step": 5635 + }, + { + "epoch": 0.02, + "learning_rate": 9.994553801144513e-05, + "loss": 2.48, + "step": 5640 + }, + { + "epoch": 0.02, + "learning_rate": 9.99454396781351e-05, + "loss": 2.5694, + "step": 5645 + }, + { + "epoch": 0.02, + "learning_rate": 9.994534125618122e-05, + "loss": 2.6, + "step": 5650 + }, + { + "epoch": 0.02, + "learning_rate": 9.994524274558365e-05, + "loss": 2.4958, + "step": 5655 + }, + { + "epoch": 0.02, + "learning_rate": 9.994514414634256e-05, + "loss": 2.5538, + "step": 5660 + }, + { + "epoch": 0.02, + "learning_rate": 9.994504545845814e-05, + "loss": 2.5116, + "step": 5665 + }, + { + "epoch": 0.02, + "learning_rate": 9.994494668193055e-05, + "loss": 2.4951, + "step": 5670 + }, + { + "epoch": 0.02, + "learning_rate": 9.994484781675998e-05, + "loss": 2.5914, + "step": 5675 + }, + { + "epoch": 0.02, + "learning_rate": 9.994474886294662e-05, + "loss": 2.4961, + "step": 5680 + }, + { + "epoch": 0.02, + "learning_rate": 9.994464982049059e-05, + "loss": 2.5933, + "step": 5685 + }, + { + "epoch": 0.02, + "learning_rate": 9.994455068939212e-05, + "loss": 2.5407, + "step": 5690 + }, + { + "epoch": 0.02, + "learning_rate": 9.994445146965137e-05, + "loss": 2.4951, + "step": 5695 + }, + { + "epoch": 0.02, + "learning_rate": 9.994435216126851e-05, + "loss": 2.5494, + "step": 5700 + }, + { + "epoch": 0.02, + "learning_rate": 9.99442527642437e-05, + "loss": 2.5019, + "step": 5705 + }, + { + "epoch": 0.02, + "learning_rate": 9.994415327857717e-05, + "loss": 2.4652, + "step": 5710 + }, + { + "epoch": 0.02, + "learning_rate": 9.994405370426905e-05, + "loss": 2.4707, + "step": 5715 + }, + { + "epoch": 0.02, + "learning_rate": 9.994395404131951e-05, + "loss": 2.5165, + "step": 5720 + }, + { + "epoch": 0.02, + "learning_rate": 9.994385428972876e-05, + "loss": 2.5089, + "step": 5725 + }, + { + "epoch": 0.02, + "learning_rate": 9.994375444949698e-05, + "loss": 2.4708, + "step": 5730 + }, + { + "epoch": 0.02, + "learning_rate": 9.99436545206243e-05, + "loss": 2.584, + "step": 5735 + }, + { + "epoch": 0.02, + "learning_rate": 9.994355450311095e-05, + "loss": 2.4077, + "step": 5740 + }, + { + "epoch": 0.02, + "learning_rate": 9.994345439695707e-05, + "loss": 2.5103, + "step": 5745 + }, + { + "epoch": 0.02, + "learning_rate": 9.994335420216285e-05, + "loss": 2.5325, + "step": 5750 + }, + { + "epoch": 0.02, + "learning_rate": 9.994325391872848e-05, + "loss": 2.4675, + "step": 5755 + }, + { + "epoch": 0.02, + "learning_rate": 9.994315354665414e-05, + "loss": 2.487, + "step": 5760 + }, + { + "epoch": 0.02, + "learning_rate": 9.994305308593997e-05, + "loss": 2.6231, + "step": 5765 + }, + { + "epoch": 0.02, + "learning_rate": 9.994295253658619e-05, + "loss": 2.4284, + "step": 5770 + }, + { + "epoch": 0.02, + "learning_rate": 9.994285189859295e-05, + "loss": 2.3793, + "step": 5775 + }, + { + "epoch": 0.02, + "learning_rate": 9.994275117196048e-05, + "loss": 2.4325, + "step": 5780 + }, + { + "epoch": 0.02, + "learning_rate": 9.994265035668889e-05, + "loss": 2.5298, + "step": 5785 + }, + { + "epoch": 0.02, + "learning_rate": 9.994254945277839e-05, + "loss": 2.4958, + "step": 5790 + }, + { + "epoch": 0.02, + "learning_rate": 9.994244846022917e-05, + "loss": 2.5888, + "step": 5795 + }, + { + "epoch": 0.02, + "learning_rate": 9.994234737904138e-05, + "loss": 2.5687, + "step": 5800 + }, + { + "epoch": 0.02, + "learning_rate": 9.994224620921525e-05, + "loss": 2.5092, + "step": 5805 + }, + { + "epoch": 0.02, + "learning_rate": 9.994214495075089e-05, + "loss": 2.609, + "step": 5810 + }, + { + "epoch": 0.02, + "learning_rate": 9.994204360364854e-05, + "loss": 2.5032, + "step": 5815 + }, + { + "epoch": 0.02, + "learning_rate": 9.994194216790834e-05, + "loss": 2.4668, + "step": 5820 + }, + { + "epoch": 0.02, + "learning_rate": 9.994184064353049e-05, + "loss": 2.4375, + "step": 5825 + }, + { + "epoch": 0.02, + "learning_rate": 9.994173903051518e-05, + "loss": 2.3863, + "step": 5830 + }, + { + "epoch": 0.02, + "learning_rate": 9.994163732886257e-05, + "loss": 2.5574, + "step": 5835 + }, + { + "epoch": 0.02, + "learning_rate": 9.994153553857284e-05, + "loss": 2.485, + "step": 5840 + }, + { + "epoch": 0.02, + "learning_rate": 9.99414336596462e-05, + "loss": 2.5437, + "step": 5845 + }, + { + "epoch": 0.02, + "learning_rate": 9.994133169208279e-05, + "loss": 2.4123, + "step": 5850 + }, + { + "epoch": 0.02, + "learning_rate": 9.994122963588282e-05, + "loss": 2.4359, + "step": 5855 + }, + { + "epoch": 0.02, + "learning_rate": 9.994112749104644e-05, + "loss": 2.5166, + "step": 5860 + }, + { + "epoch": 0.02, + "learning_rate": 9.994102525757388e-05, + "loss": 2.5201, + "step": 5865 + }, + { + "epoch": 0.02, + "learning_rate": 9.994092293546527e-05, + "loss": 2.3778, + "step": 5870 + }, + { + "epoch": 0.02, + "learning_rate": 9.994082052472084e-05, + "loss": 2.4479, + "step": 5875 + }, + { + "epoch": 0.02, + "learning_rate": 9.994071802534073e-05, + "loss": 2.6381, + "step": 5880 + }, + { + "epoch": 0.02, + "learning_rate": 9.994061543732513e-05, + "loss": 2.5196, + "step": 5885 + }, + { + "epoch": 0.02, + "learning_rate": 9.994051276067426e-05, + "loss": 2.4822, + "step": 5890 + }, + { + "epoch": 0.02, + "learning_rate": 9.994040999538825e-05, + "loss": 2.5887, + "step": 5895 + }, + { + "epoch": 0.02, + "learning_rate": 9.994030714146732e-05, + "loss": 2.5642, + "step": 5900 + }, + { + "epoch": 0.02, + "learning_rate": 9.994020419891164e-05, + "loss": 2.4682, + "step": 5905 + }, + { + "epoch": 0.02, + "learning_rate": 9.994010116772139e-05, + "loss": 2.5787, + "step": 5910 + }, + { + "epoch": 0.02, + "learning_rate": 9.993999804789675e-05, + "loss": 2.6395, + "step": 5915 + }, + { + "epoch": 0.02, + "learning_rate": 9.993989483943789e-05, + "loss": 2.4215, + "step": 5920 + }, + { + "epoch": 0.02, + "learning_rate": 9.993979154234504e-05, + "loss": 2.4835, + "step": 5925 + }, + { + "epoch": 0.02, + "learning_rate": 9.993968815661834e-05, + "loss": 2.581, + "step": 5930 + }, + { + "epoch": 0.02, + "learning_rate": 9.993958468225798e-05, + "loss": 2.656, + "step": 5935 + }, + { + "epoch": 0.02, + "learning_rate": 9.993948111926416e-05, + "loss": 2.4805, + "step": 5940 + }, + { + "epoch": 0.02, + "learning_rate": 9.993937746763704e-05, + "loss": 2.504, + "step": 5945 + }, + { + "epoch": 0.02, + "learning_rate": 9.993927372737683e-05, + "loss": 2.5601, + "step": 5950 + }, + { + "epoch": 0.02, + "learning_rate": 9.99391698984837e-05, + "loss": 2.5362, + "step": 5955 + }, + { + "epoch": 0.02, + "learning_rate": 9.993906598095786e-05, + "loss": 2.547, + "step": 5960 + }, + { + "epoch": 0.02, + "learning_rate": 9.993896197479944e-05, + "loss": 2.3443, + "step": 5965 + }, + { + "epoch": 0.02, + "learning_rate": 9.993885788000868e-05, + "loss": 2.5392, + "step": 5970 + }, + { + "epoch": 0.02, + "learning_rate": 9.993875369658572e-05, + "loss": 2.5688, + "step": 5975 + }, + { + "epoch": 0.02, + "learning_rate": 9.993864942453077e-05, + "loss": 2.3431, + "step": 5980 + }, + { + "epoch": 0.02, + "learning_rate": 9.993854506384401e-05, + "loss": 2.434, + "step": 5985 + }, + { + "epoch": 0.02, + "learning_rate": 9.993844061452564e-05, + "loss": 2.4962, + "step": 5990 + }, + { + "epoch": 0.02, + "learning_rate": 9.993833607657581e-05, + "loss": 2.4566, + "step": 5995 + }, + { + "epoch": 0.02, + "learning_rate": 9.993823144999474e-05, + "loss": 2.6142, + "step": 6000 + }, + { + "epoch": 0.02, + "learning_rate": 9.99381267347826e-05, + "loss": 2.5977, + "step": 6005 + }, + { + "epoch": 0.02, + "learning_rate": 9.993802193093957e-05, + "loss": 2.4524, + "step": 6010 + }, + { + "epoch": 0.02, + "learning_rate": 9.993791703846586e-05, + "loss": 2.4292, + "step": 6015 + }, + { + "epoch": 0.02, + "learning_rate": 9.993781205736164e-05, + "loss": 2.5696, + "step": 6020 + }, + { + "epoch": 0.02, + "learning_rate": 9.993770698762708e-05, + "loss": 2.4836, + "step": 6025 + }, + { + "epoch": 0.02, + "learning_rate": 9.99376018292624e-05, + "loss": 2.5105, + "step": 6030 + }, + { + "epoch": 0.02, + "learning_rate": 9.993749658226777e-05, + "loss": 2.5682, + "step": 6035 + }, + { + "epoch": 0.02, + "learning_rate": 9.993739124664336e-05, + "loss": 2.628, + "step": 6040 + }, + { + "epoch": 0.02, + "learning_rate": 9.993728582238939e-05, + "loss": 2.5368, + "step": 6045 + }, + { + "epoch": 0.02, + "learning_rate": 9.993718030950602e-05, + "loss": 2.3607, + "step": 6050 + }, + { + "epoch": 0.02, + "learning_rate": 9.993707470799346e-05, + "loss": 2.5052, + "step": 6055 + }, + { + "epoch": 0.02, + "learning_rate": 9.993696901785187e-05, + "loss": 2.4409, + "step": 6060 + }, + { + "epoch": 0.02, + "learning_rate": 9.993686323908145e-05, + "loss": 2.4987, + "step": 6065 + }, + { + "epoch": 0.02, + "learning_rate": 9.99367573716824e-05, + "loss": 2.4024, + "step": 6070 + }, + { + "epoch": 0.02, + "learning_rate": 9.993665141565491e-05, + "loss": 2.4625, + "step": 6075 + }, + { + "epoch": 0.02, + "learning_rate": 9.993654537099915e-05, + "loss": 2.4551, + "step": 6080 + }, + { + "epoch": 0.02, + "learning_rate": 9.99364392377153e-05, + "loss": 2.6399, + "step": 6085 + }, + { + "epoch": 0.02, + "learning_rate": 9.993633301580359e-05, + "loss": 2.5569, + "step": 6090 + }, + { + "epoch": 0.02, + "learning_rate": 9.993622670526415e-05, + "loss": 2.3773, + "step": 6095 + }, + { + "epoch": 0.02, + "learning_rate": 9.993612030609722e-05, + "loss": 2.5248, + "step": 6100 + }, + { + "epoch": 0.02, + "learning_rate": 9.993601381830297e-05, + "loss": 2.5602, + "step": 6105 + }, + { + "epoch": 0.02, + "learning_rate": 9.993590724188158e-05, + "loss": 2.5535, + "step": 6110 + }, + { + "epoch": 0.02, + "learning_rate": 9.993580057683324e-05, + "loss": 2.4729, + "step": 6115 + }, + { + "epoch": 0.02, + "learning_rate": 9.993569382315817e-05, + "loss": 2.5785, + "step": 6120 + }, + { + "epoch": 0.02, + "learning_rate": 9.993558698085652e-05, + "loss": 2.5576, + "step": 6125 + }, + { + "epoch": 0.02, + "learning_rate": 9.99354800499285e-05, + "loss": 2.4801, + "step": 6130 + }, + { + "epoch": 0.02, + "learning_rate": 9.99353730303743e-05, + "loss": 2.5147, + "step": 6135 + }, + { + "epoch": 0.02, + "learning_rate": 9.993526592219411e-05, + "loss": 2.4116, + "step": 6140 + }, + { + "epoch": 0.02, + "learning_rate": 9.99351587253881e-05, + "loss": 2.4446, + "step": 6145 + }, + { + "epoch": 0.02, + "learning_rate": 9.99350514399565e-05, + "loss": 2.4093, + "step": 6150 + }, + { + "epoch": 0.02, + "learning_rate": 9.993494406589945e-05, + "loss": 2.4862, + "step": 6155 + }, + { + "epoch": 0.02, + "learning_rate": 9.993483660321719e-05, + "loss": 2.4331, + "step": 6160 + }, + { + "epoch": 0.02, + "learning_rate": 9.993472905190986e-05, + "loss": 2.4639, + "step": 6165 + }, + { + "epoch": 0.02, + "learning_rate": 9.99346214119777e-05, + "loss": 2.4757, + "step": 6170 + }, + { + "epoch": 0.02, + "learning_rate": 9.993451368342085e-05, + "loss": 2.4746, + "step": 6175 + }, + { + "epoch": 0.02, + "learning_rate": 9.993440586623957e-05, + "loss": 2.5985, + "step": 6180 + }, + { + "epoch": 0.02, + "learning_rate": 9.9934297960434e-05, + "loss": 2.5547, + "step": 6185 + }, + { + "epoch": 0.02, + "learning_rate": 9.993418996600435e-05, + "loss": 2.431, + "step": 6190 + }, + { + "epoch": 0.02, + "learning_rate": 9.993408188295079e-05, + "loss": 2.4486, + "step": 6195 + }, + { + "epoch": 0.02, + "learning_rate": 9.993397371127352e-05, + "loss": 2.4119, + "step": 6200 + }, + { + "epoch": 0.02, + "learning_rate": 9.993386545097275e-05, + "loss": 2.543, + "step": 6205 + }, + { + "epoch": 0.02, + "learning_rate": 9.993375710204866e-05, + "loss": 2.5253, + "step": 6210 + }, + { + "epoch": 0.02, + "learning_rate": 9.993364866450146e-05, + "loss": 2.4264, + "step": 6215 + }, + { + "epoch": 0.02, + "learning_rate": 9.99335401383313e-05, + "loss": 2.6681, + "step": 6220 + }, + { + "epoch": 0.02, + "learning_rate": 9.993343152353841e-05, + "loss": 2.4156, + "step": 6225 + }, + { + "epoch": 0.02, + "learning_rate": 9.993332282012298e-05, + "loss": 2.5415, + "step": 6230 + }, + { + "epoch": 0.02, + "learning_rate": 9.993321402808518e-05, + "loss": 2.4505, + "step": 6235 + }, + { + "epoch": 0.02, + "learning_rate": 9.993310514742522e-05, + "loss": 2.59, + "step": 6240 + }, + { + "epoch": 0.02, + "learning_rate": 9.99329961781433e-05, + "loss": 2.4838, + "step": 6245 + }, + { + "epoch": 0.02, + "learning_rate": 9.99328871202396e-05, + "loss": 2.4487, + "step": 6250 + }, + { + "epoch": 0.02, + "learning_rate": 9.993277797371432e-05, + "loss": 2.4305, + "step": 6255 + }, + { + "epoch": 0.02, + "learning_rate": 9.993266873856763e-05, + "loss": 2.5547, + "step": 6260 + }, + { + "epoch": 0.02, + "learning_rate": 9.993255941479978e-05, + "loss": 2.5284, + "step": 6265 + }, + { + "epoch": 0.02, + "learning_rate": 9.993245000241091e-05, + "loss": 2.5835, + "step": 6270 + }, + { + "epoch": 0.02, + "learning_rate": 9.993234050140124e-05, + "loss": 2.5343, + "step": 6275 + }, + { + "epoch": 0.02, + "learning_rate": 9.993223091177095e-05, + "loss": 2.4057, + "step": 6280 + }, + { + "epoch": 0.02, + "learning_rate": 9.993212123352024e-05, + "loss": 2.5, + "step": 6285 + }, + { + "epoch": 0.02, + "learning_rate": 9.993201146664932e-05, + "loss": 2.489, + "step": 6290 + }, + { + "epoch": 0.02, + "learning_rate": 9.993190161115835e-05, + "loss": 2.5042, + "step": 6295 + }, + { + "epoch": 0.02, + "learning_rate": 9.993179166704755e-05, + "loss": 2.4519, + "step": 6300 + }, + { + "epoch": 0.02, + "learning_rate": 9.993168163431715e-05, + "loss": 2.4857, + "step": 6305 + }, + { + "epoch": 0.02, + "learning_rate": 9.993157151296727e-05, + "loss": 2.4246, + "step": 6310 + }, + { + "epoch": 0.02, + "learning_rate": 9.993146130299815e-05, + "loss": 2.5357, + "step": 6315 + }, + { + "epoch": 0.02, + "learning_rate": 9.993135100440997e-05, + "loss": 2.5779, + "step": 6320 + }, + { + "epoch": 0.02, + "learning_rate": 9.993124061720296e-05, + "loss": 2.4707, + "step": 6325 + }, + { + "epoch": 0.02, + "learning_rate": 9.993113014137728e-05, + "loss": 2.5165, + "step": 6330 + }, + { + "epoch": 0.02, + "learning_rate": 9.993101957693312e-05, + "loss": 2.4793, + "step": 6335 + }, + { + "epoch": 0.02, + "learning_rate": 9.99309089238707e-05, + "loss": 2.5137, + "step": 6340 + }, + { + "epoch": 0.02, + "learning_rate": 9.993079818219021e-05, + "loss": 2.4366, + "step": 6345 + }, + { + "epoch": 0.02, + "learning_rate": 9.993068735189185e-05, + "loss": 2.4102, + "step": 6350 + }, + { + "epoch": 0.02, + "learning_rate": 9.993057643297582e-05, + "loss": 2.4449, + "step": 6355 + }, + { + "epoch": 0.02, + "learning_rate": 9.993046542544229e-05, + "loss": 2.4961, + "step": 6360 + }, + { + "epoch": 0.02, + "learning_rate": 9.993035432929148e-05, + "loss": 2.4996, + "step": 6365 + }, + { + "epoch": 0.02, + "learning_rate": 9.993024314452357e-05, + "loss": 2.566, + "step": 6370 + }, + { + "epoch": 0.02, + "learning_rate": 9.99301318711388e-05, + "loss": 2.5317, + "step": 6375 + }, + { + "epoch": 0.02, + "learning_rate": 9.993002050913732e-05, + "loss": 2.5516, + "step": 6380 + }, + { + "epoch": 0.02, + "learning_rate": 9.992990905851935e-05, + "loss": 2.5118, + "step": 6385 + }, + { + "epoch": 0.02, + "learning_rate": 9.992979751928508e-05, + "loss": 2.4617, + "step": 6390 + }, + { + "epoch": 0.02, + "learning_rate": 9.992968589143471e-05, + "loss": 2.5495, + "step": 6395 + }, + { + "epoch": 0.02, + "learning_rate": 9.992957417496845e-05, + "loss": 2.5978, + "step": 6400 + }, + { + "epoch": 0.02, + "learning_rate": 9.992946236988648e-05, + "loss": 2.4981, + "step": 6405 + }, + { + "epoch": 0.02, + "learning_rate": 9.992935047618899e-05, + "loss": 2.4454, + "step": 6410 + }, + { + "epoch": 0.02, + "learning_rate": 9.99292384938762e-05, + "loss": 2.4337, + "step": 6415 + }, + { + "epoch": 0.02, + "learning_rate": 9.992912642294832e-05, + "loss": 2.5264, + "step": 6420 + }, + { + "epoch": 0.02, + "learning_rate": 9.992901426340553e-05, + "loss": 2.3756, + "step": 6425 + }, + { + "epoch": 0.02, + "learning_rate": 9.992890201524803e-05, + "loss": 2.4695, + "step": 6430 + }, + { + "epoch": 0.02, + "learning_rate": 9.992878967847601e-05, + "loss": 2.4767, + "step": 6435 + }, + { + "epoch": 0.02, + "learning_rate": 9.992867725308969e-05, + "loss": 2.4789, + "step": 6440 + }, + { + "epoch": 0.02, + "learning_rate": 9.992856473908924e-05, + "loss": 2.4447, + "step": 6445 + }, + { + "epoch": 0.02, + "learning_rate": 9.99284521364749e-05, + "loss": 2.41, + "step": 6450 + }, + { + "epoch": 0.02, + "learning_rate": 9.992833944524685e-05, + "loss": 2.5547, + "step": 6455 + }, + { + "epoch": 0.02, + "learning_rate": 9.992822666540529e-05, + "loss": 2.5471, + "step": 6460 + }, + { + "epoch": 0.02, + "learning_rate": 9.99281137969504e-05, + "loss": 2.4726, + "step": 6465 + }, + { + "epoch": 0.02, + "learning_rate": 9.992800083988241e-05, + "loss": 2.4243, + "step": 6470 + }, + { + "epoch": 0.02, + "learning_rate": 9.992788779420152e-05, + "loss": 2.4842, + "step": 6475 + }, + { + "epoch": 0.02, + "learning_rate": 9.99277746599079e-05, + "loss": 2.5104, + "step": 6480 + }, + { + "epoch": 0.02, + "learning_rate": 9.992766143700179e-05, + "loss": 2.576, + "step": 6485 + }, + { + "epoch": 0.02, + "learning_rate": 9.992754812548338e-05, + "loss": 2.5034, + "step": 6490 + }, + { + "epoch": 0.02, + "learning_rate": 9.992743472535284e-05, + "loss": 2.4899, + "step": 6495 + }, + { + "epoch": 0.02, + "learning_rate": 9.99273212366104e-05, + "loss": 2.4641, + "step": 6500 + }, + { + "epoch": 0.02, + "learning_rate": 9.992720765925625e-05, + "loss": 2.5135, + "step": 6505 + }, + { + "epoch": 0.02, + "learning_rate": 9.992709399329062e-05, + "loss": 2.6289, + "step": 6510 + }, + { + "epoch": 0.02, + "learning_rate": 9.992698023871368e-05, + "loss": 2.6081, + "step": 6515 + }, + { + "epoch": 0.02, + "learning_rate": 9.992686639552563e-05, + "loss": 2.4959, + "step": 6520 + }, + { + "epoch": 0.02, + "learning_rate": 9.99267524637267e-05, + "loss": 2.5205, + "step": 6525 + }, + { + "epoch": 0.02, + "learning_rate": 9.992663844331706e-05, + "loss": 2.4894, + "step": 6530 + }, + { + "epoch": 0.02, + "learning_rate": 9.992652433429692e-05, + "loss": 2.535, + "step": 6535 + }, + { + "epoch": 0.02, + "learning_rate": 9.992641013666651e-05, + "loss": 2.4253, + "step": 6540 + }, + { + "epoch": 0.02, + "learning_rate": 9.9926295850426e-05, + "loss": 2.4492, + "step": 6545 + }, + { + "epoch": 0.02, + "learning_rate": 9.992618147557562e-05, + "loss": 2.4499, + "step": 6550 + }, + { + "epoch": 0.02, + "learning_rate": 9.992606701211555e-05, + "loss": 2.4673, + "step": 6555 + }, + { + "epoch": 0.02, + "learning_rate": 9.992595246004602e-05, + "loss": 2.5089, + "step": 6560 + }, + { + "epoch": 0.02, + "learning_rate": 9.992583781936719e-05, + "loss": 2.4988, + "step": 6565 + }, + { + "epoch": 0.02, + "learning_rate": 9.99257230900793e-05, + "loss": 2.5587, + "step": 6570 + }, + { + "epoch": 0.02, + "learning_rate": 9.992560827218253e-05, + "loss": 2.4648, + "step": 6575 + }, + { + "epoch": 0.02, + "learning_rate": 9.992549336567712e-05, + "loss": 2.4648, + "step": 6580 + }, + { + "epoch": 0.02, + "learning_rate": 9.992537837056322e-05, + "loss": 2.5698, + "step": 6585 + }, + { + "epoch": 0.02, + "learning_rate": 9.992526328684109e-05, + "loss": 2.4249, + "step": 6590 + }, + { + "epoch": 0.02, + "learning_rate": 9.992514811451091e-05, + "loss": 2.4787, + "step": 6595 + }, + { + "epoch": 0.02, + "learning_rate": 9.992503285357287e-05, + "loss": 2.5148, + "step": 6600 + }, + { + "epoch": 0.02, + "learning_rate": 9.992491750402718e-05, + "loss": 2.5326, + "step": 6605 + }, + { + "epoch": 0.02, + "learning_rate": 9.992480206587407e-05, + "loss": 2.4483, + "step": 6610 + }, + { + "epoch": 0.02, + "learning_rate": 9.992468653911371e-05, + "loss": 2.4419, + "step": 6615 + }, + { + "epoch": 0.02, + "learning_rate": 9.992457092374634e-05, + "loss": 2.3896, + "step": 6620 + }, + { + "epoch": 0.02, + "learning_rate": 9.992445521977214e-05, + "loss": 2.4464, + "step": 6625 + }, + { + "epoch": 0.02, + "learning_rate": 9.992433942719131e-05, + "loss": 2.3422, + "step": 6630 + }, + { + "epoch": 0.02, + "learning_rate": 9.992422354600408e-05, + "loss": 2.5235, + "step": 6635 + }, + { + "epoch": 0.02, + "learning_rate": 9.992410757621065e-05, + "loss": 2.4164, + "step": 6640 + }, + { + "epoch": 0.02, + "learning_rate": 9.992399151781121e-05, + "loss": 2.4964, + "step": 6645 + }, + { + "epoch": 0.02, + "learning_rate": 9.992387537080599e-05, + "loss": 2.5693, + "step": 6650 + }, + { + "epoch": 0.02, + "learning_rate": 9.992375913519516e-05, + "loss": 2.6035, + "step": 6655 + }, + { + "epoch": 0.02, + "learning_rate": 9.992364281097895e-05, + "loss": 2.5602, + "step": 6660 + }, + { + "epoch": 0.02, + "learning_rate": 9.992352639815758e-05, + "loss": 2.3888, + "step": 6665 + }, + { + "epoch": 0.02, + "learning_rate": 9.992340989673123e-05, + "loss": 2.4818, + "step": 6670 + }, + { + "epoch": 0.02, + "learning_rate": 9.992329330670013e-05, + "loss": 2.484, + "step": 6675 + }, + { + "epoch": 0.02, + "learning_rate": 9.992317662806447e-05, + "loss": 2.6081, + "step": 6680 + }, + { + "epoch": 0.02, + "learning_rate": 9.992305986082447e-05, + "loss": 2.4047, + "step": 6685 + }, + { + "epoch": 0.02, + "learning_rate": 9.99229430049803e-05, + "loss": 2.5581, + "step": 6690 + }, + { + "epoch": 0.02, + "learning_rate": 9.992282606053223e-05, + "loss": 2.4793, + "step": 6695 + }, + { + "epoch": 0.02, + "learning_rate": 9.992270902748041e-05, + "loss": 2.5646, + "step": 6700 + }, + { + "epoch": 0.02, + "learning_rate": 9.992259190582508e-05, + "loss": 2.5994, + "step": 6705 + }, + { + "epoch": 0.02, + "learning_rate": 9.992247469556643e-05, + "loss": 2.515, + "step": 6710 + }, + { + "epoch": 0.02, + "learning_rate": 9.99223573967047e-05, + "loss": 2.5188, + "step": 6715 + }, + { + "epoch": 0.02, + "learning_rate": 9.992224000924006e-05, + "loss": 2.5379, + "step": 6720 + }, + { + "epoch": 0.02, + "learning_rate": 9.992212253317275e-05, + "loss": 2.5358, + "step": 6725 + }, + { + "epoch": 0.02, + "learning_rate": 9.992200496850294e-05, + "loss": 2.5234, + "step": 6730 + }, + { + "epoch": 0.02, + "learning_rate": 9.992188731523087e-05, + "loss": 2.5504, + "step": 6735 + }, + { + "epoch": 0.02, + "learning_rate": 9.992176957335675e-05, + "loss": 2.5583, + "step": 6740 + }, + { + "epoch": 0.02, + "learning_rate": 9.992165174288077e-05, + "loss": 2.4408, + "step": 6745 + }, + { + "epoch": 0.02, + "learning_rate": 9.992153382380315e-05, + "loss": 2.4923, + "step": 6750 + }, + { + "epoch": 0.02, + "learning_rate": 9.99214158161241e-05, + "loss": 2.4231, + "step": 6755 + }, + { + "epoch": 0.02, + "learning_rate": 9.992129771984382e-05, + "loss": 2.4993, + "step": 6760 + }, + { + "epoch": 0.02, + "learning_rate": 9.992117953496253e-05, + "loss": 2.5424, + "step": 6765 + }, + { + "epoch": 0.02, + "learning_rate": 9.992106126148045e-05, + "loss": 2.5124, + "step": 6770 + }, + { + "epoch": 0.02, + "learning_rate": 9.992094289939775e-05, + "loss": 2.4191, + "step": 6775 + }, + { + "epoch": 0.02, + "learning_rate": 9.99208244487147e-05, + "loss": 2.5457, + "step": 6780 + }, + { + "epoch": 0.02, + "learning_rate": 9.992070590943145e-05, + "loss": 2.5115, + "step": 6785 + }, + { + "epoch": 0.02, + "learning_rate": 9.992058728154824e-05, + "loss": 2.5263, + "step": 6790 + }, + { + "epoch": 0.02, + "learning_rate": 9.992046856506528e-05, + "loss": 2.5923, + "step": 6795 + }, + { + "epoch": 0.02, + "learning_rate": 9.99203497599828e-05, + "loss": 2.514, + "step": 6800 + }, + { + "epoch": 0.02, + "learning_rate": 9.992023086630096e-05, + "loss": 2.5133, + "step": 6805 + }, + { + "epoch": 0.02, + "learning_rate": 9.992011188402e-05, + "loss": 2.5589, + "step": 6810 + }, + { + "epoch": 0.02, + "learning_rate": 9.991999281314015e-05, + "loss": 2.4929, + "step": 6815 + }, + { + "epoch": 0.02, + "learning_rate": 9.991987365366158e-05, + "loss": 2.4639, + "step": 6820 + }, + { + "epoch": 0.02, + "learning_rate": 9.991975440558454e-05, + "loss": 2.5777, + "step": 6825 + }, + { + "epoch": 0.02, + "learning_rate": 9.991963506890922e-05, + "loss": 2.5107, + "step": 6830 + }, + { + "epoch": 0.02, + "learning_rate": 9.991951564363584e-05, + "loss": 2.4649, + "step": 6835 + }, + { + "epoch": 0.02, + "learning_rate": 9.991939612976461e-05, + "loss": 2.5572, + "step": 6840 + }, + { + "epoch": 0.02, + "learning_rate": 9.991927652729576e-05, + "loss": 2.5491, + "step": 6845 + }, + { + "epoch": 0.02, + "learning_rate": 9.991915683622946e-05, + "loss": 2.4988, + "step": 6850 + }, + { + "epoch": 0.02, + "learning_rate": 9.991903705656596e-05, + "loss": 2.5925, + "step": 6855 + }, + { + "epoch": 0.02, + "learning_rate": 9.991891718830544e-05, + "loss": 2.5064, + "step": 6860 + }, + { + "epoch": 0.02, + "learning_rate": 9.991879723144815e-05, + "loss": 2.3961, + "step": 6865 + }, + { + "epoch": 0.02, + "learning_rate": 9.991867718599427e-05, + "loss": 2.5329, + "step": 6870 + }, + { + "epoch": 0.02, + "learning_rate": 9.991855705194404e-05, + "loss": 2.4636, + "step": 6875 + }, + { + "epoch": 0.02, + "learning_rate": 9.991843682929764e-05, + "loss": 2.4138, + "step": 6880 + }, + { + "epoch": 0.02, + "learning_rate": 9.991831651805534e-05, + "loss": 2.4977, + "step": 6885 + }, + { + "epoch": 0.02, + "learning_rate": 9.991819611821729e-05, + "loss": 2.4877, + "step": 6890 + }, + { + "epoch": 0.02, + "learning_rate": 9.991807562978374e-05, + "loss": 2.6014, + "step": 6895 + }, + { + "epoch": 0.02, + "learning_rate": 9.99179550527549e-05, + "loss": 2.5161, + "step": 6900 + }, + { + "epoch": 0.02, + "learning_rate": 9.991783438713096e-05, + "loss": 2.4089, + "step": 6905 + }, + { + "epoch": 0.02, + "learning_rate": 9.991771363291218e-05, + "loss": 2.5049, + "step": 6910 + }, + { + "epoch": 0.02, + "learning_rate": 9.991759279009873e-05, + "loss": 2.474, + "step": 6915 + }, + { + "epoch": 0.02, + "learning_rate": 9.991747185869085e-05, + "loss": 2.5171, + "step": 6920 + }, + { + "epoch": 0.02, + "learning_rate": 9.991735083868873e-05, + "loss": 2.5643, + "step": 6925 + }, + { + "epoch": 0.02, + "learning_rate": 9.991722973009261e-05, + "loss": 2.5242, + "step": 6930 + }, + { + "epoch": 0.02, + "learning_rate": 9.991710853290272e-05, + "loss": 2.5631, + "step": 6935 + }, + { + "epoch": 0.02, + "learning_rate": 9.991698724711923e-05, + "loss": 2.5883, + "step": 6940 + }, + { + "epoch": 0.02, + "learning_rate": 9.991686587274239e-05, + "loss": 2.5191, + "step": 6945 + }, + { + "epoch": 0.02, + "learning_rate": 9.991674440977239e-05, + "loss": 2.4564, + "step": 6950 + }, + { + "epoch": 0.02, + "learning_rate": 9.991662285820945e-05, + "loss": 2.5212, + "step": 6955 + }, + { + "epoch": 0.02, + "learning_rate": 9.99165012180538e-05, + "loss": 2.4176, + "step": 6960 + }, + { + "epoch": 0.02, + "learning_rate": 9.991637948930566e-05, + "loss": 2.4187, + "step": 6965 + }, + { + "epoch": 0.02, + "learning_rate": 9.991625767196524e-05, + "loss": 2.4732, + "step": 6970 + }, + { + "epoch": 0.02, + "learning_rate": 9.991613576603273e-05, + "loss": 2.3875, + "step": 6975 + }, + { + "epoch": 0.02, + "learning_rate": 9.991601377150839e-05, + "loss": 2.5538, + "step": 6980 + }, + { + "epoch": 0.02, + "learning_rate": 9.991589168839239e-05, + "loss": 2.4546, + "step": 6985 + }, + { + "epoch": 0.02, + "learning_rate": 9.9915769516685e-05, + "loss": 2.5926, + "step": 6990 + }, + { + "epoch": 0.02, + "learning_rate": 9.991564725638638e-05, + "loss": 2.3919, + "step": 6995 + }, + { + "epoch": 0.02, + "learning_rate": 9.99155249074968e-05, + "loss": 2.5011, + "step": 7000 + }, + { + "epoch": 0.02, + "learning_rate": 9.991540247001644e-05, + "loss": 2.4821, + "step": 7005 + }, + { + "epoch": 0.02, + "learning_rate": 9.991527994394553e-05, + "loss": 2.4576, + "step": 7010 + }, + { + "epoch": 0.02, + "learning_rate": 9.991515732928429e-05, + "loss": 2.4592, + "step": 7015 + }, + { + "epoch": 0.02, + "learning_rate": 9.991503462603294e-05, + "loss": 2.4523, + "step": 7020 + }, + { + "epoch": 0.02, + "learning_rate": 9.991491183419168e-05, + "loss": 2.5185, + "step": 7025 + }, + { + "epoch": 0.02, + "learning_rate": 9.991478895376074e-05, + "loss": 2.4971, + "step": 7030 + }, + { + "epoch": 0.02, + "learning_rate": 9.991466598474036e-05, + "loss": 2.5715, + "step": 7035 + }, + { + "epoch": 0.02, + "learning_rate": 9.991454292713071e-05, + "loss": 2.5967, + "step": 7040 + }, + { + "epoch": 0.02, + "learning_rate": 9.991441978093204e-05, + "loss": 2.577, + "step": 7045 + }, + { + "epoch": 0.02, + "learning_rate": 9.991429654614457e-05, + "loss": 2.4019, + "step": 7050 + }, + { + "epoch": 0.02, + "learning_rate": 9.991417322276852e-05, + "loss": 2.5786, + "step": 7055 + }, + { + "epoch": 0.02, + "learning_rate": 9.991404981080409e-05, + "loss": 2.5297, + "step": 7060 + }, + { + "epoch": 0.02, + "learning_rate": 9.991392631025152e-05, + "loss": 2.582, + "step": 7065 + }, + { + "epoch": 0.02, + "learning_rate": 9.991380272111101e-05, + "loss": 2.4266, + "step": 7070 + }, + { + "epoch": 0.02, + "learning_rate": 9.991367904338278e-05, + "loss": 2.401, + "step": 7075 + }, + { + "epoch": 0.02, + "learning_rate": 9.991355527706707e-05, + "loss": 2.4574, + "step": 7080 + }, + { + "epoch": 0.02, + "learning_rate": 9.99134314221641e-05, + "loss": 2.4694, + "step": 7085 + }, + { + "epoch": 0.02, + "learning_rate": 9.991330747867406e-05, + "loss": 2.4436, + "step": 7090 + }, + { + "epoch": 0.02, + "learning_rate": 9.991318344659719e-05, + "loss": 2.5056, + "step": 7095 + }, + { + "epoch": 0.02, + "learning_rate": 9.991305932593372e-05, + "loss": 2.4281, + "step": 7100 + }, + { + "epoch": 0.02, + "learning_rate": 9.991293511668385e-05, + "loss": 2.4965, + "step": 7105 + }, + { + "epoch": 0.02, + "learning_rate": 9.99128108188478e-05, + "loss": 2.4363, + "step": 7110 + }, + { + "epoch": 0.02, + "learning_rate": 9.991268643242582e-05, + "loss": 2.4993, + "step": 7115 + }, + { + "epoch": 0.02, + "learning_rate": 9.991256195741809e-05, + "loss": 2.4482, + "step": 7120 + }, + { + "epoch": 0.02, + "learning_rate": 9.991243739382487e-05, + "loss": 2.4937, + "step": 7125 + }, + { + "epoch": 0.02, + "learning_rate": 9.991231274164634e-05, + "loss": 2.4623, + "step": 7130 + }, + { + "epoch": 0.02, + "learning_rate": 9.991218800088275e-05, + "loss": 2.4815, + "step": 7135 + }, + { + "epoch": 0.02, + "learning_rate": 9.991206317153433e-05, + "loss": 2.5503, + "step": 7140 + }, + { + "epoch": 0.02, + "learning_rate": 9.991193825360128e-05, + "loss": 2.4936, + "step": 7145 + }, + { + "epoch": 0.02, + "learning_rate": 9.991181324708382e-05, + "loss": 2.501, + "step": 7150 + }, + { + "epoch": 0.02, + "learning_rate": 9.991168815198219e-05, + "loss": 2.5072, + "step": 7155 + }, + { + "epoch": 0.02, + "learning_rate": 9.99115629682966e-05, + "loss": 2.5551, + "step": 7160 + }, + { + "epoch": 0.02, + "learning_rate": 9.991143769602726e-05, + "loss": 2.4484, + "step": 7165 + }, + { + "epoch": 0.02, + "learning_rate": 9.99113123351744e-05, + "loss": 2.5272, + "step": 7170 + }, + { + "epoch": 0.02, + "learning_rate": 9.991118688573829e-05, + "loss": 2.3679, + "step": 7175 + }, + { + "epoch": 0.02, + "learning_rate": 9.991106134771909e-05, + "loss": 2.3734, + "step": 7180 + }, + { + "epoch": 0.02, + "learning_rate": 9.991093572111703e-05, + "loss": 2.3481, + "step": 7185 + }, + { + "epoch": 0.02, + "learning_rate": 9.991081000593236e-05, + "loss": 2.5051, + "step": 7190 + }, + { + "epoch": 0.02, + "learning_rate": 9.99106842021653e-05, + "loss": 2.4764, + "step": 7195 + }, + { + "epoch": 0.02, + "learning_rate": 9.991055830981603e-05, + "loss": 2.5851, + "step": 7200 + }, + { + "epoch": 0.02, + "learning_rate": 9.991043232888483e-05, + "loss": 2.5048, + "step": 7205 + }, + { + "epoch": 0.02, + "learning_rate": 9.99103062593719e-05, + "loss": 2.5964, + "step": 7210 + }, + { + "epoch": 0.02, + "learning_rate": 9.991018010127748e-05, + "loss": 2.5519, + "step": 7215 + }, + { + "epoch": 0.02, + "learning_rate": 9.991005385460176e-05, + "loss": 2.4752, + "step": 7220 + }, + { + "epoch": 0.02, + "learning_rate": 9.990992751934499e-05, + "loss": 2.4413, + "step": 7225 + }, + { + "epoch": 0.02, + "learning_rate": 9.990980109550738e-05, + "loss": 2.4485, + "step": 7230 + }, + { + "epoch": 0.02, + "learning_rate": 9.990967458308916e-05, + "loss": 2.5227, + "step": 7235 + }, + { + "epoch": 0.02, + "learning_rate": 9.990954798209055e-05, + "loss": 2.435, + "step": 7240 + }, + { + "epoch": 0.02, + "learning_rate": 9.990942129251179e-05, + "loss": 2.4672, + "step": 7245 + }, + { + "epoch": 0.02, + "learning_rate": 9.990929451435309e-05, + "loss": 2.4566, + "step": 7250 + }, + { + "epoch": 0.02, + "learning_rate": 9.990916764761468e-05, + "loss": 2.4668, + "step": 7255 + }, + { + "epoch": 0.02, + "learning_rate": 9.99090406922968e-05, + "loss": 2.5705, + "step": 7260 + }, + { + "epoch": 0.02, + "learning_rate": 9.990891364839965e-05, + "loss": 2.3825, + "step": 7265 + }, + { + "epoch": 0.02, + "learning_rate": 9.990878651592345e-05, + "loss": 2.3839, + "step": 7270 + }, + { + "epoch": 0.02, + "learning_rate": 9.990865929486847e-05, + "loss": 2.5993, + "step": 7275 + }, + { + "epoch": 0.02, + "learning_rate": 9.99085319852349e-05, + "loss": 2.548, + "step": 7280 + }, + { + "epoch": 0.02, + "learning_rate": 9.990840458702298e-05, + "loss": 2.4886, + "step": 7285 + }, + { + "epoch": 0.02, + "learning_rate": 9.99082771002329e-05, + "loss": 2.3899, + "step": 7290 + }, + { + "epoch": 0.02, + "learning_rate": 9.990814952486494e-05, + "loss": 2.5297, + "step": 7295 + }, + { + "epoch": 0.02, + "learning_rate": 9.99080218609193e-05, + "loss": 2.49, + "step": 7300 + }, + { + "epoch": 0.02, + "learning_rate": 9.99078941083962e-05, + "loss": 2.3959, + "step": 7305 + }, + { + "epoch": 0.02, + "learning_rate": 9.990776626729588e-05, + "loss": 2.3687, + "step": 7310 + }, + { + "epoch": 0.02, + "learning_rate": 9.990763833761855e-05, + "loss": 2.5197, + "step": 7315 + }, + { + "epoch": 0.02, + "learning_rate": 9.990751031936447e-05, + "loss": 2.5301, + "step": 7320 + }, + { + "epoch": 0.02, + "learning_rate": 9.990738221253384e-05, + "loss": 2.5116, + "step": 7325 + }, + { + "epoch": 0.02, + "learning_rate": 9.99072540171269e-05, + "loss": 2.4939, + "step": 7330 + }, + { + "epoch": 0.02, + "learning_rate": 9.990712573314387e-05, + "loss": 2.5061, + "step": 7335 + }, + { + "epoch": 0.02, + "learning_rate": 9.990699736058496e-05, + "loss": 2.4358, + "step": 7340 + }, + { + "epoch": 0.02, + "learning_rate": 9.990686889945044e-05, + "loss": 2.6017, + "step": 7345 + }, + { + "epoch": 0.02, + "learning_rate": 9.99067403497405e-05, + "loss": 2.4977, + "step": 7350 + }, + { + "epoch": 0.02, + "learning_rate": 9.990661171145539e-05, + "loss": 2.5031, + "step": 7355 + }, + { + "epoch": 0.02, + "learning_rate": 9.990648298459533e-05, + "loss": 2.4646, + "step": 7360 + }, + { + "epoch": 0.02, + "learning_rate": 9.990635416916055e-05, + "loss": 2.5642, + "step": 7365 + }, + { + "epoch": 0.02, + "learning_rate": 9.990622526515129e-05, + "loss": 2.5477, + "step": 7370 + }, + { + "epoch": 0.02, + "learning_rate": 9.990609627256775e-05, + "loss": 2.3807, + "step": 7375 + }, + { + "epoch": 0.02, + "learning_rate": 9.990596719141019e-05, + "loss": 2.4526, + "step": 7380 + }, + { + "epoch": 0.02, + "learning_rate": 9.990583802167881e-05, + "loss": 2.4713, + "step": 7385 + }, + { + "epoch": 0.02, + "learning_rate": 9.990570876337388e-05, + "loss": 2.545, + "step": 7390 + }, + { + "epoch": 0.02, + "learning_rate": 9.990557941649559e-05, + "loss": 2.4925, + "step": 7395 + }, + { + "epoch": 0.02, + "learning_rate": 9.990544998104417e-05, + "loss": 2.5438, + "step": 7400 + }, + { + "epoch": 0.02, + "learning_rate": 9.990532045701987e-05, + "loss": 2.4221, + "step": 7405 + }, + { + "epoch": 0.02, + "learning_rate": 9.990519084442292e-05, + "loss": 2.4882, + "step": 7410 + }, + { + "epoch": 0.02, + "learning_rate": 9.990506114325356e-05, + "loss": 2.4709, + "step": 7415 + }, + { + "epoch": 0.02, + "learning_rate": 9.990493135351198e-05, + "loss": 2.5193, + "step": 7420 + }, + { + "epoch": 0.02, + "learning_rate": 9.990480147519843e-05, + "loss": 2.5678, + "step": 7425 + }, + { + "epoch": 0.02, + "learning_rate": 9.990467150831316e-05, + "loss": 2.414, + "step": 7430 + }, + { + "epoch": 0.02, + "learning_rate": 9.990454145285637e-05, + "loss": 2.4043, + "step": 7435 + }, + { + "epoch": 0.02, + "learning_rate": 9.990441130882833e-05, + "loss": 2.3485, + "step": 7440 + }, + { + "epoch": 0.02, + "learning_rate": 9.990428107622922e-05, + "loss": 2.4291, + "step": 7445 + }, + { + "epoch": 0.02, + "learning_rate": 9.990415075505931e-05, + "loss": 2.5245, + "step": 7450 + }, + { + "epoch": 0.02, + "learning_rate": 9.990402034531882e-05, + "loss": 2.4992, + "step": 7455 + }, + { + "epoch": 0.02, + "learning_rate": 9.990388984700798e-05, + "loss": 2.4564, + "step": 7460 + }, + { + "epoch": 0.02, + "learning_rate": 9.990375926012703e-05, + "loss": 2.4806, + "step": 7465 + }, + { + "epoch": 0.02, + "learning_rate": 9.990362858467618e-05, + "loss": 2.3611, + "step": 7470 + }, + { + "epoch": 0.02, + "learning_rate": 9.990349782065569e-05, + "loss": 2.3896, + "step": 7475 + }, + { + "epoch": 0.02, + "learning_rate": 9.990336696806577e-05, + "loss": 2.5089, + "step": 7480 + }, + { + "epoch": 0.02, + "learning_rate": 9.990323602690666e-05, + "loss": 2.4273, + "step": 7485 + }, + { + "epoch": 0.02, + "learning_rate": 9.990310499717859e-05, + "loss": 2.4029, + "step": 7490 + }, + { + "epoch": 0.02, + "learning_rate": 9.990297387888179e-05, + "loss": 2.5591, + "step": 7495 + }, + { + "epoch": 0.02, + "learning_rate": 9.990284267201651e-05, + "loss": 2.5342, + "step": 7500 + }, + { + "epoch": 0.02, + "learning_rate": 9.990271137658297e-05, + "loss": 2.432, + "step": 7505 + }, + { + "epoch": 0.02, + "learning_rate": 9.990257999258141e-05, + "loss": 2.4728, + "step": 7510 + }, + { + "epoch": 0.02, + "learning_rate": 9.990244852001204e-05, + "loss": 2.4883, + "step": 7515 + }, + { + "epoch": 0.02, + "learning_rate": 9.990231695887513e-05, + "loss": 2.4024, + "step": 7520 + }, + { + "epoch": 0.02, + "learning_rate": 9.990218530917089e-05, + "loss": 2.3964, + "step": 7525 + }, + { + "epoch": 0.02, + "learning_rate": 9.990205357089954e-05, + "loss": 2.4947, + "step": 7530 + }, + { + "epoch": 0.02, + "learning_rate": 9.990192174406134e-05, + "loss": 2.4646, + "step": 7535 + }, + { + "epoch": 0.02, + "learning_rate": 9.990178982865652e-05, + "loss": 2.5044, + "step": 7540 + }, + { + "epoch": 0.02, + "learning_rate": 9.99016578246853e-05, + "loss": 2.4765, + "step": 7545 + }, + { + "epoch": 0.02, + "learning_rate": 9.990152573214794e-05, + "loss": 2.6132, + "step": 7550 + }, + { + "epoch": 0.02, + "learning_rate": 9.990139355104464e-05, + "loss": 2.538, + "step": 7555 + }, + { + "epoch": 0.02, + "learning_rate": 9.990126128137566e-05, + "loss": 2.3061, + "step": 7560 + }, + { + "epoch": 0.02, + "learning_rate": 9.990112892314124e-05, + "loss": 2.5848, + "step": 7565 + }, + { + "epoch": 0.02, + "learning_rate": 9.990099647634159e-05, + "loss": 2.4529, + "step": 7570 + }, + { + "epoch": 0.02, + "learning_rate": 9.990086394097696e-05, + "loss": 2.405, + "step": 7575 + }, + { + "epoch": 0.02, + "learning_rate": 9.990073131704759e-05, + "loss": 2.4112, + "step": 7580 + }, + { + "epoch": 0.02, + "learning_rate": 9.990059860455369e-05, + "loss": 2.4378, + "step": 7585 + }, + { + "epoch": 0.02, + "learning_rate": 9.990046580349554e-05, + "loss": 2.6024, + "step": 7590 + }, + { + "epoch": 0.02, + "learning_rate": 9.990033291387332e-05, + "loss": 2.6397, + "step": 7595 + }, + { + "epoch": 0.02, + "learning_rate": 9.990019993568731e-05, + "loss": 2.5278, + "step": 7600 + }, + { + "epoch": 0.02, + "learning_rate": 9.990006686893773e-05, + "loss": 2.4834, + "step": 7605 + }, + { + "epoch": 0.02, + "learning_rate": 9.989993371362482e-05, + "loss": 2.5141, + "step": 7610 + }, + { + "epoch": 0.02, + "learning_rate": 9.989980046974882e-05, + "loss": 2.5069, + "step": 7615 + }, + { + "epoch": 0.02, + "learning_rate": 9.989966713730995e-05, + "loss": 2.5923, + "step": 7620 + }, + { + "epoch": 0.02, + "learning_rate": 9.989953371630847e-05, + "loss": 2.4649, + "step": 7625 + }, + { + "epoch": 0.02, + "learning_rate": 9.989940020674459e-05, + "loss": 2.41, + "step": 7630 + }, + { + "epoch": 0.02, + "learning_rate": 9.989926660861856e-05, + "loss": 2.3131, + "step": 7635 + }, + { + "epoch": 0.02, + "learning_rate": 9.989913292193062e-05, + "loss": 2.5408, + "step": 7640 + }, + { + "epoch": 0.02, + "learning_rate": 9.989899914668101e-05, + "loss": 2.5665, + "step": 7645 + }, + { + "epoch": 0.02, + "learning_rate": 9.989886528286996e-05, + "loss": 2.6185, + "step": 7650 + }, + { + "epoch": 0.02, + "learning_rate": 9.989873133049769e-05, + "loss": 2.4146, + "step": 7655 + }, + { + "epoch": 0.02, + "learning_rate": 9.98985972895645e-05, + "loss": 2.4565, + "step": 7660 + }, + { + "epoch": 0.02, + "learning_rate": 9.989846316007056e-05, + "loss": 2.5376, + "step": 7665 + }, + { + "epoch": 0.02, + "learning_rate": 9.989832894201613e-05, + "loss": 2.4443, + "step": 7670 + }, + { + "epoch": 0.02, + "learning_rate": 9.989819463540146e-05, + "loss": 2.545, + "step": 7675 + }, + { + "epoch": 0.02, + "learning_rate": 9.989806024022677e-05, + "loss": 2.5788, + "step": 7680 + }, + { + "epoch": 0.02, + "learning_rate": 9.989792575649232e-05, + "loss": 2.4895, + "step": 7685 + }, + { + "epoch": 0.02, + "learning_rate": 9.989779118419834e-05, + "loss": 2.3747, + "step": 7690 + }, + { + "epoch": 0.02, + "learning_rate": 9.989765652334506e-05, + "loss": 2.5251, + "step": 7695 + }, + { + "epoch": 0.02, + "learning_rate": 9.989752177393272e-05, + "loss": 2.5356, + "step": 7700 + }, + { + "epoch": 0.02, + "learning_rate": 9.989738693596158e-05, + "loss": 2.5201, + "step": 7705 + }, + { + "epoch": 0.02, + "learning_rate": 9.989725200943185e-05, + "loss": 2.4912, + "step": 7710 + }, + { + "epoch": 0.02, + "learning_rate": 9.989711699434379e-05, + "loss": 2.5684, + "step": 7715 + }, + { + "epoch": 0.02, + "learning_rate": 9.989698189069762e-05, + "loss": 2.4848, + "step": 7720 + }, + { + "epoch": 0.02, + "learning_rate": 9.989684669849362e-05, + "loss": 2.5004, + "step": 7725 + }, + { + "epoch": 0.02, + "learning_rate": 9.989671141773198e-05, + "loss": 2.4596, + "step": 7730 + }, + { + "epoch": 0.02, + "learning_rate": 9.989657604841298e-05, + "loss": 2.368, + "step": 7735 + }, + { + "epoch": 0.02, + "learning_rate": 9.989644059053683e-05, + "loss": 2.4034, + "step": 7740 + }, + { + "epoch": 0.02, + "learning_rate": 9.98963050441038e-05, + "loss": 2.5986, + "step": 7745 + }, + { + "epoch": 0.02, + "learning_rate": 9.989616940911408e-05, + "loss": 2.4261, + "step": 7750 + }, + { + "epoch": 0.02, + "learning_rate": 9.989603368556797e-05, + "loss": 2.5341, + "step": 7755 + }, + { + "epoch": 0.02, + "learning_rate": 9.989589787346567e-05, + "loss": 2.457, + "step": 7760 + }, + { + "epoch": 0.02, + "learning_rate": 9.989576197280746e-05, + "loss": 2.5336, + "step": 7765 + }, + { + "epoch": 0.02, + "learning_rate": 9.989562598359354e-05, + "loss": 2.5514, + "step": 7770 + }, + { + "epoch": 0.02, + "learning_rate": 9.989548990582418e-05, + "loss": 2.4985, + "step": 7775 + }, + { + "epoch": 0.02, + "learning_rate": 9.98953537394996e-05, + "loss": 2.4851, + "step": 7780 + }, + { + "epoch": 0.02, + "learning_rate": 9.989521748462005e-05, + "loss": 2.5555, + "step": 7785 + }, + { + "epoch": 0.02, + "learning_rate": 9.989508114118579e-05, + "loss": 2.4159, + "step": 7790 + }, + { + "epoch": 0.02, + "learning_rate": 9.989494470919704e-05, + "loss": 2.5428, + "step": 7795 + }, + { + "epoch": 0.02, + "learning_rate": 9.989480818865404e-05, + "loss": 2.5478, + "step": 7800 + }, + { + "epoch": 0.02, + "learning_rate": 9.989467157955705e-05, + "loss": 2.3666, + "step": 7805 + }, + { + "epoch": 0.02, + "learning_rate": 9.98945348819063e-05, + "loss": 2.552, + "step": 7810 + }, + { + "epoch": 0.02, + "learning_rate": 9.989439809570204e-05, + "loss": 2.303, + "step": 7815 + }, + { + "epoch": 0.02, + "learning_rate": 9.98942612209445e-05, + "loss": 2.5052, + "step": 7820 + }, + { + "epoch": 0.02, + "learning_rate": 9.989412425763394e-05, + "loss": 2.5412, + "step": 7825 + }, + { + "epoch": 0.02, + "learning_rate": 9.98939872057706e-05, + "loss": 2.5548, + "step": 7830 + }, + { + "epoch": 0.02, + "learning_rate": 9.989385006535468e-05, + "loss": 2.5238, + "step": 7835 + }, + { + "epoch": 0.02, + "learning_rate": 9.98937128363865e-05, + "loss": 2.5093, + "step": 7840 + }, + { + "epoch": 0.02, + "learning_rate": 9.989357551886625e-05, + "loss": 2.496, + "step": 7845 + }, + { + "epoch": 0.02, + "learning_rate": 9.989343811279419e-05, + "loss": 2.4603, + "step": 7850 + }, + { + "epoch": 0.02, + "learning_rate": 9.989330061817058e-05, + "loss": 2.531, + "step": 7855 + }, + { + "epoch": 0.02, + "learning_rate": 9.989316303499562e-05, + "loss": 2.5637, + "step": 7860 + }, + { + "epoch": 0.02, + "learning_rate": 9.989302536326959e-05, + "loss": 2.4265, + "step": 7865 + }, + { + "epoch": 0.02, + "learning_rate": 9.989288760299272e-05, + "loss": 2.5183, + "step": 7870 + }, + { + "epoch": 0.02, + "learning_rate": 9.989274975416527e-05, + "loss": 2.5046, + "step": 7875 + }, + { + "epoch": 0.02, + "learning_rate": 9.989261181678746e-05, + "loss": 2.4336, + "step": 7880 + }, + { + "epoch": 0.02, + "learning_rate": 9.989247379085956e-05, + "loss": 2.5289, + "step": 7885 + }, + { + "epoch": 0.02, + "learning_rate": 9.98923356763818e-05, + "loss": 2.5446, + "step": 7890 + }, + { + "epoch": 0.02, + "learning_rate": 9.989219747335443e-05, + "loss": 2.4338, + "step": 7895 + }, + { + "epoch": 0.02, + "learning_rate": 9.989205918177771e-05, + "loss": 2.4107, + "step": 7900 + }, + { + "epoch": 0.02, + "learning_rate": 9.989192080165184e-05, + "loss": 2.3474, + "step": 7905 + }, + { + "epoch": 0.02, + "learning_rate": 9.989178233297711e-05, + "loss": 2.4565, + "step": 7910 + }, + { + "epoch": 0.02, + "learning_rate": 9.989164377575376e-05, + "loss": 2.4976, + "step": 7915 + }, + { + "epoch": 0.02, + "learning_rate": 9.989150512998201e-05, + "loss": 2.4772, + "step": 7920 + }, + { + "epoch": 0.02, + "learning_rate": 9.989136639566214e-05, + "loss": 2.5097, + "step": 7925 + }, + { + "epoch": 0.02, + "learning_rate": 9.989122757279437e-05, + "loss": 2.3779, + "step": 7930 + }, + { + "epoch": 0.02, + "learning_rate": 9.989108866137895e-05, + "loss": 2.4403, + "step": 7935 + }, + { + "epoch": 0.02, + "learning_rate": 9.989094966141614e-05, + "loss": 2.6136, + "step": 7940 + }, + { + "epoch": 0.02, + "learning_rate": 9.989081057290617e-05, + "loss": 2.4324, + "step": 7945 + }, + { + "epoch": 0.02, + "learning_rate": 9.98906713958493e-05, + "loss": 2.3579, + "step": 7950 + }, + { + "epoch": 0.02, + "learning_rate": 9.989053213024577e-05, + "loss": 2.4483, + "step": 7955 + }, + { + "epoch": 0.02, + "learning_rate": 9.989039277609585e-05, + "loss": 2.362, + "step": 7960 + }, + { + "epoch": 0.02, + "learning_rate": 9.989025333339974e-05, + "loss": 2.5609, + "step": 7965 + }, + { + "epoch": 0.02, + "learning_rate": 9.989011380215773e-05, + "loss": 2.3987, + "step": 7970 + }, + { + "epoch": 0.02, + "learning_rate": 9.988997418237005e-05, + "loss": 2.4965, + "step": 7975 + }, + { + "epoch": 0.02, + "learning_rate": 9.988983447403695e-05, + "loss": 2.4355, + "step": 7980 + }, + { + "epoch": 0.02, + "learning_rate": 9.988969467715867e-05, + "loss": 2.4576, + "step": 7985 + }, + { + "epoch": 0.02, + "learning_rate": 9.988955479173548e-05, + "loss": 2.5025, + "step": 7990 + }, + { + "epoch": 0.02, + "learning_rate": 9.988941481776758e-05, + "loss": 2.52, + "step": 7995 + }, + { + "epoch": 0.02, + "learning_rate": 9.98892747552553e-05, + "loss": 2.4163, + "step": 8000 + }, + { + "epoch": 0.02, + "learning_rate": 9.988913460419882e-05, + "loss": 2.5192, + "step": 8005 + }, + { + "epoch": 0.02, + "learning_rate": 9.988899436459841e-05, + "loss": 2.511, + "step": 8010 + }, + { + "epoch": 0.02, + "learning_rate": 9.988885403645431e-05, + "loss": 2.5385, + "step": 8015 + }, + { + "epoch": 0.02, + "learning_rate": 9.98887136197668e-05, + "loss": 2.363, + "step": 8020 + }, + { + "epoch": 0.02, + "learning_rate": 9.98885731145361e-05, + "loss": 2.5103, + "step": 8025 + }, + { + "epoch": 0.02, + "learning_rate": 9.988843252076245e-05, + "loss": 2.3957, + "step": 8030 + }, + { + "epoch": 0.02, + "learning_rate": 9.988829183844614e-05, + "loss": 2.5535, + "step": 8035 + }, + { + "epoch": 0.02, + "learning_rate": 9.988815106758737e-05, + "loss": 2.4477, + "step": 8040 + }, + { + "epoch": 0.02, + "learning_rate": 9.988801020818643e-05, + "loss": 2.2735, + "step": 8045 + }, + { + "epoch": 0.02, + "learning_rate": 9.988786926024357e-05, + "loss": 2.5367, + "step": 8050 + }, + { + "epoch": 0.02, + "learning_rate": 9.9887728223759e-05, + "loss": 2.512, + "step": 8055 + }, + { + "epoch": 0.02, + "learning_rate": 9.988758709873302e-05, + "loss": 2.4843, + "step": 8060 + }, + { + "epoch": 0.02, + "learning_rate": 9.988744588516584e-05, + "loss": 2.43, + "step": 8065 + }, + { + "epoch": 0.02, + "learning_rate": 9.988730458305774e-05, + "loss": 2.5498, + "step": 8070 + }, + { + "epoch": 0.02, + "learning_rate": 9.988716319240896e-05, + "loss": 2.4359, + "step": 8075 + }, + { + "epoch": 0.02, + "learning_rate": 9.988702171321972e-05, + "loss": 2.4999, + "step": 8080 + }, + { + "epoch": 0.02, + "learning_rate": 9.988688014549033e-05, + "loss": 2.4768, + "step": 8085 + }, + { + "epoch": 0.02, + "learning_rate": 9.9886738489221e-05, + "loss": 2.3458, + "step": 8090 + }, + { + "epoch": 0.02, + "learning_rate": 9.988659674441198e-05, + "loss": 2.5988, + "step": 8095 + }, + { + "epoch": 0.02, + "learning_rate": 9.988645491106357e-05, + "loss": 2.6221, + "step": 8100 + }, + { + "epoch": 0.02, + "learning_rate": 9.988631298917596e-05, + "loss": 2.526, + "step": 8105 + }, + { + "epoch": 0.02, + "learning_rate": 9.988617097874944e-05, + "loss": 2.4848, + "step": 8110 + }, + { + "epoch": 0.02, + "learning_rate": 9.988602887978424e-05, + "loss": 2.5134, + "step": 8115 + }, + { + "epoch": 0.02, + "learning_rate": 9.988588669228063e-05, + "loss": 2.5292, + "step": 8120 + }, + { + "epoch": 0.02, + "learning_rate": 9.988574441623884e-05, + "loss": 2.4933, + "step": 8125 + }, + { + "epoch": 0.02, + "learning_rate": 9.988560205165916e-05, + "loss": 2.5536, + "step": 8130 + }, + { + "epoch": 0.02, + "learning_rate": 9.988545959854178e-05, + "loss": 2.3924, + "step": 8135 + }, + { + "epoch": 0.02, + "learning_rate": 9.988531705688703e-05, + "loss": 2.352, + "step": 8140 + }, + { + "epoch": 0.02, + "learning_rate": 9.988517442669512e-05, + "loss": 2.6236, + "step": 8145 + }, + { + "epoch": 0.02, + "learning_rate": 9.98850317079663e-05, + "loss": 2.5479, + "step": 8150 + }, + { + "epoch": 0.02, + "learning_rate": 9.988488890070084e-05, + "loss": 2.4564, + "step": 8155 + }, + { + "epoch": 0.02, + "learning_rate": 9.988474600489899e-05, + "loss": 2.4058, + "step": 8160 + }, + { + "epoch": 0.02, + "learning_rate": 9.988460302056097e-05, + "loss": 2.4988, + "step": 8165 + }, + { + "epoch": 0.02, + "learning_rate": 9.98844599476871e-05, + "loss": 2.4781, + "step": 8170 + }, + { + "epoch": 0.02, + "learning_rate": 9.988431678627757e-05, + "loss": 2.4738, + "step": 8175 + }, + { + "epoch": 0.02, + "learning_rate": 9.988417353633268e-05, + "loss": 2.5769, + "step": 8180 + }, + { + "epoch": 0.02, + "learning_rate": 9.988403019785266e-05, + "loss": 2.4806, + "step": 8185 + }, + { + "epoch": 0.02, + "learning_rate": 9.988388677083776e-05, + "loss": 2.4131, + "step": 8190 + }, + { + "epoch": 0.02, + "learning_rate": 9.988374325528825e-05, + "loss": 2.5275, + "step": 8195 + }, + { + "epoch": 0.02, + "learning_rate": 9.988359965120438e-05, + "loss": 2.536, + "step": 8200 + }, + { + "epoch": 0.02, + "learning_rate": 9.98834559585864e-05, + "loss": 2.5712, + "step": 8205 + }, + { + "epoch": 0.02, + "learning_rate": 9.988331217743457e-05, + "loss": 2.502, + "step": 8210 + }, + { + "epoch": 0.02, + "learning_rate": 9.988316830774916e-05, + "loss": 2.4251, + "step": 8215 + }, + { + "epoch": 0.02, + "learning_rate": 9.988302434953039e-05, + "loss": 2.3634, + "step": 8220 + }, + { + "epoch": 0.02, + "learning_rate": 9.988288030277854e-05, + "loss": 2.5289, + "step": 8225 + }, + { + "epoch": 0.02, + "learning_rate": 9.988273616749386e-05, + "loss": 2.5266, + "step": 8230 + }, + { + "epoch": 0.02, + "learning_rate": 9.98825919436766e-05, + "loss": 2.3857, + "step": 8235 + }, + { + "epoch": 0.02, + "learning_rate": 9.988244763132704e-05, + "loss": 2.5545, + "step": 8240 + }, + { + "epoch": 0.02, + "learning_rate": 9.98823032304454e-05, + "loss": 2.415, + "step": 8245 + }, + { + "epoch": 0.02, + "learning_rate": 9.988215874103195e-05, + "loss": 2.5154, + "step": 8250 + }, + { + "epoch": 0.02, + "learning_rate": 9.988201416308698e-05, + "loss": 2.4339, + "step": 8255 + }, + { + "epoch": 0.02, + "learning_rate": 9.988186949661069e-05, + "loss": 2.4905, + "step": 8260 + }, + { + "epoch": 0.02, + "learning_rate": 9.988172474160337e-05, + "loss": 2.4694, + "step": 8265 + }, + { + "epoch": 0.02, + "learning_rate": 9.988157989806528e-05, + "loss": 2.4547, + "step": 8270 + }, + { + "epoch": 0.02, + "learning_rate": 9.988143496599667e-05, + "loss": 2.4443, + "step": 8275 + }, + { + "epoch": 0.02, + "learning_rate": 9.988128994539778e-05, + "loss": 2.5147, + "step": 8280 + }, + { + "epoch": 0.02, + "learning_rate": 9.988114483626889e-05, + "loss": 2.3995, + "step": 8285 + }, + { + "epoch": 0.02, + "learning_rate": 9.988099963861026e-05, + "loss": 2.422, + "step": 8290 + }, + { + "epoch": 0.02, + "learning_rate": 9.988085435242214e-05, + "loss": 2.5605, + "step": 8295 + }, + { + "epoch": 0.02, + "learning_rate": 9.988070897770478e-05, + "loss": 2.4225, + "step": 8300 + }, + { + "epoch": 0.02, + "learning_rate": 9.988056351445843e-05, + "loss": 2.5222, + "step": 8305 + }, + { + "epoch": 0.02, + "learning_rate": 9.988041796268337e-05, + "loss": 2.4287, + "step": 8310 + }, + { + "epoch": 0.02, + "learning_rate": 9.988027232237987e-05, + "loss": 2.4121, + "step": 8315 + }, + { + "epoch": 0.02, + "learning_rate": 9.988012659354816e-05, + "loss": 2.4439, + "step": 8320 + }, + { + "epoch": 0.02, + "learning_rate": 9.987998077618849e-05, + "loss": 2.4972, + "step": 8325 + }, + { + "epoch": 0.02, + "learning_rate": 9.987983487030114e-05, + "loss": 2.4814, + "step": 8330 + }, + { + "epoch": 0.02, + "learning_rate": 9.987968887588639e-05, + "loss": 2.5392, + "step": 8335 + }, + { + "epoch": 0.02, + "learning_rate": 9.987954279294445e-05, + "loss": 2.3642, + "step": 8340 + }, + { + "epoch": 0.02, + "learning_rate": 9.987939662147561e-05, + "loss": 2.3606, + "step": 8345 + }, + { + "epoch": 0.02, + "learning_rate": 9.987925036148012e-05, + "loss": 2.5234, + "step": 8350 + }, + { + "epoch": 0.02, + "learning_rate": 9.987910401295826e-05, + "loss": 2.4955, + "step": 8355 + }, + { + "epoch": 0.02, + "learning_rate": 9.987895757591026e-05, + "loss": 2.5131, + "step": 8360 + }, + { + "epoch": 0.02, + "learning_rate": 9.987881105033638e-05, + "loss": 2.4294, + "step": 8365 + }, + { + "epoch": 0.02, + "learning_rate": 9.987866443623691e-05, + "loss": 2.462, + "step": 8370 + }, + { + "epoch": 0.02, + "learning_rate": 9.98785177336121e-05, + "loss": 2.6067, + "step": 8375 + }, + { + "epoch": 0.02, + "learning_rate": 9.987837094246218e-05, + "loss": 2.4861, + "step": 8380 + }, + { + "epoch": 0.02, + "learning_rate": 9.987822406278746e-05, + "loss": 2.4926, + "step": 8385 + }, + { + "epoch": 0.02, + "learning_rate": 9.987807709458815e-05, + "loss": 2.4444, + "step": 8390 + }, + { + "epoch": 0.02, + "learning_rate": 9.987793003786454e-05, + "loss": 2.5358, + "step": 8395 + }, + { + "epoch": 0.02, + "learning_rate": 9.987778289261688e-05, + "loss": 2.3867, + "step": 8400 + }, + { + "epoch": 0.02, + "learning_rate": 9.987763565884547e-05, + "loss": 2.4705, + "step": 8405 + }, + { + "epoch": 0.02, + "learning_rate": 9.987748833655051e-05, + "loss": 2.6979, + "step": 8410 + }, + { + "epoch": 0.02, + "learning_rate": 9.98773409257323e-05, + "loss": 2.5092, + "step": 8415 + }, + { + "epoch": 0.02, + "learning_rate": 9.98771934263911e-05, + "loss": 2.5734, + "step": 8420 + }, + { + "epoch": 0.02, + "learning_rate": 9.987704583852715e-05, + "loss": 2.344, + "step": 8425 + }, + { + "epoch": 0.02, + "learning_rate": 9.987689816214072e-05, + "loss": 2.473, + "step": 8430 + }, + { + "epoch": 0.02, + "learning_rate": 9.98767503972321e-05, + "loss": 2.4836, + "step": 8435 + }, + { + "epoch": 0.02, + "learning_rate": 9.98766025438015e-05, + "loss": 2.3179, + "step": 8440 + }, + { + "epoch": 0.02, + "learning_rate": 9.987645460184924e-05, + "loss": 2.4941, + "step": 8445 + }, + { + "epoch": 0.02, + "learning_rate": 9.987630657137555e-05, + "loss": 2.5246, + "step": 8450 + }, + { + "epoch": 0.02, + "learning_rate": 9.987615845238067e-05, + "loss": 2.4092, + "step": 8455 + }, + { + "epoch": 0.02, + "learning_rate": 9.987601024486492e-05, + "loss": 2.4981, + "step": 8460 + }, + { + "epoch": 0.02, + "learning_rate": 9.987586194882852e-05, + "loss": 2.4549, + "step": 8465 + }, + { + "epoch": 0.02, + "learning_rate": 9.987571356427176e-05, + "loss": 2.5528, + "step": 8470 + }, + { + "epoch": 0.02, + "learning_rate": 9.987556509119487e-05, + "loss": 2.3504, + "step": 8475 + }, + { + "epoch": 0.02, + "learning_rate": 9.987541652959815e-05, + "loss": 2.5082, + "step": 8480 + }, + { + "epoch": 0.02, + "learning_rate": 9.987526787948185e-05, + "loss": 2.4003, + "step": 8485 + }, + { + "epoch": 0.02, + "learning_rate": 9.987511914084622e-05, + "loss": 2.5257, + "step": 8490 + }, + { + "epoch": 0.02, + "learning_rate": 9.987497031369152e-05, + "loss": 2.4806, + "step": 8495 + }, + { + "epoch": 0.02, + "learning_rate": 9.987482139801806e-05, + "loss": 2.4108, + "step": 8500 + }, + { + "epoch": 0.02, + "learning_rate": 9.987467239382607e-05, + "loss": 2.4843, + "step": 8505 + }, + { + "epoch": 0.02, + "learning_rate": 9.987452330111579e-05, + "loss": 2.4841, + "step": 8510 + }, + { + "epoch": 0.02, + "learning_rate": 9.987437411988753e-05, + "loss": 2.4653, + "step": 8515 + }, + { + "epoch": 0.02, + "learning_rate": 9.987422485014156e-05, + "loss": 2.4048, + "step": 8520 + }, + { + "epoch": 0.02, + "learning_rate": 9.987407549187809e-05, + "loss": 2.449, + "step": 8525 + }, + { + "epoch": 0.02, + "learning_rate": 9.987392604509743e-05, + "loss": 2.5413, + "step": 8530 + }, + { + "epoch": 0.02, + "learning_rate": 9.987377650979983e-05, + "loss": 2.5017, + "step": 8535 + }, + { + "epoch": 0.02, + "learning_rate": 9.987362688598556e-05, + "loss": 2.4562, + "step": 8540 + }, + { + "epoch": 0.02, + "learning_rate": 9.987347717365489e-05, + "loss": 2.4042, + "step": 8545 + }, + { + "epoch": 0.02, + "learning_rate": 9.987332737280806e-05, + "loss": 2.4821, + "step": 8550 + }, + { + "epoch": 0.02, + "learning_rate": 9.987317748344536e-05, + "loss": 2.3108, + "step": 8555 + }, + { + "epoch": 0.02, + "learning_rate": 9.987302750556706e-05, + "loss": 2.3111, + "step": 8560 + }, + { + "epoch": 0.02, + "learning_rate": 9.987287743917342e-05, + "loss": 2.4532, + "step": 8565 + }, + { + "epoch": 0.02, + "learning_rate": 9.98727272842647e-05, + "loss": 2.5729, + "step": 8570 + }, + { + "epoch": 0.02, + "learning_rate": 9.987257704084116e-05, + "loss": 2.3807, + "step": 8575 + }, + { + "epoch": 0.02, + "learning_rate": 9.987242670890308e-05, + "loss": 2.5064, + "step": 8580 + }, + { + "epoch": 0.02, + "learning_rate": 9.987227628845074e-05, + "loss": 2.3993, + "step": 8585 + }, + { + "epoch": 0.02, + "learning_rate": 9.987212577948438e-05, + "loss": 2.4402, + "step": 8590 + }, + { + "epoch": 0.02, + "learning_rate": 9.987197518200429e-05, + "loss": 2.521, + "step": 8595 + }, + { + "epoch": 0.02, + "learning_rate": 9.98718244960107e-05, + "loss": 2.5654, + "step": 8600 + }, + { + "epoch": 0.02, + "learning_rate": 9.987167372150393e-05, + "loss": 2.4265, + "step": 8605 + }, + { + "epoch": 0.02, + "learning_rate": 9.98715228584842e-05, + "loss": 2.4337, + "step": 8610 + }, + { + "epoch": 0.02, + "learning_rate": 9.987137190695179e-05, + "loss": 2.3833, + "step": 8615 + }, + { + "epoch": 0.02, + "learning_rate": 9.987122086690699e-05, + "loss": 2.5078, + "step": 8620 + }, + { + "epoch": 0.02, + "learning_rate": 9.987106973835006e-05, + "loss": 2.6926, + "step": 8625 + }, + { + "epoch": 0.02, + "learning_rate": 9.987091852128126e-05, + "loss": 2.5078, + "step": 8630 + }, + { + "epoch": 0.02, + "learning_rate": 9.987076721570085e-05, + "loss": 2.3126, + "step": 8635 + }, + { + "epoch": 0.02, + "learning_rate": 9.98706158216091e-05, + "loss": 2.5993, + "step": 8640 + }, + { + "epoch": 0.02, + "learning_rate": 9.987046433900631e-05, + "loss": 2.5391, + "step": 8645 + }, + { + "epoch": 0.02, + "learning_rate": 9.987031276789271e-05, + "loss": 2.5046, + "step": 8650 + }, + { + "epoch": 0.02, + "learning_rate": 9.98701611082686e-05, + "loss": 2.5297, + "step": 8655 + }, + { + "epoch": 0.02, + "learning_rate": 9.987000936013421e-05, + "loss": 2.5405, + "step": 8660 + }, + { + "epoch": 0.02, + "learning_rate": 9.986985752348987e-05, + "loss": 2.4772, + "step": 8665 + }, + { + "epoch": 0.02, + "learning_rate": 9.98697055983358e-05, + "loss": 2.5204, + "step": 8670 + }, + { + "epoch": 0.02, + "learning_rate": 9.986955358467227e-05, + "loss": 2.3737, + "step": 8675 + }, + { + "epoch": 0.02, + "learning_rate": 9.986940148249956e-05, + "loss": 2.4422, + "step": 8680 + }, + { + "epoch": 0.02, + "learning_rate": 9.986924929181796e-05, + "loss": 2.4601, + "step": 8685 + }, + { + "epoch": 0.02, + "learning_rate": 9.986909701262772e-05, + "loss": 2.4012, + "step": 8690 + }, + { + "epoch": 0.02, + "learning_rate": 9.98689446449291e-05, + "loss": 2.4782, + "step": 8695 + }, + { + "epoch": 0.02, + "learning_rate": 9.98687921887224e-05, + "loss": 2.3629, + "step": 8700 + }, + { + "epoch": 0.02, + "learning_rate": 9.986863964400787e-05, + "loss": 2.4797, + "step": 8705 + }, + { + "epoch": 0.02, + "learning_rate": 9.986848701078578e-05, + "loss": 2.471, + "step": 8710 + }, + { + "epoch": 0.02, + "learning_rate": 9.986833428905641e-05, + "loss": 2.5387, + "step": 8715 + }, + { + "epoch": 0.02, + "learning_rate": 9.986818147882002e-05, + "loss": 2.464, + "step": 8720 + }, + { + "epoch": 0.02, + "learning_rate": 9.98680285800769e-05, + "loss": 2.4412, + "step": 8725 + }, + { + "epoch": 0.02, + "learning_rate": 9.986787559282731e-05, + "loss": 2.4044, + "step": 8730 + }, + { + "epoch": 0.02, + "learning_rate": 9.98677225170715e-05, + "loss": 2.4306, + "step": 8735 + }, + { + "epoch": 0.02, + "learning_rate": 9.986756935280978e-05, + "loss": 2.5957, + "step": 8740 + }, + { + "epoch": 0.02, + "learning_rate": 9.98674161000424e-05, + "loss": 2.4116, + "step": 8745 + }, + { + "epoch": 0.02, + "learning_rate": 9.986726275876963e-05, + "loss": 2.4961, + "step": 8750 + }, + { + "epoch": 0.02, + "learning_rate": 9.986710932899177e-05, + "loss": 2.4535, + "step": 8755 + }, + { + "epoch": 0.02, + "learning_rate": 9.986695581070905e-05, + "loss": 2.4896, + "step": 8760 + }, + { + "epoch": 0.02, + "learning_rate": 9.986680220392178e-05, + "loss": 2.4931, + "step": 8765 + }, + { + "epoch": 0.02, + "learning_rate": 9.986664850863019e-05, + "loss": 2.4075, + "step": 8770 + }, + { + "epoch": 0.02, + "learning_rate": 9.986649472483461e-05, + "loss": 2.4292, + "step": 8775 + }, + { + "epoch": 0.02, + "learning_rate": 9.986634085253525e-05, + "loss": 2.4299, + "step": 8780 + }, + { + "epoch": 0.02, + "learning_rate": 9.986618689173243e-05, + "loss": 2.5201, + "step": 8785 + }, + { + "epoch": 0.02, + "learning_rate": 9.986603284242642e-05, + "loss": 2.435, + "step": 8790 + }, + { + "epoch": 0.02, + "learning_rate": 9.986587870461747e-05, + "loss": 2.4275, + "step": 8795 + }, + { + "epoch": 0.02, + "learning_rate": 9.986572447830587e-05, + "loss": 2.4883, + "step": 8800 + }, + { + "epoch": 0.02, + "learning_rate": 9.986557016349188e-05, + "loss": 2.3872, + "step": 8805 + }, + { + "epoch": 0.02, + "learning_rate": 9.986541576017578e-05, + "loss": 2.5036, + "step": 8810 + }, + { + "epoch": 0.02, + "learning_rate": 9.986526126835787e-05, + "loss": 2.4406, + "step": 8815 + }, + { + "epoch": 0.02, + "learning_rate": 9.986510668803839e-05, + "loss": 2.3808, + "step": 8820 + }, + { + "epoch": 0.02, + "learning_rate": 9.986495201921761e-05, + "loss": 2.4058, + "step": 8825 + }, + { + "epoch": 0.02, + "learning_rate": 9.986479726189584e-05, + "loss": 2.4376, + "step": 8830 + }, + { + "epoch": 0.02, + "learning_rate": 9.986464241607332e-05, + "loss": 2.6113, + "step": 8835 + }, + { + "epoch": 0.02, + "learning_rate": 9.986448748175035e-05, + "loss": 2.4405, + "step": 8840 + }, + { + "epoch": 0.02, + "learning_rate": 9.986433245892719e-05, + "loss": 2.2891, + "step": 8845 + }, + { + "epoch": 0.02, + "learning_rate": 9.986417734760411e-05, + "loss": 2.5023, + "step": 8850 + }, + { + "epoch": 0.02, + "learning_rate": 9.986402214778142e-05, + "loss": 2.342, + "step": 8855 + }, + { + "epoch": 0.02, + "learning_rate": 9.986386685945934e-05, + "loss": 2.483, + "step": 8860 + }, + { + "epoch": 0.02, + "learning_rate": 9.986371148263817e-05, + "loss": 2.3338, + "step": 8865 + }, + { + "epoch": 0.02, + "learning_rate": 9.986355601731822e-05, + "loss": 2.556, + "step": 8870 + }, + { + "epoch": 0.02, + "learning_rate": 9.986340046349974e-05, + "loss": 2.5108, + "step": 8875 + }, + { + "epoch": 0.02, + "learning_rate": 9.986324482118298e-05, + "loss": 2.3342, + "step": 8880 + }, + { + "epoch": 0.02, + "learning_rate": 9.986308909036825e-05, + "loss": 2.386, + "step": 8885 + }, + { + "epoch": 0.02, + "learning_rate": 9.986293327105583e-05, + "loss": 2.4325, + "step": 8890 + }, + { + "epoch": 0.02, + "learning_rate": 9.986277736324596e-05, + "loss": 2.4784, + "step": 8895 + }, + { + "epoch": 0.02, + "learning_rate": 9.986262136693895e-05, + "loss": 2.5338, + "step": 8900 + }, + { + "epoch": 0.02, + "learning_rate": 9.986246528213506e-05, + "loss": 2.3976, + "step": 8905 + }, + { + "epoch": 0.02, + "learning_rate": 9.986230910883458e-05, + "loss": 2.4278, + "step": 8910 + }, + { + "epoch": 0.02, + "learning_rate": 9.986215284703779e-05, + "loss": 2.4742, + "step": 8915 + }, + { + "epoch": 0.02, + "learning_rate": 9.986199649674494e-05, + "loss": 2.4735, + "step": 8920 + }, + { + "epoch": 0.02, + "learning_rate": 9.986184005795634e-05, + "loss": 2.5195, + "step": 8925 + }, + { + "epoch": 0.02, + "learning_rate": 9.986168353067225e-05, + "loss": 2.6046, + "step": 8930 + }, + { + "epoch": 0.02, + "learning_rate": 9.986152691489293e-05, + "loss": 2.5487, + "step": 8935 + }, + { + "epoch": 0.02, + "learning_rate": 9.98613702106187e-05, + "loss": 2.4722, + "step": 8940 + }, + { + "epoch": 0.02, + "learning_rate": 9.986121341784981e-05, + "loss": 2.4304, + "step": 8945 + }, + { + "epoch": 0.02, + "learning_rate": 9.986105653658655e-05, + "loss": 2.3868, + "step": 8950 + }, + { + "epoch": 0.02, + "learning_rate": 9.986089956682919e-05, + "loss": 2.4523, + "step": 8955 + }, + { + "epoch": 0.02, + "learning_rate": 9.986074250857801e-05, + "loss": 2.3762, + "step": 8960 + }, + { + "epoch": 0.02, + "learning_rate": 9.986058536183329e-05, + "loss": 2.5011, + "step": 8965 + }, + { + "epoch": 0.02, + "learning_rate": 9.98604281265953e-05, + "loss": 2.4191, + "step": 8970 + }, + { + "epoch": 0.02, + "learning_rate": 9.986027080286435e-05, + "loss": 2.4476, + "step": 8975 + }, + { + "epoch": 0.02, + "learning_rate": 9.986011339064069e-05, + "loss": 2.3832, + "step": 8980 + }, + { + "epoch": 0.02, + "learning_rate": 9.985995588992461e-05, + "loss": 2.4525, + "step": 8985 + }, + { + "epoch": 0.02, + "learning_rate": 9.985979830071639e-05, + "loss": 2.4027, + "step": 8990 + }, + { + "epoch": 0.02, + "learning_rate": 9.98596406230163e-05, + "loss": 2.5448, + "step": 8995 + }, + { + "epoch": 0.02, + "learning_rate": 9.985948285682463e-05, + "loss": 2.4909, + "step": 9000 + }, + { + "epoch": 0.02, + "learning_rate": 9.985932500214165e-05, + "loss": 2.4826, + "step": 9005 + }, + { + "epoch": 0.02, + "learning_rate": 9.985916705896765e-05, + "loss": 2.3156, + "step": 9010 + }, + { + "epoch": 0.02, + "learning_rate": 9.98590090273029e-05, + "loss": 2.5207, + "step": 9015 + }, + { + "epoch": 0.02, + "learning_rate": 9.98588509071477e-05, + "loss": 2.3622, + "step": 9020 + }, + { + "epoch": 0.02, + "learning_rate": 9.985869269850232e-05, + "loss": 2.3789, + "step": 9025 + }, + { + "epoch": 0.02, + "learning_rate": 9.985853440136704e-05, + "loss": 2.515, + "step": 9030 + }, + { + "epoch": 0.02, + "learning_rate": 9.985837601574212e-05, + "loss": 2.5304, + "step": 9035 + }, + { + "epoch": 0.02, + "learning_rate": 9.985821754162788e-05, + "loss": 2.3694, + "step": 9040 + }, + { + "epoch": 0.02, + "learning_rate": 9.985805897902456e-05, + "loss": 2.5389, + "step": 9045 + }, + { + "epoch": 0.02, + "learning_rate": 9.985790032793249e-05, + "loss": 2.5611, + "step": 9050 + }, + { + "epoch": 0.02, + "learning_rate": 9.98577415883519e-05, + "loss": 2.3245, + "step": 9055 + }, + { + "epoch": 0.02, + "learning_rate": 9.985758276028312e-05, + "loss": 2.5178, + "step": 9060 + }, + { + "epoch": 0.02, + "learning_rate": 9.985742384372641e-05, + "loss": 2.4901, + "step": 9065 + }, + { + "epoch": 0.02, + "learning_rate": 9.985726483868203e-05, + "loss": 2.4701, + "step": 9070 + }, + { + "epoch": 0.02, + "learning_rate": 9.98571057451503e-05, + "loss": 2.4783, + "step": 9075 + }, + { + "epoch": 0.02, + "learning_rate": 9.985694656313147e-05, + "loss": 2.5865, + "step": 9080 + }, + { + "epoch": 0.02, + "learning_rate": 9.985678729262584e-05, + "loss": 2.4755, + "step": 9085 + }, + { + "epoch": 0.02, + "learning_rate": 9.985662793363371e-05, + "loss": 2.4627, + "step": 9090 + }, + { + "epoch": 0.02, + "learning_rate": 9.985646848615533e-05, + "loss": 2.4488, + "step": 9095 + }, + { + "epoch": 0.02, + "learning_rate": 9.9856308950191e-05, + "loss": 2.4607, + "step": 9100 + }, + { + "epoch": 0.02, + "learning_rate": 9.985614932574098e-05, + "loss": 2.4979, + "step": 9105 + }, + { + "epoch": 0.02, + "learning_rate": 9.98559896128056e-05, + "loss": 2.548, + "step": 9110 + }, + { + "epoch": 0.02, + "learning_rate": 9.98558298113851e-05, + "loss": 2.4701, + "step": 9115 + }, + { + "epoch": 0.02, + "learning_rate": 9.985566992147979e-05, + "loss": 2.524, + "step": 9120 + }, + { + "epoch": 0.02, + "learning_rate": 9.985550994308994e-05, + "loss": 2.5026, + "step": 9125 + }, + { + "epoch": 0.02, + "learning_rate": 9.985534987621585e-05, + "loss": 2.5203, + "step": 9130 + }, + { + "epoch": 0.02, + "learning_rate": 9.985518972085778e-05, + "loss": 2.439, + "step": 9135 + }, + { + "epoch": 0.02, + "learning_rate": 9.985502947701601e-05, + "loss": 2.5438, + "step": 9140 + }, + { + "epoch": 0.02, + "learning_rate": 9.985486914469087e-05, + "loss": 2.4281, + "step": 9145 + }, + { + "epoch": 0.02, + "learning_rate": 9.985470872388261e-05, + "loss": 2.4506, + "step": 9150 + }, + { + "epoch": 0.02, + "learning_rate": 9.98545482145915e-05, + "loss": 2.5052, + "step": 9155 + }, + { + "epoch": 0.02, + "learning_rate": 9.985438761681786e-05, + "loss": 2.4716, + "step": 9160 + }, + { + "epoch": 0.02, + "learning_rate": 9.985422693056196e-05, + "loss": 2.5822, + "step": 9165 + }, + { + "epoch": 0.02, + "learning_rate": 9.985406615582406e-05, + "loss": 2.4804, + "step": 9170 + }, + { + "epoch": 0.02, + "learning_rate": 9.98539052926045e-05, + "loss": 2.3079, + "step": 9175 + }, + { + "epoch": 0.02, + "learning_rate": 9.985374434090354e-05, + "loss": 2.3622, + "step": 9180 + }, + { + "epoch": 0.02, + "learning_rate": 9.985358330072143e-05, + "loss": 2.4638, + "step": 9185 + }, + { + "epoch": 0.02, + "learning_rate": 9.985342217205851e-05, + "loss": 2.4362, + "step": 9190 + }, + { + "epoch": 0.02, + "learning_rate": 9.985326095491505e-05, + "loss": 2.3367, + "step": 9195 + }, + { + "epoch": 0.02, + "learning_rate": 9.985309964929131e-05, + "loss": 2.3821, + "step": 9200 + }, + { + "epoch": 0.02, + "learning_rate": 9.98529382551876e-05, + "loss": 2.4378, + "step": 9205 + }, + { + "epoch": 0.02, + "learning_rate": 9.985277677260421e-05, + "loss": 2.4468, + "step": 9210 + }, + { + "epoch": 0.02, + "learning_rate": 9.985261520154141e-05, + "loss": 2.4591, + "step": 9215 + }, + { + "epoch": 0.02, + "learning_rate": 9.985245354199951e-05, + "loss": 2.3816, + "step": 9220 + }, + { + "epoch": 0.02, + "learning_rate": 9.985229179397876e-05, + "loss": 2.3712, + "step": 9225 + }, + { + "epoch": 0.02, + "learning_rate": 9.985212995747948e-05, + "loss": 2.5059, + "step": 9230 + }, + { + "epoch": 0.02, + "learning_rate": 9.985196803250194e-05, + "loss": 2.4592, + "step": 9235 + }, + { + "epoch": 0.02, + "learning_rate": 9.985180601904645e-05, + "loss": 2.4276, + "step": 9240 + }, + { + "epoch": 0.02, + "learning_rate": 9.985164391711326e-05, + "loss": 2.5192, + "step": 9245 + }, + { + "epoch": 0.02, + "learning_rate": 9.98514817267027e-05, + "loss": 2.3132, + "step": 9250 + }, + { + "epoch": 0.02, + "learning_rate": 9.985131944781503e-05, + "loss": 2.6721, + "step": 9255 + }, + { + "epoch": 0.02, + "learning_rate": 9.985115708045054e-05, + "loss": 2.4242, + "step": 9260 + }, + { + "epoch": 0.02, + "learning_rate": 9.985099462460952e-05, + "loss": 2.4215, + "step": 9265 + }, + { + "epoch": 0.02, + "learning_rate": 9.985083208029227e-05, + "loss": 2.512, + "step": 9270 + }, + { + "epoch": 0.02, + "learning_rate": 9.985066944749908e-05, + "loss": 2.4288, + "step": 9275 + }, + { + "epoch": 0.02, + "learning_rate": 9.98505067262302e-05, + "loss": 2.3966, + "step": 9280 + }, + { + "epoch": 0.02, + "learning_rate": 9.985034391648598e-05, + "loss": 2.472, + "step": 9285 + }, + { + "epoch": 0.02, + "learning_rate": 9.985018101826665e-05, + "loss": 2.4558, + "step": 9290 + }, + { + "epoch": 0.02, + "learning_rate": 9.985001803157255e-05, + "loss": 2.5374, + "step": 9295 + }, + { + "epoch": 0.02, + "learning_rate": 9.984985495640392e-05, + "loss": 2.5462, + "step": 9300 + }, + { + "epoch": 0.02, + "learning_rate": 9.98496917927611e-05, + "loss": 2.4317, + "step": 9305 + }, + { + "epoch": 0.02, + "learning_rate": 9.984952854064433e-05, + "loss": 2.5057, + "step": 9310 + }, + { + "epoch": 0.02, + "learning_rate": 9.984936520005393e-05, + "loss": 2.5094, + "step": 9315 + }, + { + "epoch": 0.02, + "learning_rate": 9.984920177099019e-05, + "loss": 2.4443, + "step": 9320 + }, + { + "epoch": 0.03, + "learning_rate": 9.984903825345338e-05, + "loss": 2.5095, + "step": 9325 + }, + { + "epoch": 0.03, + "learning_rate": 9.984887464744381e-05, + "loss": 2.5756, + "step": 9330 + }, + { + "epoch": 0.03, + "learning_rate": 9.984871095296178e-05, + "loss": 2.4142, + "step": 9335 + }, + { + "epoch": 0.03, + "learning_rate": 9.984854717000756e-05, + "loss": 2.5243, + "step": 9340 + }, + { + "epoch": 0.03, + "learning_rate": 9.984838329858143e-05, + "loss": 2.4183, + "step": 9345 + }, + { + "epoch": 0.03, + "learning_rate": 9.98482193386837e-05, + "loss": 2.5297, + "step": 9350 + }, + { + "epoch": 0.03, + "learning_rate": 9.984805529031467e-05, + "loss": 2.3832, + "step": 9355 + }, + { + "epoch": 0.03, + "learning_rate": 9.98478911534746e-05, + "loss": 2.543, + "step": 9360 + }, + { + "epoch": 0.03, + "learning_rate": 9.98477269281638e-05, + "loss": 2.4648, + "step": 9365 + }, + { + "epoch": 0.03, + "learning_rate": 9.984756261438258e-05, + "loss": 2.4402, + "step": 9370 + }, + { + "epoch": 0.03, + "learning_rate": 9.98473982121312e-05, + "loss": 2.4325, + "step": 9375 + }, + { + "epoch": 0.03, + "learning_rate": 9.984723372140998e-05, + "loss": 2.4815, + "step": 9380 + }, + { + "epoch": 0.03, + "learning_rate": 9.984706914221918e-05, + "loss": 2.4602, + "step": 9385 + }, + { + "epoch": 0.03, + "learning_rate": 9.984690447455911e-05, + "loss": 2.5752, + "step": 9390 + }, + { + "epoch": 0.03, + "learning_rate": 9.984673971843006e-05, + "loss": 2.5765, + "step": 9395 + }, + { + "epoch": 0.03, + "learning_rate": 9.984657487383233e-05, + "loss": 2.4383, + "step": 9400 + }, + { + "epoch": 0.03, + "learning_rate": 9.98464099407662e-05, + "loss": 2.3614, + "step": 9405 + }, + { + "epoch": 0.03, + "learning_rate": 9.984624491923198e-05, + "loss": 2.4591, + "step": 9410 + }, + { + "epoch": 0.03, + "learning_rate": 9.984607980922994e-05, + "loss": 2.5138, + "step": 9415 + }, + { + "epoch": 0.03, + "learning_rate": 9.984591461076038e-05, + "loss": 2.4036, + "step": 9420 + }, + { + "epoch": 0.03, + "learning_rate": 9.98457493238236e-05, + "loss": 2.4063, + "step": 9425 + }, + { + "epoch": 0.03, + "learning_rate": 9.98455839484199e-05, + "loss": 2.4901, + "step": 9430 + }, + { + "epoch": 0.03, + "learning_rate": 9.984541848454956e-05, + "loss": 2.4468, + "step": 9435 + }, + { + "epoch": 0.03, + "learning_rate": 9.984525293221287e-05, + "loss": 2.4536, + "step": 9440 + }, + { + "epoch": 0.03, + "learning_rate": 9.984508729141016e-05, + "loss": 2.446, + "step": 9445 + }, + { + "epoch": 0.03, + "learning_rate": 9.984492156214166e-05, + "loss": 2.5611, + "step": 9450 + }, + { + "epoch": 0.03, + "learning_rate": 9.984475574440773e-05, + "loss": 2.5075, + "step": 9455 + }, + { + "epoch": 0.03, + "learning_rate": 9.984458983820862e-05, + "loss": 2.4967, + "step": 9460 + }, + { + "epoch": 0.03, + "learning_rate": 9.984442384354463e-05, + "loss": 2.3602, + "step": 9465 + }, + { + "epoch": 0.03, + "learning_rate": 9.984425776041607e-05, + "loss": 2.4785, + "step": 9470 + }, + { + "epoch": 0.03, + "learning_rate": 9.984409158882324e-05, + "loss": 2.4571, + "step": 9475 + }, + { + "epoch": 0.03, + "learning_rate": 9.984392532876641e-05, + "loss": 2.502, + "step": 9480 + }, + { + "epoch": 0.03, + "learning_rate": 9.984375898024589e-05, + "loss": 2.4618, + "step": 9485 + }, + { + "epoch": 0.03, + "learning_rate": 9.984359254326197e-05, + "loss": 2.4712, + "step": 9490 + }, + { + "epoch": 0.03, + "learning_rate": 9.984342601781495e-05, + "loss": 2.4962, + "step": 9495 + }, + { + "epoch": 0.03, + "learning_rate": 9.984325940390515e-05, + "loss": 2.4307, + "step": 9500 + }, + { + "epoch": 0.03, + "learning_rate": 9.984309270153281e-05, + "loss": 2.4335, + "step": 9505 + }, + { + "epoch": 0.03, + "learning_rate": 9.984292591069826e-05, + "loss": 2.4572, + "step": 9510 + }, + { + "epoch": 0.03, + "learning_rate": 9.984275903140182e-05, + "loss": 2.4895, + "step": 9515 + }, + { + "epoch": 0.03, + "learning_rate": 9.984259206364373e-05, + "loss": 2.5949, + "step": 9520 + }, + { + "epoch": 0.03, + "learning_rate": 9.984242500742431e-05, + "loss": 2.5211, + "step": 9525 + }, + { + "epoch": 0.03, + "learning_rate": 9.984225786274388e-05, + "loss": 2.3449, + "step": 9530 + }, + { + "epoch": 0.03, + "learning_rate": 9.98420906296027e-05, + "loss": 2.4474, + "step": 9535 + }, + { + "epoch": 0.03, + "learning_rate": 9.98419233080011e-05, + "loss": 2.5128, + "step": 9540 + }, + { + "epoch": 0.03, + "learning_rate": 9.984175589793934e-05, + "loss": 2.5037, + "step": 9545 + }, + { + "epoch": 0.03, + "learning_rate": 9.984158839941777e-05, + "loss": 2.4236, + "step": 9550 + }, + { + "epoch": 0.03, + "learning_rate": 9.984142081243664e-05, + "loss": 2.4476, + "step": 9555 + }, + { + "epoch": 0.03, + "learning_rate": 9.984125313699626e-05, + "loss": 2.2806, + "step": 9560 + }, + { + "epoch": 0.03, + "learning_rate": 9.984108537309693e-05, + "loss": 2.4772, + "step": 9565 + }, + { + "epoch": 0.03, + "learning_rate": 9.984091752073896e-05, + "loss": 2.5701, + "step": 9570 + }, + { + "epoch": 0.03, + "learning_rate": 9.984074957992262e-05, + "loss": 2.498, + "step": 9575 + }, + { + "epoch": 0.03, + "learning_rate": 9.984058155064823e-05, + "loss": 2.5761, + "step": 9580 + }, + { + "epoch": 0.03, + "learning_rate": 9.98404134329161e-05, + "loss": 2.5601, + "step": 9585 + }, + { + "epoch": 0.03, + "learning_rate": 9.984024522672648e-05, + "loss": 2.4912, + "step": 9590 + }, + { + "epoch": 0.03, + "learning_rate": 9.984007693207973e-05, + "loss": 2.3881, + "step": 9595 + }, + { + "epoch": 0.03, + "learning_rate": 9.98399085489761e-05, + "loss": 2.5198, + "step": 9600 + }, + { + "epoch": 0.03, + "learning_rate": 9.983974007741591e-05, + "loss": 2.4521, + "step": 9605 + }, + { + "epoch": 0.03, + "learning_rate": 9.983957151739946e-05, + "loss": 2.4362, + "step": 9610 + }, + { + "epoch": 0.03, + "learning_rate": 9.983940286892705e-05, + "loss": 2.4996, + "step": 9615 + }, + { + "epoch": 0.03, + "learning_rate": 9.983923413199896e-05, + "loss": 2.3693, + "step": 9620 + }, + { + "epoch": 0.03, + "learning_rate": 9.983906530661552e-05, + "loss": 2.5336, + "step": 9625 + }, + { + "epoch": 0.03, + "learning_rate": 9.983889639277701e-05, + "loss": 2.4834, + "step": 9630 + }, + { + "epoch": 0.03, + "learning_rate": 9.983872739048372e-05, + "loss": 2.4068, + "step": 9635 + }, + { + "epoch": 0.03, + "learning_rate": 9.983855829973598e-05, + "loss": 2.4613, + "step": 9640 + }, + { + "epoch": 0.03, + "learning_rate": 9.983838912053407e-05, + "loss": 2.4582, + "step": 9645 + }, + { + "epoch": 0.03, + "learning_rate": 9.98382198528783e-05, + "loss": 2.4204, + "step": 9650 + }, + { + "epoch": 0.03, + "learning_rate": 9.983805049676896e-05, + "loss": 2.4056, + "step": 9655 + }, + { + "epoch": 0.03, + "learning_rate": 9.983788105220634e-05, + "loss": 2.4815, + "step": 9660 + }, + { + "epoch": 0.03, + "learning_rate": 9.983771151919077e-05, + "loss": 2.3954, + "step": 9665 + }, + { + "epoch": 0.03, + "learning_rate": 9.983754189772254e-05, + "loss": 2.4104, + "step": 9670 + }, + { + "epoch": 0.03, + "learning_rate": 9.983737218780194e-05, + "loss": 2.4315, + "step": 9675 + }, + { + "epoch": 0.03, + "learning_rate": 9.983720238942929e-05, + "loss": 2.4372, + "step": 9680 + }, + { + "epoch": 0.03, + "learning_rate": 9.983703250260486e-05, + "loss": 2.4498, + "step": 9685 + }, + { + "epoch": 0.03, + "learning_rate": 9.983686252732898e-05, + "loss": 2.4869, + "step": 9690 + }, + { + "epoch": 0.03, + "learning_rate": 9.983669246360195e-05, + "loss": 2.5492, + "step": 9695 + }, + { + "epoch": 0.03, + "learning_rate": 9.983652231142406e-05, + "loss": 2.4308, + "step": 9700 + }, + { + "epoch": 0.03, + "learning_rate": 9.983635207079562e-05, + "loss": 2.4237, + "step": 9705 + }, + { + "epoch": 0.03, + "learning_rate": 9.983618174171692e-05, + "loss": 2.3146, + "step": 9710 + }, + { + "epoch": 0.03, + "learning_rate": 9.983601132418829e-05, + "loss": 2.4073, + "step": 9715 + }, + { + "epoch": 0.03, + "learning_rate": 9.983584081821e-05, + "loss": 2.5213, + "step": 9720 + }, + { + "epoch": 0.03, + "learning_rate": 9.983567022378238e-05, + "loss": 2.4501, + "step": 9725 + }, + { + "epoch": 0.03, + "learning_rate": 9.983549954090569e-05, + "loss": 2.2896, + "step": 9730 + }, + { + "epoch": 0.03, + "learning_rate": 9.983532876958028e-05, + "loss": 2.4413, + "step": 9735 + }, + { + "epoch": 0.03, + "learning_rate": 9.983515790980644e-05, + "loss": 2.4906, + "step": 9740 + }, + { + "epoch": 0.03, + "learning_rate": 9.983498696158447e-05, + "loss": 2.4717, + "step": 9745 + }, + { + "epoch": 0.03, + "learning_rate": 9.983481592491469e-05, + "loss": 2.4593, + "step": 9750 + }, + { + "epoch": 0.03, + "learning_rate": 9.983464479979734e-05, + "loss": 2.4452, + "step": 9755 + }, + { + "epoch": 0.03, + "learning_rate": 9.98344735862328e-05, + "loss": 2.432, + "step": 9760 + }, + { + "epoch": 0.03, + "learning_rate": 9.983430228422134e-05, + "loss": 2.3781, + "step": 9765 + }, + { + "epoch": 0.03, + "learning_rate": 9.983413089376327e-05, + "loss": 2.5103, + "step": 9770 + }, + { + "epoch": 0.03, + "learning_rate": 9.983395941485888e-05, + "loss": 2.5856, + "step": 9775 + }, + { + "epoch": 0.03, + "learning_rate": 9.98337878475085e-05, + "loss": 2.4591, + "step": 9780 + }, + { + "epoch": 0.03, + "learning_rate": 9.98336161917124e-05, + "loss": 2.5292, + "step": 9785 + }, + { + "epoch": 0.03, + "learning_rate": 9.983344444747092e-05, + "loss": 2.3433, + "step": 9790 + }, + { + "epoch": 0.03, + "learning_rate": 9.983327261478435e-05, + "loss": 2.4542, + "step": 9795 + }, + { + "epoch": 0.03, + "learning_rate": 9.983310069365299e-05, + "loss": 2.3612, + "step": 9800 + }, + { + "epoch": 0.03, + "learning_rate": 9.983292868407715e-05, + "loss": 2.4351, + "step": 9805 + }, + { + "epoch": 0.03, + "learning_rate": 9.983275658605715e-05, + "loss": 2.4913, + "step": 9810 + }, + { + "epoch": 0.03, + "learning_rate": 9.983258439959327e-05, + "loss": 2.3785, + "step": 9815 + }, + { + "epoch": 0.03, + "learning_rate": 9.983241212468582e-05, + "loss": 2.477, + "step": 9820 + }, + { + "epoch": 0.03, + "learning_rate": 9.983223976133511e-05, + "loss": 2.4861, + "step": 9825 + }, + { + "epoch": 0.03, + "learning_rate": 9.983206730954146e-05, + "loss": 2.4676, + "step": 9830 + }, + { + "epoch": 0.03, + "learning_rate": 9.983189476930516e-05, + "loss": 2.375, + "step": 9835 + }, + { + "epoch": 0.03, + "learning_rate": 9.983172214062651e-05, + "loss": 2.5316, + "step": 9840 + }, + { + "epoch": 0.03, + "learning_rate": 9.983154942350584e-05, + "loss": 2.4656, + "step": 9845 + }, + { + "epoch": 0.03, + "learning_rate": 9.983137661794344e-05, + "loss": 2.4571, + "step": 9850 + }, + { + "epoch": 0.03, + "learning_rate": 9.983120372393961e-05, + "loss": 2.3989, + "step": 9855 + }, + { + "epoch": 0.03, + "learning_rate": 9.983103074149469e-05, + "loss": 2.5957, + "step": 9860 + }, + { + "epoch": 0.03, + "learning_rate": 9.983085767060893e-05, + "loss": 2.4681, + "step": 9865 + }, + { + "epoch": 0.03, + "learning_rate": 9.98306845112827e-05, + "loss": 2.3617, + "step": 9870 + }, + { + "epoch": 0.03, + "learning_rate": 9.983051126351626e-05, + "loss": 2.4276, + "step": 9875 + }, + { + "epoch": 0.03, + "learning_rate": 9.983033792730995e-05, + "loss": 2.3265, + "step": 9880 + }, + { + "epoch": 0.03, + "learning_rate": 9.983016450266405e-05, + "loss": 2.4753, + "step": 9885 + }, + { + "epoch": 0.03, + "learning_rate": 9.98299909895789e-05, + "loss": 2.3881, + "step": 9890 + }, + { + "epoch": 0.03, + "learning_rate": 9.982981738805477e-05, + "loss": 2.41, + "step": 9895 + }, + { + "epoch": 0.03, + "learning_rate": 9.9829643698092e-05, + "loss": 2.6656, + "step": 9900 + }, + { + "epoch": 0.03, + "learning_rate": 9.982946991969088e-05, + "loss": 2.4487, + "step": 9905 + }, + { + "epoch": 0.03, + "learning_rate": 9.982929605285173e-05, + "loss": 2.4905, + "step": 9910 + }, + { + "epoch": 0.03, + "learning_rate": 9.982912209757484e-05, + "loss": 2.4122, + "step": 9915 + }, + { + "epoch": 0.03, + "learning_rate": 9.982894805386053e-05, + "loss": 2.3822, + "step": 9920 + }, + { + "epoch": 0.03, + "learning_rate": 9.982877392170913e-05, + "loss": 2.4824, + "step": 9925 + }, + { + "epoch": 0.03, + "learning_rate": 9.982859970112092e-05, + "loss": 2.5501, + "step": 9930 + }, + { + "epoch": 0.03, + "learning_rate": 9.98284253920962e-05, + "loss": 2.4739, + "step": 9935 + }, + { + "epoch": 0.03, + "learning_rate": 9.982825099463533e-05, + "loss": 2.4931, + "step": 9940 + }, + { + "epoch": 0.03, + "learning_rate": 9.982807650873856e-05, + "loss": 2.3806, + "step": 9945 + }, + { + "epoch": 0.03, + "learning_rate": 9.982790193440625e-05, + "loss": 2.4026, + "step": 9950 + }, + { + "epoch": 0.03, + "learning_rate": 9.982772727163867e-05, + "loss": 2.4286, + "step": 9955 + }, + { + "epoch": 0.03, + "learning_rate": 9.982755252043615e-05, + "loss": 2.3445, + "step": 9960 + }, + { + "epoch": 0.03, + "learning_rate": 9.9827377680799e-05, + "loss": 2.4538, + "step": 9965 + }, + { + "epoch": 0.03, + "learning_rate": 9.982720275272752e-05, + "loss": 2.3908, + "step": 9970 + }, + { + "epoch": 0.03, + "learning_rate": 9.982702773622203e-05, + "loss": 2.4843, + "step": 9975 + }, + { + "epoch": 0.03, + "learning_rate": 9.982685263128284e-05, + "loss": 2.4972, + "step": 9980 + }, + { + "epoch": 0.03, + "learning_rate": 9.982667743791027e-05, + "loss": 2.5529, + "step": 9985 + }, + { + "epoch": 0.03, + "learning_rate": 9.982650215610461e-05, + "loss": 2.4919, + "step": 9990 + }, + { + "epoch": 0.03, + "learning_rate": 9.982632678586618e-05, + "loss": 2.4596, + "step": 9995 + }, + { + "epoch": 0.03, + "learning_rate": 9.982615132719528e-05, + "loss": 2.5557, + "step": 10000 + }, + { + "epoch": 0.03, + "learning_rate": 9.982597578009225e-05, + "loss": 2.541, + "step": 10005 + }, + { + "epoch": 0.03, + "learning_rate": 9.98258001445574e-05, + "loss": 2.4642, + "step": 10010 + }, + { + "epoch": 0.03, + "learning_rate": 9.9825624420591e-05, + "loss": 2.5251, + "step": 10015 + }, + { + "epoch": 0.03, + "learning_rate": 9.98254486081934e-05, + "loss": 2.6081, + "step": 10020 + }, + { + "epoch": 0.03, + "learning_rate": 9.98252727073649e-05, + "loss": 2.4746, + "step": 10025 + }, + { + "epoch": 0.03, + "learning_rate": 9.982509671810581e-05, + "loss": 2.4518, + "step": 10030 + }, + { + "epoch": 0.03, + "learning_rate": 9.982492064041645e-05, + "loss": 2.4759, + "step": 10035 + }, + { + "epoch": 0.03, + "learning_rate": 9.982474447429712e-05, + "loss": 2.3112, + "step": 10040 + }, + { + "epoch": 0.03, + "learning_rate": 9.982456821974815e-05, + "loss": 2.345, + "step": 10045 + }, + { + "epoch": 0.03, + "learning_rate": 9.982439187676984e-05, + "loss": 2.5058, + "step": 10050 + }, + { + "epoch": 0.03, + "learning_rate": 9.982421544536252e-05, + "loss": 2.3389, + "step": 10055 + }, + { + "epoch": 0.03, + "learning_rate": 9.982403892552647e-05, + "loss": 2.5035, + "step": 10060 + }, + { + "epoch": 0.03, + "learning_rate": 9.982386231726203e-05, + "loss": 2.5297, + "step": 10065 + }, + { + "epoch": 0.03, + "learning_rate": 9.982368562056951e-05, + "loss": 2.3595, + "step": 10070 + }, + { + "epoch": 0.03, + "learning_rate": 9.982350883544922e-05, + "loss": 2.3826, + "step": 10075 + }, + { + "epoch": 0.03, + "learning_rate": 9.982333196190146e-05, + "loss": 2.4686, + "step": 10080 + }, + { + "epoch": 0.03, + "learning_rate": 9.982315499992657e-05, + "loss": 2.3624, + "step": 10085 + }, + { + "epoch": 0.03, + "learning_rate": 9.982297794952485e-05, + "loss": 2.3977, + "step": 10090 + }, + { + "epoch": 0.03, + "learning_rate": 9.982280081069661e-05, + "loss": 2.5415, + "step": 10095 + }, + { + "epoch": 0.03, + "learning_rate": 9.982262358344219e-05, + "loss": 2.388, + "step": 10100 + }, + { + "epoch": 0.03, + "learning_rate": 9.982244626776187e-05, + "loss": 2.4966, + "step": 10105 + }, + { + "epoch": 0.03, + "learning_rate": 9.982226886365598e-05, + "loss": 2.5819, + "step": 10110 + }, + { + "epoch": 0.03, + "learning_rate": 9.982209137112483e-05, + "loss": 2.4634, + "step": 10115 + }, + { + "epoch": 0.03, + "learning_rate": 9.982191379016874e-05, + "loss": 2.4885, + "step": 10120 + }, + { + "epoch": 0.03, + "learning_rate": 9.982173612078803e-05, + "loss": 2.3958, + "step": 10125 + }, + { + "epoch": 0.03, + "learning_rate": 9.982155836298301e-05, + "loss": 2.4284, + "step": 10130 + }, + { + "epoch": 0.03, + "learning_rate": 9.982138051675399e-05, + "loss": 2.4197, + "step": 10135 + }, + { + "epoch": 0.03, + "learning_rate": 9.98212025821013e-05, + "loss": 2.3913, + "step": 10140 + }, + { + "epoch": 0.03, + "learning_rate": 9.982102455902524e-05, + "loss": 2.4295, + "step": 10145 + }, + { + "epoch": 0.03, + "learning_rate": 9.982084644752614e-05, + "loss": 2.4746, + "step": 10150 + }, + { + "epoch": 0.03, + "learning_rate": 9.98206682476043e-05, + "loss": 2.486, + "step": 10155 + }, + { + "epoch": 0.03, + "learning_rate": 9.982048995926006e-05, + "loss": 2.5059, + "step": 10160 + }, + { + "epoch": 0.03, + "learning_rate": 9.98203115824937e-05, + "loss": 2.4409, + "step": 10165 + }, + { + "epoch": 0.03, + "learning_rate": 9.982013311730557e-05, + "loss": 2.4253, + "step": 10170 + }, + { + "epoch": 0.03, + "learning_rate": 9.981995456369598e-05, + "loss": 2.5289, + "step": 10175 + }, + { + "epoch": 0.03, + "learning_rate": 9.981977592166523e-05, + "loss": 2.4485, + "step": 10180 + }, + { + "epoch": 0.03, + "learning_rate": 9.981959719121365e-05, + "loss": 2.3595, + "step": 10185 + }, + { + "epoch": 0.03, + "learning_rate": 9.981941837234157e-05, + "loss": 2.4802, + "step": 10190 + }, + { + "epoch": 0.03, + "learning_rate": 9.981923946504927e-05, + "loss": 2.4398, + "step": 10195 + }, + { + "epoch": 0.03, + "learning_rate": 9.981906046933711e-05, + "loss": 2.3061, + "step": 10200 + }, + { + "epoch": 0.03, + "learning_rate": 9.981888138520539e-05, + "loss": 2.4811, + "step": 10205 + }, + { + "epoch": 0.03, + "learning_rate": 9.981870221265441e-05, + "loss": 2.4269, + "step": 10210 + }, + { + "epoch": 0.03, + "learning_rate": 9.981852295168452e-05, + "loss": 2.4303, + "step": 10215 + }, + { + "epoch": 0.03, + "learning_rate": 9.981834360229603e-05, + "loss": 2.5303, + "step": 10220 + }, + { + "epoch": 0.03, + "learning_rate": 9.981816416448924e-05, + "loss": 2.4251, + "step": 10225 + }, + { + "epoch": 0.03, + "learning_rate": 9.981798463826447e-05, + "loss": 2.3671, + "step": 10230 + }, + { + "epoch": 0.03, + "learning_rate": 9.981780502362206e-05, + "loss": 2.3821, + "step": 10235 + }, + { + "epoch": 0.03, + "learning_rate": 9.981762532056231e-05, + "loss": 2.4917, + "step": 10240 + }, + { + "epoch": 0.03, + "learning_rate": 9.981744552908556e-05, + "loss": 2.4494, + "step": 10245 + }, + { + "epoch": 0.03, + "learning_rate": 9.98172656491921e-05, + "loss": 2.5207, + "step": 10250 + }, + { + "epoch": 0.03, + "learning_rate": 9.981708568088227e-05, + "loss": 2.4631, + "step": 10255 + }, + { + "epoch": 0.03, + "learning_rate": 9.981690562415638e-05, + "loss": 2.5461, + "step": 10260 + }, + { + "epoch": 0.03, + "learning_rate": 9.981672547901475e-05, + "loss": 2.4878, + "step": 10265 + }, + { + "epoch": 0.03, + "learning_rate": 9.981654524545772e-05, + "loss": 2.4466, + "step": 10270 + }, + { + "epoch": 0.03, + "learning_rate": 9.981636492348558e-05, + "loss": 2.4305, + "step": 10275 + }, + { + "epoch": 0.03, + "learning_rate": 9.981618451309867e-05, + "loss": 2.4592, + "step": 10280 + }, + { + "epoch": 0.03, + "learning_rate": 9.98160040142973e-05, + "loss": 2.4345, + "step": 10285 + }, + { + "epoch": 0.03, + "learning_rate": 9.981582342708177e-05, + "loss": 2.4638, + "step": 10290 + }, + { + "epoch": 0.03, + "learning_rate": 9.981564275145245e-05, + "loss": 2.4701, + "step": 10295 + }, + { + "epoch": 0.03, + "learning_rate": 9.981546198740964e-05, + "loss": 2.2569, + "step": 10300 + }, + { + "epoch": 0.03, + "learning_rate": 9.981528113495365e-05, + "loss": 2.4981, + "step": 10305 + }, + { + "epoch": 0.03, + "learning_rate": 9.981510019408479e-05, + "loss": 2.6479, + "step": 10310 + }, + { + "epoch": 0.03, + "learning_rate": 9.981491916480341e-05, + "loss": 2.3557, + "step": 10315 + }, + { + "epoch": 0.03, + "learning_rate": 9.981473804710981e-05, + "loss": 2.5265, + "step": 10320 + }, + { + "epoch": 0.03, + "learning_rate": 9.981455684100434e-05, + "loss": 2.435, + "step": 10325 + }, + { + "epoch": 0.03, + "learning_rate": 9.981437554648728e-05, + "loss": 2.3656, + "step": 10330 + }, + { + "epoch": 0.03, + "learning_rate": 9.981419416355899e-05, + "loss": 2.4834, + "step": 10335 + }, + { + "epoch": 0.03, + "learning_rate": 9.981401269221976e-05, + "loss": 2.4969, + "step": 10340 + }, + { + "epoch": 0.03, + "learning_rate": 9.981383113246994e-05, + "loss": 2.4649, + "step": 10345 + }, + { + "epoch": 0.03, + "learning_rate": 9.981364948430984e-05, + "loss": 2.4877, + "step": 10350 + }, + { + "epoch": 0.03, + "learning_rate": 9.981346774773977e-05, + "loss": 2.5526, + "step": 10355 + }, + { + "epoch": 0.03, + "learning_rate": 9.981328592276007e-05, + "loss": 2.4556, + "step": 10360 + }, + { + "epoch": 0.03, + "learning_rate": 9.981310400937106e-05, + "loss": 2.5164, + "step": 10365 + }, + { + "epoch": 0.03, + "learning_rate": 9.981292200757306e-05, + "loss": 2.4615, + "step": 10370 + }, + { + "epoch": 0.03, + "learning_rate": 9.981273991736639e-05, + "loss": 2.3917, + "step": 10375 + }, + { + "epoch": 0.03, + "learning_rate": 9.98125577387514e-05, + "loss": 2.458, + "step": 10380 + }, + { + "epoch": 0.03, + "learning_rate": 9.981237547172836e-05, + "loss": 2.3409, + "step": 10385 + }, + { + "epoch": 0.03, + "learning_rate": 9.981219311629764e-05, + "loss": 2.3321, + "step": 10390 + }, + { + "epoch": 0.03, + "learning_rate": 9.981201067245954e-05, + "loss": 2.4315, + "step": 10395 + }, + { + "epoch": 0.03, + "learning_rate": 9.981182814021439e-05, + "loss": 2.4391, + "step": 10400 + }, + { + "epoch": 0.03, + "learning_rate": 9.981164551956251e-05, + "loss": 2.4694, + "step": 10405 + }, + { + "epoch": 0.03, + "learning_rate": 9.981146281050425e-05, + "loss": 2.4851, + "step": 10410 + }, + { + "epoch": 0.03, + "learning_rate": 9.981128001303989e-05, + "loss": 2.3934, + "step": 10415 + }, + { + "epoch": 0.03, + "learning_rate": 9.981109712716979e-05, + "loss": 2.4163, + "step": 10420 + }, + { + "epoch": 0.03, + "learning_rate": 9.981091415289425e-05, + "loss": 2.4758, + "step": 10425 + }, + { + "epoch": 0.03, + "learning_rate": 9.981073109021362e-05, + "loss": 2.3813, + "step": 10430 + }, + { + "epoch": 0.03, + "learning_rate": 9.98105479391282e-05, + "loss": 2.532, + "step": 10435 + }, + { + "epoch": 0.03, + "learning_rate": 9.981036469963834e-05, + "loss": 2.414, + "step": 10440 + }, + { + "epoch": 0.03, + "learning_rate": 9.981018137174433e-05, + "loss": 2.5457, + "step": 10445 + }, + { + "epoch": 0.03, + "learning_rate": 9.980999795544654e-05, + "loss": 2.4775, + "step": 10450 + }, + { + "epoch": 0.03, + "learning_rate": 9.980981445074528e-05, + "loss": 2.5315, + "step": 10455 + }, + { + "epoch": 0.03, + "learning_rate": 9.980963085764086e-05, + "loss": 2.3417, + "step": 10460 + }, + { + "epoch": 0.03, + "learning_rate": 9.980944717613361e-05, + "loss": 2.5114, + "step": 10465 + }, + { + "epoch": 0.03, + "learning_rate": 9.980926340622385e-05, + "loss": 2.4127, + "step": 10470 + }, + { + "epoch": 0.03, + "learning_rate": 9.980907954791192e-05, + "loss": 2.5222, + "step": 10475 + }, + { + "epoch": 0.03, + "learning_rate": 9.980889560119816e-05, + "loss": 2.3759, + "step": 10480 + }, + { + "epoch": 0.03, + "learning_rate": 9.980871156608287e-05, + "loss": 2.4514, + "step": 10485 + }, + { + "epoch": 0.03, + "learning_rate": 9.980852744256639e-05, + "loss": 2.3674, + "step": 10490 + }, + { + "epoch": 0.03, + "learning_rate": 9.980834323064903e-05, + "loss": 2.4072, + "step": 10495 + }, + { + "epoch": 0.03, + "learning_rate": 9.980815893033114e-05, + "loss": 2.493, + "step": 10500 + }, + { + "epoch": 0.03, + "learning_rate": 9.980797454161303e-05, + "loss": 2.4457, + "step": 10505 + }, + { + "epoch": 0.03, + "learning_rate": 9.980779006449506e-05, + "loss": 2.4285, + "step": 10510 + }, + { + "epoch": 0.03, + "learning_rate": 9.98076054989775e-05, + "loss": 2.3942, + "step": 10515 + }, + { + "epoch": 0.03, + "learning_rate": 9.980742084506072e-05, + "loss": 2.5572, + "step": 10520 + }, + { + "epoch": 0.03, + "learning_rate": 9.980723610274504e-05, + "loss": 2.4894, + "step": 10525 + }, + { + "epoch": 0.03, + "learning_rate": 9.980705127203079e-05, + "loss": 2.4145, + "step": 10530 + }, + { + "epoch": 0.03, + "learning_rate": 9.980686635291826e-05, + "loss": 2.3478, + "step": 10535 + }, + { + "epoch": 0.03, + "learning_rate": 9.980668134540785e-05, + "loss": 2.5357, + "step": 10540 + }, + { + "epoch": 0.03, + "learning_rate": 9.980649624949984e-05, + "loss": 2.4038, + "step": 10545 + }, + { + "epoch": 0.03, + "learning_rate": 9.980631106519456e-05, + "loss": 2.4915, + "step": 10550 + }, + { + "epoch": 0.03, + "learning_rate": 9.980612579249236e-05, + "loss": 2.506, + "step": 10555 + }, + { + "epoch": 0.03, + "learning_rate": 9.980594043139354e-05, + "loss": 2.4229, + "step": 10560 + }, + { + "epoch": 0.03, + "learning_rate": 9.980575498189845e-05, + "loss": 2.4172, + "step": 10565 + }, + { + "epoch": 0.03, + "learning_rate": 9.980556944400741e-05, + "loss": 2.3974, + "step": 10570 + }, + { + "epoch": 0.03, + "learning_rate": 9.980538381772075e-05, + "loss": 2.4178, + "step": 10575 + }, + { + "epoch": 0.03, + "learning_rate": 9.980519810303881e-05, + "loss": 2.5312, + "step": 10580 + }, + { + "epoch": 0.03, + "learning_rate": 9.980501229996191e-05, + "loss": 2.391, + "step": 10585 + }, + { + "epoch": 0.03, + "learning_rate": 9.980482640849038e-05, + "loss": 2.4986, + "step": 10590 + }, + { + "epoch": 0.03, + "learning_rate": 9.980464042862457e-05, + "loss": 2.4477, + "step": 10595 + }, + { + "epoch": 0.03, + "learning_rate": 9.980445436036476e-05, + "loss": 2.5485, + "step": 10600 + }, + { + "epoch": 0.03, + "learning_rate": 9.980426820371134e-05, + "loss": 2.5002, + "step": 10605 + }, + { + "epoch": 0.03, + "learning_rate": 9.98040819586646e-05, + "loss": 2.4999, + "step": 10610 + }, + { + "epoch": 0.03, + "learning_rate": 9.980389562522487e-05, + "loss": 2.3618, + "step": 10615 + }, + { + "epoch": 0.03, + "learning_rate": 9.98037092033925e-05, + "loss": 2.3669, + "step": 10620 + }, + { + "epoch": 0.03, + "learning_rate": 9.980352269316783e-05, + "loss": 2.504, + "step": 10625 + }, + { + "epoch": 0.03, + "learning_rate": 9.980333609455118e-05, + "loss": 2.3761, + "step": 10630 + }, + { + "epoch": 0.03, + "learning_rate": 9.980314940754285e-05, + "loss": 2.3396, + "step": 10635 + }, + { + "epoch": 0.03, + "learning_rate": 9.98029626321432e-05, + "loss": 2.3799, + "step": 10640 + }, + { + "epoch": 0.03, + "learning_rate": 9.980277576835257e-05, + "loss": 2.4054, + "step": 10645 + }, + { + "epoch": 0.03, + "learning_rate": 9.980258881617128e-05, + "loss": 2.5248, + "step": 10650 + }, + { + "epoch": 0.03, + "learning_rate": 9.980240177559968e-05, + "loss": 2.3307, + "step": 10655 + }, + { + "epoch": 0.03, + "learning_rate": 9.980221464663806e-05, + "loss": 2.5453, + "step": 10660 + }, + { + "epoch": 0.03, + "learning_rate": 9.980202742928678e-05, + "loss": 2.4417, + "step": 10665 + }, + { + "epoch": 0.03, + "learning_rate": 9.980184012354617e-05, + "loss": 2.4376, + "step": 10670 + }, + { + "epoch": 0.03, + "learning_rate": 9.980165272941657e-05, + "loss": 2.5612, + "step": 10675 + }, + { + "epoch": 0.03, + "learning_rate": 9.98014652468983e-05, + "loss": 2.4572, + "step": 10680 + }, + { + "epoch": 0.03, + "learning_rate": 9.98012776759917e-05, + "loss": 2.4076, + "step": 10685 + }, + { + "epoch": 0.03, + "learning_rate": 9.98010900166971e-05, + "loss": 2.4799, + "step": 10690 + }, + { + "epoch": 0.03, + "learning_rate": 9.980090226901482e-05, + "loss": 2.4674, + "step": 10695 + }, + { + "epoch": 0.03, + "learning_rate": 9.980071443294522e-05, + "loss": 2.3989, + "step": 10700 + }, + { + "epoch": 0.03, + "learning_rate": 9.980052650848861e-05, + "loss": 2.5223, + "step": 10705 + }, + { + "epoch": 0.03, + "learning_rate": 9.980033849564535e-05, + "loss": 2.4506, + "step": 10710 + }, + { + "epoch": 0.03, + "learning_rate": 9.980015039441574e-05, + "loss": 2.463, + "step": 10715 + }, + { + "epoch": 0.03, + "learning_rate": 9.979996220480013e-05, + "loss": 2.4412, + "step": 10720 + }, + { + "epoch": 0.03, + "learning_rate": 9.979977392679887e-05, + "loss": 2.4534, + "step": 10725 + }, + { + "epoch": 0.03, + "learning_rate": 9.979958556041226e-05, + "loss": 2.2679, + "step": 10730 + }, + { + "epoch": 0.03, + "learning_rate": 9.979939710564067e-05, + "loss": 2.5339, + "step": 10735 + }, + { + "epoch": 0.03, + "learning_rate": 9.979920856248442e-05, + "loss": 2.4797, + "step": 10740 + }, + { + "epoch": 0.03, + "learning_rate": 9.979901993094383e-05, + "loss": 2.5306, + "step": 10745 + }, + { + "epoch": 0.03, + "learning_rate": 9.979883121101924e-05, + "loss": 2.5069, + "step": 10750 + }, + { + "epoch": 0.03, + "learning_rate": 9.979864240271099e-05, + "loss": 2.4442, + "step": 10755 + }, + { + "epoch": 0.03, + "learning_rate": 9.979845350601944e-05, + "loss": 2.3553, + "step": 10760 + }, + { + "epoch": 0.03, + "learning_rate": 9.97982645209449e-05, + "loss": 2.4307, + "step": 10765 + }, + { + "epoch": 0.03, + "learning_rate": 9.979807544748769e-05, + "loss": 2.5324, + "step": 10770 + }, + { + "epoch": 0.03, + "learning_rate": 9.979788628564816e-05, + "loss": 2.4991, + "step": 10775 + }, + { + "epoch": 0.03, + "learning_rate": 9.979769703542667e-05, + "loss": 2.4616, + "step": 10780 + }, + { + "epoch": 0.03, + "learning_rate": 9.979750769682353e-05, + "loss": 2.4064, + "step": 10785 + }, + { + "epoch": 0.03, + "learning_rate": 9.979731826983907e-05, + "loss": 2.3875, + "step": 10790 + }, + { + "epoch": 0.03, + "learning_rate": 9.979712875447364e-05, + "loss": 2.4587, + "step": 10795 + }, + { + "epoch": 0.03, + "learning_rate": 9.979693915072758e-05, + "loss": 2.3513, + "step": 10800 + }, + { + "epoch": 0.03, + "learning_rate": 9.979674945860122e-05, + "loss": 2.3306, + "step": 10805 + }, + { + "epoch": 0.03, + "learning_rate": 9.979655967809489e-05, + "loss": 2.394, + "step": 10810 + }, + { + "epoch": 0.03, + "learning_rate": 9.979636980920894e-05, + "loss": 2.3921, + "step": 10815 + }, + { + "epoch": 0.03, + "learning_rate": 9.97961798519437e-05, + "loss": 2.4554, + "step": 10820 + }, + { + "epoch": 0.03, + "learning_rate": 9.979598980629949e-05, + "loss": 2.4895, + "step": 10825 + }, + { + "epoch": 0.03, + "learning_rate": 9.979579967227668e-05, + "loss": 2.3783, + "step": 10830 + }, + { + "epoch": 0.03, + "learning_rate": 9.97956094498756e-05, + "loss": 2.4345, + "step": 10835 + }, + { + "epoch": 0.03, + "learning_rate": 9.979541913909657e-05, + "loss": 2.4629, + "step": 10840 + }, + { + "epoch": 0.03, + "learning_rate": 9.979522873993994e-05, + "loss": 2.3946, + "step": 10845 + }, + { + "epoch": 0.03, + "learning_rate": 9.979503825240605e-05, + "loss": 2.5383, + "step": 10850 + }, + { + "epoch": 0.03, + "learning_rate": 9.979484767649523e-05, + "loss": 2.4518, + "step": 10855 + }, + { + "epoch": 0.03, + "learning_rate": 9.979465701220781e-05, + "loss": 2.305, + "step": 10860 + }, + { + "epoch": 0.03, + "learning_rate": 9.979446625954416e-05, + "loss": 2.4731, + "step": 10865 + }, + { + "epoch": 0.03, + "learning_rate": 9.979427541850459e-05, + "loss": 2.4577, + "step": 10870 + }, + { + "epoch": 0.03, + "learning_rate": 9.979408448908944e-05, + "loss": 2.4935, + "step": 10875 + }, + { + "epoch": 0.03, + "learning_rate": 9.979389347129907e-05, + "loss": 2.4029, + "step": 10880 + }, + { + "epoch": 0.03, + "learning_rate": 9.97937023651338e-05, + "loss": 2.455, + "step": 10885 + }, + { + "epoch": 0.03, + "learning_rate": 9.979351117059397e-05, + "loss": 2.3458, + "step": 10890 + }, + { + "epoch": 0.03, + "learning_rate": 9.979331988767993e-05, + "loss": 2.5941, + "step": 10895 + }, + { + "epoch": 0.03, + "learning_rate": 9.979312851639203e-05, + "loss": 2.3824, + "step": 10900 + }, + { + "epoch": 0.03, + "learning_rate": 9.979293705673056e-05, + "loss": 2.4305, + "step": 10905 + }, + { + "epoch": 0.03, + "learning_rate": 9.979274550869591e-05, + "loss": 2.511, + "step": 10910 + }, + { + "epoch": 0.03, + "learning_rate": 9.97925538722884e-05, + "loss": 2.4498, + "step": 10915 + }, + { + "epoch": 0.03, + "learning_rate": 9.979236214750837e-05, + "loss": 2.4157, + "step": 10920 + }, + { + "epoch": 0.03, + "learning_rate": 9.979217033435617e-05, + "loss": 2.508, + "step": 10925 + }, + { + "epoch": 0.03, + "learning_rate": 9.979197843283212e-05, + "loss": 2.5995, + "step": 10930 + }, + { + "epoch": 0.03, + "learning_rate": 9.979178644293659e-05, + "loss": 2.5279, + "step": 10935 + }, + { + "epoch": 0.03, + "learning_rate": 9.97915943646699e-05, + "loss": 2.5239, + "step": 10940 + }, + { + "epoch": 0.03, + "learning_rate": 9.97914021980324e-05, + "loss": 2.4266, + "step": 10945 + }, + { + "epoch": 0.03, + "learning_rate": 9.979120994302443e-05, + "loss": 2.5462, + "step": 10950 + }, + { + "epoch": 0.03, + "learning_rate": 9.97910175996463e-05, + "loss": 2.4417, + "step": 10955 + }, + { + "epoch": 0.03, + "learning_rate": 9.979082516789841e-05, + "loss": 2.5134, + "step": 10960 + }, + { + "epoch": 0.03, + "learning_rate": 9.979063264778106e-05, + "loss": 2.3999, + "step": 10965 + }, + { + "epoch": 0.03, + "learning_rate": 9.979044003929461e-05, + "loss": 2.6629, + "step": 10970 + }, + { + "epoch": 0.03, + "learning_rate": 9.97902473424394e-05, + "loss": 2.4275, + "step": 10975 + }, + { + "epoch": 0.03, + "learning_rate": 9.979005455721575e-05, + "loss": 2.3655, + "step": 10980 + }, + { + "epoch": 0.03, + "learning_rate": 9.978986168362403e-05, + "loss": 2.4918, + "step": 10985 + }, + { + "epoch": 0.03, + "learning_rate": 9.978966872166455e-05, + "loss": 2.4647, + "step": 10990 + }, + { + "epoch": 0.03, + "learning_rate": 9.978947567133769e-05, + "loss": 2.5938, + "step": 10995 + }, + { + "epoch": 0.03, + "learning_rate": 9.97892825326438e-05, + "loss": 2.5192, + "step": 11000 + }, + { + "epoch": 0.03, + "learning_rate": 9.978908930558315e-05, + "loss": 2.4353, + "step": 11005 + }, + { + "epoch": 0.03, + "learning_rate": 9.978889599015617e-05, + "loss": 2.5045, + "step": 11010 + }, + { + "epoch": 0.03, + "learning_rate": 9.978870258636315e-05, + "loss": 2.5052, + "step": 11015 + }, + { + "epoch": 0.03, + "learning_rate": 9.978850909420446e-05, + "loss": 2.5531, + "step": 11020 + }, + { + "epoch": 0.03, + "learning_rate": 9.978831551368042e-05, + "loss": 2.4485, + "step": 11025 + }, + { + "epoch": 0.03, + "learning_rate": 9.978812184479138e-05, + "loss": 2.4458, + "step": 11030 + }, + { + "epoch": 0.03, + "learning_rate": 9.97879280875377e-05, + "loss": 2.6005, + "step": 11035 + }, + { + "epoch": 0.03, + "learning_rate": 9.978773424191972e-05, + "loss": 2.3568, + "step": 11040 + }, + { + "epoch": 0.03, + "learning_rate": 9.978754030793775e-05, + "loss": 2.4012, + "step": 11045 + }, + { + "epoch": 0.03, + "learning_rate": 9.978734628559219e-05, + "loss": 2.3939, + "step": 11050 + }, + { + "epoch": 0.03, + "learning_rate": 9.978715217488333e-05, + "loss": 2.515, + "step": 11055 + }, + { + "epoch": 0.03, + "learning_rate": 9.978695797581156e-05, + "loss": 2.4067, + "step": 11060 + }, + { + "epoch": 0.03, + "learning_rate": 9.978676368837721e-05, + "loss": 2.5186, + "step": 11065 + }, + { + "epoch": 0.03, + "learning_rate": 9.97865693125806e-05, + "loss": 2.5086, + "step": 11070 + }, + { + "epoch": 0.03, + "learning_rate": 9.97863748484221e-05, + "loss": 2.33, + "step": 11075 + }, + { + "epoch": 0.03, + "learning_rate": 9.978618029590205e-05, + "loss": 2.5419, + "step": 11080 + }, + { + "epoch": 0.03, + "learning_rate": 9.97859856550208e-05, + "loss": 2.3864, + "step": 11085 + }, + { + "epoch": 0.03, + "learning_rate": 9.978579092577868e-05, + "loss": 2.4028, + "step": 11090 + }, + { + "epoch": 0.03, + "learning_rate": 9.978559610817607e-05, + "loss": 2.3991, + "step": 11095 + }, + { + "epoch": 0.03, + "learning_rate": 9.978540120221326e-05, + "loss": 2.4616, + "step": 11100 + }, + { + "epoch": 0.03, + "learning_rate": 9.978520620789064e-05, + "loss": 2.5089, + "step": 11105 + }, + { + "epoch": 0.03, + "learning_rate": 9.978501112520855e-05, + "loss": 2.4838, + "step": 11110 + }, + { + "epoch": 0.03, + "learning_rate": 9.978481595416733e-05, + "loss": 2.4518, + "step": 11115 + }, + { + "epoch": 0.03, + "learning_rate": 9.978462069476733e-05, + "loss": 2.4606, + "step": 11120 + }, + { + "epoch": 0.03, + "learning_rate": 9.978442534700887e-05, + "loss": 2.5134, + "step": 11125 + }, + { + "epoch": 0.03, + "learning_rate": 9.978422991089233e-05, + "loss": 2.455, + "step": 11130 + }, + { + "epoch": 0.03, + "learning_rate": 9.978403438641806e-05, + "loss": 2.4655, + "step": 11135 + }, + { + "epoch": 0.03, + "learning_rate": 9.97838387735864e-05, + "loss": 2.5264, + "step": 11140 + }, + { + "epoch": 0.03, + "learning_rate": 9.978364307239768e-05, + "loss": 2.4925, + "step": 11145 + }, + { + "epoch": 0.03, + "learning_rate": 9.978344728285224e-05, + "loss": 2.4683, + "step": 11150 + }, + { + "epoch": 0.03, + "learning_rate": 9.978325140495046e-05, + "loss": 2.3233, + "step": 11155 + }, + { + "epoch": 0.03, + "learning_rate": 9.978305543869268e-05, + "loss": 2.4655, + "step": 11160 + }, + { + "epoch": 0.03, + "learning_rate": 9.978285938407923e-05, + "loss": 2.5205, + "step": 11165 + }, + { + "epoch": 0.03, + "learning_rate": 9.978266324111047e-05, + "loss": 2.4004, + "step": 11170 + }, + { + "epoch": 0.03, + "learning_rate": 9.978246700978675e-05, + "loss": 2.419, + "step": 11175 + }, + { + "epoch": 0.03, + "learning_rate": 9.978227069010842e-05, + "loss": 2.4019, + "step": 11180 + }, + { + "epoch": 0.03, + "learning_rate": 9.978207428207581e-05, + "loss": 2.3716, + "step": 11185 + }, + { + "epoch": 0.03, + "learning_rate": 9.97818777856893e-05, + "loss": 2.3605, + "step": 11190 + }, + { + "epoch": 0.03, + "learning_rate": 9.97816812009492e-05, + "loss": 2.3647, + "step": 11195 + }, + { + "epoch": 0.03, + "learning_rate": 9.978148452785589e-05, + "loss": 2.5104, + "step": 11200 + }, + { + "epoch": 0.03, + "learning_rate": 9.978128776640971e-05, + "loss": 2.4851, + "step": 11205 + }, + { + "epoch": 0.03, + "learning_rate": 9.978109091661102e-05, + "loss": 2.2988, + "step": 11210 + }, + { + "epoch": 0.03, + "learning_rate": 9.978089397846014e-05, + "loss": 2.4219, + "step": 11215 + }, + { + "epoch": 0.03, + "learning_rate": 9.978069695195745e-05, + "loss": 2.4359, + "step": 11220 + }, + { + "epoch": 0.03, + "learning_rate": 9.978049983710328e-05, + "loss": 2.4711, + "step": 11225 + }, + { + "epoch": 0.03, + "learning_rate": 9.978030263389799e-05, + "loss": 2.4838, + "step": 11230 + }, + { + "epoch": 0.03, + "learning_rate": 9.978010534234193e-05, + "loss": 2.4172, + "step": 11235 + }, + { + "epoch": 0.03, + "learning_rate": 9.977990796243544e-05, + "loss": 2.4411, + "step": 11240 + }, + { + "epoch": 0.03, + "learning_rate": 9.977971049417887e-05, + "loss": 2.4207, + "step": 11245 + }, + { + "epoch": 0.03, + "learning_rate": 9.977951293757259e-05, + "loss": 2.4363, + "step": 11250 + }, + { + "epoch": 0.03, + "learning_rate": 9.977931529261692e-05, + "loss": 2.3781, + "step": 11255 + }, + { + "epoch": 0.03, + "learning_rate": 9.977911755931227e-05, + "loss": 2.4884, + "step": 11260 + }, + { + "epoch": 0.03, + "learning_rate": 9.977891973765891e-05, + "loss": 2.4071, + "step": 11265 + }, + { + "epoch": 0.03, + "learning_rate": 9.977872182765724e-05, + "loss": 2.443, + "step": 11270 + }, + { + "epoch": 0.03, + "learning_rate": 9.977852382930761e-05, + "loss": 2.3874, + "step": 11275 + }, + { + "epoch": 0.03, + "learning_rate": 9.977832574261037e-05, + "loss": 2.4024, + "step": 11280 + }, + { + "epoch": 0.03, + "learning_rate": 9.977812756756585e-05, + "loss": 2.5742, + "step": 11285 + }, + { + "epoch": 0.03, + "learning_rate": 9.977792930417443e-05, + "loss": 2.393, + "step": 11290 + }, + { + "epoch": 0.03, + "learning_rate": 9.977773095243643e-05, + "loss": 2.5272, + "step": 11295 + }, + { + "epoch": 0.03, + "learning_rate": 9.977753251235224e-05, + "loss": 2.4544, + "step": 11300 + }, + { + "epoch": 0.03, + "learning_rate": 9.977733398392219e-05, + "loss": 2.454, + "step": 11305 + }, + { + "epoch": 0.03, + "learning_rate": 9.977713536714663e-05, + "loss": 2.5138, + "step": 11310 + }, + { + "epoch": 0.03, + "learning_rate": 9.977693666202591e-05, + "loss": 2.4543, + "step": 11315 + }, + { + "epoch": 0.03, + "learning_rate": 9.977673786856042e-05, + "loss": 2.4011, + "step": 11320 + }, + { + "epoch": 0.03, + "learning_rate": 9.977653898675044e-05, + "loss": 2.3837, + "step": 11325 + }, + { + "epoch": 0.03, + "learning_rate": 9.97763400165964e-05, + "loss": 2.554, + "step": 11330 + }, + { + "epoch": 0.03, + "learning_rate": 9.97761409580986e-05, + "loss": 2.3322, + "step": 11335 + }, + { + "epoch": 0.03, + "learning_rate": 9.977594181125743e-05, + "loss": 2.5074, + "step": 11340 + }, + { + "epoch": 0.03, + "learning_rate": 9.977574257607322e-05, + "loss": 2.3276, + "step": 11345 + }, + { + "epoch": 0.03, + "learning_rate": 9.977554325254633e-05, + "loss": 2.363, + "step": 11350 + }, + { + "epoch": 0.03, + "learning_rate": 9.977534384067711e-05, + "loss": 2.6495, + "step": 11355 + }, + { + "epoch": 0.03, + "learning_rate": 9.977514434046592e-05, + "loss": 2.4329, + "step": 11360 + }, + { + "epoch": 0.03, + "learning_rate": 9.977494475191312e-05, + "loss": 2.4815, + "step": 11365 + }, + { + "epoch": 0.03, + "learning_rate": 9.977474507501904e-05, + "loss": 2.5078, + "step": 11370 + }, + { + "epoch": 0.03, + "learning_rate": 9.977454530978405e-05, + "loss": 2.4275, + "step": 11375 + }, + { + "epoch": 0.03, + "learning_rate": 9.977434545620852e-05, + "loss": 2.5241, + "step": 11380 + }, + { + "epoch": 0.03, + "learning_rate": 9.977414551429278e-05, + "loss": 2.3951, + "step": 11385 + }, + { + "epoch": 0.03, + "learning_rate": 9.97739454840372e-05, + "loss": 2.3995, + "step": 11390 + }, + { + "epoch": 0.03, + "learning_rate": 9.977374536544213e-05, + "loss": 2.5284, + "step": 11395 + }, + { + "epoch": 0.03, + "learning_rate": 9.977354515850792e-05, + "loss": 2.4793, + "step": 11400 + }, + { + "epoch": 0.03, + "learning_rate": 9.977334486323493e-05, + "loss": 2.4808, + "step": 11405 + }, + { + "epoch": 0.03, + "learning_rate": 9.977314447962352e-05, + "loss": 2.5574, + "step": 11410 + }, + { + "epoch": 0.03, + "learning_rate": 9.977294400767403e-05, + "loss": 2.4792, + "step": 11415 + }, + { + "epoch": 0.03, + "learning_rate": 9.977274344738683e-05, + "loss": 2.5175, + "step": 11420 + }, + { + "epoch": 0.03, + "learning_rate": 9.977254279876228e-05, + "loss": 2.3211, + "step": 11425 + }, + { + "epoch": 0.03, + "learning_rate": 9.977234206180072e-05, + "loss": 2.3862, + "step": 11430 + }, + { + "epoch": 0.03, + "learning_rate": 9.977214123650253e-05, + "loss": 2.3303, + "step": 11435 + }, + { + "epoch": 0.03, + "learning_rate": 9.977194032286805e-05, + "loss": 2.4421, + "step": 11440 + }, + { + "epoch": 0.03, + "learning_rate": 9.977173932089762e-05, + "loss": 2.4059, + "step": 11445 + }, + { + "epoch": 0.03, + "learning_rate": 9.977153823059164e-05, + "loss": 2.4179, + "step": 11450 + }, + { + "epoch": 0.03, + "learning_rate": 9.977133705195041e-05, + "loss": 2.4698, + "step": 11455 + }, + { + "epoch": 0.03, + "learning_rate": 9.977113578497435e-05, + "loss": 2.3684, + "step": 11460 + }, + { + "epoch": 0.03, + "learning_rate": 9.977093442966376e-05, + "loss": 2.4052, + "step": 11465 + }, + { + "epoch": 0.03, + "learning_rate": 9.977073298601905e-05, + "loss": 2.4347, + "step": 11470 + }, + { + "epoch": 0.03, + "learning_rate": 9.977053145404054e-05, + "loss": 2.41, + "step": 11475 + }, + { + "epoch": 0.03, + "learning_rate": 9.97703298337286e-05, + "loss": 2.407, + "step": 11480 + }, + { + "epoch": 0.03, + "learning_rate": 9.977012812508357e-05, + "loss": 2.377, + "step": 11485 + }, + { + "epoch": 0.03, + "learning_rate": 9.976992632810584e-05, + "loss": 2.5089, + "step": 11490 + }, + { + "epoch": 0.03, + "learning_rate": 9.976972444279575e-05, + "loss": 2.3174, + "step": 11495 + }, + { + "epoch": 0.03, + "learning_rate": 9.976952246915366e-05, + "loss": 2.42, + "step": 11500 + }, + { + "epoch": 0.03, + "learning_rate": 9.976932040717993e-05, + "loss": 2.4653, + "step": 11505 + }, + { + "epoch": 0.03, + "learning_rate": 9.976911825687491e-05, + "loss": 2.4791, + "step": 11510 + }, + { + "epoch": 0.03, + "learning_rate": 9.976891601823899e-05, + "loss": 2.4856, + "step": 11515 + }, + { + "epoch": 0.03, + "learning_rate": 9.97687136912725e-05, + "loss": 2.4363, + "step": 11520 + }, + { + "epoch": 0.03, + "learning_rate": 9.976851127597579e-05, + "loss": 2.4069, + "step": 11525 + }, + { + "epoch": 0.03, + "learning_rate": 9.976830877234924e-05, + "loss": 2.3747, + "step": 11530 + }, + { + "epoch": 0.03, + "learning_rate": 9.976810618039321e-05, + "loss": 2.4318, + "step": 11535 + }, + { + "epoch": 0.03, + "learning_rate": 9.976790350010804e-05, + "loss": 2.5407, + "step": 11540 + }, + { + "epoch": 0.03, + "learning_rate": 9.97677007314941e-05, + "loss": 2.4144, + "step": 11545 + }, + { + "epoch": 0.03, + "learning_rate": 9.976749787455178e-05, + "loss": 2.4072, + "step": 11550 + }, + { + "epoch": 0.03, + "learning_rate": 9.976729492928139e-05, + "loss": 2.2635, + "step": 11555 + }, + { + "epoch": 0.03, + "learning_rate": 9.976709189568332e-05, + "loss": 2.5441, + "step": 11560 + }, + { + "epoch": 0.03, + "learning_rate": 9.976688877375791e-05, + "loss": 2.4058, + "step": 11565 + }, + { + "epoch": 0.03, + "learning_rate": 9.976668556350555e-05, + "loss": 2.4061, + "step": 11570 + }, + { + "epoch": 0.03, + "learning_rate": 9.976648226492658e-05, + "loss": 2.4027, + "step": 11575 + }, + { + "epoch": 0.03, + "learning_rate": 9.976627887802136e-05, + "loss": 2.4868, + "step": 11580 + }, + { + "epoch": 0.03, + "learning_rate": 9.976607540279027e-05, + "loss": 2.5428, + "step": 11585 + }, + { + "epoch": 0.03, + "learning_rate": 9.976587183923365e-05, + "loss": 2.3345, + "step": 11590 + }, + { + "epoch": 0.03, + "learning_rate": 9.976566818735186e-05, + "loss": 2.5059, + "step": 11595 + }, + { + "epoch": 0.03, + "learning_rate": 9.976546444714528e-05, + "loss": 2.4263, + "step": 11600 + }, + { + "epoch": 0.03, + "learning_rate": 9.976526061861426e-05, + "loss": 2.4167, + "step": 11605 + }, + { + "epoch": 0.03, + "learning_rate": 9.976505670175918e-05, + "loss": 2.4434, + "step": 11610 + }, + { + "epoch": 0.03, + "learning_rate": 9.976485269658035e-05, + "loss": 2.3844, + "step": 11615 + }, + { + "epoch": 0.03, + "learning_rate": 9.976464860307819e-05, + "loss": 2.4274, + "step": 11620 + }, + { + "epoch": 0.03, + "learning_rate": 9.976444442125303e-05, + "loss": 2.5138, + "step": 11625 + }, + { + "epoch": 0.03, + "learning_rate": 9.976424015110526e-05, + "loss": 2.4883, + "step": 11630 + }, + { + "epoch": 0.03, + "learning_rate": 9.97640357926352e-05, + "loss": 2.4556, + "step": 11635 + }, + { + "epoch": 0.03, + "learning_rate": 9.976383134584327e-05, + "loss": 2.4813, + "step": 11640 + }, + { + "epoch": 0.03, + "learning_rate": 9.976362681072977e-05, + "loss": 2.5362, + "step": 11645 + }, + { + "epoch": 0.03, + "learning_rate": 9.976342218729509e-05, + "loss": 2.51, + "step": 11650 + }, + { + "epoch": 0.03, + "learning_rate": 9.976321747553961e-05, + "loss": 2.4551, + "step": 11655 + }, + { + "epoch": 0.03, + "learning_rate": 9.976301267546368e-05, + "loss": 2.5745, + "step": 11660 + }, + { + "epoch": 0.03, + "learning_rate": 9.976280778706768e-05, + "loss": 2.3944, + "step": 11665 + }, + { + "epoch": 0.03, + "learning_rate": 9.976260281035193e-05, + "loss": 2.4361, + "step": 11670 + }, + { + "epoch": 0.03, + "learning_rate": 9.976239774531682e-05, + "loss": 2.4815, + "step": 11675 + }, + { + "epoch": 0.03, + "learning_rate": 9.976219259196273e-05, + "loss": 2.3611, + "step": 11680 + }, + { + "epoch": 0.03, + "learning_rate": 9.976198735029e-05, + "loss": 2.4707, + "step": 11685 + }, + { + "epoch": 0.03, + "learning_rate": 9.9761782020299e-05, + "loss": 2.4828, + "step": 11690 + }, + { + "epoch": 0.03, + "learning_rate": 9.976157660199011e-05, + "loss": 2.4327, + "step": 11695 + }, + { + "epoch": 0.03, + "learning_rate": 9.976137109536368e-05, + "loss": 2.3754, + "step": 11700 + }, + { + "epoch": 0.03, + "learning_rate": 9.976116550042007e-05, + "loss": 2.3877, + "step": 11705 + }, + { + "epoch": 0.03, + "learning_rate": 9.976095981715965e-05, + "loss": 2.4268, + "step": 11710 + }, + { + "epoch": 0.03, + "learning_rate": 9.976075404558278e-05, + "loss": 2.4275, + "step": 11715 + }, + { + "epoch": 0.03, + "learning_rate": 9.976054818568986e-05, + "loss": 2.4387, + "step": 11720 + }, + { + "epoch": 0.03, + "learning_rate": 9.97603422374812e-05, + "loss": 2.3572, + "step": 11725 + }, + { + "epoch": 0.03, + "learning_rate": 9.97601362009572e-05, + "loss": 2.4748, + "step": 11730 + }, + { + "epoch": 0.03, + "learning_rate": 9.975993007611822e-05, + "loss": 2.4144, + "step": 11735 + }, + { + "epoch": 0.03, + "learning_rate": 9.975972386296461e-05, + "loss": 2.4347, + "step": 11740 + }, + { + "epoch": 0.03, + "learning_rate": 9.975951756149677e-05, + "loss": 2.4868, + "step": 11745 + }, + { + "epoch": 0.03, + "learning_rate": 9.975931117171504e-05, + "loss": 2.4278, + "step": 11750 + }, + { + "epoch": 0.03, + "learning_rate": 9.97591046936198e-05, + "loss": 2.3423, + "step": 11755 + }, + { + "epoch": 0.03, + "learning_rate": 9.97588981272114e-05, + "loss": 2.4602, + "step": 11760 + }, + { + "epoch": 0.03, + "learning_rate": 9.975869147249022e-05, + "loss": 2.5092, + "step": 11765 + }, + { + "epoch": 0.03, + "learning_rate": 9.975848472945661e-05, + "loss": 2.4346, + "step": 11770 + }, + { + "epoch": 0.03, + "learning_rate": 9.975827789811097e-05, + "loss": 2.49, + "step": 11775 + }, + { + "epoch": 0.03, + "learning_rate": 9.975807097845363e-05, + "loss": 2.4454, + "step": 11780 + }, + { + "epoch": 0.03, + "learning_rate": 9.975786397048499e-05, + "loss": 2.395, + "step": 11785 + }, + { + "epoch": 0.03, + "learning_rate": 9.975765687420538e-05, + "loss": 2.4819, + "step": 11790 + }, + { + "epoch": 0.03, + "learning_rate": 9.97574496896152e-05, + "loss": 2.4167, + "step": 11795 + }, + { + "epoch": 0.03, + "learning_rate": 9.975724241671481e-05, + "loss": 2.469, + "step": 11800 + }, + { + "epoch": 0.03, + "learning_rate": 9.975703505550456e-05, + "loss": 2.324, + "step": 11805 + }, + { + "epoch": 0.03, + "learning_rate": 9.975682760598485e-05, + "loss": 2.4607, + "step": 11810 + }, + { + "epoch": 0.03, + "learning_rate": 9.975662006815603e-05, + "loss": 2.546, + "step": 11815 + }, + { + "epoch": 0.03, + "learning_rate": 9.975641244201846e-05, + "loss": 2.5182, + "step": 11820 + }, + { + "epoch": 0.03, + "learning_rate": 9.975620472757253e-05, + "loss": 2.5408, + "step": 11825 + }, + { + "epoch": 0.03, + "learning_rate": 9.975599692481856e-05, + "loss": 2.5695, + "step": 11830 + }, + { + "epoch": 0.03, + "learning_rate": 9.9755789033757e-05, + "loss": 2.3097, + "step": 11835 + }, + { + "epoch": 0.03, + "learning_rate": 9.975558105438814e-05, + "loss": 2.5618, + "step": 11840 + }, + { + "epoch": 0.03, + "learning_rate": 9.975537298671241e-05, + "loss": 2.4257, + "step": 11845 + }, + { + "epoch": 0.03, + "learning_rate": 9.975516483073013e-05, + "loss": 2.4918, + "step": 11850 + }, + { + "epoch": 0.03, + "learning_rate": 9.97549565864417e-05, + "loss": 2.4966, + "step": 11855 + }, + { + "epoch": 0.03, + "learning_rate": 9.975474825384748e-05, + "loss": 2.4043, + "step": 11860 + }, + { + "epoch": 0.03, + "learning_rate": 9.975453983294784e-05, + "loss": 2.3602, + "step": 11865 + }, + { + "epoch": 0.03, + "learning_rate": 9.975433132374316e-05, + "loss": 2.5422, + "step": 11870 + }, + { + "epoch": 0.03, + "learning_rate": 9.975412272623379e-05, + "loss": 2.3511, + "step": 11875 + }, + { + "epoch": 0.03, + "learning_rate": 9.975391404042011e-05, + "loss": 2.4486, + "step": 11880 + }, + { + "epoch": 0.03, + "learning_rate": 9.97537052663025e-05, + "loss": 2.3194, + "step": 11885 + }, + { + "epoch": 0.03, + "learning_rate": 9.975349640388132e-05, + "loss": 2.4319, + "step": 11890 + }, + { + "epoch": 0.03, + "learning_rate": 9.975328745315695e-05, + "loss": 2.4517, + "step": 11895 + }, + { + "epoch": 0.03, + "learning_rate": 9.975307841412974e-05, + "loss": 2.3556, + "step": 11900 + }, + { + "epoch": 0.03, + "learning_rate": 9.975286928680007e-05, + "loss": 2.4942, + "step": 11905 + }, + { + "epoch": 0.03, + "learning_rate": 9.975266007116834e-05, + "loss": 2.4753, + "step": 11910 + }, + { + "epoch": 0.03, + "learning_rate": 9.975245076723488e-05, + "loss": 2.4234, + "step": 11915 + }, + { + "epoch": 0.03, + "learning_rate": 9.975224137500008e-05, + "loss": 2.4043, + "step": 11920 + }, + { + "epoch": 0.03, + "learning_rate": 9.975203189446431e-05, + "loss": 2.4467, + "step": 11925 + }, + { + "epoch": 0.03, + "learning_rate": 9.975182232562793e-05, + "loss": 2.413, + "step": 11930 + }, + { + "epoch": 0.03, + "learning_rate": 9.975161266849135e-05, + "loss": 2.3998, + "step": 11935 + }, + { + "epoch": 0.03, + "learning_rate": 9.975140292305492e-05, + "loss": 2.3542, + "step": 11940 + }, + { + "epoch": 0.03, + "learning_rate": 9.975119308931898e-05, + "loss": 2.3328, + "step": 11945 + }, + { + "epoch": 0.03, + "learning_rate": 9.975098316728394e-05, + "loss": 2.4392, + "step": 11950 + }, + { + "epoch": 0.03, + "learning_rate": 9.975077315695016e-05, + "loss": 2.373, + "step": 11955 + }, + { + "epoch": 0.03, + "learning_rate": 9.975056305831803e-05, + "loss": 2.4873, + "step": 11960 + }, + { + "epoch": 0.03, + "learning_rate": 9.97503528713879e-05, + "loss": 2.3784, + "step": 11965 + }, + { + "epoch": 0.03, + "learning_rate": 9.975014259616015e-05, + "loss": 2.3521, + "step": 11970 + }, + { + "epoch": 0.03, + "learning_rate": 9.974993223263516e-05, + "loss": 2.4584, + "step": 11975 + }, + { + "epoch": 0.03, + "learning_rate": 9.97497217808133e-05, + "loss": 2.4563, + "step": 11980 + }, + { + "epoch": 0.03, + "learning_rate": 9.974951124069494e-05, + "loss": 2.3378, + "step": 11985 + }, + { + "epoch": 0.03, + "learning_rate": 9.974930061228045e-05, + "loss": 2.4841, + "step": 11990 + }, + { + "epoch": 0.03, + "learning_rate": 9.974908989557022e-05, + "loss": 2.4129, + "step": 11995 + }, + { + "epoch": 0.03, + "learning_rate": 9.97488790905646e-05, + "loss": 2.4169, + "step": 12000 + }, + { + "epoch": 0.03, + "learning_rate": 9.974866819726398e-05, + "loss": 2.4435, + "step": 12005 + }, + { + "epoch": 0.03, + "learning_rate": 9.974845721566875e-05, + "loss": 2.4721, + "step": 12010 + }, + { + "epoch": 0.03, + "learning_rate": 9.974824614577925e-05, + "loss": 2.5498, + "step": 12015 + }, + { + "epoch": 0.03, + "learning_rate": 9.974803498759588e-05, + "loss": 2.3655, + "step": 12020 + }, + { + "epoch": 0.03, + "learning_rate": 9.9747823741119e-05, + "loss": 2.5594, + "step": 12025 + }, + { + "epoch": 0.03, + "learning_rate": 9.974761240634897e-05, + "loss": 2.5672, + "step": 12030 + }, + { + "epoch": 0.03, + "learning_rate": 9.974740098328621e-05, + "loss": 2.528, + "step": 12035 + }, + { + "epoch": 0.03, + "learning_rate": 9.974718947193107e-05, + "loss": 2.4832, + "step": 12040 + }, + { + "epoch": 0.03, + "learning_rate": 9.974697787228394e-05, + "loss": 2.4477, + "step": 12045 + }, + { + "epoch": 0.03, + "learning_rate": 9.974676618434515e-05, + "loss": 2.4431, + "step": 12050 + }, + { + "epoch": 0.03, + "learning_rate": 9.974655440811514e-05, + "loss": 2.4149, + "step": 12055 + }, + { + "epoch": 0.03, + "learning_rate": 9.974634254359423e-05, + "loss": 2.4189, + "step": 12060 + }, + { + "epoch": 0.03, + "learning_rate": 9.974613059078282e-05, + "loss": 2.501, + "step": 12065 + }, + { + "epoch": 0.03, + "learning_rate": 9.974591854968129e-05, + "loss": 2.4434, + "step": 12070 + }, + { + "epoch": 0.03, + "learning_rate": 9.974570642029002e-05, + "loss": 2.4826, + "step": 12075 + }, + { + "epoch": 0.03, + "learning_rate": 9.974549420260937e-05, + "loss": 2.3639, + "step": 12080 + }, + { + "epoch": 0.03, + "learning_rate": 9.974528189663971e-05, + "loss": 2.4374, + "step": 12085 + }, + { + "epoch": 0.03, + "learning_rate": 9.974506950238147e-05, + "loss": 2.4266, + "step": 12090 + }, + { + "epoch": 0.03, + "learning_rate": 9.974485701983495e-05, + "loss": 2.5574, + "step": 12095 + }, + { + "epoch": 0.03, + "learning_rate": 9.97446444490006e-05, + "loss": 2.4736, + "step": 12100 + }, + { + "epoch": 0.03, + "learning_rate": 9.974443178987875e-05, + "loss": 2.2792, + "step": 12105 + }, + { + "epoch": 0.03, + "learning_rate": 9.974421904246979e-05, + "loss": 2.442, + "step": 12110 + }, + { + "epoch": 0.03, + "learning_rate": 9.97440062067741e-05, + "loss": 2.4352, + "step": 12115 + }, + { + "epoch": 0.03, + "learning_rate": 9.974379328279205e-05, + "loss": 2.4494, + "step": 12120 + }, + { + "epoch": 0.03, + "learning_rate": 9.974358027052401e-05, + "loss": 2.5174, + "step": 12125 + }, + { + "epoch": 0.03, + "learning_rate": 9.974336716997041e-05, + "loss": 2.3242, + "step": 12130 + }, + { + "epoch": 0.03, + "learning_rate": 9.974315398113157e-05, + "loss": 2.2626, + "step": 12135 + }, + { + "epoch": 0.03, + "learning_rate": 9.974294070400788e-05, + "loss": 2.3068, + "step": 12140 + }, + { + "epoch": 0.03, + "learning_rate": 9.974272733859974e-05, + "loss": 2.3979, + "step": 12145 + }, + { + "epoch": 0.03, + "learning_rate": 9.974251388490751e-05, + "loss": 2.4697, + "step": 12150 + }, + { + "epoch": 0.03, + "learning_rate": 9.974230034293158e-05, + "loss": 2.3522, + "step": 12155 + }, + { + "epoch": 0.03, + "learning_rate": 9.974208671267232e-05, + "loss": 2.3906, + "step": 12160 + }, + { + "epoch": 0.03, + "learning_rate": 9.974187299413012e-05, + "loss": 2.3656, + "step": 12165 + }, + { + "epoch": 0.03, + "learning_rate": 9.974165918730533e-05, + "loss": 2.3918, + "step": 12170 + }, + { + "epoch": 0.03, + "learning_rate": 9.974144529219837e-05, + "loss": 2.477, + "step": 12175 + }, + { + "epoch": 0.03, + "learning_rate": 9.97412313088096e-05, + "loss": 2.4862, + "step": 12180 + }, + { + "epoch": 0.03, + "learning_rate": 9.97410172371394e-05, + "loss": 2.4742, + "step": 12185 + }, + { + "epoch": 0.03, + "learning_rate": 9.974080307718814e-05, + "loss": 2.5369, + "step": 12190 + }, + { + "epoch": 0.03, + "learning_rate": 9.974058882895623e-05, + "loss": 2.4225, + "step": 12195 + }, + { + "epoch": 0.03, + "learning_rate": 9.974037449244402e-05, + "loss": 2.427, + "step": 12200 + }, + { + "epoch": 0.03, + "learning_rate": 9.97401600676519e-05, + "loss": 2.525, + "step": 12205 + }, + { + "epoch": 0.03, + "learning_rate": 9.973994555458026e-05, + "loss": 2.5258, + "step": 12210 + }, + { + "epoch": 0.03, + "learning_rate": 9.973973095322946e-05, + "loss": 2.4241, + "step": 12215 + }, + { + "epoch": 0.03, + "learning_rate": 9.97395162635999e-05, + "loss": 2.2398, + "step": 12220 + }, + { + "epoch": 0.03, + "learning_rate": 9.973930148569198e-05, + "loss": 2.4354, + "step": 12225 + }, + { + "epoch": 0.03, + "learning_rate": 9.973908661950602e-05, + "loss": 2.5834, + "step": 12230 + }, + { + "epoch": 0.03, + "learning_rate": 9.973887166504246e-05, + "loss": 2.4993, + "step": 12235 + }, + { + "epoch": 0.03, + "learning_rate": 9.973865662230164e-05, + "loss": 2.5146, + "step": 12240 + }, + { + "epoch": 0.03, + "learning_rate": 9.973844149128397e-05, + "loss": 2.3881, + "step": 12245 + }, + { + "epoch": 0.03, + "learning_rate": 9.973822627198982e-05, + "loss": 2.457, + "step": 12250 + }, + { + "epoch": 0.03, + "learning_rate": 9.973801096441958e-05, + "loss": 2.4899, + "step": 12255 + }, + { + "epoch": 0.03, + "learning_rate": 9.973779556857362e-05, + "loss": 2.3026, + "step": 12260 + }, + { + "epoch": 0.03, + "learning_rate": 9.973758008445233e-05, + "loss": 2.4724, + "step": 12265 + }, + { + "epoch": 0.03, + "learning_rate": 9.97373645120561e-05, + "loss": 2.3537, + "step": 12270 + }, + { + "epoch": 0.03, + "learning_rate": 9.97371488513853e-05, + "loss": 2.4278, + "step": 12275 + }, + { + "epoch": 0.03, + "learning_rate": 9.973693310244032e-05, + "loss": 2.4481, + "step": 12280 + }, + { + "epoch": 0.03, + "learning_rate": 9.973671726522154e-05, + "loss": 2.4334, + "step": 12285 + }, + { + "epoch": 0.03, + "learning_rate": 9.973650133972933e-05, + "loss": 2.4553, + "step": 12290 + }, + { + "epoch": 0.03, + "learning_rate": 9.973628532596409e-05, + "loss": 2.3771, + "step": 12295 + }, + { + "epoch": 0.03, + "learning_rate": 9.97360692239262e-05, + "loss": 2.42, + "step": 12300 + }, + { + "epoch": 0.03, + "learning_rate": 9.973585303361604e-05, + "loss": 2.3764, + "step": 12305 + }, + { + "epoch": 0.03, + "learning_rate": 9.973563675503401e-05, + "loss": 2.4719, + "step": 12310 + }, + { + "epoch": 0.03, + "learning_rate": 9.973542038818046e-05, + "loss": 2.4091, + "step": 12315 + }, + { + "epoch": 0.03, + "learning_rate": 9.97352039330558e-05, + "loss": 2.4757, + "step": 12320 + }, + { + "epoch": 0.03, + "learning_rate": 9.973498738966042e-05, + "loss": 2.3922, + "step": 12325 + }, + { + "epoch": 0.03, + "learning_rate": 9.973477075799469e-05, + "loss": 2.2517, + "step": 12330 + }, + { + "epoch": 0.03, + "learning_rate": 9.9734554038059e-05, + "loss": 2.4951, + "step": 12335 + }, + { + "epoch": 0.03, + "learning_rate": 9.973433722985372e-05, + "loss": 2.2686, + "step": 12340 + }, + { + "epoch": 0.03, + "learning_rate": 9.973412033337925e-05, + "loss": 2.4197, + "step": 12345 + }, + { + "epoch": 0.03, + "learning_rate": 9.973390334863598e-05, + "loss": 2.4085, + "step": 12350 + }, + { + "epoch": 0.03, + "learning_rate": 9.973368627562429e-05, + "loss": 2.3885, + "step": 12355 + }, + { + "epoch": 0.03, + "learning_rate": 9.973346911434454e-05, + "loss": 2.4082, + "step": 12360 + }, + { + "epoch": 0.03, + "learning_rate": 9.973325186479715e-05, + "loss": 2.4784, + "step": 12365 + }, + { + "epoch": 0.03, + "learning_rate": 9.97330345269825e-05, + "loss": 2.5146, + "step": 12370 + }, + { + "epoch": 0.03, + "learning_rate": 9.973281710090096e-05, + "loss": 2.4444, + "step": 12375 + }, + { + "epoch": 0.03, + "learning_rate": 9.973259958655293e-05, + "loss": 2.4766, + "step": 12380 + }, + { + "epoch": 0.03, + "learning_rate": 9.97323819839388e-05, + "loss": 2.3983, + "step": 12385 + }, + { + "epoch": 0.03, + "learning_rate": 9.973216429305892e-05, + "loss": 2.3604, + "step": 12390 + }, + { + "epoch": 0.03, + "learning_rate": 9.973194651391373e-05, + "loss": 2.3738, + "step": 12395 + }, + { + "epoch": 0.03, + "learning_rate": 9.973172864650358e-05, + "loss": 2.4197, + "step": 12400 + }, + { + "epoch": 0.03, + "learning_rate": 9.973151069082886e-05, + "loss": 2.4443, + "step": 12405 + }, + { + "epoch": 0.03, + "learning_rate": 9.973129264688997e-05, + "loss": 2.3848, + "step": 12410 + }, + { + "epoch": 0.03, + "learning_rate": 9.973107451468728e-05, + "loss": 2.3454, + "step": 12415 + }, + { + "epoch": 0.03, + "learning_rate": 9.97308562942212e-05, + "loss": 2.3429, + "step": 12420 + }, + { + "epoch": 0.03, + "learning_rate": 9.97306379854921e-05, + "loss": 2.42, + "step": 12425 + }, + { + "epoch": 0.03, + "learning_rate": 9.973041958850037e-05, + "loss": 2.3959, + "step": 12430 + }, + { + "epoch": 0.03, + "learning_rate": 9.97302011032464e-05, + "loss": 2.2707, + "step": 12435 + }, + { + "epoch": 0.03, + "learning_rate": 9.972998252973058e-05, + "loss": 2.4679, + "step": 12440 + }, + { + "epoch": 0.03, + "learning_rate": 9.97297638679533e-05, + "loss": 2.5144, + "step": 12445 + }, + { + "epoch": 0.03, + "learning_rate": 9.972954511791494e-05, + "loss": 2.4359, + "step": 12450 + }, + { + "epoch": 0.03, + "learning_rate": 9.972932627961588e-05, + "loss": 2.4681, + "step": 12455 + }, + { + "epoch": 0.03, + "learning_rate": 9.972910735305655e-05, + "loss": 2.4048, + "step": 12460 + }, + { + "epoch": 0.03, + "learning_rate": 9.972888833823728e-05, + "loss": 2.3055, + "step": 12465 + }, + { + "epoch": 0.03, + "learning_rate": 9.972866923515849e-05, + "loss": 2.4518, + "step": 12470 + }, + { + "epoch": 0.03, + "learning_rate": 9.972845004382059e-05, + "loss": 2.4344, + "step": 12475 + }, + { + "epoch": 0.03, + "learning_rate": 9.972823076422392e-05, + "loss": 2.4246, + "step": 12480 + }, + { + "epoch": 0.03, + "learning_rate": 9.97280113963689e-05, + "loss": 2.4404, + "step": 12485 + }, + { + "epoch": 0.03, + "learning_rate": 9.972779194025592e-05, + "loss": 2.4495, + "step": 12490 + }, + { + "epoch": 0.03, + "learning_rate": 9.972757239588537e-05, + "loss": 2.5006, + "step": 12495 + }, + { + "epoch": 0.03, + "learning_rate": 9.972735276325763e-05, + "loss": 2.4964, + "step": 12500 + }, + { + "epoch": 0.03, + "learning_rate": 9.972713304237306e-05, + "loss": 2.242, + "step": 12505 + }, + { + "epoch": 0.03, + "learning_rate": 9.972691323323212e-05, + "loss": 2.338, + "step": 12510 + }, + { + "epoch": 0.03, + "learning_rate": 9.972669333583516e-05, + "loss": 2.4505, + "step": 12515 + }, + { + "epoch": 0.03, + "learning_rate": 9.972647335018256e-05, + "loss": 2.4537, + "step": 12520 + }, + { + "epoch": 0.03, + "learning_rate": 9.972625327627474e-05, + "loss": 2.4458, + "step": 12525 + }, + { + "epoch": 0.03, + "learning_rate": 9.972603311411207e-05, + "loss": 2.4604, + "step": 12530 + }, + { + "epoch": 0.03, + "learning_rate": 9.972581286369492e-05, + "loss": 2.5101, + "step": 12535 + }, + { + "epoch": 0.03, + "learning_rate": 9.972559252502374e-05, + "loss": 2.4877, + "step": 12540 + }, + { + "epoch": 0.03, + "learning_rate": 9.972537209809887e-05, + "loss": 2.3746, + "step": 12545 + }, + { + "epoch": 0.03, + "learning_rate": 9.97251515829207e-05, + "loss": 2.5392, + "step": 12550 + }, + { + "epoch": 0.03, + "learning_rate": 9.972493097948968e-05, + "loss": 2.5175, + "step": 12555 + }, + { + "epoch": 0.03, + "learning_rate": 9.972471028780613e-05, + "loss": 2.4486, + "step": 12560 + }, + { + "epoch": 0.03, + "learning_rate": 9.972448950787047e-05, + "loss": 2.4229, + "step": 12565 + }, + { + "epoch": 0.03, + "learning_rate": 9.972426863968311e-05, + "loss": 2.3956, + "step": 12570 + }, + { + "epoch": 0.03, + "learning_rate": 9.972404768324443e-05, + "loss": 2.4448, + "step": 12575 + }, + { + "epoch": 0.03, + "learning_rate": 9.972382663855482e-05, + "loss": 2.3778, + "step": 12580 + }, + { + "epoch": 0.03, + "learning_rate": 9.972360550561466e-05, + "loss": 2.4846, + "step": 12585 + }, + { + "epoch": 0.03, + "learning_rate": 9.972338428442436e-05, + "loss": 2.4764, + "step": 12590 + }, + { + "epoch": 0.03, + "learning_rate": 9.97231629749843e-05, + "loss": 2.4327, + "step": 12595 + }, + { + "epoch": 0.03, + "learning_rate": 9.97229415772949e-05, + "loss": 2.4999, + "step": 12600 + }, + { + "epoch": 0.03, + "learning_rate": 9.97227200913565e-05, + "loss": 2.4639, + "step": 12605 + }, + { + "epoch": 0.03, + "learning_rate": 9.972249851716955e-05, + "loss": 2.4519, + "step": 12610 + }, + { + "epoch": 0.03, + "learning_rate": 9.972227685473441e-05, + "loss": 2.5954, + "step": 12615 + }, + { + "epoch": 0.03, + "learning_rate": 9.972205510405147e-05, + "loss": 2.4321, + "step": 12620 + }, + { + "epoch": 0.03, + "learning_rate": 9.972183326512114e-05, + "loss": 2.3753, + "step": 12625 + }, + { + "epoch": 0.03, + "learning_rate": 9.972161133794383e-05, + "loss": 2.384, + "step": 12630 + }, + { + "epoch": 0.03, + "learning_rate": 9.972138932251989e-05, + "loss": 2.3985, + "step": 12635 + }, + { + "epoch": 0.03, + "learning_rate": 9.972116721884975e-05, + "loss": 2.4723, + "step": 12640 + }, + { + "epoch": 0.03, + "learning_rate": 9.972094502693377e-05, + "loss": 2.4175, + "step": 12645 + }, + { + "epoch": 0.03, + "learning_rate": 9.972072274677238e-05, + "loss": 2.3395, + "step": 12650 + }, + { + "epoch": 0.03, + "learning_rate": 9.972050037836597e-05, + "loss": 2.4498, + "step": 12655 + }, + { + "epoch": 0.03, + "learning_rate": 9.972027792171491e-05, + "loss": 2.5632, + "step": 12660 + }, + { + "epoch": 0.03, + "learning_rate": 9.972005537681961e-05, + "loss": 2.5249, + "step": 12665 + }, + { + "epoch": 0.03, + "learning_rate": 9.971983274368048e-05, + "loss": 2.3206, + "step": 12670 + }, + { + "epoch": 0.03, + "learning_rate": 9.971961002229788e-05, + "loss": 2.4224, + "step": 12675 + }, + { + "epoch": 0.03, + "learning_rate": 9.971938721267223e-05, + "loss": 2.378, + "step": 12680 + }, + { + "epoch": 0.03, + "learning_rate": 9.971916431480392e-05, + "loss": 2.4081, + "step": 12685 + }, + { + "epoch": 0.03, + "learning_rate": 9.971894132869335e-05, + "loss": 2.4658, + "step": 12690 + }, + { + "epoch": 0.03, + "learning_rate": 9.971871825434091e-05, + "loss": 2.5203, + "step": 12695 + }, + { + "epoch": 0.03, + "learning_rate": 9.9718495091747e-05, + "loss": 2.4901, + "step": 12700 + }, + { + "epoch": 0.03, + "learning_rate": 9.971827184091201e-05, + "loss": 2.4353, + "step": 12705 + }, + { + "epoch": 0.03, + "learning_rate": 9.971804850183635e-05, + "loss": 2.4505, + "step": 12710 + }, + { + "epoch": 0.03, + "learning_rate": 9.971782507452039e-05, + "loss": 2.3637, + "step": 12715 + }, + { + "epoch": 0.03, + "learning_rate": 9.971760155896454e-05, + "loss": 2.4072, + "step": 12720 + }, + { + "epoch": 0.03, + "learning_rate": 9.971737795516922e-05, + "loss": 2.4605, + "step": 12725 + }, + { + "epoch": 0.03, + "learning_rate": 9.97171542631348e-05, + "loss": 2.4422, + "step": 12730 + }, + { + "epoch": 0.03, + "learning_rate": 9.971693048286168e-05, + "loss": 2.3385, + "step": 12735 + }, + { + "epoch": 0.03, + "learning_rate": 9.971670661435025e-05, + "loss": 2.2586, + "step": 12740 + }, + { + "epoch": 0.03, + "learning_rate": 9.971648265760093e-05, + "loss": 2.472, + "step": 12745 + }, + { + "epoch": 0.03, + "learning_rate": 9.971625861261412e-05, + "loss": 2.3358, + "step": 12750 + }, + { + "epoch": 0.03, + "learning_rate": 9.971603447939017e-05, + "loss": 2.4538, + "step": 12755 + }, + { + "epoch": 0.03, + "learning_rate": 9.971581025792953e-05, + "loss": 2.4598, + "step": 12760 + }, + { + "epoch": 0.03, + "learning_rate": 9.971558594823256e-05, + "loss": 2.5641, + "step": 12765 + }, + { + "epoch": 0.03, + "learning_rate": 9.97153615502997e-05, + "loss": 2.4165, + "step": 12770 + }, + { + "epoch": 0.03, + "learning_rate": 9.971513706413132e-05, + "loss": 2.3998, + "step": 12775 + }, + { + "epoch": 0.03, + "learning_rate": 9.971491248972782e-05, + "loss": 2.3522, + "step": 12780 + }, + { + "epoch": 0.03, + "learning_rate": 9.97146878270896e-05, + "loss": 2.5132, + "step": 12785 + }, + { + "epoch": 0.03, + "learning_rate": 9.971446307621706e-05, + "loss": 2.3417, + "step": 12790 + }, + { + "epoch": 0.03, + "learning_rate": 9.97142382371106e-05, + "loss": 2.4331, + "step": 12795 + }, + { + "epoch": 0.03, + "learning_rate": 9.971401330977063e-05, + "loss": 2.3398, + "step": 12800 + }, + { + "epoch": 0.03, + "learning_rate": 9.971378829419754e-05, + "loss": 2.4329, + "step": 12805 + }, + { + "epoch": 0.03, + "learning_rate": 9.97135631903917e-05, + "loss": 2.4217, + "step": 12810 + }, + { + "epoch": 0.03, + "learning_rate": 9.971333799835356e-05, + "loss": 2.4449, + "step": 12815 + }, + { + "epoch": 0.03, + "learning_rate": 9.971311271808348e-05, + "loss": 2.4041, + "step": 12820 + }, + { + "epoch": 0.03, + "learning_rate": 9.971288734958188e-05, + "loss": 2.4957, + "step": 12825 + }, + { + "epoch": 0.03, + "learning_rate": 9.971266189284917e-05, + "loss": 2.3872, + "step": 12830 + }, + { + "epoch": 0.03, + "learning_rate": 9.971243634788573e-05, + "loss": 2.4896, + "step": 12835 + }, + { + "epoch": 0.03, + "learning_rate": 9.971221071469196e-05, + "loss": 2.4578, + "step": 12840 + }, + { + "epoch": 0.03, + "learning_rate": 9.971198499326829e-05, + "loss": 2.4635, + "step": 12845 + }, + { + "epoch": 0.03, + "learning_rate": 9.971175918361506e-05, + "loss": 2.4472, + "step": 12850 + }, + { + "epoch": 0.03, + "learning_rate": 9.971153328573273e-05, + "loss": 2.3751, + "step": 12855 + }, + { + "epoch": 0.03, + "learning_rate": 9.971130729962168e-05, + "loss": 2.3525, + "step": 12860 + }, + { + "epoch": 0.03, + "learning_rate": 9.971108122528233e-05, + "loss": 2.4764, + "step": 12865 + }, + { + "epoch": 0.03, + "learning_rate": 9.971085506271503e-05, + "loss": 2.3957, + "step": 12870 + }, + { + "epoch": 0.03, + "learning_rate": 9.971062881192021e-05, + "loss": 2.4581, + "step": 12875 + }, + { + "epoch": 0.03, + "learning_rate": 9.971040247289829e-05, + "loss": 2.4716, + "step": 12880 + }, + { + "epoch": 0.03, + "learning_rate": 9.971017604564966e-05, + "loss": 2.3754, + "step": 12885 + }, + { + "epoch": 0.03, + "learning_rate": 9.970994953017472e-05, + "loss": 2.3966, + "step": 12890 + }, + { + "epoch": 0.03, + "learning_rate": 9.970972292647386e-05, + "loss": 2.3474, + "step": 12895 + }, + { + "epoch": 0.03, + "learning_rate": 9.970949623454749e-05, + "loss": 2.5362, + "step": 12900 + }, + { + "epoch": 0.03, + "learning_rate": 9.970926945439603e-05, + "loss": 2.4547, + "step": 12905 + }, + { + "epoch": 0.03, + "learning_rate": 9.970904258601985e-05, + "loss": 2.4591, + "step": 12910 + }, + { + "epoch": 0.03, + "learning_rate": 9.970881562941938e-05, + "loss": 2.4762, + "step": 12915 + }, + { + "epoch": 0.03, + "learning_rate": 9.970858858459502e-05, + "loss": 2.4571, + "step": 12920 + }, + { + "epoch": 0.03, + "learning_rate": 9.970836145154715e-05, + "loss": 2.3062, + "step": 12925 + }, + { + "epoch": 0.03, + "learning_rate": 9.97081342302762e-05, + "loss": 2.4554, + "step": 12930 + }, + { + "epoch": 0.03, + "learning_rate": 9.970790692078256e-05, + "loss": 2.4254, + "step": 12935 + }, + { + "epoch": 0.03, + "learning_rate": 9.970767952306663e-05, + "loss": 2.363, + "step": 12940 + }, + { + "epoch": 0.03, + "learning_rate": 9.970745203712882e-05, + "loss": 2.4069, + "step": 12945 + }, + { + "epoch": 0.03, + "learning_rate": 9.970722446296955e-05, + "loss": 2.5299, + "step": 12950 + }, + { + "epoch": 0.03, + "learning_rate": 9.970699680058918e-05, + "loss": 2.3369, + "step": 12955 + }, + { + "epoch": 0.03, + "learning_rate": 9.970676904998814e-05, + "loss": 2.5304, + "step": 12960 + }, + { + "epoch": 0.03, + "learning_rate": 9.970654121116688e-05, + "loss": 2.3946, + "step": 12965 + }, + { + "epoch": 0.03, + "learning_rate": 9.970631328412572e-05, + "loss": 2.4243, + "step": 12970 + }, + { + "epoch": 0.03, + "learning_rate": 9.97060852688651e-05, + "loss": 2.2856, + "step": 12975 + }, + { + "epoch": 0.03, + "learning_rate": 9.970585716538545e-05, + "loss": 2.4607, + "step": 12980 + }, + { + "epoch": 0.03, + "learning_rate": 9.970562897368713e-05, + "loss": 2.225, + "step": 12985 + }, + { + "epoch": 0.03, + "learning_rate": 9.970540069377058e-05, + "loss": 2.4779, + "step": 12990 + }, + { + "epoch": 0.03, + "learning_rate": 9.97051723256362e-05, + "loss": 2.3667, + "step": 12995 + }, + { + "epoch": 0.03, + "learning_rate": 9.97049438692844e-05, + "loss": 2.3231, + "step": 13000 + }, + { + "epoch": 0.03, + "learning_rate": 9.970471532471555e-05, + "loss": 2.4219, + "step": 13005 + }, + { + "epoch": 0.03, + "learning_rate": 9.970448669193009e-05, + "loss": 2.2448, + "step": 13010 + }, + { + "epoch": 0.03, + "learning_rate": 9.970425797092841e-05, + "loss": 2.5069, + "step": 13015 + }, + { + "epoch": 0.03, + "learning_rate": 9.970402916171093e-05, + "loss": 2.5212, + "step": 13020 + }, + { + "epoch": 0.03, + "learning_rate": 9.970380026427805e-05, + "loss": 2.4804, + "step": 13025 + }, + { + "epoch": 0.03, + "learning_rate": 9.970357127863017e-05, + "loss": 2.4149, + "step": 13030 + }, + { + "epoch": 0.03, + "learning_rate": 9.970334220476771e-05, + "loss": 2.437, + "step": 13035 + }, + { + "epoch": 0.03, + "learning_rate": 9.970311304269106e-05, + "loss": 2.4438, + "step": 13040 + }, + { + "epoch": 0.03, + "learning_rate": 9.970288379240062e-05, + "loss": 2.3858, + "step": 13045 + }, + { + "epoch": 0.03, + "learning_rate": 9.970265445389683e-05, + "loss": 2.4477, + "step": 13050 + }, + { + "epoch": 0.04, + "learning_rate": 9.970242502718007e-05, + "loss": 2.4482, + "step": 13055 + }, + { + "epoch": 0.04, + "learning_rate": 9.970219551225075e-05, + "loss": 2.435, + "step": 13060 + }, + { + "epoch": 0.04, + "learning_rate": 9.97019659091093e-05, + "loss": 2.3391, + "step": 13065 + }, + { + "epoch": 0.04, + "learning_rate": 9.97017362177561e-05, + "loss": 2.4659, + "step": 13070 + }, + { + "epoch": 0.04, + "learning_rate": 9.970150643819157e-05, + "loss": 2.3778, + "step": 13075 + }, + { + "epoch": 0.04, + "learning_rate": 9.970127657041609e-05, + "loss": 2.4515, + "step": 13080 + }, + { + "epoch": 0.04, + "learning_rate": 9.970104661443012e-05, + "loss": 2.4047, + "step": 13085 + }, + { + "epoch": 0.04, + "learning_rate": 9.970081657023402e-05, + "loss": 2.3692, + "step": 13090 + }, + { + "epoch": 0.04, + "learning_rate": 9.970058643782824e-05, + "loss": 2.4037, + "step": 13095 + }, + { + "epoch": 0.04, + "learning_rate": 9.970035621721317e-05, + "loss": 2.5833, + "step": 13100 + }, + { + "epoch": 0.04, + "learning_rate": 9.970012590838921e-05, + "loss": 2.5252, + "step": 13105 + }, + { + "epoch": 0.04, + "learning_rate": 9.969989551135677e-05, + "loss": 2.4064, + "step": 13110 + }, + { + "epoch": 0.04, + "learning_rate": 9.969966502611627e-05, + "loss": 2.4749, + "step": 13115 + }, + { + "epoch": 0.04, + "learning_rate": 9.969943445266812e-05, + "loss": 2.4303, + "step": 13120 + }, + { + "epoch": 0.04, + "learning_rate": 9.96992037910127e-05, + "loss": 2.3617, + "step": 13125 + }, + { + "epoch": 0.04, + "learning_rate": 9.969897304115045e-05, + "loss": 2.4673, + "step": 13130 + }, + { + "epoch": 0.04, + "learning_rate": 9.96987422030818e-05, + "loss": 2.3399, + "step": 13135 + }, + { + "epoch": 0.04, + "learning_rate": 9.96985112768071e-05, + "loss": 2.5453, + "step": 13140 + }, + { + "epoch": 0.04, + "learning_rate": 9.969828026232681e-05, + "loss": 2.4891, + "step": 13145 + }, + { + "epoch": 0.04, + "learning_rate": 9.96980491596413e-05, + "loss": 2.3006, + "step": 13150 + }, + { + "epoch": 0.04, + "learning_rate": 9.969781796875102e-05, + "loss": 2.4902, + "step": 13155 + }, + { + "epoch": 0.04, + "learning_rate": 9.969758668965635e-05, + "loss": 2.4024, + "step": 13160 + }, + { + "epoch": 0.04, + "learning_rate": 9.969735532235772e-05, + "loss": 2.402, + "step": 13165 + }, + { + "epoch": 0.04, + "learning_rate": 9.969712386685552e-05, + "loss": 2.3572, + "step": 13170 + }, + { + "epoch": 0.04, + "learning_rate": 9.969689232315019e-05, + "loss": 2.3012, + "step": 13175 + }, + { + "epoch": 0.04, + "learning_rate": 9.969666069124213e-05, + "loss": 2.4131, + "step": 13180 + }, + { + "epoch": 0.04, + "learning_rate": 9.969642897113172e-05, + "loss": 2.4682, + "step": 13185 + }, + { + "epoch": 0.04, + "learning_rate": 9.969619716281941e-05, + "loss": 2.371, + "step": 13190 + }, + { + "epoch": 0.04, + "learning_rate": 9.96959652663056e-05, + "loss": 2.4252, + "step": 13195 + }, + { + "epoch": 0.04, + "learning_rate": 9.96957332815907e-05, + "loss": 2.4677, + "step": 13200 + }, + { + "epoch": 0.04, + "learning_rate": 9.969550120867512e-05, + "loss": 2.4917, + "step": 13205 + }, + { + "epoch": 0.04, + "learning_rate": 9.969526904755928e-05, + "loss": 2.3736, + "step": 13210 + }, + { + "epoch": 0.04, + "learning_rate": 9.969503679824357e-05, + "loss": 2.3488, + "step": 13215 + }, + { + "epoch": 0.04, + "learning_rate": 9.969480446072843e-05, + "loss": 2.3821, + "step": 13220 + }, + { + "epoch": 0.04, + "learning_rate": 9.969457203501426e-05, + "loss": 2.2072, + "step": 13225 + }, + { + "epoch": 0.04, + "learning_rate": 9.969433952110147e-05, + "loss": 2.4671, + "step": 13230 + }, + { + "epoch": 0.04, + "learning_rate": 9.969410691899046e-05, + "loss": 2.3643, + "step": 13235 + }, + { + "epoch": 0.04, + "learning_rate": 9.969387422868168e-05, + "loss": 2.4271, + "step": 13240 + }, + { + "epoch": 0.04, + "learning_rate": 9.969364145017552e-05, + "loss": 2.4329, + "step": 13245 + }, + { + "epoch": 0.04, + "learning_rate": 9.969340858347238e-05, + "loss": 2.3996, + "step": 13250 + }, + { + "epoch": 0.04, + "learning_rate": 9.96931756285727e-05, + "loss": 2.2554, + "step": 13255 + }, + { + "epoch": 0.04, + "learning_rate": 9.969294258547688e-05, + "loss": 2.4284, + "step": 13260 + }, + { + "epoch": 0.04, + "learning_rate": 9.969270945418532e-05, + "loss": 2.3821, + "step": 13265 + }, + { + "epoch": 0.04, + "learning_rate": 9.969247623469847e-05, + "loss": 2.4524, + "step": 13270 + }, + { + "epoch": 0.04, + "learning_rate": 9.969224292701671e-05, + "loss": 2.4197, + "step": 13275 + }, + { + "epoch": 0.04, + "learning_rate": 9.969200953114046e-05, + "loss": 2.3887, + "step": 13280 + }, + { + "epoch": 0.04, + "learning_rate": 9.969177604707016e-05, + "loss": 2.3458, + "step": 13285 + }, + { + "epoch": 0.04, + "learning_rate": 9.969154247480618e-05, + "loss": 2.4173, + "step": 13290 + }, + { + "epoch": 0.04, + "learning_rate": 9.969130881434898e-05, + "loss": 2.3853, + "step": 13295 + }, + { + "epoch": 0.04, + "learning_rate": 9.969107506569894e-05, + "loss": 2.3636, + "step": 13300 + }, + { + "epoch": 0.04, + "learning_rate": 9.96908412288565e-05, + "loss": 2.4965, + "step": 13305 + }, + { + "epoch": 0.04, + "learning_rate": 9.969060730382205e-05, + "loss": 2.5012, + "step": 13310 + }, + { + "epoch": 0.04, + "learning_rate": 9.969037329059603e-05, + "loss": 2.5095, + "step": 13315 + }, + { + "epoch": 0.04, + "learning_rate": 9.969013918917885e-05, + "loss": 2.3618, + "step": 13320 + }, + { + "epoch": 0.04, + "learning_rate": 9.968990499957092e-05, + "loss": 2.5303, + "step": 13325 + }, + { + "epoch": 0.04, + "learning_rate": 9.968967072177264e-05, + "loss": 2.4305, + "step": 13330 + }, + { + "epoch": 0.04, + "learning_rate": 9.968943635578445e-05, + "loss": 2.425, + "step": 13335 + }, + { + "epoch": 0.04, + "learning_rate": 9.968920190160675e-05, + "loss": 2.4997, + "step": 13340 + }, + { + "epoch": 0.04, + "learning_rate": 9.968896735923998e-05, + "loss": 2.3925, + "step": 13345 + }, + { + "epoch": 0.04, + "learning_rate": 9.968873272868453e-05, + "loss": 2.5453, + "step": 13350 + }, + { + "epoch": 0.04, + "learning_rate": 9.968849800994082e-05, + "loss": 2.3475, + "step": 13355 + }, + { + "epoch": 0.04, + "learning_rate": 9.968826320300929e-05, + "loss": 2.3714, + "step": 13360 + }, + { + "epoch": 0.04, + "learning_rate": 9.968802830789032e-05, + "loss": 2.4607, + "step": 13365 + }, + { + "epoch": 0.04, + "learning_rate": 9.968779332458435e-05, + "loss": 2.5018, + "step": 13370 + }, + { + "epoch": 0.04, + "learning_rate": 9.968755825309181e-05, + "loss": 2.3526, + "step": 13375 + }, + { + "epoch": 0.04, + "learning_rate": 9.968732309341309e-05, + "loss": 2.4298, + "step": 13380 + }, + { + "epoch": 0.04, + "learning_rate": 9.968708784554861e-05, + "loss": 2.4728, + "step": 13385 + }, + { + "epoch": 0.04, + "learning_rate": 9.96868525094988e-05, + "loss": 2.4432, + "step": 13390 + }, + { + "epoch": 0.04, + "learning_rate": 9.96866170852641e-05, + "loss": 2.4342, + "step": 13395 + }, + { + "epoch": 0.04, + "learning_rate": 9.968638157284486e-05, + "loss": 2.3941, + "step": 13400 + }, + { + "epoch": 0.04, + "learning_rate": 9.968614597224156e-05, + "loss": 2.3221, + "step": 13405 + }, + { + "epoch": 0.04, + "learning_rate": 9.96859102834546e-05, + "loss": 2.4454, + "step": 13410 + }, + { + "epoch": 0.04, + "learning_rate": 9.968567450648439e-05, + "loss": 2.4677, + "step": 13415 + }, + { + "epoch": 0.04, + "learning_rate": 9.968543864133135e-05, + "loss": 2.3756, + "step": 13420 + }, + { + "epoch": 0.04, + "learning_rate": 9.968520268799591e-05, + "loss": 2.421, + "step": 13425 + }, + { + "epoch": 0.04, + "learning_rate": 9.968496664647848e-05, + "loss": 2.5024, + "step": 13430 + }, + { + "epoch": 0.04, + "learning_rate": 9.968473051677946e-05, + "loss": 2.5432, + "step": 13435 + }, + { + "epoch": 0.04, + "learning_rate": 9.968449429889933e-05, + "loss": 2.4407, + "step": 13440 + }, + { + "epoch": 0.04, + "learning_rate": 9.968425799283845e-05, + "loss": 2.4296, + "step": 13445 + }, + { + "epoch": 0.04, + "learning_rate": 9.968402159859725e-05, + "loss": 2.4814, + "step": 13450 + }, + { + "epoch": 0.04, + "learning_rate": 9.968378511617615e-05, + "loss": 2.3624, + "step": 13455 + }, + { + "epoch": 0.04, + "learning_rate": 9.96835485455756e-05, + "loss": 2.3314, + "step": 13460 + }, + { + "epoch": 0.04, + "learning_rate": 9.968331188679599e-05, + "loss": 2.466, + "step": 13465 + }, + { + "epoch": 0.04, + "learning_rate": 9.968307513983774e-05, + "loss": 2.4457, + "step": 13470 + }, + { + "epoch": 0.04, + "learning_rate": 9.968283830470128e-05, + "loss": 2.4506, + "step": 13475 + }, + { + "epoch": 0.04, + "learning_rate": 9.968260138138703e-05, + "loss": 2.5557, + "step": 13480 + }, + { + "epoch": 0.04, + "learning_rate": 9.968236436989539e-05, + "loss": 2.4706, + "step": 13485 + }, + { + "epoch": 0.04, + "learning_rate": 9.968212727022683e-05, + "loss": 2.3931, + "step": 13490 + }, + { + "epoch": 0.04, + "learning_rate": 9.968189008238172e-05, + "loss": 2.5638, + "step": 13495 + }, + { + "epoch": 0.04, + "learning_rate": 9.968165280636052e-05, + "loss": 2.3497, + "step": 13500 + }, + { + "epoch": 0.04, + "learning_rate": 9.96814154421636e-05, + "loss": 2.5864, + "step": 13505 + }, + { + "epoch": 0.04, + "learning_rate": 9.968117798979144e-05, + "loss": 2.3964, + "step": 13510 + }, + { + "epoch": 0.04, + "learning_rate": 9.968094044924443e-05, + "loss": 2.5434, + "step": 13515 + }, + { + "epoch": 0.04, + "learning_rate": 9.968070282052299e-05, + "loss": 2.4058, + "step": 13520 + }, + { + "epoch": 0.04, + "learning_rate": 9.968046510362755e-05, + "loss": 2.4877, + "step": 13525 + }, + { + "epoch": 0.04, + "learning_rate": 9.968022729855852e-05, + "loss": 2.4332, + "step": 13530 + }, + { + "epoch": 0.04, + "learning_rate": 9.967998940531635e-05, + "loss": 2.3576, + "step": 13535 + }, + { + "epoch": 0.04, + "learning_rate": 9.967975142390143e-05, + "loss": 2.5375, + "step": 13540 + }, + { + "epoch": 0.04, + "learning_rate": 9.96795133543142e-05, + "loss": 2.4287, + "step": 13545 + }, + { + "epoch": 0.04, + "learning_rate": 9.967927519655508e-05, + "loss": 2.4387, + "step": 13550 + }, + { + "epoch": 0.04, + "learning_rate": 9.96790369506245e-05, + "loss": 2.5504, + "step": 13555 + }, + { + "epoch": 0.04, + "learning_rate": 9.967879861652288e-05, + "loss": 2.3424, + "step": 13560 + }, + { + "epoch": 0.04, + "learning_rate": 9.967856019425061e-05, + "loss": 2.4973, + "step": 13565 + }, + { + "epoch": 0.04, + "learning_rate": 9.967832168380815e-05, + "loss": 2.3955, + "step": 13570 + }, + { + "epoch": 0.04, + "learning_rate": 9.967808308519593e-05, + "loss": 2.5045, + "step": 13575 + }, + { + "epoch": 0.04, + "learning_rate": 9.967784439841435e-05, + "loss": 2.4064, + "step": 13580 + }, + { + "epoch": 0.04, + "learning_rate": 9.967760562346381e-05, + "loss": 2.3535, + "step": 13585 + }, + { + "epoch": 0.04, + "learning_rate": 9.96773667603448e-05, + "loss": 2.4612, + "step": 13590 + }, + { + "epoch": 0.04, + "learning_rate": 9.967712780905772e-05, + "loss": 2.3253, + "step": 13595 + }, + { + "epoch": 0.04, + "learning_rate": 9.967688876960294e-05, + "loss": 2.3859, + "step": 13600 + }, + { + "epoch": 0.04, + "learning_rate": 9.967664964198096e-05, + "loss": 2.4061, + "step": 13605 + }, + { + "epoch": 0.04, + "learning_rate": 9.967641042619216e-05, + "loss": 2.4336, + "step": 13610 + }, + { + "epoch": 0.04, + "learning_rate": 9.967617112223697e-05, + "loss": 2.434, + "step": 13615 + }, + { + "epoch": 0.04, + "learning_rate": 9.967593173011584e-05, + "loss": 2.2418, + "step": 13620 + }, + { + "epoch": 0.04, + "learning_rate": 9.967569224982915e-05, + "loss": 2.4886, + "step": 13625 + }, + { + "epoch": 0.04, + "learning_rate": 9.967545268137738e-05, + "loss": 2.3764, + "step": 13630 + }, + { + "epoch": 0.04, + "learning_rate": 9.96752130247609e-05, + "loss": 2.4769, + "step": 13635 + }, + { + "epoch": 0.04, + "learning_rate": 9.967497327998018e-05, + "loss": 2.4836, + "step": 13640 + }, + { + "epoch": 0.04, + "learning_rate": 9.967473344703562e-05, + "loss": 2.3111, + "step": 13645 + }, + { + "epoch": 0.04, + "learning_rate": 9.967449352592765e-05, + "loss": 2.439, + "step": 13650 + }, + { + "epoch": 0.04, + "learning_rate": 9.967425351665669e-05, + "loss": 2.4785, + "step": 13655 + }, + { + "epoch": 0.04, + "learning_rate": 9.967401341922319e-05, + "loss": 2.4289, + "step": 13660 + }, + { + "epoch": 0.04, + "learning_rate": 9.967377323362755e-05, + "loss": 2.3939, + "step": 13665 + }, + { + "epoch": 0.04, + "learning_rate": 9.967353295987021e-05, + "loss": 2.4521, + "step": 13670 + }, + { + "epoch": 0.04, + "learning_rate": 9.96732925979516e-05, + "loss": 2.4248, + "step": 13675 + }, + { + "epoch": 0.04, + "learning_rate": 9.967305214787213e-05, + "loss": 2.3261, + "step": 13680 + }, + { + "epoch": 0.04, + "learning_rate": 9.967281160963225e-05, + "loss": 2.4782, + "step": 13685 + }, + { + "epoch": 0.04, + "learning_rate": 9.967257098323235e-05, + "loss": 2.4283, + "step": 13690 + }, + { + "epoch": 0.04, + "learning_rate": 9.967233026867289e-05, + "loss": 2.3171, + "step": 13695 + }, + { + "epoch": 0.04, + "learning_rate": 9.96720894659543e-05, + "loss": 2.4215, + "step": 13700 + }, + { + "epoch": 0.04, + "learning_rate": 9.967184857507697e-05, + "loss": 2.6154, + "step": 13705 + }, + { + "epoch": 0.04, + "learning_rate": 9.967160759604137e-05, + "loss": 2.3821, + "step": 13710 + }, + { + "epoch": 0.04, + "learning_rate": 9.967136652884791e-05, + "loss": 2.3455, + "step": 13715 + }, + { + "epoch": 0.04, + "learning_rate": 9.967112537349701e-05, + "loss": 2.3946, + "step": 13720 + }, + { + "epoch": 0.04, + "learning_rate": 9.967088412998912e-05, + "loss": 2.4, + "step": 13725 + }, + { + "epoch": 0.04, + "learning_rate": 9.967064279832465e-05, + "loss": 2.4146, + "step": 13730 + }, + { + "epoch": 0.04, + "learning_rate": 9.967040137850401e-05, + "loss": 2.4055, + "step": 13735 + }, + { + "epoch": 0.04, + "learning_rate": 9.967015987052768e-05, + "loss": 2.5137, + "step": 13740 + }, + { + "epoch": 0.04, + "learning_rate": 9.966991827439604e-05, + "loss": 2.3818, + "step": 13745 + }, + { + "epoch": 0.04, + "learning_rate": 9.966967659010955e-05, + "loss": 2.4505, + "step": 13750 + }, + { + "epoch": 0.04, + "learning_rate": 9.966943481766864e-05, + "loss": 2.5929, + "step": 13755 + }, + { + "epoch": 0.04, + "learning_rate": 9.966919295707371e-05, + "loss": 2.44, + "step": 13760 + }, + { + "epoch": 0.04, + "learning_rate": 9.96689510083252e-05, + "loss": 2.3333, + "step": 13765 + }, + { + "epoch": 0.04, + "learning_rate": 9.966870897142356e-05, + "loss": 2.4756, + "step": 13770 + }, + { + "epoch": 0.04, + "learning_rate": 9.966846684636919e-05, + "loss": 2.4267, + "step": 13775 + }, + { + "epoch": 0.04, + "learning_rate": 9.966822463316254e-05, + "loss": 2.5296, + "step": 13780 + }, + { + "epoch": 0.04, + "learning_rate": 9.966798233180403e-05, + "loss": 2.5856, + "step": 13785 + }, + { + "epoch": 0.04, + "learning_rate": 9.96677399422941e-05, + "loss": 2.4743, + "step": 13790 + }, + { + "epoch": 0.04, + "learning_rate": 9.966749746463317e-05, + "loss": 2.2569, + "step": 13795 + }, + { + "epoch": 0.04, + "learning_rate": 9.966725489882169e-05, + "loss": 2.5109, + "step": 13800 + }, + { + "epoch": 0.04, + "learning_rate": 9.966701224486005e-05, + "loss": 2.3925, + "step": 13805 + }, + { + "epoch": 0.04, + "learning_rate": 9.966676950274873e-05, + "loss": 2.437, + "step": 13810 + }, + { + "epoch": 0.04, + "learning_rate": 9.966652667248812e-05, + "loss": 2.361, + "step": 13815 + }, + { + "epoch": 0.04, + "learning_rate": 9.966628375407867e-05, + "loss": 2.5492, + "step": 13820 + }, + { + "epoch": 0.04, + "learning_rate": 9.966604074752081e-05, + "loss": 2.445, + "step": 13825 + }, + { + "epoch": 0.04, + "learning_rate": 9.966579765281499e-05, + "loss": 2.377, + "step": 13830 + }, + { + "epoch": 0.04, + "learning_rate": 9.966555446996159e-05, + "loss": 2.4243, + "step": 13835 + }, + { + "epoch": 0.04, + "learning_rate": 9.966531119896109e-05, + "loss": 2.4349, + "step": 13840 + }, + { + "epoch": 0.04, + "learning_rate": 9.966506783981391e-05, + "loss": 2.3738, + "step": 13845 + }, + { + "epoch": 0.04, + "learning_rate": 9.966482439252048e-05, + "loss": 2.507, + "step": 13850 + }, + { + "epoch": 0.04, + "learning_rate": 9.966458085708121e-05, + "loss": 2.4916, + "step": 13855 + }, + { + "epoch": 0.04, + "learning_rate": 9.966433723349658e-05, + "loss": 2.4135, + "step": 13860 + }, + { + "epoch": 0.04, + "learning_rate": 9.966409352176696e-05, + "loss": 2.3829, + "step": 13865 + }, + { + "epoch": 0.04, + "learning_rate": 9.966384972189285e-05, + "loss": 2.5196, + "step": 13870 + }, + { + "epoch": 0.04, + "learning_rate": 9.966360583387462e-05, + "loss": 2.3872, + "step": 13875 + }, + { + "epoch": 0.04, + "learning_rate": 9.966336185771275e-05, + "loss": 2.4799, + "step": 13880 + }, + { + "epoch": 0.04, + "learning_rate": 9.966311779340763e-05, + "loss": 2.3924, + "step": 13885 + }, + { + "epoch": 0.04, + "learning_rate": 9.966287364095974e-05, + "loss": 2.4582, + "step": 13890 + }, + { + "epoch": 0.04, + "learning_rate": 9.966262940036949e-05, + "loss": 2.5159, + "step": 13895 + }, + { + "epoch": 0.04, + "learning_rate": 9.96623850716373e-05, + "loss": 2.3764, + "step": 13900 + }, + { + "epoch": 0.04, + "learning_rate": 9.966214065476364e-05, + "loss": 2.3646, + "step": 13905 + }, + { + "epoch": 0.04, + "learning_rate": 9.966189614974891e-05, + "loss": 2.4458, + "step": 13910 + }, + { + "epoch": 0.04, + "learning_rate": 9.966165155659356e-05, + "loss": 2.5042, + "step": 13915 + }, + { + "epoch": 0.04, + "learning_rate": 9.966140687529801e-05, + "loss": 2.3222, + "step": 13920 + }, + { + "epoch": 0.04, + "learning_rate": 9.966116210586272e-05, + "loss": 2.3708, + "step": 13925 + }, + { + "epoch": 0.04, + "learning_rate": 9.96609172482881e-05, + "loss": 2.5198, + "step": 13930 + }, + { + "epoch": 0.04, + "learning_rate": 9.966067230257459e-05, + "loss": 2.5285, + "step": 13935 + }, + { + "epoch": 0.04, + "learning_rate": 9.966042726872265e-05, + "loss": 2.4138, + "step": 13940 + }, + { + "epoch": 0.04, + "learning_rate": 9.966018214673268e-05, + "loss": 2.4918, + "step": 13945 + }, + { + "epoch": 0.04, + "learning_rate": 9.965993693660511e-05, + "loss": 2.4291, + "step": 13950 + }, + { + "epoch": 0.04, + "learning_rate": 9.96596916383404e-05, + "loss": 2.4099, + "step": 13955 + }, + { + "epoch": 0.04, + "learning_rate": 9.9659446251939e-05, + "loss": 2.2645, + "step": 13960 + }, + { + "epoch": 0.04, + "learning_rate": 9.965920077740131e-05, + "loss": 2.4929, + "step": 13965 + }, + { + "epoch": 0.04, + "learning_rate": 9.965895521472778e-05, + "loss": 2.3393, + "step": 13970 + }, + { + "epoch": 0.04, + "learning_rate": 9.965870956391885e-05, + "loss": 2.4541, + "step": 13975 + }, + { + "epoch": 0.04, + "learning_rate": 9.965846382497494e-05, + "loss": 2.4024, + "step": 13980 + }, + { + "epoch": 0.04, + "learning_rate": 9.965821799789651e-05, + "loss": 2.3329, + "step": 13985 + }, + { + "epoch": 0.04, + "learning_rate": 9.965797208268399e-05, + "loss": 2.4645, + "step": 13990 + }, + { + "epoch": 0.04, + "learning_rate": 9.965772607933779e-05, + "loss": 2.4528, + "step": 13995 + }, + { + "epoch": 0.04, + "learning_rate": 9.965747998785838e-05, + "loss": 2.5011, + "step": 14000 + }, + { + "epoch": 0.04, + "learning_rate": 9.965723380824619e-05, + "loss": 2.4139, + "step": 14005 + }, + { + "epoch": 0.04, + "learning_rate": 9.965698754050164e-05, + "loss": 2.4643, + "step": 14010 + }, + { + "epoch": 0.04, + "learning_rate": 9.965674118462519e-05, + "loss": 2.3094, + "step": 14015 + }, + { + "epoch": 0.04, + "learning_rate": 9.965649474061724e-05, + "loss": 2.3905, + "step": 14020 + }, + { + "epoch": 0.04, + "learning_rate": 9.965624820847828e-05, + "loss": 2.4447, + "step": 14025 + }, + { + "epoch": 0.04, + "learning_rate": 9.965600158820871e-05, + "loss": 2.3812, + "step": 14030 + }, + { + "epoch": 0.04, + "learning_rate": 9.965575487980898e-05, + "loss": 2.3842, + "step": 14035 + }, + { + "epoch": 0.04, + "learning_rate": 9.965550808327953e-05, + "loss": 2.4367, + "step": 14040 + }, + { + "epoch": 0.04, + "learning_rate": 9.965526119862078e-05, + "loss": 2.5105, + "step": 14045 + }, + { + "epoch": 0.04, + "learning_rate": 9.965501422583319e-05, + "loss": 2.4167, + "step": 14050 + }, + { + "epoch": 0.04, + "learning_rate": 9.965476716491718e-05, + "loss": 2.4956, + "step": 14055 + }, + { + "epoch": 0.04, + "learning_rate": 9.965452001587321e-05, + "loss": 2.4704, + "step": 14060 + }, + { + "epoch": 0.04, + "learning_rate": 9.96542727787017e-05, + "loss": 2.3768, + "step": 14065 + }, + { + "epoch": 0.04, + "learning_rate": 9.965402545340312e-05, + "loss": 2.2985, + "step": 14070 + }, + { + "epoch": 0.04, + "learning_rate": 9.965377803997787e-05, + "loss": 2.4935, + "step": 14075 + }, + { + "epoch": 0.04, + "learning_rate": 9.965353053842639e-05, + "loss": 2.487, + "step": 14080 + }, + { + "epoch": 0.04, + "learning_rate": 9.965328294874914e-05, + "loss": 2.3958, + "step": 14085 + }, + { + "epoch": 0.04, + "learning_rate": 9.965303527094655e-05, + "loss": 2.3543, + "step": 14090 + }, + { + "epoch": 0.04, + "learning_rate": 9.965278750501908e-05, + "loss": 2.2703, + "step": 14095 + }, + { + "epoch": 0.04, + "learning_rate": 9.965253965096714e-05, + "loss": 2.4955, + "step": 14100 + }, + { + "epoch": 0.04, + "learning_rate": 9.96522917087912e-05, + "loss": 2.5037, + "step": 14105 + }, + { + "epoch": 0.04, + "learning_rate": 9.965204367849165e-05, + "loss": 2.3872, + "step": 14110 + }, + { + "epoch": 0.04, + "learning_rate": 9.965179556006898e-05, + "loss": 2.5054, + "step": 14115 + }, + { + "epoch": 0.04, + "learning_rate": 9.965154735352363e-05, + "loss": 2.3164, + "step": 14120 + }, + { + "epoch": 0.04, + "learning_rate": 9.965129905885599e-05, + "loss": 2.3496, + "step": 14125 + }, + { + "epoch": 0.04, + "learning_rate": 9.965105067606656e-05, + "loss": 2.4244, + "step": 14130 + }, + { + "epoch": 0.04, + "learning_rate": 9.965080220515575e-05, + "loss": 2.3248, + "step": 14135 + }, + { + "epoch": 0.04, + "learning_rate": 9.965055364612399e-05, + "loss": 2.3181, + "step": 14140 + }, + { + "epoch": 0.04, + "learning_rate": 9.965030499897175e-05, + "loss": 2.4587, + "step": 14145 + }, + { + "epoch": 0.04, + "learning_rate": 9.965005626369945e-05, + "loss": 2.4367, + "step": 14150 + }, + { + "epoch": 0.04, + "learning_rate": 9.964980744030756e-05, + "loss": 2.4807, + "step": 14155 + }, + { + "epoch": 0.04, + "learning_rate": 9.964955852879648e-05, + "loss": 2.3716, + "step": 14160 + }, + { + "epoch": 0.04, + "learning_rate": 9.96493095291667e-05, + "loss": 2.5385, + "step": 14165 + }, + { + "epoch": 0.04, + "learning_rate": 9.96490604414186e-05, + "loss": 2.4713, + "step": 14170 + }, + { + "epoch": 0.04, + "learning_rate": 9.964881126555268e-05, + "loss": 2.3878, + "step": 14175 + }, + { + "epoch": 0.04, + "learning_rate": 9.964856200156935e-05, + "loss": 2.3459, + "step": 14180 + }, + { + "epoch": 0.04, + "learning_rate": 9.964831264946905e-05, + "loss": 2.4775, + "step": 14185 + }, + { + "epoch": 0.04, + "learning_rate": 9.964806320925226e-05, + "loss": 2.402, + "step": 14190 + }, + { + "epoch": 0.04, + "learning_rate": 9.964781368091937e-05, + "loss": 2.4737, + "step": 14195 + }, + { + "epoch": 0.04, + "learning_rate": 9.964756406447088e-05, + "loss": 2.4934, + "step": 14200 + }, + { + "epoch": 0.04, + "learning_rate": 9.964731435990718e-05, + "loss": 2.2771, + "step": 14205 + }, + { + "epoch": 0.04, + "learning_rate": 9.964706456722874e-05, + "loss": 2.3055, + "step": 14210 + }, + { + "epoch": 0.04, + "learning_rate": 9.964681468643602e-05, + "loss": 2.5162, + "step": 14215 + }, + { + "epoch": 0.04, + "learning_rate": 9.964656471752943e-05, + "loss": 2.4123, + "step": 14220 + }, + { + "epoch": 0.04, + "learning_rate": 9.964631466050941e-05, + "loss": 2.3655, + "step": 14225 + }, + { + "epoch": 0.04, + "learning_rate": 9.964606451537644e-05, + "loss": 2.3623, + "step": 14230 + }, + { + "epoch": 0.04, + "learning_rate": 9.964581428213094e-05, + "loss": 2.4312, + "step": 14235 + }, + { + "epoch": 0.04, + "learning_rate": 9.964556396077336e-05, + "loss": 2.394, + "step": 14240 + }, + { + "epoch": 0.04, + "learning_rate": 9.964531355130413e-05, + "loss": 2.517, + "step": 14245 + }, + { + "epoch": 0.04, + "learning_rate": 9.964506305372371e-05, + "loss": 2.4923, + "step": 14250 + }, + { + "epoch": 0.04, + "learning_rate": 9.964481246803256e-05, + "loss": 2.5004, + "step": 14255 + }, + { + "epoch": 0.04, + "learning_rate": 9.96445617942311e-05, + "loss": 2.5037, + "step": 14260 + }, + { + "epoch": 0.04, + "learning_rate": 9.964431103231977e-05, + "loss": 2.4519, + "step": 14265 + }, + { + "epoch": 0.04, + "learning_rate": 9.964406018229904e-05, + "loss": 2.492, + "step": 14270 + }, + { + "epoch": 0.04, + "learning_rate": 9.964380924416932e-05, + "loss": 2.4718, + "step": 14275 + }, + { + "epoch": 0.04, + "learning_rate": 9.96435582179311e-05, + "loss": 2.4411, + "step": 14280 + }, + { + "epoch": 0.04, + "learning_rate": 9.96433071035848e-05, + "loss": 2.3875, + "step": 14285 + }, + { + "epoch": 0.04, + "learning_rate": 9.964305590113086e-05, + "loss": 2.4722, + "step": 14290 + }, + { + "epoch": 0.04, + "learning_rate": 9.964280461056973e-05, + "loss": 2.4636, + "step": 14295 + }, + { + "epoch": 0.04, + "learning_rate": 9.964255323190186e-05, + "loss": 2.5074, + "step": 14300 + }, + { + "epoch": 0.04, + "learning_rate": 9.964230176512772e-05, + "loss": 2.504, + "step": 14305 + }, + { + "epoch": 0.04, + "learning_rate": 9.964205021024771e-05, + "loss": 2.4543, + "step": 14310 + }, + { + "epoch": 0.04, + "learning_rate": 9.96417985672623e-05, + "loss": 2.2679, + "step": 14315 + }, + { + "epoch": 0.04, + "learning_rate": 9.964154683617194e-05, + "loss": 2.4733, + "step": 14320 + }, + { + "epoch": 0.04, + "learning_rate": 9.964129501697708e-05, + "loss": 2.4399, + "step": 14325 + }, + { + "epoch": 0.04, + "learning_rate": 9.964104310967814e-05, + "loss": 2.4665, + "step": 14330 + }, + { + "epoch": 0.04, + "learning_rate": 9.96407911142756e-05, + "loss": 2.4261, + "step": 14335 + }, + { + "epoch": 0.04, + "learning_rate": 9.964053903076989e-05, + "loss": 2.3265, + "step": 14340 + }, + { + "epoch": 0.04, + "learning_rate": 9.964028685916147e-05, + "loss": 2.4733, + "step": 14345 + }, + { + "epoch": 0.04, + "learning_rate": 9.964003459945077e-05, + "loss": 2.4773, + "step": 14350 + }, + { + "epoch": 0.04, + "learning_rate": 9.963978225163825e-05, + "loss": 2.4845, + "step": 14355 + }, + { + "epoch": 0.04, + "learning_rate": 9.963952981572436e-05, + "loss": 2.4275, + "step": 14360 + }, + { + "epoch": 0.04, + "learning_rate": 9.963927729170952e-05, + "loss": 2.4255, + "step": 14365 + }, + { + "epoch": 0.04, + "learning_rate": 9.963902467959422e-05, + "loss": 2.5272, + "step": 14370 + }, + { + "epoch": 0.04, + "learning_rate": 9.963877197937887e-05, + "loss": 2.5198, + "step": 14375 + }, + { + "epoch": 0.04, + "learning_rate": 9.963851919106396e-05, + "loss": 2.3942, + "step": 14380 + }, + { + "epoch": 0.04, + "learning_rate": 9.963826631464991e-05, + "loss": 2.3855, + "step": 14385 + }, + { + "epoch": 0.04, + "learning_rate": 9.963801335013717e-05, + "loss": 2.4544, + "step": 14390 + }, + { + "epoch": 0.04, + "learning_rate": 9.96377602975262e-05, + "loss": 2.376, + "step": 14395 + }, + { + "epoch": 0.04, + "learning_rate": 9.963750715681743e-05, + "loss": 2.4503, + "step": 14400 + }, + { + "epoch": 0.04, + "learning_rate": 9.963725392801133e-05, + "loss": 2.4902, + "step": 14405 + }, + { + "epoch": 0.04, + "learning_rate": 9.963700061110834e-05, + "loss": 2.491, + "step": 14410 + }, + { + "epoch": 0.04, + "learning_rate": 9.963674720610891e-05, + "loss": 2.4817, + "step": 14415 + }, + { + "epoch": 0.04, + "learning_rate": 9.96364937130135e-05, + "loss": 2.2152, + "step": 14420 + }, + { + "epoch": 0.04, + "learning_rate": 9.963624013182254e-05, + "loss": 2.3726, + "step": 14425 + }, + { + "epoch": 0.04, + "learning_rate": 9.96359864625365e-05, + "loss": 2.4642, + "step": 14430 + }, + { + "epoch": 0.04, + "learning_rate": 9.963573270515582e-05, + "loss": 2.5176, + "step": 14435 + }, + { + "epoch": 0.04, + "learning_rate": 9.963547885968094e-05, + "loss": 2.3843, + "step": 14440 + }, + { + "epoch": 0.04, + "learning_rate": 9.963522492611233e-05, + "loss": 2.4041, + "step": 14445 + }, + { + "epoch": 0.04, + "learning_rate": 9.963497090445044e-05, + "loss": 2.4299, + "step": 14450 + }, + { + "epoch": 0.04, + "learning_rate": 9.963471679469569e-05, + "loss": 2.3855, + "step": 14455 + }, + { + "epoch": 0.04, + "learning_rate": 9.963446259684859e-05, + "loss": 2.3706, + "step": 14460 + }, + { + "epoch": 0.04, + "learning_rate": 9.963420831090953e-05, + "loss": 2.4383, + "step": 14465 + }, + { + "epoch": 0.04, + "learning_rate": 9.963395393687899e-05, + "loss": 2.4517, + "step": 14470 + }, + { + "epoch": 0.04, + "learning_rate": 9.963369947475743e-05, + "loss": 2.3466, + "step": 14475 + }, + { + "epoch": 0.04, + "learning_rate": 9.963344492454527e-05, + "loss": 2.4438, + "step": 14480 + }, + { + "epoch": 0.04, + "learning_rate": 9.9633190286243e-05, + "loss": 2.543, + "step": 14485 + }, + { + "epoch": 0.04, + "learning_rate": 9.963293555985105e-05, + "loss": 2.5156, + "step": 14490 + }, + { + "epoch": 0.04, + "learning_rate": 9.963268074536986e-05, + "loss": 2.5131, + "step": 14495 + }, + { + "epoch": 0.04, + "learning_rate": 9.963242584279992e-05, + "loss": 2.4781, + "step": 14500 + }, + { + "epoch": 0.04, + "learning_rate": 9.963217085214165e-05, + "loss": 2.3772, + "step": 14505 + }, + { + "epoch": 0.04, + "learning_rate": 9.96319157733955e-05, + "loss": 2.3468, + "step": 14510 + }, + { + "epoch": 0.04, + "learning_rate": 9.963166060656195e-05, + "loss": 2.4399, + "step": 14515 + }, + { + "epoch": 0.04, + "learning_rate": 9.963140535164144e-05, + "loss": 2.5023, + "step": 14520 + }, + { + "epoch": 0.04, + "learning_rate": 9.963115000863443e-05, + "loss": 2.3132, + "step": 14525 + }, + { + "epoch": 0.04, + "learning_rate": 9.963089457754135e-05, + "loss": 2.4167, + "step": 14530 + }, + { + "epoch": 0.04, + "learning_rate": 9.963063905836267e-05, + "loss": 2.4457, + "step": 14535 + }, + { + "epoch": 0.04, + "learning_rate": 9.963038345109886e-05, + "loss": 2.4412, + "step": 14540 + }, + { + "epoch": 0.04, + "learning_rate": 9.963012775575033e-05, + "loss": 2.4037, + "step": 14545 + }, + { + "epoch": 0.04, + "learning_rate": 9.962987197231757e-05, + "loss": 2.5052, + "step": 14550 + }, + { + "epoch": 0.04, + "learning_rate": 9.962961610080101e-05, + "loss": 2.3712, + "step": 14555 + }, + { + "epoch": 0.04, + "learning_rate": 9.962936014120114e-05, + "loss": 2.4211, + "step": 14560 + }, + { + "epoch": 0.04, + "learning_rate": 9.962910409351838e-05, + "loss": 2.2523, + "step": 14565 + }, + { + "epoch": 0.04, + "learning_rate": 9.96288479577532e-05, + "loss": 2.3616, + "step": 14570 + }, + { + "epoch": 0.04, + "learning_rate": 9.962859173390604e-05, + "loss": 2.4002, + "step": 14575 + }, + { + "epoch": 0.04, + "learning_rate": 9.962833542197739e-05, + "loss": 2.421, + "step": 14580 + }, + { + "epoch": 0.04, + "learning_rate": 9.962807902196764e-05, + "loss": 2.5345, + "step": 14585 + }, + { + "epoch": 0.04, + "learning_rate": 9.962782253387733e-05, + "loss": 2.4816, + "step": 14590 + }, + { + "epoch": 0.04, + "learning_rate": 9.962756595770683e-05, + "loss": 2.4156, + "step": 14595 + }, + { + "epoch": 0.04, + "learning_rate": 9.962730929345666e-05, + "loss": 2.4085, + "step": 14600 + }, + { + "epoch": 0.04, + "learning_rate": 9.962705254112725e-05, + "loss": 2.2961, + "step": 14605 + }, + { + "epoch": 0.04, + "learning_rate": 9.962679570071904e-05, + "loss": 2.3914, + "step": 14610 + }, + { + "epoch": 0.04, + "learning_rate": 9.962653877223254e-05, + "loss": 2.3885, + "step": 14615 + }, + { + "epoch": 0.04, + "learning_rate": 9.962628175566812e-05, + "loss": 2.5115, + "step": 14620 + }, + { + "epoch": 0.04, + "learning_rate": 9.962602465102632e-05, + "loss": 2.4812, + "step": 14625 + }, + { + "epoch": 0.04, + "learning_rate": 9.962576745830757e-05, + "loss": 2.4566, + "step": 14630 + }, + { + "epoch": 0.04, + "learning_rate": 9.962551017751228e-05, + "loss": 2.5117, + "step": 14635 + }, + { + "epoch": 0.04, + "learning_rate": 9.962525280864097e-05, + "loss": 2.3814, + "step": 14640 + }, + { + "epoch": 0.04, + "learning_rate": 9.962499535169406e-05, + "loss": 2.3733, + "step": 14645 + }, + { + "epoch": 0.04, + "learning_rate": 9.962473780667203e-05, + "loss": 2.4634, + "step": 14650 + }, + { + "epoch": 0.04, + "learning_rate": 9.962448017357532e-05, + "loss": 2.43, + "step": 14655 + }, + { + "epoch": 0.04, + "learning_rate": 9.962422245240439e-05, + "loss": 2.534, + "step": 14660 + }, + { + "epoch": 0.04, + "learning_rate": 9.962396464315971e-05, + "loss": 2.5123, + "step": 14665 + }, + { + "epoch": 0.04, + "learning_rate": 9.96237067458417e-05, + "loss": 2.5148, + "step": 14670 + }, + { + "epoch": 0.04, + "learning_rate": 9.962344876045087e-05, + "loss": 2.2761, + "step": 14675 + }, + { + "epoch": 0.04, + "learning_rate": 9.962319068698765e-05, + "loss": 2.5174, + "step": 14680 + }, + { + "epoch": 0.04, + "learning_rate": 9.962293252545249e-05, + "loss": 2.4202, + "step": 14685 + }, + { + "epoch": 0.04, + "learning_rate": 9.962267427584587e-05, + "loss": 2.4617, + "step": 14690 + }, + { + "epoch": 0.04, + "learning_rate": 9.962241593816823e-05, + "loss": 2.4636, + "step": 14695 + }, + { + "epoch": 0.04, + "learning_rate": 9.962215751242006e-05, + "loss": 2.4202, + "step": 14700 + }, + { + "epoch": 0.04, + "learning_rate": 9.962189899860177e-05, + "loss": 2.4203, + "step": 14705 + }, + { + "epoch": 0.04, + "learning_rate": 9.962164039671385e-05, + "loss": 2.4585, + "step": 14710 + }, + { + "epoch": 0.04, + "learning_rate": 9.962138170675674e-05, + "loss": 2.4625, + "step": 14715 + }, + { + "epoch": 0.04, + "learning_rate": 9.962112292873094e-05, + "loss": 2.4077, + "step": 14720 + }, + { + "epoch": 0.04, + "learning_rate": 9.962086406263687e-05, + "loss": 2.5139, + "step": 14725 + }, + { + "epoch": 0.04, + "learning_rate": 9.962060510847499e-05, + "loss": 2.4554, + "step": 14730 + }, + { + "epoch": 0.04, + "learning_rate": 9.962034606624577e-05, + "loss": 2.522, + "step": 14735 + }, + { + "epoch": 0.04, + "learning_rate": 9.962008693594969e-05, + "loss": 2.2708, + "step": 14740 + }, + { + "epoch": 0.04, + "learning_rate": 9.961982771758716e-05, + "loss": 2.3892, + "step": 14745 + }, + { + "epoch": 0.04, + "learning_rate": 9.961956841115869e-05, + "loss": 2.4193, + "step": 14750 + }, + { + "epoch": 0.04, + "learning_rate": 9.961930901666472e-05, + "loss": 2.4364, + "step": 14755 + }, + { + "epoch": 0.04, + "learning_rate": 9.96190495341057e-05, + "loss": 2.507, + "step": 14760 + }, + { + "epoch": 0.04, + "learning_rate": 9.96187899634821e-05, + "loss": 2.3564, + "step": 14765 + }, + { + "epoch": 0.04, + "learning_rate": 9.96185303047944e-05, + "loss": 2.3183, + "step": 14770 + }, + { + "epoch": 0.04, + "learning_rate": 9.961827055804303e-05, + "loss": 2.4842, + "step": 14775 + }, + { + "epoch": 0.04, + "learning_rate": 9.961801072322846e-05, + "loss": 2.4267, + "step": 14780 + }, + { + "epoch": 0.04, + "learning_rate": 9.961775080035116e-05, + "loss": 2.5022, + "step": 14785 + }, + { + "epoch": 0.04, + "learning_rate": 9.961749078941159e-05, + "loss": 2.3206, + "step": 14790 + }, + { + "epoch": 0.04, + "learning_rate": 9.961723069041021e-05, + "loss": 2.3807, + "step": 14795 + }, + { + "epoch": 0.04, + "learning_rate": 9.961697050334746e-05, + "loss": 2.5148, + "step": 14800 + }, + { + "epoch": 0.04, + "learning_rate": 9.961671022822382e-05, + "loss": 2.3002, + "step": 14805 + }, + { + "epoch": 0.04, + "learning_rate": 9.961644986503977e-05, + "loss": 2.3872, + "step": 14810 + }, + { + "epoch": 0.04, + "learning_rate": 9.961618941379575e-05, + "loss": 2.2984, + "step": 14815 + }, + { + "epoch": 0.04, + "learning_rate": 9.961592887449223e-05, + "loss": 2.3364, + "step": 14820 + }, + { + "epoch": 0.04, + "learning_rate": 9.961566824712967e-05, + "loss": 2.4108, + "step": 14825 + }, + { + "epoch": 0.04, + "learning_rate": 9.961540753170852e-05, + "loss": 2.4898, + "step": 14830 + }, + { + "epoch": 0.04, + "learning_rate": 9.961514672822927e-05, + "loss": 2.3746, + "step": 14835 + }, + { + "epoch": 0.04, + "learning_rate": 9.961488583669237e-05, + "loss": 2.5012, + "step": 14840 + }, + { + "epoch": 0.04, + "learning_rate": 9.961462485709827e-05, + "loss": 2.4138, + "step": 14845 + }, + { + "epoch": 0.04, + "learning_rate": 9.961436378944744e-05, + "loss": 2.3792, + "step": 14850 + }, + { + "epoch": 0.04, + "learning_rate": 9.961410263374036e-05, + "loss": 2.3634, + "step": 14855 + }, + { + "epoch": 0.04, + "learning_rate": 9.961384138997746e-05, + "loss": 2.5265, + "step": 14860 + }, + { + "epoch": 0.04, + "learning_rate": 9.961358005815924e-05, + "loss": 2.3826, + "step": 14865 + }, + { + "epoch": 0.04, + "learning_rate": 9.961331863828616e-05, + "loss": 2.4156, + "step": 14870 + }, + { + "epoch": 0.04, + "learning_rate": 9.961305713035864e-05, + "loss": 2.4348, + "step": 14875 + }, + { + "epoch": 0.04, + "learning_rate": 9.961279553437722e-05, + "loss": 2.4336, + "step": 14880 + }, + { + "epoch": 0.04, + "learning_rate": 9.961253385034228e-05, + "loss": 2.4214, + "step": 14885 + }, + { + "epoch": 0.04, + "learning_rate": 9.961227207825434e-05, + "loss": 2.3832, + "step": 14890 + }, + { + "epoch": 0.04, + "learning_rate": 9.961201021811385e-05, + "loss": 2.4505, + "step": 14895 + }, + { + "epoch": 0.04, + "learning_rate": 9.961174826992128e-05, + "loss": 2.4605, + "step": 14900 + }, + { + "epoch": 0.04, + "learning_rate": 9.961148623367709e-05, + "loss": 2.4751, + "step": 14905 + }, + { + "epoch": 0.04, + "learning_rate": 9.961122410938173e-05, + "loss": 2.5535, + "step": 14910 + }, + { + "epoch": 0.04, + "learning_rate": 9.961096189703569e-05, + "loss": 2.3319, + "step": 14915 + }, + { + "epoch": 0.04, + "learning_rate": 9.961069959663943e-05, + "loss": 2.3426, + "step": 14920 + }, + { + "epoch": 0.04, + "learning_rate": 9.961043720819339e-05, + "loss": 2.4084, + "step": 14925 + }, + { + "epoch": 0.04, + "learning_rate": 9.961017473169807e-05, + "loss": 2.3987, + "step": 14930 + }, + { + "epoch": 0.04, + "learning_rate": 9.960991216715392e-05, + "loss": 2.3951, + "step": 14935 + }, + { + "epoch": 0.04, + "learning_rate": 9.960964951456143e-05, + "loss": 2.2484, + "step": 14940 + }, + { + "epoch": 0.04, + "learning_rate": 9.960938677392101e-05, + "loss": 2.4274, + "step": 14945 + }, + { + "epoch": 0.04, + "learning_rate": 9.960912394523317e-05, + "loss": 2.4307, + "step": 14950 + }, + { + "epoch": 0.04, + "learning_rate": 9.960886102849838e-05, + "loss": 2.437, + "step": 14955 + }, + { + "epoch": 0.04, + "learning_rate": 9.960859802371708e-05, + "loss": 2.3634, + "step": 14960 + }, + { + "epoch": 0.04, + "learning_rate": 9.960833493088974e-05, + "loss": 2.3541, + "step": 14965 + }, + { + "epoch": 0.04, + "learning_rate": 9.960807175001687e-05, + "loss": 2.4334, + "step": 14970 + }, + { + "epoch": 0.04, + "learning_rate": 9.960780848109889e-05, + "loss": 2.3931, + "step": 14975 + }, + { + "epoch": 0.04, + "learning_rate": 9.960754512413628e-05, + "loss": 2.4572, + "step": 14980 + }, + { + "epoch": 0.04, + "learning_rate": 9.96072816791295e-05, + "loss": 2.4712, + "step": 14985 + }, + { + "epoch": 0.04, + "learning_rate": 9.960701814607905e-05, + "loss": 2.3772, + "step": 14990 + }, + { + "epoch": 0.04, + "learning_rate": 9.960675452498535e-05, + "loss": 2.4522, + "step": 14995 + }, + { + "epoch": 0.04, + "learning_rate": 9.96064908158489e-05, + "loss": 2.3357, + "step": 15000 + }, + { + "epoch": 0.04, + "learning_rate": 9.960622701867016e-05, + "loss": 2.4017, + "step": 15005 + }, + { + "epoch": 0.04, + "learning_rate": 9.96059631334496e-05, + "loss": 2.3865, + "step": 15010 + }, + { + "epoch": 0.04, + "learning_rate": 9.960569916018769e-05, + "loss": 2.4009, + "step": 15015 + }, + { + "epoch": 0.04, + "learning_rate": 9.960543509888489e-05, + "loss": 2.4812, + "step": 15020 + }, + { + "epoch": 0.04, + "learning_rate": 9.960517094954166e-05, + "loss": 2.4356, + "step": 15025 + }, + { + "epoch": 0.04, + "learning_rate": 9.960490671215852e-05, + "loss": 2.41, + "step": 15030 + }, + { + "epoch": 0.04, + "learning_rate": 9.960464238673588e-05, + "loss": 2.3893, + "step": 15035 + }, + { + "epoch": 0.04, + "learning_rate": 9.960437797327424e-05, + "loss": 2.3472, + "step": 15040 + }, + { + "epoch": 0.04, + "learning_rate": 9.960411347177404e-05, + "loss": 2.4539, + "step": 15045 + }, + { + "epoch": 0.04, + "learning_rate": 9.960384888223579e-05, + "loss": 2.4734, + "step": 15050 + }, + { + "epoch": 0.04, + "learning_rate": 9.960358420465992e-05, + "loss": 2.4608, + "step": 15055 + }, + { + "epoch": 0.04, + "learning_rate": 9.960331943904694e-05, + "loss": 2.3494, + "step": 15060 + }, + { + "epoch": 0.04, + "learning_rate": 9.96030545853973e-05, + "loss": 2.4225, + "step": 15065 + }, + { + "epoch": 0.04, + "learning_rate": 9.960278964371145e-05, + "loss": 2.4369, + "step": 15070 + }, + { + "epoch": 0.04, + "learning_rate": 9.960252461398989e-05, + "loss": 2.5832, + "step": 15075 + }, + { + "epoch": 0.04, + "learning_rate": 9.960225949623308e-05, + "loss": 2.4307, + "step": 15080 + }, + { + "epoch": 0.04, + "learning_rate": 9.960199429044148e-05, + "loss": 2.5375, + "step": 15085 + }, + { + "epoch": 0.04, + "learning_rate": 9.960172899661558e-05, + "loss": 2.3475, + "step": 15090 + }, + { + "epoch": 0.04, + "learning_rate": 9.960146361475584e-05, + "loss": 2.5972, + "step": 15095 + }, + { + "epoch": 0.04, + "learning_rate": 9.960119814486272e-05, + "loss": 2.3441, + "step": 15100 + }, + { + "epoch": 0.04, + "learning_rate": 9.960093258693672e-05, + "loss": 2.3557, + "step": 15105 + }, + { + "epoch": 0.04, + "learning_rate": 9.96006669409783e-05, + "loss": 2.4371, + "step": 15110 + }, + { + "epoch": 0.04, + "learning_rate": 9.960040120698792e-05, + "loss": 2.4374, + "step": 15115 + }, + { + "epoch": 0.04, + "learning_rate": 9.960013538496604e-05, + "loss": 2.3653, + "step": 15120 + }, + { + "epoch": 0.04, + "learning_rate": 9.959986947491318e-05, + "loss": 2.4516, + "step": 15125 + }, + { + "epoch": 0.04, + "learning_rate": 9.959960347682977e-05, + "loss": 2.4644, + "step": 15130 + }, + { + "epoch": 0.04, + "learning_rate": 9.959933739071628e-05, + "loss": 2.3751, + "step": 15135 + }, + { + "epoch": 0.04, + "learning_rate": 9.959907121657321e-05, + "loss": 2.4023, + "step": 15140 + }, + { + "epoch": 0.04, + "learning_rate": 9.959880495440102e-05, + "loss": 2.4057, + "step": 15145 + }, + { + "epoch": 0.04, + "learning_rate": 9.959853860420018e-05, + "loss": 2.3435, + "step": 15150 + }, + { + "epoch": 0.04, + "learning_rate": 9.959827216597116e-05, + "loss": 2.3989, + "step": 15155 + }, + { + "epoch": 0.04, + "learning_rate": 9.959800563971445e-05, + "loss": 2.5335, + "step": 15160 + }, + { + "epoch": 0.04, + "learning_rate": 9.95977390254305e-05, + "loss": 2.4008, + "step": 15165 + }, + { + "epoch": 0.04, + "learning_rate": 9.959747232311977e-05, + "loss": 2.3043, + "step": 15170 + }, + { + "epoch": 0.04, + "learning_rate": 9.95972055327828e-05, + "loss": 2.4306, + "step": 15175 + }, + { + "epoch": 0.04, + "learning_rate": 9.959693865441999e-05, + "loss": 2.4455, + "step": 15180 + }, + { + "epoch": 0.04, + "learning_rate": 9.959667168803187e-05, + "loss": 2.4947, + "step": 15185 + }, + { + "epoch": 0.04, + "learning_rate": 9.959640463361886e-05, + "loss": 2.2681, + "step": 15190 + }, + { + "epoch": 0.04, + "learning_rate": 9.959613749118146e-05, + "loss": 2.3881, + "step": 15195 + }, + { + "epoch": 0.04, + "learning_rate": 9.959587026072017e-05, + "loss": 2.3813, + "step": 15200 + }, + { + "epoch": 0.04, + "learning_rate": 9.959560294223542e-05, + "loss": 2.4878, + "step": 15205 + }, + { + "epoch": 0.04, + "learning_rate": 9.959533553572771e-05, + "loss": 2.4052, + "step": 15210 + }, + { + "epoch": 0.04, + "learning_rate": 9.959506804119751e-05, + "loss": 2.4657, + "step": 15215 + }, + { + "epoch": 0.04, + "learning_rate": 9.95948004586453e-05, + "loss": 2.4142, + "step": 15220 + }, + { + "epoch": 0.04, + "learning_rate": 9.959453278807154e-05, + "loss": 2.4423, + "step": 15225 + }, + { + "epoch": 0.04, + "learning_rate": 9.959426502947671e-05, + "loss": 2.4678, + "step": 15230 + }, + { + "epoch": 0.04, + "learning_rate": 9.95939971828613e-05, + "loss": 2.3439, + "step": 15235 + }, + { + "epoch": 0.04, + "learning_rate": 9.959372924822576e-05, + "loss": 2.4069, + "step": 15240 + }, + { + "epoch": 0.04, + "learning_rate": 9.95934612255706e-05, + "loss": 2.41, + "step": 15245 + }, + { + "epoch": 0.04, + "learning_rate": 9.959319311489626e-05, + "loss": 2.3063, + "step": 15250 + }, + { + "epoch": 0.04, + "learning_rate": 9.959292491620324e-05, + "loss": 2.3927, + "step": 15255 + }, + { + "epoch": 0.04, + "learning_rate": 9.9592656629492e-05, + "loss": 2.3238, + "step": 15260 + }, + { + "epoch": 0.04, + "learning_rate": 9.959238825476302e-05, + "loss": 2.4396, + "step": 15265 + }, + { + "epoch": 0.04, + "learning_rate": 9.959211979201679e-05, + "loss": 2.4433, + "step": 15270 + }, + { + "epoch": 0.04, + "learning_rate": 9.959185124125377e-05, + "loss": 2.4468, + "step": 15275 + }, + { + "epoch": 0.04, + "learning_rate": 9.959158260247446e-05, + "loss": 2.4772, + "step": 15280 + }, + { + "epoch": 0.04, + "learning_rate": 9.959131387567929e-05, + "loss": 2.4971, + "step": 15285 + }, + { + "epoch": 0.04, + "learning_rate": 9.959104506086878e-05, + "loss": 2.4234, + "step": 15290 + }, + { + "epoch": 0.04, + "learning_rate": 9.95907761580434e-05, + "loss": 2.4669, + "step": 15295 + }, + { + "epoch": 0.04, + "learning_rate": 9.959050716720362e-05, + "loss": 2.4606, + "step": 15300 + }, + { + "epoch": 0.04, + "learning_rate": 9.959023808834991e-05, + "loss": 2.497, + "step": 15305 + }, + { + "epoch": 0.04, + "learning_rate": 9.958996892148276e-05, + "loss": 2.2454, + "step": 15310 + }, + { + "epoch": 0.04, + "learning_rate": 9.958969966660265e-05, + "loss": 2.3747, + "step": 15315 + }, + { + "epoch": 0.04, + "learning_rate": 9.958943032371004e-05, + "loss": 2.4751, + "step": 15320 + }, + { + "epoch": 0.04, + "learning_rate": 9.958916089280542e-05, + "loss": 2.5435, + "step": 15325 + }, + { + "epoch": 0.04, + "learning_rate": 9.958889137388927e-05, + "loss": 2.4097, + "step": 15330 + }, + { + "epoch": 0.04, + "learning_rate": 9.958862176696206e-05, + "loss": 2.3647, + "step": 15335 + }, + { + "epoch": 0.04, + "learning_rate": 9.95883520720243e-05, + "loss": 2.4289, + "step": 15340 + }, + { + "epoch": 0.04, + "learning_rate": 9.958808228907641e-05, + "loss": 2.4545, + "step": 15345 + }, + { + "epoch": 0.04, + "learning_rate": 9.958781241811892e-05, + "loss": 2.4483, + "step": 15350 + }, + { + "epoch": 0.04, + "learning_rate": 9.95875424591523e-05, + "loss": 2.5652, + "step": 15355 + }, + { + "epoch": 0.04, + "learning_rate": 9.9587272412177e-05, + "loss": 2.3415, + "step": 15360 + }, + { + "epoch": 0.04, + "learning_rate": 9.958700227719354e-05, + "loss": 2.3435, + "step": 15365 + }, + { + "epoch": 0.04, + "learning_rate": 9.958673205420236e-05, + "loss": 2.4363, + "step": 15370 + }, + { + "epoch": 0.04, + "learning_rate": 9.958646174320397e-05, + "loss": 2.479, + "step": 15375 + }, + { + "epoch": 0.04, + "learning_rate": 9.958619134419883e-05, + "loss": 2.3812, + "step": 15380 + }, + { + "epoch": 0.04, + "learning_rate": 9.958592085718744e-05, + "loss": 2.3862, + "step": 15385 + }, + { + "epoch": 0.04, + "learning_rate": 9.958565028217025e-05, + "loss": 2.4482, + "step": 15390 + }, + { + "epoch": 0.04, + "learning_rate": 9.958537961914778e-05, + "loss": 2.4404, + "step": 15395 + }, + { + "epoch": 0.04, + "learning_rate": 9.958510886812047e-05, + "loss": 2.4652, + "step": 15400 + }, + { + "epoch": 0.04, + "learning_rate": 9.958483802908882e-05, + "loss": 2.5212, + "step": 15405 + }, + { + "epoch": 0.04, + "learning_rate": 9.958456710205333e-05, + "loss": 2.4323, + "step": 15410 + }, + { + "epoch": 0.04, + "learning_rate": 9.958429608701444e-05, + "loss": 2.4477, + "step": 15415 + }, + { + "epoch": 0.04, + "learning_rate": 9.958402498397266e-05, + "loss": 2.4568, + "step": 15420 + }, + { + "epoch": 0.04, + "learning_rate": 9.958375379292845e-05, + "loss": 2.3377, + "step": 15425 + }, + { + "epoch": 0.04, + "learning_rate": 9.958348251388232e-05, + "loss": 2.4048, + "step": 15430 + }, + { + "epoch": 0.04, + "learning_rate": 9.958321114683474e-05, + "loss": 2.4474, + "step": 15435 + }, + { + "epoch": 0.04, + "learning_rate": 9.958293969178619e-05, + "loss": 2.3564, + "step": 15440 + }, + { + "epoch": 0.04, + "learning_rate": 9.958266814873714e-05, + "loss": 2.3788, + "step": 15445 + }, + { + "epoch": 0.04, + "learning_rate": 9.958239651768806e-05, + "loss": 2.3806, + "step": 15450 + }, + { + "epoch": 0.04, + "learning_rate": 9.958212479863949e-05, + "loss": 2.3157, + "step": 15455 + }, + { + "epoch": 0.04, + "learning_rate": 9.958185299159185e-05, + "loss": 2.4401, + "step": 15460 + }, + { + "epoch": 0.04, + "learning_rate": 9.958158109654565e-05, + "loss": 2.3908, + "step": 15465 + }, + { + "epoch": 0.04, + "learning_rate": 9.958130911350139e-05, + "loss": 2.4014, + "step": 15470 + }, + { + "epoch": 0.04, + "learning_rate": 9.958103704245951e-05, + "loss": 2.4404, + "step": 15475 + }, + { + "epoch": 0.04, + "learning_rate": 9.958076488342054e-05, + "loss": 2.4492, + "step": 15480 + }, + { + "epoch": 0.04, + "learning_rate": 9.958049263638491e-05, + "loss": 2.4339, + "step": 15485 + }, + { + "epoch": 0.04, + "learning_rate": 9.958022030135316e-05, + "loss": 2.3219, + "step": 15490 + }, + { + "epoch": 0.04, + "learning_rate": 9.957994787832573e-05, + "loss": 2.3326, + "step": 15495 + }, + { + "epoch": 0.04, + "learning_rate": 9.957967536730313e-05, + "loss": 2.439, + "step": 15500 + }, + { + "epoch": 0.04, + "learning_rate": 9.957940276828581e-05, + "loss": 2.4194, + "step": 15505 + }, + { + "epoch": 0.04, + "learning_rate": 9.95791300812743e-05, + "loss": 2.465, + "step": 15510 + }, + { + "epoch": 0.04, + "learning_rate": 9.957885730626905e-05, + "loss": 2.3268, + "step": 15515 + }, + { + "epoch": 0.04, + "learning_rate": 9.957858444327055e-05, + "loss": 2.4943, + "step": 15520 + }, + { + "epoch": 0.04, + "learning_rate": 9.957831149227929e-05, + "loss": 2.4771, + "step": 15525 + }, + { + "epoch": 0.04, + "learning_rate": 9.957803845329578e-05, + "loss": 2.3819, + "step": 15530 + }, + { + "epoch": 0.04, + "learning_rate": 9.957776532632045e-05, + "loss": 2.469, + "step": 15535 + }, + { + "epoch": 0.04, + "learning_rate": 9.957749211135382e-05, + "loss": 2.415, + "step": 15540 + }, + { + "epoch": 0.04, + "learning_rate": 9.957721880839637e-05, + "loss": 2.4315, + "step": 15545 + }, + { + "epoch": 0.04, + "learning_rate": 9.957694541744858e-05, + "loss": 2.4069, + "step": 15550 + }, + { + "epoch": 0.04, + "learning_rate": 9.957667193851093e-05, + "loss": 2.4881, + "step": 15555 + }, + { + "epoch": 0.04, + "learning_rate": 9.957639837158393e-05, + "loss": 2.4414, + "step": 15560 + }, + { + "epoch": 0.04, + "learning_rate": 9.957612471666803e-05, + "loss": 2.4406, + "step": 15565 + }, + { + "epoch": 0.04, + "learning_rate": 9.957585097376375e-05, + "loss": 2.4614, + "step": 15570 + }, + { + "epoch": 0.04, + "learning_rate": 9.957557714287156e-05, + "loss": 2.4368, + "step": 15575 + }, + { + "epoch": 0.04, + "learning_rate": 9.957530322399193e-05, + "loss": 2.2788, + "step": 15580 + }, + { + "epoch": 0.04, + "learning_rate": 9.957502921712538e-05, + "loss": 2.3992, + "step": 15585 + }, + { + "epoch": 0.04, + "learning_rate": 9.957475512227236e-05, + "loss": 2.3301, + "step": 15590 + }, + { + "epoch": 0.04, + "learning_rate": 9.95744809394334e-05, + "loss": 2.4622, + "step": 15595 + }, + { + "epoch": 0.04, + "learning_rate": 9.957420666860895e-05, + "loss": 2.4575, + "step": 15600 + }, + { + "epoch": 0.04, + "learning_rate": 9.957393230979952e-05, + "loss": 2.3855, + "step": 15605 + }, + { + "epoch": 0.04, + "learning_rate": 9.957365786300557e-05, + "loss": 2.4342, + "step": 15610 + }, + { + "epoch": 0.04, + "learning_rate": 9.95733833282276e-05, + "loss": 2.4232, + "step": 15615 + }, + { + "epoch": 0.04, + "learning_rate": 9.95731087054661e-05, + "loss": 2.5029, + "step": 15620 + }, + { + "epoch": 0.04, + "learning_rate": 9.957283399472156e-05, + "loss": 2.4403, + "step": 15625 + }, + { + "epoch": 0.04, + "learning_rate": 9.957255919599448e-05, + "loss": 2.4816, + "step": 15630 + }, + { + "epoch": 0.04, + "learning_rate": 9.957228430928533e-05, + "loss": 2.5397, + "step": 15635 + }, + { + "epoch": 0.04, + "learning_rate": 9.957200933459458e-05, + "loss": 2.3216, + "step": 15640 + }, + { + "epoch": 0.04, + "learning_rate": 9.957173427192275e-05, + "loss": 2.433, + "step": 15645 + }, + { + "epoch": 0.04, + "learning_rate": 9.957145912127033e-05, + "loss": 2.5144, + "step": 15650 + }, + { + "epoch": 0.04, + "learning_rate": 9.957118388263778e-05, + "loss": 2.369, + "step": 15655 + }, + { + "epoch": 0.04, + "learning_rate": 9.95709085560256e-05, + "loss": 2.3527, + "step": 15660 + }, + { + "epoch": 0.04, + "learning_rate": 9.957063314143428e-05, + "loss": 2.4044, + "step": 15665 + }, + { + "epoch": 0.04, + "learning_rate": 9.957035763886433e-05, + "loss": 2.4504, + "step": 15670 + }, + { + "epoch": 0.04, + "learning_rate": 9.957008204831622e-05, + "loss": 2.3499, + "step": 15675 + }, + { + "epoch": 0.04, + "learning_rate": 9.956980636979041e-05, + "loss": 2.4292, + "step": 15680 + }, + { + "epoch": 0.04, + "learning_rate": 9.956953060328744e-05, + "loss": 2.4347, + "step": 15685 + }, + { + "epoch": 0.04, + "learning_rate": 9.956925474880778e-05, + "loss": 2.5549, + "step": 15690 + }, + { + "epoch": 0.04, + "learning_rate": 9.956897880635191e-05, + "loss": 2.2784, + "step": 15695 + }, + { + "epoch": 0.04, + "learning_rate": 9.956870277592034e-05, + "loss": 2.2873, + "step": 15700 + }, + { + "epoch": 0.04, + "learning_rate": 9.956842665751352e-05, + "loss": 2.3916, + "step": 15705 + }, + { + "epoch": 0.04, + "learning_rate": 9.956815045113199e-05, + "loss": 2.3634, + "step": 15710 + }, + { + "epoch": 0.04, + "learning_rate": 9.956787415677621e-05, + "loss": 2.5076, + "step": 15715 + }, + { + "epoch": 0.04, + "learning_rate": 9.956759777444667e-05, + "loss": 2.4272, + "step": 15720 + }, + { + "epoch": 0.04, + "learning_rate": 9.956732130414389e-05, + "loss": 2.4206, + "step": 15725 + }, + { + "epoch": 0.04, + "learning_rate": 9.956704474586832e-05, + "loss": 2.5201, + "step": 15730 + }, + { + "epoch": 0.04, + "learning_rate": 9.956676809962046e-05, + "loss": 2.3536, + "step": 15735 + }, + { + "epoch": 0.04, + "learning_rate": 9.956649136540084e-05, + "loss": 2.4565, + "step": 15740 + }, + { + "epoch": 0.04, + "learning_rate": 9.956621454320991e-05, + "loss": 2.3465, + "step": 15745 + }, + { + "epoch": 0.04, + "learning_rate": 9.956593763304816e-05, + "loss": 2.3438, + "step": 15750 + }, + { + "epoch": 0.04, + "learning_rate": 9.956566063491612e-05, + "loss": 2.5235, + "step": 15755 + }, + { + "epoch": 0.04, + "learning_rate": 9.956538354881424e-05, + "loss": 2.4308, + "step": 15760 + }, + { + "epoch": 0.04, + "learning_rate": 9.956510637474304e-05, + "loss": 2.3267, + "step": 15765 + }, + { + "epoch": 0.04, + "learning_rate": 9.956482911270297e-05, + "loss": 2.4128, + "step": 15770 + }, + { + "epoch": 0.04, + "learning_rate": 9.956455176269459e-05, + "loss": 2.4355, + "step": 15775 + }, + { + "epoch": 0.04, + "learning_rate": 9.956427432471833e-05, + "loss": 2.3365, + "step": 15780 + }, + { + "epoch": 0.04, + "learning_rate": 9.95639967987747e-05, + "loss": 2.4141, + "step": 15785 + }, + { + "epoch": 0.04, + "learning_rate": 9.956371918486422e-05, + "loss": 2.4169, + "step": 15790 + }, + { + "epoch": 0.04, + "learning_rate": 9.956344148298736e-05, + "loss": 2.3838, + "step": 15795 + }, + { + "epoch": 0.04, + "learning_rate": 9.95631636931446e-05, + "loss": 2.4971, + "step": 15800 + }, + { + "epoch": 0.04, + "learning_rate": 9.956288581533645e-05, + "loss": 2.2822, + "step": 15805 + }, + { + "epoch": 0.04, + "learning_rate": 9.956260784956341e-05, + "loss": 2.364, + "step": 15810 + }, + { + "epoch": 0.04, + "learning_rate": 9.956232979582598e-05, + "loss": 2.4486, + "step": 15815 + }, + { + "epoch": 0.04, + "learning_rate": 9.956205165412463e-05, + "loss": 2.3413, + "step": 15820 + }, + { + "epoch": 0.04, + "learning_rate": 9.956177342445983e-05, + "loss": 2.3289, + "step": 15825 + }, + { + "epoch": 0.04, + "learning_rate": 9.956149510683214e-05, + "loss": 2.2923, + "step": 15830 + }, + { + "epoch": 0.04, + "learning_rate": 9.956121670124201e-05, + "loss": 2.4738, + "step": 15835 + }, + { + "epoch": 0.04, + "learning_rate": 9.956093820768994e-05, + "loss": 2.3427, + "step": 15840 + }, + { + "epoch": 0.04, + "learning_rate": 9.956065962617642e-05, + "loss": 2.45, + "step": 15845 + }, + { + "epoch": 0.04, + "learning_rate": 9.956038095670196e-05, + "loss": 2.4636, + "step": 15850 + }, + { + "epoch": 0.04, + "learning_rate": 9.956010219926705e-05, + "loss": 2.4012, + "step": 15855 + }, + { + "epoch": 0.04, + "learning_rate": 9.955982335387217e-05, + "loss": 2.4408, + "step": 15860 + }, + { + "epoch": 0.04, + "learning_rate": 9.955954442051785e-05, + "loss": 2.2955, + "step": 15865 + }, + { + "epoch": 0.04, + "learning_rate": 9.955926539920454e-05, + "loss": 2.3822, + "step": 15870 + }, + { + "epoch": 0.04, + "learning_rate": 9.955898628993277e-05, + "loss": 2.4664, + "step": 15875 + }, + { + "epoch": 0.04, + "learning_rate": 9.9558707092703e-05, + "loss": 2.301, + "step": 15880 + }, + { + "epoch": 0.04, + "learning_rate": 9.955842780751577e-05, + "loss": 2.3435, + "step": 15885 + }, + { + "epoch": 0.04, + "learning_rate": 9.955814843437155e-05, + "loss": 2.3413, + "step": 15890 + }, + { + "epoch": 0.04, + "learning_rate": 9.955786897327083e-05, + "loss": 2.5438, + "step": 15895 + }, + { + "epoch": 0.04, + "learning_rate": 9.955758942421412e-05, + "loss": 2.2569, + "step": 15900 + }, + { + "epoch": 0.04, + "learning_rate": 9.955730978720191e-05, + "loss": 2.4442, + "step": 15905 + }, + { + "epoch": 0.04, + "learning_rate": 9.955703006223471e-05, + "loss": 2.6238, + "step": 15910 + }, + { + "epoch": 0.04, + "learning_rate": 9.955675024931299e-05, + "loss": 2.4228, + "step": 15915 + }, + { + "epoch": 0.04, + "learning_rate": 9.955647034843726e-05, + "loss": 2.2595, + "step": 15920 + }, + { + "epoch": 0.04, + "learning_rate": 9.955619035960802e-05, + "loss": 2.3281, + "step": 15925 + }, + { + "epoch": 0.04, + "learning_rate": 9.955591028282577e-05, + "loss": 2.3724, + "step": 15930 + }, + { + "epoch": 0.04, + "learning_rate": 9.9555630118091e-05, + "loss": 2.5657, + "step": 15935 + }, + { + "epoch": 0.04, + "learning_rate": 9.955534986540419e-05, + "loss": 2.3564, + "step": 15940 + }, + { + "epoch": 0.04, + "learning_rate": 9.955506952476587e-05, + "loss": 2.459, + "step": 15945 + }, + { + "epoch": 0.04, + "learning_rate": 9.955478909617652e-05, + "loss": 2.5006, + "step": 15950 + }, + { + "epoch": 0.04, + "learning_rate": 9.955450857963665e-05, + "loss": 2.3799, + "step": 15955 + }, + { + "epoch": 0.04, + "learning_rate": 9.955422797514675e-05, + "loss": 2.3376, + "step": 15960 + }, + { + "epoch": 0.04, + "learning_rate": 9.95539472827073e-05, + "loss": 2.433, + "step": 15965 + }, + { + "epoch": 0.04, + "learning_rate": 9.955366650231883e-05, + "loss": 2.3499, + "step": 15970 + }, + { + "epoch": 0.04, + "learning_rate": 9.955338563398182e-05, + "loss": 2.3335, + "step": 15975 + }, + { + "epoch": 0.04, + "learning_rate": 9.955310467769678e-05, + "loss": 2.4371, + "step": 15980 + }, + { + "epoch": 0.04, + "learning_rate": 9.955282363346419e-05, + "loss": 2.3893, + "step": 15985 + }, + { + "epoch": 0.04, + "learning_rate": 9.955254250128456e-05, + "loss": 2.3944, + "step": 15990 + }, + { + "epoch": 0.04, + "learning_rate": 9.95522612811584e-05, + "loss": 2.4308, + "step": 15995 + }, + { + "epoch": 0.04, + "learning_rate": 9.955197997308618e-05, + "loss": 2.5478, + "step": 16000 + }, + { + "epoch": 0.04, + "learning_rate": 9.955169857706843e-05, + "loss": 2.4469, + "step": 16005 + }, + { + "epoch": 0.04, + "learning_rate": 9.955141709310564e-05, + "loss": 2.4419, + "step": 16010 + }, + { + "epoch": 0.04, + "learning_rate": 9.95511355211983e-05, + "loss": 2.5621, + "step": 16015 + }, + { + "epoch": 0.04, + "learning_rate": 9.955085386134691e-05, + "loss": 2.4585, + "step": 16020 + }, + { + "epoch": 0.04, + "learning_rate": 9.955057211355199e-05, + "loss": 2.3729, + "step": 16025 + }, + { + "epoch": 0.04, + "learning_rate": 9.9550290277814e-05, + "loss": 2.3361, + "step": 16030 + }, + { + "epoch": 0.04, + "learning_rate": 9.955000835413349e-05, + "loss": 2.4185, + "step": 16035 + }, + { + "epoch": 0.04, + "learning_rate": 9.954972634251092e-05, + "loss": 2.3419, + "step": 16040 + }, + { + "epoch": 0.04, + "learning_rate": 9.954944424294682e-05, + "loss": 2.3891, + "step": 16045 + }, + { + "epoch": 0.04, + "learning_rate": 9.954916205544168e-05, + "loss": 2.4023, + "step": 16050 + }, + { + "epoch": 0.04, + "learning_rate": 9.954887977999598e-05, + "loss": 2.3507, + "step": 16055 + }, + { + "epoch": 0.04, + "learning_rate": 9.954859741661024e-05, + "loss": 2.2414, + "step": 16060 + }, + { + "epoch": 0.04, + "learning_rate": 9.954831496528499e-05, + "loss": 2.4957, + "step": 16065 + }, + { + "epoch": 0.04, + "learning_rate": 9.954803242602067e-05, + "loss": 2.3959, + "step": 16070 + }, + { + "epoch": 0.04, + "learning_rate": 9.954774979881783e-05, + "loss": 2.374, + "step": 16075 + }, + { + "epoch": 0.04, + "learning_rate": 9.954746708367695e-05, + "loss": 2.5393, + "step": 16080 + }, + { + "epoch": 0.04, + "learning_rate": 9.954718428059853e-05, + "loss": 2.2671, + "step": 16085 + }, + { + "epoch": 0.04, + "learning_rate": 9.954690138958308e-05, + "loss": 2.4342, + "step": 16090 + }, + { + "epoch": 0.04, + "learning_rate": 9.95466184106311e-05, + "loss": 2.4716, + "step": 16095 + }, + { + "epoch": 0.04, + "learning_rate": 9.95463353437431e-05, + "loss": 2.4383, + "step": 16100 + }, + { + "epoch": 0.04, + "learning_rate": 9.954605218891957e-05, + "loss": 2.4708, + "step": 16105 + }, + { + "epoch": 0.04, + "learning_rate": 9.954576894616103e-05, + "loss": 2.4748, + "step": 16110 + }, + { + "epoch": 0.04, + "learning_rate": 9.954548561546795e-05, + "loss": 2.4195, + "step": 16115 + }, + { + "epoch": 0.04, + "learning_rate": 9.954520219684086e-05, + "loss": 2.3715, + "step": 16120 + }, + { + "epoch": 0.04, + "learning_rate": 9.954491869028027e-05, + "loss": 2.3265, + "step": 16125 + }, + { + "epoch": 0.04, + "learning_rate": 9.954463509578665e-05, + "loss": 2.4003, + "step": 16130 + }, + { + "epoch": 0.04, + "learning_rate": 9.954435141336054e-05, + "loss": 2.4165, + "step": 16135 + }, + { + "epoch": 0.04, + "learning_rate": 9.954406764300241e-05, + "loss": 2.488, + "step": 16140 + }, + { + "epoch": 0.04, + "learning_rate": 9.954378378471278e-05, + "loss": 2.3671, + "step": 16145 + }, + { + "epoch": 0.04, + "learning_rate": 9.954349983849218e-05, + "loss": 2.5331, + "step": 16150 + }, + { + "epoch": 0.04, + "learning_rate": 9.954321580434105e-05, + "loss": 2.367, + "step": 16155 + }, + { + "epoch": 0.04, + "learning_rate": 9.954293168225996e-05, + "loss": 2.5488, + "step": 16160 + }, + { + "epoch": 0.04, + "learning_rate": 9.954264747224938e-05, + "loss": 2.2094, + "step": 16165 + }, + { + "epoch": 0.04, + "learning_rate": 9.95423631743098e-05, + "loss": 2.4081, + "step": 16170 + }, + { + "epoch": 0.04, + "learning_rate": 9.954207878844177e-05, + "loss": 2.5121, + "step": 16175 + }, + { + "epoch": 0.04, + "learning_rate": 9.954179431464577e-05, + "loss": 2.4163, + "step": 16180 + }, + { + "epoch": 0.04, + "learning_rate": 9.954150975292229e-05, + "loss": 2.3544, + "step": 16185 + }, + { + "epoch": 0.04, + "learning_rate": 9.954122510327185e-05, + "loss": 2.4006, + "step": 16190 + }, + { + "epoch": 0.04, + "learning_rate": 9.954094036569497e-05, + "loss": 2.4179, + "step": 16195 + }, + { + "epoch": 0.04, + "learning_rate": 9.954065554019212e-05, + "loss": 2.4721, + "step": 16200 + }, + { + "epoch": 0.04, + "learning_rate": 9.954037062676384e-05, + "loss": 2.3969, + "step": 16205 + }, + { + "epoch": 0.04, + "learning_rate": 9.954008562541062e-05, + "loss": 2.4202, + "step": 16210 + }, + { + "epoch": 0.04, + "learning_rate": 9.953980053613297e-05, + "loss": 2.5058, + "step": 16215 + }, + { + "epoch": 0.04, + "learning_rate": 9.953951535893138e-05, + "loss": 2.3865, + "step": 16220 + }, + { + "epoch": 0.04, + "learning_rate": 9.953923009380638e-05, + "loss": 2.3818, + "step": 16225 + }, + { + "epoch": 0.04, + "learning_rate": 9.953894474075846e-05, + "loss": 2.4681, + "step": 16230 + }, + { + "epoch": 0.04, + "learning_rate": 9.953865929978814e-05, + "loss": 2.3231, + "step": 16235 + }, + { + "epoch": 0.04, + "learning_rate": 9.953837377089592e-05, + "loss": 2.3606, + "step": 16240 + }, + { + "epoch": 0.04, + "learning_rate": 9.953808815408228e-05, + "loss": 2.3341, + "step": 16245 + }, + { + "epoch": 0.04, + "learning_rate": 9.953780244934778e-05, + "loss": 2.3056, + "step": 16250 + }, + { + "epoch": 0.04, + "learning_rate": 9.953751665669288e-05, + "loss": 2.3494, + "step": 16255 + }, + { + "epoch": 0.04, + "learning_rate": 9.953723077611812e-05, + "loss": 2.4723, + "step": 16260 + }, + { + "epoch": 0.04, + "learning_rate": 9.953694480762399e-05, + "loss": 2.3713, + "step": 16265 + }, + { + "epoch": 0.04, + "learning_rate": 9.953665875121101e-05, + "loss": 2.4346, + "step": 16270 + }, + { + "epoch": 0.04, + "learning_rate": 9.953637260687967e-05, + "loss": 2.323, + "step": 16275 + }, + { + "epoch": 0.04, + "learning_rate": 9.953608637463049e-05, + "loss": 2.4094, + "step": 16280 + }, + { + "epoch": 0.04, + "learning_rate": 9.953580005446396e-05, + "loss": 2.3991, + "step": 16285 + }, + { + "epoch": 0.04, + "learning_rate": 9.953551364638063e-05, + "loss": 2.4092, + "step": 16290 + }, + { + "epoch": 0.04, + "learning_rate": 9.953522715038097e-05, + "loss": 2.3667, + "step": 16295 + }, + { + "epoch": 0.04, + "learning_rate": 9.953494056646549e-05, + "loss": 2.3331, + "step": 16300 + }, + { + "epoch": 0.04, + "learning_rate": 9.953465389463473e-05, + "loss": 2.3309, + "step": 16305 + }, + { + "epoch": 0.04, + "learning_rate": 9.953436713488917e-05, + "loss": 2.5188, + "step": 16310 + }, + { + "epoch": 0.04, + "learning_rate": 9.953408028722931e-05, + "loss": 2.5383, + "step": 16315 + }, + { + "epoch": 0.04, + "learning_rate": 9.953379335165569e-05, + "loss": 2.3738, + "step": 16320 + }, + { + "epoch": 0.04, + "learning_rate": 9.95335063281688e-05, + "loss": 2.4612, + "step": 16325 + }, + { + "epoch": 0.04, + "learning_rate": 9.953321921676915e-05, + "loss": 2.4696, + "step": 16330 + }, + { + "epoch": 0.04, + "learning_rate": 9.953293201745725e-05, + "loss": 2.4629, + "step": 16335 + }, + { + "epoch": 0.04, + "learning_rate": 9.953264473023363e-05, + "loss": 2.2635, + "step": 16340 + }, + { + "epoch": 0.04, + "learning_rate": 9.953235735509878e-05, + "loss": 2.4255, + "step": 16345 + }, + { + "epoch": 0.04, + "learning_rate": 9.953206989205321e-05, + "loss": 2.3857, + "step": 16350 + }, + { + "epoch": 0.04, + "learning_rate": 9.953178234109743e-05, + "loss": 2.383, + "step": 16355 + }, + { + "epoch": 0.04, + "learning_rate": 9.953149470223197e-05, + "loss": 2.3532, + "step": 16360 + }, + { + "epoch": 0.04, + "learning_rate": 9.95312069754573e-05, + "loss": 2.4983, + "step": 16365 + }, + { + "epoch": 0.04, + "learning_rate": 9.953091916077397e-05, + "loss": 2.4435, + "step": 16370 + }, + { + "epoch": 0.04, + "learning_rate": 9.953063125818248e-05, + "loss": 2.3565, + "step": 16375 + }, + { + "epoch": 0.04, + "learning_rate": 9.953034326768334e-05, + "loss": 2.449, + "step": 16380 + }, + { + "epoch": 0.04, + "learning_rate": 9.953005518927703e-05, + "loss": 2.5514, + "step": 16385 + }, + { + "epoch": 0.04, + "learning_rate": 9.952976702296412e-05, + "loss": 2.4719, + "step": 16390 + }, + { + "epoch": 0.04, + "learning_rate": 9.952947876874508e-05, + "loss": 2.3794, + "step": 16395 + }, + { + "epoch": 0.04, + "learning_rate": 9.952919042662044e-05, + "loss": 2.4797, + "step": 16400 + }, + { + "epoch": 0.04, + "learning_rate": 9.952890199659069e-05, + "loss": 2.3918, + "step": 16405 + }, + { + "epoch": 0.04, + "learning_rate": 9.952861347865636e-05, + "loss": 2.4095, + "step": 16410 + }, + { + "epoch": 0.04, + "learning_rate": 9.952832487281795e-05, + "loss": 2.5049, + "step": 16415 + }, + { + "epoch": 0.04, + "learning_rate": 9.952803617907598e-05, + "loss": 2.4871, + "step": 16420 + }, + { + "epoch": 0.04, + "learning_rate": 9.952774739743098e-05, + "loss": 2.3785, + "step": 16425 + }, + { + "epoch": 0.04, + "learning_rate": 9.952745852788344e-05, + "loss": 2.3448, + "step": 16430 + }, + { + "epoch": 0.04, + "learning_rate": 9.952716957043388e-05, + "loss": 2.3401, + "step": 16435 + }, + { + "epoch": 0.04, + "learning_rate": 9.95268805250828e-05, + "loss": 2.4368, + "step": 16440 + }, + { + "epoch": 0.04, + "learning_rate": 9.952659139183074e-05, + "loss": 2.3966, + "step": 16445 + }, + { + "epoch": 0.04, + "learning_rate": 9.952630217067818e-05, + "loss": 2.4107, + "step": 16450 + }, + { + "epoch": 0.04, + "learning_rate": 9.952601286162565e-05, + "loss": 2.2579, + "step": 16455 + }, + { + "epoch": 0.04, + "learning_rate": 9.952572346467369e-05, + "loss": 2.3653, + "step": 16460 + }, + { + "epoch": 0.04, + "learning_rate": 9.952543397982275e-05, + "loss": 2.4081, + "step": 16465 + }, + { + "epoch": 0.04, + "learning_rate": 9.95251444070734e-05, + "loss": 2.5255, + "step": 16470 + }, + { + "epoch": 0.04, + "learning_rate": 9.952485474642614e-05, + "loss": 2.4276, + "step": 16475 + }, + { + "epoch": 0.04, + "learning_rate": 9.952456499788148e-05, + "loss": 2.4789, + "step": 16480 + }, + { + "epoch": 0.04, + "learning_rate": 9.952427516143992e-05, + "loss": 2.5249, + "step": 16485 + }, + { + "epoch": 0.04, + "learning_rate": 9.952398523710199e-05, + "loss": 2.6285, + "step": 16490 + }, + { + "epoch": 0.04, + "learning_rate": 9.952369522486821e-05, + "loss": 2.3599, + "step": 16495 + }, + { + "epoch": 0.04, + "learning_rate": 9.952340512473908e-05, + "loss": 2.3207, + "step": 16500 + }, + { + "epoch": 0.04, + "learning_rate": 9.952311493671512e-05, + "loss": 2.5273, + "step": 16505 + }, + { + "epoch": 0.04, + "learning_rate": 9.952282466079686e-05, + "loss": 2.2995, + "step": 16510 + }, + { + "epoch": 0.04, + "learning_rate": 9.952253429698479e-05, + "loss": 2.2392, + "step": 16515 + }, + { + "epoch": 0.04, + "learning_rate": 9.952224384527944e-05, + "loss": 2.3767, + "step": 16520 + }, + { + "epoch": 0.04, + "learning_rate": 9.952195330568133e-05, + "loss": 2.4462, + "step": 16525 + }, + { + "epoch": 0.04, + "learning_rate": 9.952166267819097e-05, + "loss": 2.549, + "step": 16530 + }, + { + "epoch": 0.04, + "learning_rate": 9.952137196280886e-05, + "loss": 2.3829, + "step": 16535 + }, + { + "epoch": 0.04, + "learning_rate": 9.952108115953555e-05, + "loss": 2.3631, + "step": 16540 + }, + { + "epoch": 0.04, + "learning_rate": 9.952079026837152e-05, + "loss": 2.4992, + "step": 16545 + }, + { + "epoch": 0.04, + "learning_rate": 9.95204992893173e-05, + "loss": 2.5208, + "step": 16550 + }, + { + "epoch": 0.04, + "learning_rate": 9.952020822237342e-05, + "loss": 2.4687, + "step": 16555 + }, + { + "epoch": 0.04, + "learning_rate": 9.951991706754039e-05, + "loss": 2.4295, + "step": 16560 + }, + { + "epoch": 0.04, + "learning_rate": 9.95196258248187e-05, + "loss": 2.5275, + "step": 16565 + }, + { + "epoch": 0.04, + "learning_rate": 9.951933449420891e-05, + "loss": 2.4752, + "step": 16570 + }, + { + "epoch": 0.04, + "learning_rate": 9.951904307571152e-05, + "loss": 2.3974, + "step": 16575 + }, + { + "epoch": 0.04, + "learning_rate": 9.951875156932704e-05, + "loss": 2.5591, + "step": 16580 + }, + { + "epoch": 0.04, + "learning_rate": 9.951845997505599e-05, + "loss": 2.4223, + "step": 16585 + }, + { + "epoch": 0.04, + "learning_rate": 9.951816829289888e-05, + "loss": 2.3084, + "step": 16590 + }, + { + "epoch": 0.04, + "learning_rate": 9.951787652285624e-05, + "loss": 2.4311, + "step": 16595 + }, + { + "epoch": 0.04, + "learning_rate": 9.95175846649286e-05, + "loss": 2.373, + "step": 16600 + }, + { + "epoch": 0.04, + "learning_rate": 9.951729271911645e-05, + "loss": 2.3906, + "step": 16605 + }, + { + "epoch": 0.04, + "learning_rate": 9.951700068542031e-05, + "loss": 2.3, + "step": 16610 + }, + { + "epoch": 0.04, + "learning_rate": 9.951670856384073e-05, + "loss": 2.5279, + "step": 16615 + }, + { + "epoch": 0.04, + "learning_rate": 9.951641635437818e-05, + "loss": 2.4593, + "step": 16620 + }, + { + "epoch": 0.04, + "learning_rate": 9.951612405703323e-05, + "loss": 2.2768, + "step": 16625 + }, + { + "epoch": 0.04, + "learning_rate": 9.951583167180636e-05, + "loss": 2.5326, + "step": 16630 + }, + { + "epoch": 0.04, + "learning_rate": 9.95155391986981e-05, + "loss": 2.479, + "step": 16635 + }, + { + "epoch": 0.04, + "learning_rate": 9.951524663770899e-05, + "loss": 2.3952, + "step": 16640 + }, + { + "epoch": 0.04, + "learning_rate": 9.951495398883953e-05, + "loss": 2.4937, + "step": 16645 + }, + { + "epoch": 0.04, + "learning_rate": 9.951466125209023e-05, + "loss": 2.1906, + "step": 16650 + }, + { + "epoch": 0.04, + "learning_rate": 9.951436842746162e-05, + "loss": 2.3502, + "step": 16655 + }, + { + "epoch": 0.04, + "learning_rate": 9.951407551495422e-05, + "loss": 2.4567, + "step": 16660 + }, + { + "epoch": 0.04, + "learning_rate": 9.951378251456856e-05, + "loss": 2.201, + "step": 16665 + }, + { + "epoch": 0.04, + "learning_rate": 9.951348942630514e-05, + "loss": 2.3683, + "step": 16670 + }, + { + "epoch": 0.04, + "learning_rate": 9.951319625016451e-05, + "loss": 2.3575, + "step": 16675 + }, + { + "epoch": 0.04, + "learning_rate": 9.951290298614715e-05, + "loss": 2.3512, + "step": 16680 + }, + { + "epoch": 0.04, + "learning_rate": 9.95126096342536e-05, + "loss": 2.384, + "step": 16685 + }, + { + "epoch": 0.04, + "learning_rate": 9.95123161944844e-05, + "loss": 2.4607, + "step": 16690 + }, + { + "epoch": 0.04, + "learning_rate": 9.951202266684005e-05, + "loss": 2.4511, + "step": 16695 + }, + { + "epoch": 0.04, + "learning_rate": 9.951172905132105e-05, + "loss": 2.2742, + "step": 16700 + }, + { + "epoch": 0.04, + "learning_rate": 9.951143534792797e-05, + "loss": 2.4185, + "step": 16705 + }, + { + "epoch": 0.04, + "learning_rate": 9.951114155666131e-05, + "loss": 2.2616, + "step": 16710 + }, + { + "epoch": 0.04, + "learning_rate": 9.951084767752157e-05, + "loss": 2.4048, + "step": 16715 + }, + { + "epoch": 0.04, + "learning_rate": 9.951055371050927e-05, + "loss": 2.4546, + "step": 16720 + }, + { + "epoch": 0.04, + "learning_rate": 9.951025965562499e-05, + "loss": 2.5384, + "step": 16725 + }, + { + "epoch": 0.04, + "learning_rate": 9.95099655128692e-05, + "loss": 2.5192, + "step": 16730 + }, + { + "epoch": 0.04, + "learning_rate": 9.950967128224244e-05, + "loss": 2.2678, + "step": 16735 + }, + { + "epoch": 0.04, + "learning_rate": 9.950937696374521e-05, + "loss": 2.5068, + "step": 16740 + }, + { + "epoch": 0.04, + "learning_rate": 9.950908255737805e-05, + "loss": 2.3799, + "step": 16745 + }, + { + "epoch": 0.04, + "learning_rate": 9.95087880631415e-05, + "loss": 2.3437, + "step": 16750 + }, + { + "epoch": 0.04, + "learning_rate": 9.950849348103606e-05, + "loss": 2.2544, + "step": 16755 + }, + { + "epoch": 0.04, + "learning_rate": 9.950819881106224e-05, + "loss": 2.4145, + "step": 16760 + }, + { + "epoch": 0.04, + "learning_rate": 9.95079040532206e-05, + "loss": 2.5036, + "step": 16765 + }, + { + "epoch": 0.04, + "learning_rate": 9.950760920751163e-05, + "loss": 2.4387, + "step": 16770 + }, + { + "epoch": 0.04, + "learning_rate": 9.950731427393588e-05, + "loss": 2.4669, + "step": 16775 + }, + { + "epoch": 0.04, + "learning_rate": 9.950701925249387e-05, + "loss": 2.3248, + "step": 16780 + }, + { + "epoch": 0.05, + "learning_rate": 9.950672414318608e-05, + "loss": 2.3281, + "step": 16785 + }, + { + "epoch": 0.05, + "learning_rate": 9.950642894601309e-05, + "loss": 2.4068, + "step": 16790 + }, + { + "epoch": 0.05, + "learning_rate": 9.950613366097539e-05, + "loss": 2.5725, + "step": 16795 + }, + { + "epoch": 0.05, + "learning_rate": 9.950583828807352e-05, + "loss": 2.344, + "step": 16800 + }, + { + "epoch": 0.05, + "learning_rate": 9.950554282730801e-05, + "loss": 2.3686, + "step": 16805 + }, + { + "epoch": 0.05, + "learning_rate": 9.950524727867936e-05, + "loss": 2.5656, + "step": 16810 + }, + { + "epoch": 0.05, + "learning_rate": 9.950495164218812e-05, + "loss": 2.4383, + "step": 16815 + }, + { + "epoch": 0.05, + "learning_rate": 9.95046559178348e-05, + "loss": 2.4212, + "step": 16820 + }, + { + "epoch": 0.05, + "learning_rate": 9.950436010561993e-05, + "loss": 2.4943, + "step": 16825 + }, + { + "epoch": 0.05, + "learning_rate": 9.950406420554403e-05, + "loss": 2.4746, + "step": 16830 + }, + { + "epoch": 0.05, + "learning_rate": 9.950376821760762e-05, + "loss": 2.3322, + "step": 16835 + }, + { + "epoch": 0.05, + "learning_rate": 9.950347214181124e-05, + "loss": 2.4254, + "step": 16840 + }, + { + "epoch": 0.05, + "learning_rate": 9.950317597815541e-05, + "loss": 2.34, + "step": 16845 + }, + { + "epoch": 0.05, + "learning_rate": 9.950287972664065e-05, + "loss": 2.3883, + "step": 16850 + }, + { + "epoch": 0.05, + "learning_rate": 9.95025833872675e-05, + "loss": 2.4494, + "step": 16855 + }, + { + "epoch": 0.05, + "learning_rate": 9.950228696003647e-05, + "loss": 2.4425, + "step": 16860 + }, + { + "epoch": 0.05, + "learning_rate": 9.95019904449481e-05, + "loss": 2.3244, + "step": 16865 + }, + { + "epoch": 0.05, + "learning_rate": 9.950169384200292e-05, + "loss": 2.3631, + "step": 16870 + }, + { + "epoch": 0.05, + "learning_rate": 9.950139715120142e-05, + "loss": 2.4709, + "step": 16875 + }, + { + "epoch": 0.05, + "learning_rate": 9.950110037254416e-05, + "loss": 2.586, + "step": 16880 + }, + { + "epoch": 0.05, + "learning_rate": 9.950080350603167e-05, + "loss": 2.4592, + "step": 16885 + }, + { + "epoch": 0.05, + "learning_rate": 9.950050655166446e-05, + "loss": 2.3432, + "step": 16890 + }, + { + "epoch": 0.05, + "learning_rate": 9.950020950944306e-05, + "loss": 2.4162, + "step": 16895 + }, + { + "epoch": 0.05, + "learning_rate": 9.9499912379368e-05, + "loss": 2.3705, + "step": 16900 + }, + { + "epoch": 0.05, + "learning_rate": 9.949961516143979e-05, + "loss": 2.5355, + "step": 16905 + }, + { + "epoch": 0.05, + "learning_rate": 9.9499317855659e-05, + "loss": 2.4221, + "step": 16910 + }, + { + "epoch": 0.05, + "learning_rate": 9.949902046202612e-05, + "loss": 2.3671, + "step": 16915 + }, + { + "epoch": 0.05, + "learning_rate": 9.94987229805417e-05, + "loss": 2.3481, + "step": 16920 + }, + { + "epoch": 0.05, + "learning_rate": 9.949842541120626e-05, + "loss": 2.3874, + "step": 16925 + }, + { + "epoch": 0.05, + "learning_rate": 9.949812775402032e-05, + "loss": 2.5642, + "step": 16930 + }, + { + "epoch": 0.05, + "learning_rate": 9.949783000898441e-05, + "loss": 2.3655, + "step": 16935 + }, + { + "epoch": 0.05, + "learning_rate": 9.949753217609907e-05, + "loss": 2.5004, + "step": 16940 + }, + { + "epoch": 0.05, + "learning_rate": 9.949723425536481e-05, + "loss": 2.4646, + "step": 16945 + }, + { + "epoch": 0.05, + "learning_rate": 9.949693624678218e-05, + "loss": 2.4507, + "step": 16950 + }, + { + "epoch": 0.05, + "learning_rate": 9.94966381503517e-05, + "loss": 2.4299, + "step": 16955 + }, + { + "epoch": 0.05, + "learning_rate": 9.94963399660739e-05, + "loss": 2.4724, + "step": 16960 + }, + { + "epoch": 0.05, + "learning_rate": 9.94960416939493e-05, + "loss": 2.5149, + "step": 16965 + }, + { + "epoch": 0.05, + "learning_rate": 9.949574333397844e-05, + "loss": 2.4216, + "step": 16970 + }, + { + "epoch": 0.05, + "learning_rate": 9.949544488616185e-05, + "loss": 2.4776, + "step": 16975 + }, + { + "epoch": 0.05, + "learning_rate": 9.949514635050006e-05, + "loss": 2.427, + "step": 16980 + }, + { + "epoch": 0.05, + "learning_rate": 9.949484772699358e-05, + "loss": 2.4671, + "step": 16985 + }, + { + "epoch": 0.05, + "learning_rate": 9.949454901564297e-05, + "loss": 2.5272, + "step": 16990 + }, + { + "epoch": 0.05, + "learning_rate": 9.949425021644874e-05, + "loss": 2.5133, + "step": 16995 + }, + { + "epoch": 0.05, + "learning_rate": 9.949395132941142e-05, + "loss": 2.4549, + "step": 17000 + }, + { + "epoch": 0.05, + "learning_rate": 9.949365235453156e-05, + "loss": 2.3524, + "step": 17005 + }, + { + "epoch": 0.05, + "learning_rate": 9.949335329180967e-05, + "loss": 2.4029, + "step": 17010 + }, + { + "epoch": 0.05, + "learning_rate": 9.94930541412463e-05, + "loss": 2.4075, + "step": 17015 + }, + { + "epoch": 0.05, + "learning_rate": 9.949275490284195e-05, + "loss": 2.4646, + "step": 17020 + }, + { + "epoch": 0.05, + "learning_rate": 9.949245557659719e-05, + "loss": 2.4025, + "step": 17025 + }, + { + "epoch": 0.05, + "learning_rate": 9.949215616251252e-05, + "loss": 2.424, + "step": 17030 + }, + { + "epoch": 0.05, + "learning_rate": 9.94918566605885e-05, + "loss": 2.4248, + "step": 17035 + }, + { + "epoch": 0.05, + "learning_rate": 9.949155707082561e-05, + "loss": 2.4708, + "step": 17040 + }, + { + "epoch": 0.05, + "learning_rate": 9.949125739322445e-05, + "loss": 2.555, + "step": 17045 + }, + { + "epoch": 0.05, + "learning_rate": 9.94909576277855e-05, + "loss": 2.3204, + "step": 17050 + }, + { + "epoch": 0.05, + "learning_rate": 9.949065777450931e-05, + "loss": 2.2844, + "step": 17055 + }, + { + "epoch": 0.05, + "learning_rate": 9.949035783339642e-05, + "loss": 2.5334, + "step": 17060 + }, + { + "epoch": 0.05, + "learning_rate": 9.949005780444736e-05, + "loss": 2.488, + "step": 17065 + }, + { + "epoch": 0.05, + "learning_rate": 9.948975768766265e-05, + "loss": 2.3613, + "step": 17070 + }, + { + "epoch": 0.05, + "learning_rate": 9.948945748304282e-05, + "loss": 2.4421, + "step": 17075 + }, + { + "epoch": 0.05, + "learning_rate": 9.948915719058842e-05, + "loss": 2.4315, + "step": 17080 + }, + { + "epoch": 0.05, + "learning_rate": 9.948885681029998e-05, + "loss": 2.4005, + "step": 17085 + }, + { + "epoch": 0.05, + "learning_rate": 9.948855634217802e-05, + "loss": 2.4114, + "step": 17090 + }, + { + "epoch": 0.05, + "learning_rate": 9.94882557862231e-05, + "loss": 2.4609, + "step": 17095 + }, + { + "epoch": 0.05, + "learning_rate": 9.948795514243572e-05, + "loss": 2.529, + "step": 17100 + }, + { + "epoch": 0.05, + "learning_rate": 9.948765441081643e-05, + "loss": 2.5689, + "step": 17105 + }, + { + "epoch": 0.05, + "learning_rate": 9.948735359136576e-05, + "loss": 2.3787, + "step": 17110 + }, + { + "epoch": 0.05, + "learning_rate": 9.948705268408426e-05, + "loss": 2.4665, + "step": 17115 + }, + { + "epoch": 0.05, + "learning_rate": 9.948675168897244e-05, + "loss": 2.3535, + "step": 17120 + }, + { + "epoch": 0.05, + "learning_rate": 9.948645060603086e-05, + "loss": 2.3566, + "step": 17125 + }, + { + "epoch": 0.05, + "learning_rate": 9.948614943526003e-05, + "loss": 2.3652, + "step": 17130 + }, + { + "epoch": 0.05, + "learning_rate": 9.948584817666048e-05, + "loss": 2.3779, + "step": 17135 + }, + { + "epoch": 0.05, + "learning_rate": 9.948554683023276e-05, + "loss": 2.402, + "step": 17140 + }, + { + "epoch": 0.05, + "learning_rate": 9.948524539597742e-05, + "loss": 2.2815, + "step": 17145 + }, + { + "epoch": 0.05, + "learning_rate": 9.948494387389498e-05, + "loss": 2.2299, + "step": 17150 + }, + { + "epoch": 0.05, + "learning_rate": 9.948464226398596e-05, + "loss": 2.4435, + "step": 17155 + }, + { + "epoch": 0.05, + "learning_rate": 9.948434056625091e-05, + "loss": 2.3958, + "step": 17160 + }, + { + "epoch": 0.05, + "learning_rate": 9.948403878069037e-05, + "loss": 2.4052, + "step": 17165 + }, + { + "epoch": 0.05, + "learning_rate": 9.948373690730487e-05, + "loss": 2.466, + "step": 17170 + }, + { + "epoch": 0.05, + "learning_rate": 9.948343494609494e-05, + "loss": 2.3414, + "step": 17175 + }, + { + "epoch": 0.05, + "learning_rate": 9.948313289706112e-05, + "loss": 2.4626, + "step": 17180 + }, + { + "epoch": 0.05, + "learning_rate": 9.948283076020398e-05, + "loss": 2.576, + "step": 17185 + }, + { + "epoch": 0.05, + "learning_rate": 9.948252853552398e-05, + "loss": 2.4307, + "step": 17190 + }, + { + "epoch": 0.05, + "learning_rate": 9.948222622302173e-05, + "loss": 2.439, + "step": 17195 + }, + { + "epoch": 0.05, + "learning_rate": 9.948192382269772e-05, + "loss": 2.3564, + "step": 17200 + }, + { + "epoch": 0.05, + "learning_rate": 9.94816213345525e-05, + "loss": 2.4648, + "step": 17205 + }, + { + "epoch": 0.05, + "learning_rate": 9.948131875858662e-05, + "loss": 2.2711, + "step": 17210 + }, + { + "epoch": 0.05, + "learning_rate": 9.94810160948006e-05, + "loss": 2.4882, + "step": 17215 + }, + { + "epoch": 0.05, + "learning_rate": 9.9480713343195e-05, + "loss": 2.5318, + "step": 17220 + }, + { + "epoch": 0.05, + "learning_rate": 9.948041050377033e-05, + "loss": 2.3633, + "step": 17225 + }, + { + "epoch": 0.05, + "learning_rate": 9.948010757652714e-05, + "loss": 2.3059, + "step": 17230 + }, + { + "epoch": 0.05, + "learning_rate": 9.947980456146597e-05, + "loss": 2.3354, + "step": 17235 + }, + { + "epoch": 0.05, + "learning_rate": 9.947950145858736e-05, + "loss": 2.6357, + "step": 17240 + }, + { + "epoch": 0.05, + "learning_rate": 9.947919826789182e-05, + "loss": 2.4347, + "step": 17245 + }, + { + "epoch": 0.05, + "learning_rate": 9.947889498937994e-05, + "loss": 2.4296, + "step": 17250 + }, + { + "epoch": 0.05, + "learning_rate": 9.947859162305222e-05, + "loss": 2.3641, + "step": 17255 + }, + { + "epoch": 0.05, + "learning_rate": 9.947828816890921e-05, + "loss": 2.4704, + "step": 17260 + }, + { + "epoch": 0.05, + "learning_rate": 9.947798462695142e-05, + "loss": 2.3117, + "step": 17265 + }, + { + "epoch": 0.05, + "learning_rate": 9.947768099717945e-05, + "loss": 2.3889, + "step": 17270 + }, + { + "epoch": 0.05, + "learning_rate": 9.947737727959379e-05, + "loss": 2.4501, + "step": 17275 + }, + { + "epoch": 0.05, + "learning_rate": 9.947707347419499e-05, + "loss": 2.3265, + "step": 17280 + }, + { + "epoch": 0.05, + "learning_rate": 9.947676958098358e-05, + "loss": 2.339, + "step": 17285 + }, + { + "epoch": 0.05, + "learning_rate": 9.947646559996014e-05, + "loss": 2.4204, + "step": 17290 + }, + { + "epoch": 0.05, + "learning_rate": 9.947616153112515e-05, + "loss": 2.3437, + "step": 17295 + }, + { + "epoch": 0.05, + "learning_rate": 9.947585737447921e-05, + "loss": 2.3827, + "step": 17300 + }, + { + "epoch": 0.05, + "learning_rate": 9.947555313002281e-05, + "loss": 2.4705, + "step": 17305 + }, + { + "epoch": 0.05, + "learning_rate": 9.947524879775652e-05, + "loss": 2.4207, + "step": 17310 + }, + { + "epoch": 0.05, + "learning_rate": 9.947494437768086e-05, + "loss": 2.2751, + "step": 17315 + }, + { + "epoch": 0.05, + "learning_rate": 9.94746398697964e-05, + "loss": 2.4043, + "step": 17320 + }, + { + "epoch": 0.05, + "learning_rate": 9.947433527410365e-05, + "loss": 2.446, + "step": 17325 + }, + { + "epoch": 0.05, + "learning_rate": 9.947403059060316e-05, + "loss": 2.2267, + "step": 17330 + }, + { + "epoch": 0.05, + "learning_rate": 9.947372581929547e-05, + "loss": 2.3062, + "step": 17335 + }, + { + "epoch": 0.05, + "learning_rate": 9.947342096018113e-05, + "loss": 2.3664, + "step": 17340 + }, + { + "epoch": 0.05, + "learning_rate": 9.947311601326067e-05, + "loss": 2.4146, + "step": 17345 + }, + { + "epoch": 0.05, + "learning_rate": 9.947281097853465e-05, + "loss": 2.342, + "step": 17350 + }, + { + "epoch": 0.05, + "learning_rate": 9.947250585600357e-05, + "loss": 2.3758, + "step": 17355 + }, + { + "epoch": 0.05, + "learning_rate": 9.947220064566802e-05, + "loss": 2.4671, + "step": 17360 + }, + { + "epoch": 0.05, + "learning_rate": 9.947189534752852e-05, + "loss": 2.3401, + "step": 17365 + }, + { + "epoch": 0.05, + "learning_rate": 9.94715899615856e-05, + "loss": 2.318, + "step": 17370 + }, + { + "epoch": 0.05, + "learning_rate": 9.947128448783982e-05, + "loss": 2.4301, + "step": 17375 + }, + { + "epoch": 0.05, + "learning_rate": 9.947097892629171e-05, + "loss": 2.4687, + "step": 17380 + }, + { + "epoch": 0.05, + "learning_rate": 9.947067327694184e-05, + "loss": 2.4457, + "step": 17385 + }, + { + "epoch": 0.05, + "learning_rate": 9.947036753979071e-05, + "loss": 2.5051, + "step": 17390 + }, + { + "epoch": 0.05, + "learning_rate": 9.94700617148389e-05, + "loss": 2.3834, + "step": 17395 + }, + { + "epoch": 0.05, + "learning_rate": 9.946975580208691e-05, + "loss": 2.3444, + "step": 17400 + }, + { + "epoch": 0.05, + "learning_rate": 9.946944980153534e-05, + "loss": 2.5079, + "step": 17405 + }, + { + "epoch": 0.05, + "learning_rate": 9.946914371318467e-05, + "loss": 2.4533, + "step": 17410 + }, + { + "epoch": 0.05, + "learning_rate": 9.946883753703551e-05, + "loss": 2.3354, + "step": 17415 + }, + { + "epoch": 0.05, + "learning_rate": 9.946853127308835e-05, + "loss": 2.3894, + "step": 17420 + }, + { + "epoch": 0.05, + "learning_rate": 9.946822492134376e-05, + "loss": 2.3407, + "step": 17425 + }, + { + "epoch": 0.05, + "learning_rate": 9.946791848180228e-05, + "loss": 2.5024, + "step": 17430 + }, + { + "epoch": 0.05, + "learning_rate": 9.946761195446444e-05, + "loss": 2.3496, + "step": 17435 + }, + { + "epoch": 0.05, + "learning_rate": 9.946730533933079e-05, + "loss": 2.3158, + "step": 17440 + }, + { + "epoch": 0.05, + "learning_rate": 9.94669986364019e-05, + "loss": 2.4499, + "step": 17445 + }, + { + "epoch": 0.05, + "learning_rate": 9.946669184567827e-05, + "loss": 2.5135, + "step": 17450 + }, + { + "epoch": 0.05, + "learning_rate": 9.946638496716048e-05, + "loss": 2.3557, + "step": 17455 + }, + { + "epoch": 0.05, + "learning_rate": 9.946607800084907e-05, + "loss": 2.4299, + "step": 17460 + }, + { + "epoch": 0.05, + "learning_rate": 9.946577094674456e-05, + "loss": 2.3841, + "step": 17465 + }, + { + "epoch": 0.05, + "learning_rate": 9.946546380484751e-05, + "loss": 2.3787, + "step": 17470 + }, + { + "epoch": 0.05, + "learning_rate": 9.946515657515849e-05, + "loss": 2.3172, + "step": 17475 + }, + { + "epoch": 0.05, + "learning_rate": 9.946484925767801e-05, + "loss": 2.4994, + "step": 17480 + }, + { + "epoch": 0.05, + "learning_rate": 9.946454185240662e-05, + "loss": 2.4736, + "step": 17485 + }, + { + "epoch": 0.05, + "learning_rate": 9.946423435934489e-05, + "loss": 2.3645, + "step": 17490 + }, + { + "epoch": 0.05, + "learning_rate": 9.946392677849332e-05, + "loss": 2.3749, + "step": 17495 + }, + { + "epoch": 0.05, + "learning_rate": 9.946361910985252e-05, + "loss": 2.3824, + "step": 17500 + }, + { + "epoch": 0.05, + "learning_rate": 9.946331135342298e-05, + "loss": 2.5578, + "step": 17505 + }, + { + "epoch": 0.05, + "learning_rate": 9.946300350920526e-05, + "loss": 2.3384, + "step": 17510 + }, + { + "epoch": 0.05, + "learning_rate": 9.946269557719994e-05, + "loss": 2.3046, + "step": 17515 + }, + { + "epoch": 0.05, + "learning_rate": 9.946238755740752e-05, + "loss": 2.4128, + "step": 17520 + }, + { + "epoch": 0.05, + "learning_rate": 9.946207944982857e-05, + "loss": 2.3563, + "step": 17525 + }, + { + "epoch": 0.05, + "learning_rate": 9.946177125446363e-05, + "loss": 2.4813, + "step": 17530 + }, + { + "epoch": 0.05, + "learning_rate": 9.946146297131327e-05, + "loss": 2.3289, + "step": 17535 + }, + { + "epoch": 0.05, + "learning_rate": 9.946115460037799e-05, + "loss": 2.4769, + "step": 17540 + }, + { + "epoch": 0.05, + "learning_rate": 9.946084614165838e-05, + "loss": 2.3651, + "step": 17545 + }, + { + "epoch": 0.05, + "learning_rate": 9.946053759515498e-05, + "loss": 2.5058, + "step": 17550 + }, + { + "epoch": 0.05, + "learning_rate": 9.946022896086832e-05, + "loss": 2.3224, + "step": 17555 + }, + { + "epoch": 0.05, + "learning_rate": 9.945992023879895e-05, + "loss": 2.4991, + "step": 17560 + }, + { + "epoch": 0.05, + "learning_rate": 9.945961142894744e-05, + "loss": 2.3643, + "step": 17565 + }, + { + "epoch": 0.05, + "learning_rate": 9.945930253131433e-05, + "loss": 2.4754, + "step": 17570 + }, + { + "epoch": 0.05, + "learning_rate": 9.945899354590014e-05, + "loss": 2.4209, + "step": 17575 + }, + { + "epoch": 0.05, + "learning_rate": 9.945868447270545e-05, + "loss": 2.3924, + "step": 17580 + }, + { + "epoch": 0.05, + "learning_rate": 9.945837531173081e-05, + "loss": 2.4149, + "step": 17585 + }, + { + "epoch": 0.05, + "learning_rate": 9.945806606297674e-05, + "loss": 2.2281, + "step": 17590 + }, + { + "epoch": 0.05, + "learning_rate": 9.945775672644382e-05, + "loss": 2.4634, + "step": 17595 + }, + { + "epoch": 0.05, + "learning_rate": 9.945744730213258e-05, + "loss": 2.4117, + "step": 17600 + }, + { + "epoch": 0.05, + "learning_rate": 9.945713779004357e-05, + "loss": 2.3002, + "step": 17605 + }, + { + "epoch": 0.05, + "learning_rate": 9.945682819017736e-05, + "loss": 2.4223, + "step": 17610 + }, + { + "epoch": 0.05, + "learning_rate": 9.945651850253448e-05, + "loss": 2.3736, + "step": 17615 + }, + { + "epoch": 0.05, + "learning_rate": 9.945620872711547e-05, + "loss": 2.3223, + "step": 17620 + }, + { + "epoch": 0.05, + "learning_rate": 9.94558988639209e-05, + "loss": 2.4006, + "step": 17625 + }, + { + "epoch": 0.05, + "learning_rate": 9.945558891295131e-05, + "loss": 2.4093, + "step": 17630 + }, + { + "epoch": 0.05, + "learning_rate": 9.945527887420726e-05, + "loss": 2.4036, + "step": 17635 + }, + { + "epoch": 0.05, + "learning_rate": 9.945496874768929e-05, + "loss": 2.4506, + "step": 17640 + }, + { + "epoch": 0.05, + "learning_rate": 9.945465853339795e-05, + "loss": 2.408, + "step": 17645 + }, + { + "epoch": 0.05, + "learning_rate": 9.94543482313338e-05, + "loss": 2.3717, + "step": 17650 + }, + { + "epoch": 0.05, + "learning_rate": 9.945403784149738e-05, + "loss": 2.5136, + "step": 17655 + }, + { + "epoch": 0.05, + "learning_rate": 9.945372736388925e-05, + "loss": 2.413, + "step": 17660 + }, + { + "epoch": 0.05, + "learning_rate": 9.945341679850996e-05, + "loss": 2.3495, + "step": 17665 + }, + { + "epoch": 0.05, + "learning_rate": 9.945310614536005e-05, + "loss": 2.5071, + "step": 17670 + }, + { + "epoch": 0.05, + "learning_rate": 9.945279540444008e-05, + "loss": 2.2793, + "step": 17675 + }, + { + "epoch": 0.05, + "learning_rate": 9.94524845757506e-05, + "loss": 2.4764, + "step": 17680 + }, + { + "epoch": 0.05, + "learning_rate": 9.945217365929218e-05, + "loss": 2.375, + "step": 17685 + }, + { + "epoch": 0.05, + "learning_rate": 9.945186265506534e-05, + "loss": 2.3987, + "step": 17690 + }, + { + "epoch": 0.05, + "learning_rate": 9.945155156307064e-05, + "loss": 2.3464, + "step": 17695 + }, + { + "epoch": 0.05, + "learning_rate": 9.945124038330867e-05, + "loss": 2.4192, + "step": 17700 + }, + { + "epoch": 0.05, + "learning_rate": 9.945092911577991e-05, + "loss": 2.4793, + "step": 17705 + }, + { + "epoch": 0.05, + "learning_rate": 9.945061776048497e-05, + "loss": 2.3858, + "step": 17710 + }, + { + "epoch": 0.05, + "learning_rate": 9.945030631742439e-05, + "loss": 2.5161, + "step": 17715 + }, + { + "epoch": 0.05, + "learning_rate": 9.944999478659872e-05, + "loss": 2.3934, + "step": 17720 + }, + { + "epoch": 0.05, + "learning_rate": 9.944968316800851e-05, + "loss": 2.4567, + "step": 17725 + }, + { + "epoch": 0.05, + "learning_rate": 9.944937146165431e-05, + "loss": 2.4346, + "step": 17730 + }, + { + "epoch": 0.05, + "learning_rate": 9.944905966753667e-05, + "loss": 2.498, + "step": 17735 + }, + { + "epoch": 0.05, + "learning_rate": 9.944874778565616e-05, + "loss": 2.3151, + "step": 17740 + }, + { + "epoch": 0.05, + "learning_rate": 9.944843581601334e-05, + "loss": 2.4115, + "step": 17745 + }, + { + "epoch": 0.05, + "learning_rate": 9.944812375860874e-05, + "loss": 2.441, + "step": 17750 + }, + { + "epoch": 0.05, + "learning_rate": 9.944781161344291e-05, + "loss": 2.4472, + "step": 17755 + }, + { + "epoch": 0.05, + "learning_rate": 9.944749938051643e-05, + "loss": 2.3757, + "step": 17760 + }, + { + "epoch": 0.05, + "learning_rate": 9.944718705982982e-05, + "loss": 2.3994, + "step": 17765 + }, + { + "epoch": 0.05, + "learning_rate": 9.944687465138368e-05, + "loss": 2.4618, + "step": 17770 + }, + { + "epoch": 0.05, + "learning_rate": 9.944656215517854e-05, + "loss": 2.3433, + "step": 17775 + }, + { + "epoch": 0.05, + "learning_rate": 9.944624957121493e-05, + "loss": 2.352, + "step": 17780 + }, + { + "epoch": 0.05, + "learning_rate": 9.944593689949345e-05, + "loss": 2.4393, + "step": 17785 + }, + { + "epoch": 0.05, + "learning_rate": 9.944562414001461e-05, + "loss": 2.4827, + "step": 17790 + }, + { + "epoch": 0.05, + "learning_rate": 9.944531129277901e-05, + "loss": 2.3896, + "step": 17795 + }, + { + "epoch": 0.05, + "learning_rate": 9.944499835778718e-05, + "loss": 2.3616, + "step": 17800 + }, + { + "epoch": 0.05, + "learning_rate": 9.944468533503967e-05, + "loss": 2.413, + "step": 17805 + }, + { + "epoch": 0.05, + "learning_rate": 9.944437222453706e-05, + "loss": 2.4895, + "step": 17810 + }, + { + "epoch": 0.05, + "learning_rate": 9.944405902627988e-05, + "loss": 2.2538, + "step": 17815 + }, + { + "epoch": 0.05, + "learning_rate": 9.944374574026871e-05, + "loss": 2.4288, + "step": 17820 + }, + { + "epoch": 0.05, + "learning_rate": 9.944343236650407e-05, + "loss": 2.3802, + "step": 17825 + }, + { + "epoch": 0.05, + "learning_rate": 9.944311890498655e-05, + "loss": 2.4979, + "step": 17830 + }, + { + "epoch": 0.05, + "learning_rate": 9.94428053557167e-05, + "loss": 2.5149, + "step": 17835 + }, + { + "epoch": 0.05, + "learning_rate": 9.944249171869506e-05, + "loss": 2.4137, + "step": 17840 + }, + { + "epoch": 0.05, + "learning_rate": 9.944217799392221e-05, + "loss": 2.4312, + "step": 17845 + }, + { + "epoch": 0.05, + "learning_rate": 9.94418641813987e-05, + "loss": 2.3362, + "step": 17850 + }, + { + "epoch": 0.05, + "learning_rate": 9.944155028112507e-05, + "loss": 2.5291, + "step": 17855 + }, + { + "epoch": 0.05, + "learning_rate": 9.944123629310189e-05, + "loss": 2.3761, + "step": 17860 + }, + { + "epoch": 0.05, + "learning_rate": 9.94409222173297e-05, + "loss": 2.4351, + "step": 17865 + }, + { + "epoch": 0.05, + "learning_rate": 9.944060805380911e-05, + "loss": 2.4568, + "step": 17870 + }, + { + "epoch": 0.05, + "learning_rate": 9.944029380254062e-05, + "loss": 2.3473, + "step": 17875 + }, + { + "epoch": 0.05, + "learning_rate": 9.943997946352481e-05, + "loss": 2.4116, + "step": 17880 + }, + { + "epoch": 0.05, + "learning_rate": 9.943966503676223e-05, + "loss": 2.2752, + "step": 17885 + }, + { + "epoch": 0.05, + "learning_rate": 9.943935052225346e-05, + "loss": 2.5216, + "step": 17890 + }, + { + "epoch": 0.05, + "learning_rate": 9.943903591999904e-05, + "loss": 2.4012, + "step": 17895 + }, + { + "epoch": 0.05, + "learning_rate": 9.943872122999953e-05, + "loss": 2.5509, + "step": 17900 + }, + { + "epoch": 0.05, + "learning_rate": 9.943840645225548e-05, + "loss": 2.3366, + "step": 17905 + }, + { + "epoch": 0.05, + "learning_rate": 9.943809158676747e-05, + "loss": 2.3359, + "step": 17910 + }, + { + "epoch": 0.05, + "learning_rate": 9.943777663353604e-05, + "loss": 2.2678, + "step": 17915 + }, + { + "epoch": 0.05, + "learning_rate": 9.943746159256177e-05, + "loss": 2.3057, + "step": 17920 + }, + { + "epoch": 0.05, + "learning_rate": 9.943714646384519e-05, + "loss": 2.2306, + "step": 17925 + }, + { + "epoch": 0.05, + "learning_rate": 9.94368312473869e-05, + "loss": 2.4883, + "step": 17930 + }, + { + "epoch": 0.05, + "learning_rate": 9.943651594318741e-05, + "loss": 2.4405, + "step": 17935 + }, + { + "epoch": 0.05, + "learning_rate": 9.943620055124732e-05, + "loss": 2.3819, + "step": 17940 + }, + { + "epoch": 0.05, + "learning_rate": 9.943588507156716e-05, + "loss": 2.5254, + "step": 17945 + }, + { + "epoch": 0.05, + "learning_rate": 9.94355695041475e-05, + "loss": 2.4936, + "step": 17950 + }, + { + "epoch": 0.05, + "learning_rate": 9.943525384898891e-05, + "loss": 2.3829, + "step": 17955 + }, + { + "epoch": 0.05, + "learning_rate": 9.943493810609195e-05, + "loss": 2.3506, + "step": 17960 + }, + { + "epoch": 0.05, + "learning_rate": 9.943462227545717e-05, + "loss": 2.3526, + "step": 17965 + }, + { + "epoch": 0.05, + "learning_rate": 9.943430635708514e-05, + "loss": 2.4527, + "step": 17970 + }, + { + "epoch": 0.05, + "learning_rate": 9.94339903509764e-05, + "loss": 2.2089, + "step": 17975 + }, + { + "epoch": 0.05, + "learning_rate": 9.943367425713155e-05, + "loss": 2.2831, + "step": 17980 + }, + { + "epoch": 0.05, + "learning_rate": 9.943335807555112e-05, + "loss": 2.3073, + "step": 17985 + }, + { + "epoch": 0.05, + "learning_rate": 9.943304180623567e-05, + "loss": 2.2933, + "step": 17990 + }, + { + "epoch": 0.05, + "learning_rate": 9.943272544918578e-05, + "loss": 2.3328, + "step": 17995 + }, + { + "epoch": 0.05, + "learning_rate": 9.943240900440198e-05, + "loss": 2.3144, + "step": 18000 + }, + { + "epoch": 0.05, + "learning_rate": 9.943209247188488e-05, + "loss": 2.3611, + "step": 18005 + }, + { + "epoch": 0.05, + "learning_rate": 9.943177585163502e-05, + "loss": 2.4547, + "step": 18010 + }, + { + "epoch": 0.05, + "learning_rate": 9.943145914365292e-05, + "loss": 2.4098, + "step": 18015 + }, + { + "epoch": 0.05, + "learning_rate": 9.94311423479392e-05, + "loss": 2.3547, + "step": 18020 + }, + { + "epoch": 0.05, + "learning_rate": 9.943082546449441e-05, + "loss": 2.3888, + "step": 18025 + }, + { + "epoch": 0.05, + "learning_rate": 9.94305084933191e-05, + "loss": 2.5256, + "step": 18030 + }, + { + "epoch": 0.05, + "learning_rate": 9.943019143441383e-05, + "loss": 2.4426, + "step": 18035 + }, + { + "epoch": 0.05, + "learning_rate": 9.942987428777916e-05, + "loss": 2.2359, + "step": 18040 + }, + { + "epoch": 0.05, + "learning_rate": 9.942955705341568e-05, + "loss": 2.32, + "step": 18045 + }, + { + "epoch": 0.05, + "learning_rate": 9.942923973132392e-05, + "loss": 2.3858, + "step": 18050 + }, + { + "epoch": 0.05, + "learning_rate": 9.942892232150447e-05, + "loss": 2.3961, + "step": 18055 + }, + { + "epoch": 0.05, + "learning_rate": 9.942860482395787e-05, + "loss": 2.3415, + "step": 18060 + }, + { + "epoch": 0.05, + "learning_rate": 9.94282872386847e-05, + "loss": 2.5286, + "step": 18065 + }, + { + "epoch": 0.05, + "learning_rate": 9.942796956568552e-05, + "loss": 2.3813, + "step": 18070 + }, + { + "epoch": 0.05, + "learning_rate": 9.94276518049609e-05, + "loss": 2.5562, + "step": 18075 + }, + { + "epoch": 0.05, + "learning_rate": 9.942733395651138e-05, + "loss": 2.3693, + "step": 18080 + }, + { + "epoch": 0.05, + "learning_rate": 9.942701602033754e-05, + "loss": 2.3792, + "step": 18085 + }, + { + "epoch": 0.05, + "learning_rate": 9.942669799643996e-05, + "loss": 2.4453, + "step": 18090 + }, + { + "epoch": 0.05, + "learning_rate": 9.942637988481919e-05, + "loss": 2.4525, + "step": 18095 + }, + { + "epoch": 0.05, + "learning_rate": 9.942606168547579e-05, + "loss": 2.3389, + "step": 18100 + }, + { + "epoch": 0.05, + "learning_rate": 9.942574339841031e-05, + "loss": 2.5049, + "step": 18105 + }, + { + "epoch": 0.05, + "learning_rate": 9.942542502362336e-05, + "loss": 2.4416, + "step": 18110 + }, + { + "epoch": 0.05, + "learning_rate": 9.942510656111546e-05, + "loss": 2.3387, + "step": 18115 + }, + { + "epoch": 0.05, + "learning_rate": 9.942478801088719e-05, + "loss": 2.2757, + "step": 18120 + }, + { + "epoch": 0.05, + "learning_rate": 9.942446937293915e-05, + "loss": 2.3216, + "step": 18125 + }, + { + "epoch": 0.05, + "learning_rate": 9.942415064727184e-05, + "loss": 2.4087, + "step": 18130 + }, + { + "epoch": 0.05, + "learning_rate": 9.942383183388588e-05, + "loss": 2.3837, + "step": 18135 + }, + { + "epoch": 0.05, + "learning_rate": 9.94235129327818e-05, + "loss": 2.339, + "step": 18140 + }, + { + "epoch": 0.05, + "learning_rate": 9.94231939439602e-05, + "loss": 2.5038, + "step": 18145 + }, + { + "epoch": 0.05, + "learning_rate": 9.942287486742162e-05, + "loss": 2.4867, + "step": 18150 + }, + { + "epoch": 0.05, + "learning_rate": 9.942255570316663e-05, + "loss": 2.3146, + "step": 18155 + }, + { + "epoch": 0.05, + "learning_rate": 9.94222364511958e-05, + "loss": 2.3495, + "step": 18160 + }, + { + "epoch": 0.05, + "learning_rate": 9.942191711150971e-05, + "loss": 2.3808, + "step": 18165 + }, + { + "epoch": 0.05, + "learning_rate": 9.94215976841089e-05, + "loss": 2.3568, + "step": 18170 + }, + { + "epoch": 0.05, + "learning_rate": 9.942127816899397e-05, + "loss": 2.3786, + "step": 18175 + }, + { + "epoch": 0.05, + "learning_rate": 9.942095856616544e-05, + "loss": 2.2863, + "step": 18180 + }, + { + "epoch": 0.05, + "learning_rate": 9.942063887562393e-05, + "loss": 2.3676, + "step": 18185 + }, + { + "epoch": 0.05, + "learning_rate": 9.942031909736997e-05, + "loss": 2.4012, + "step": 18190 + }, + { + "epoch": 0.05, + "learning_rate": 9.941999923140415e-05, + "loss": 2.3255, + "step": 18195 + }, + { + "epoch": 0.05, + "learning_rate": 9.941967927772702e-05, + "loss": 2.3638, + "step": 18200 + }, + { + "epoch": 0.05, + "learning_rate": 9.941935923633917e-05, + "loss": 2.3214, + "step": 18205 + }, + { + "epoch": 0.05, + "learning_rate": 9.941903910724113e-05, + "loss": 2.5143, + "step": 18210 + }, + { + "epoch": 0.05, + "learning_rate": 9.941871889043351e-05, + "loss": 2.3332, + "step": 18215 + }, + { + "epoch": 0.05, + "learning_rate": 9.941839858591687e-05, + "loss": 2.4047, + "step": 18220 + }, + { + "epoch": 0.05, + "learning_rate": 9.941807819369174e-05, + "loss": 2.4749, + "step": 18225 + }, + { + "epoch": 0.05, + "learning_rate": 9.941775771375873e-05, + "loss": 2.3459, + "step": 18230 + }, + { + "epoch": 0.05, + "learning_rate": 9.94174371461184e-05, + "loss": 2.4647, + "step": 18235 + }, + { + "epoch": 0.05, + "learning_rate": 9.941711649077132e-05, + "loss": 2.5107, + "step": 18240 + }, + { + "epoch": 0.05, + "learning_rate": 9.941679574771806e-05, + "loss": 2.4054, + "step": 18245 + }, + { + "epoch": 0.05, + "learning_rate": 9.941647491695915e-05, + "loss": 2.3999, + "step": 18250 + }, + { + "epoch": 0.05, + "learning_rate": 9.941615399849523e-05, + "loss": 2.4056, + "step": 18255 + }, + { + "epoch": 0.05, + "learning_rate": 9.941583299232682e-05, + "loss": 2.3269, + "step": 18260 + }, + { + "epoch": 0.05, + "learning_rate": 9.94155118984545e-05, + "loss": 2.4932, + "step": 18265 + }, + { + "epoch": 0.05, + "learning_rate": 9.941519071687885e-05, + "loss": 2.4202, + "step": 18270 + }, + { + "epoch": 0.05, + "learning_rate": 9.941486944760042e-05, + "loss": 2.4219, + "step": 18275 + }, + { + "epoch": 0.05, + "learning_rate": 9.941454809061981e-05, + "loss": 2.3875, + "step": 18280 + }, + { + "epoch": 0.05, + "learning_rate": 9.941422664593755e-05, + "loss": 2.3763, + "step": 18285 + }, + { + "epoch": 0.05, + "learning_rate": 9.941390511355425e-05, + "loss": 2.3459, + "step": 18290 + }, + { + "epoch": 0.05, + "learning_rate": 9.941358349347045e-05, + "loss": 2.4805, + "step": 18295 + }, + { + "epoch": 0.05, + "learning_rate": 9.941326178568675e-05, + "loss": 2.4417, + "step": 18300 + }, + { + "epoch": 0.05, + "learning_rate": 9.94129399902037e-05, + "loss": 2.5541, + "step": 18305 + }, + { + "epoch": 0.05, + "learning_rate": 9.941261810702188e-05, + "loss": 2.4326, + "step": 18310 + }, + { + "epoch": 0.05, + "learning_rate": 9.941229613614186e-05, + "loss": 2.3806, + "step": 18315 + }, + { + "epoch": 0.05, + "learning_rate": 9.94119740775642e-05, + "loss": 2.4456, + "step": 18320 + }, + { + "epoch": 0.05, + "learning_rate": 9.94116519312895e-05, + "loss": 2.419, + "step": 18325 + }, + { + "epoch": 0.05, + "learning_rate": 9.941132969731829e-05, + "loss": 2.4034, + "step": 18330 + }, + { + "epoch": 0.05, + "learning_rate": 9.941100737565117e-05, + "loss": 2.4987, + "step": 18335 + }, + { + "epoch": 0.05, + "learning_rate": 9.941068496628871e-05, + "loss": 2.3845, + "step": 18340 + }, + { + "epoch": 0.05, + "learning_rate": 9.941036246923148e-05, + "loss": 2.451, + "step": 18345 + }, + { + "epoch": 0.05, + "learning_rate": 9.941003988448005e-05, + "loss": 2.4623, + "step": 18350 + }, + { + "epoch": 0.05, + "learning_rate": 9.9409717212035e-05, + "loss": 2.446, + "step": 18355 + }, + { + "epoch": 0.05, + "learning_rate": 9.94093944518969e-05, + "loss": 2.4195, + "step": 18360 + }, + { + "epoch": 0.05, + "learning_rate": 9.940907160406631e-05, + "loss": 2.529, + "step": 18365 + }, + { + "epoch": 0.05, + "learning_rate": 9.940874866854382e-05, + "loss": 2.5587, + "step": 18370 + }, + { + "epoch": 0.05, + "learning_rate": 9.940842564533e-05, + "loss": 2.3887, + "step": 18375 + }, + { + "epoch": 0.05, + "learning_rate": 9.94081025344254e-05, + "loss": 2.5519, + "step": 18380 + }, + { + "epoch": 0.05, + "learning_rate": 9.940777933583063e-05, + "loss": 2.3761, + "step": 18385 + }, + { + "epoch": 0.05, + "learning_rate": 9.940745604954624e-05, + "loss": 2.4091, + "step": 18390 + }, + { + "epoch": 0.05, + "learning_rate": 9.94071326755728e-05, + "loss": 2.3609, + "step": 18395 + }, + { + "epoch": 0.05, + "learning_rate": 9.94068092139109e-05, + "loss": 2.377, + "step": 18400 + }, + { + "epoch": 0.05, + "learning_rate": 9.940648566456113e-05, + "loss": 2.364, + "step": 18405 + }, + { + "epoch": 0.05, + "learning_rate": 9.940616202752402e-05, + "loss": 2.4924, + "step": 18410 + }, + { + "epoch": 0.05, + "learning_rate": 9.940583830280017e-05, + "loss": 2.5202, + "step": 18415 + }, + { + "epoch": 0.05, + "learning_rate": 9.940551449039015e-05, + "loss": 2.4637, + "step": 18420 + }, + { + "epoch": 0.05, + "learning_rate": 9.940519059029454e-05, + "loss": 2.399, + "step": 18425 + }, + { + "epoch": 0.05, + "learning_rate": 9.94048666025139e-05, + "loss": 2.4099, + "step": 18430 + }, + { + "epoch": 0.05, + "learning_rate": 9.940454252704882e-05, + "loss": 2.2852, + "step": 18435 + }, + { + "epoch": 0.05, + "learning_rate": 9.940421836389988e-05, + "loss": 2.3394, + "step": 18440 + }, + { + "epoch": 0.05, + "learning_rate": 9.940389411306764e-05, + "loss": 2.502, + "step": 18445 + }, + { + "epoch": 0.05, + "learning_rate": 9.940356977455269e-05, + "loss": 2.3646, + "step": 18450 + }, + { + "epoch": 0.05, + "learning_rate": 9.940324534835559e-05, + "loss": 2.4183, + "step": 18455 + }, + { + "epoch": 0.05, + "learning_rate": 9.940292083447692e-05, + "loss": 2.3838, + "step": 18460 + }, + { + "epoch": 0.05, + "learning_rate": 9.940259623291725e-05, + "loss": 2.3648, + "step": 18465 + }, + { + "epoch": 0.05, + "learning_rate": 9.940227154367719e-05, + "loss": 2.4077, + "step": 18470 + }, + { + "epoch": 0.05, + "learning_rate": 9.940194676675728e-05, + "loss": 2.4954, + "step": 18475 + }, + { + "epoch": 0.05, + "learning_rate": 9.940162190215811e-05, + "loss": 2.3673, + "step": 18480 + }, + { + "epoch": 0.05, + "learning_rate": 9.940129694988025e-05, + "loss": 2.4975, + "step": 18485 + }, + { + "epoch": 0.05, + "learning_rate": 9.940097190992428e-05, + "loss": 2.3068, + "step": 18490 + }, + { + "epoch": 0.05, + "learning_rate": 9.940064678229079e-05, + "loss": 2.275, + "step": 18495 + }, + { + "epoch": 0.05, + "learning_rate": 9.940032156698034e-05, + "loss": 2.3577, + "step": 18500 + }, + { + "epoch": 0.05, + "learning_rate": 9.93999962639935e-05, + "loss": 2.3843, + "step": 18505 + }, + { + "epoch": 0.05, + "learning_rate": 9.939967087333086e-05, + "loss": 2.4186, + "step": 18510 + }, + { + "epoch": 0.05, + "learning_rate": 9.939934539499302e-05, + "loss": 2.3325, + "step": 18515 + }, + { + "epoch": 0.05, + "learning_rate": 9.939901982898053e-05, + "loss": 2.2715, + "step": 18520 + }, + { + "epoch": 0.05, + "learning_rate": 9.939869417529395e-05, + "loss": 2.4006, + "step": 18525 + }, + { + "epoch": 0.05, + "learning_rate": 9.93983684339339e-05, + "loss": 2.3545, + "step": 18530 + }, + { + "epoch": 0.05, + "learning_rate": 9.939804260490093e-05, + "loss": 2.4738, + "step": 18535 + }, + { + "epoch": 0.05, + "learning_rate": 9.939771668819563e-05, + "loss": 2.3666, + "step": 18540 + }, + { + "epoch": 0.05, + "learning_rate": 9.939739068381857e-05, + "loss": 2.4636, + "step": 18545 + }, + { + "epoch": 0.05, + "learning_rate": 9.939706459177034e-05, + "loss": 2.3631, + "step": 18550 + }, + { + "epoch": 0.05, + "learning_rate": 9.939673841205152e-05, + "loss": 2.4909, + "step": 18555 + }, + { + "epoch": 0.05, + "learning_rate": 9.939641214466267e-05, + "loss": 2.4701, + "step": 18560 + }, + { + "epoch": 0.05, + "learning_rate": 9.939608578960439e-05, + "loss": 2.5102, + "step": 18565 + }, + { + "epoch": 0.05, + "learning_rate": 9.939575934687724e-05, + "loss": 2.529, + "step": 18570 + }, + { + "epoch": 0.05, + "learning_rate": 9.939543281648183e-05, + "loss": 2.4345, + "step": 18575 + }, + { + "epoch": 0.05, + "learning_rate": 9.939510619841869e-05, + "loss": 2.4181, + "step": 18580 + }, + { + "epoch": 0.05, + "learning_rate": 9.939477949268845e-05, + "loss": 2.3192, + "step": 18585 + }, + { + "epoch": 0.05, + "learning_rate": 9.939445269929166e-05, + "loss": 2.4963, + "step": 18590 + }, + { + "epoch": 0.05, + "learning_rate": 9.939412581822891e-05, + "loss": 2.401, + "step": 18595 + }, + { + "epoch": 0.05, + "learning_rate": 9.939379884950078e-05, + "loss": 2.4824, + "step": 18600 + }, + { + "epoch": 0.05, + "learning_rate": 9.939347179310785e-05, + "loss": 2.4048, + "step": 18605 + }, + { + "epoch": 0.05, + "learning_rate": 9.93931446490507e-05, + "loss": 2.3699, + "step": 18610 + }, + { + "epoch": 0.05, + "learning_rate": 9.93928174173299e-05, + "loss": 2.495, + "step": 18615 + }, + { + "epoch": 0.05, + "learning_rate": 9.939249009794605e-05, + "loss": 2.2857, + "step": 18620 + }, + { + "epoch": 0.05, + "learning_rate": 9.939216269089972e-05, + "loss": 2.5412, + "step": 18625 + }, + { + "epoch": 0.05, + "learning_rate": 9.939183519619149e-05, + "loss": 2.5209, + "step": 18630 + }, + { + "epoch": 0.05, + "learning_rate": 9.939150761382195e-05, + "loss": 2.4226, + "step": 18635 + }, + { + "epoch": 0.05, + "learning_rate": 9.939117994379166e-05, + "loss": 2.5167, + "step": 18640 + }, + { + "epoch": 0.05, + "learning_rate": 9.939085218610124e-05, + "loss": 2.3741, + "step": 18645 + }, + { + "epoch": 0.05, + "learning_rate": 9.939052434075123e-05, + "loss": 2.3291, + "step": 18650 + }, + { + "epoch": 0.05, + "learning_rate": 9.939019640774224e-05, + "loss": 2.4266, + "step": 18655 + }, + { + "epoch": 0.05, + "learning_rate": 9.938986838707485e-05, + "loss": 2.495, + "step": 18660 + }, + { + "epoch": 0.05, + "learning_rate": 9.938954027874963e-05, + "loss": 2.4337, + "step": 18665 + }, + { + "epoch": 0.05, + "learning_rate": 9.938921208276715e-05, + "loss": 2.4905, + "step": 18670 + }, + { + "epoch": 0.05, + "learning_rate": 9.938888379912803e-05, + "loss": 2.4159, + "step": 18675 + }, + { + "epoch": 0.05, + "learning_rate": 9.938855542783281e-05, + "loss": 2.4044, + "step": 18680 + }, + { + "epoch": 0.05, + "learning_rate": 9.938822696888213e-05, + "loss": 2.5617, + "step": 18685 + }, + { + "epoch": 0.05, + "learning_rate": 9.938789842227652e-05, + "loss": 2.4, + "step": 18690 + }, + { + "epoch": 0.05, + "learning_rate": 9.938756978801658e-05, + "loss": 2.4153, + "step": 18695 + }, + { + "epoch": 0.05, + "learning_rate": 9.938724106610289e-05, + "loss": 2.4912, + "step": 18700 + }, + { + "epoch": 0.05, + "learning_rate": 9.938691225653603e-05, + "loss": 2.3643, + "step": 18705 + }, + { + "epoch": 0.05, + "learning_rate": 9.938658335931661e-05, + "loss": 2.4038, + "step": 18710 + }, + { + "epoch": 0.05, + "learning_rate": 9.938625437444519e-05, + "loss": 2.3758, + "step": 18715 + }, + { + "epoch": 0.05, + "learning_rate": 9.938592530192235e-05, + "loss": 2.4592, + "step": 18720 + }, + { + "epoch": 0.05, + "learning_rate": 9.938559614174869e-05, + "loss": 2.3149, + "step": 18725 + }, + { + "epoch": 0.05, + "learning_rate": 9.938526689392479e-05, + "loss": 2.3825, + "step": 18730 + }, + { + "epoch": 0.05, + "learning_rate": 9.938493755845123e-05, + "loss": 2.3977, + "step": 18735 + }, + { + "epoch": 0.05, + "learning_rate": 9.938460813532859e-05, + "loss": 2.4159, + "step": 18740 + }, + { + "epoch": 0.05, + "learning_rate": 9.938427862455746e-05, + "loss": 2.4155, + "step": 18745 + }, + { + "epoch": 0.05, + "learning_rate": 9.938394902613843e-05, + "loss": 2.4109, + "step": 18750 + }, + { + "epoch": 0.05, + "learning_rate": 9.93836193400721e-05, + "loss": 2.5042, + "step": 18755 + }, + { + "epoch": 0.05, + "learning_rate": 9.9383289566359e-05, + "loss": 2.309, + "step": 18760 + }, + { + "epoch": 0.05, + "learning_rate": 9.938295970499977e-05, + "loss": 2.4273, + "step": 18765 + }, + { + "epoch": 0.05, + "learning_rate": 9.9382629755995e-05, + "loss": 2.3752, + "step": 18770 + }, + { + "epoch": 0.05, + "learning_rate": 9.93822997193452e-05, + "loss": 2.4716, + "step": 18775 + }, + { + "epoch": 0.05, + "learning_rate": 9.938196959505105e-05, + "loss": 2.4193, + "step": 18780 + }, + { + "epoch": 0.05, + "learning_rate": 9.938163938311308e-05, + "loss": 2.3424, + "step": 18785 + }, + { + "epoch": 0.05, + "learning_rate": 9.93813090835319e-05, + "loss": 2.3909, + "step": 18790 + }, + { + "epoch": 0.05, + "learning_rate": 9.938097869630807e-05, + "loss": 2.4211, + "step": 18795 + }, + { + "epoch": 0.05, + "learning_rate": 9.93806482214422e-05, + "loss": 2.4396, + "step": 18800 + }, + { + "epoch": 0.05, + "learning_rate": 9.938031765893486e-05, + "loss": 2.4263, + "step": 18805 + }, + { + "epoch": 0.05, + "learning_rate": 9.937998700878664e-05, + "loss": 2.4767, + "step": 18810 + }, + { + "epoch": 0.05, + "learning_rate": 9.937965627099814e-05, + "loss": 2.4072, + "step": 18815 + }, + { + "epoch": 0.05, + "learning_rate": 9.937932544556996e-05, + "loss": 2.3451, + "step": 18820 + }, + { + "epoch": 0.05, + "learning_rate": 9.937899453250264e-05, + "loss": 2.4227, + "step": 18825 + }, + { + "epoch": 0.05, + "learning_rate": 9.937866353179682e-05, + "loss": 2.3192, + "step": 18830 + }, + { + "epoch": 0.05, + "learning_rate": 9.937833244345302e-05, + "loss": 2.2082, + "step": 18835 + }, + { + "epoch": 0.05, + "learning_rate": 9.93780012674719e-05, + "loss": 2.4443, + "step": 18840 + }, + { + "epoch": 0.05, + "learning_rate": 9.937767000385401e-05, + "loss": 2.4912, + "step": 18845 + }, + { + "epoch": 0.05, + "learning_rate": 9.937733865259995e-05, + "loss": 2.4612, + "step": 18850 + }, + { + "epoch": 0.05, + "learning_rate": 9.937700721371031e-05, + "loss": 2.3515, + "step": 18855 + }, + { + "epoch": 0.05, + "learning_rate": 9.937667568718565e-05, + "loss": 2.4559, + "step": 18860 + }, + { + "epoch": 0.05, + "learning_rate": 9.937634407302659e-05, + "loss": 2.3374, + "step": 18865 + }, + { + "epoch": 0.05, + "learning_rate": 9.937601237123371e-05, + "loss": 2.3884, + "step": 18870 + }, + { + "epoch": 0.05, + "learning_rate": 9.93756805818076e-05, + "loss": 2.334, + "step": 18875 + }, + { + "epoch": 0.05, + "learning_rate": 9.937534870474882e-05, + "loss": 2.4217, + "step": 18880 + }, + { + "epoch": 0.05, + "learning_rate": 9.937501674005802e-05, + "loss": 2.4721, + "step": 18885 + }, + { + "epoch": 0.05, + "learning_rate": 9.937468468773574e-05, + "loss": 2.4556, + "step": 18890 + }, + { + "epoch": 0.05, + "learning_rate": 9.937435254778257e-05, + "loss": 2.51, + "step": 18895 + }, + { + "epoch": 0.05, + "learning_rate": 9.937402032019912e-05, + "loss": 2.5116, + "step": 18900 + }, + { + "epoch": 0.05, + "learning_rate": 9.937368800498598e-05, + "loss": 2.4214, + "step": 18905 + }, + { + "epoch": 0.05, + "learning_rate": 9.937335560214373e-05, + "loss": 2.4426, + "step": 18910 + }, + { + "epoch": 0.05, + "learning_rate": 9.937302311167296e-05, + "loss": 2.2393, + "step": 18915 + }, + { + "epoch": 0.05, + "learning_rate": 9.937269053357425e-05, + "loss": 2.3614, + "step": 18920 + }, + { + "epoch": 0.05, + "learning_rate": 9.937235786784823e-05, + "loss": 2.4739, + "step": 18925 + }, + { + "epoch": 0.05, + "learning_rate": 9.937202511449544e-05, + "loss": 2.4965, + "step": 18930 + }, + { + "epoch": 0.05, + "learning_rate": 9.937169227351649e-05, + "loss": 2.3988, + "step": 18935 + }, + { + "epoch": 0.05, + "learning_rate": 9.937135934491199e-05, + "loss": 2.4044, + "step": 18940 + }, + { + "epoch": 0.05, + "learning_rate": 9.93710263286825e-05, + "loss": 2.4157, + "step": 18945 + }, + { + "epoch": 0.05, + "learning_rate": 9.937069322482863e-05, + "loss": 2.3333, + "step": 18950 + }, + { + "epoch": 0.05, + "learning_rate": 9.937036003335098e-05, + "loss": 2.356, + "step": 18955 + }, + { + "epoch": 0.05, + "learning_rate": 9.937002675425011e-05, + "loss": 2.3963, + "step": 18960 + }, + { + "epoch": 0.05, + "learning_rate": 9.936969338752664e-05, + "loss": 2.4154, + "step": 18965 + }, + { + "epoch": 0.05, + "learning_rate": 9.936935993318116e-05, + "loss": 2.4303, + "step": 18970 + }, + { + "epoch": 0.05, + "learning_rate": 9.936902639121425e-05, + "loss": 2.3697, + "step": 18975 + }, + { + "epoch": 0.05, + "learning_rate": 9.936869276162651e-05, + "loss": 2.4573, + "step": 18980 + }, + { + "epoch": 0.05, + "learning_rate": 9.93683590444185e-05, + "loss": 2.5026, + "step": 18985 + }, + { + "epoch": 0.05, + "learning_rate": 9.936802523959085e-05, + "loss": 2.4219, + "step": 18990 + }, + { + "epoch": 0.05, + "learning_rate": 9.936769134714416e-05, + "loss": 2.4013, + "step": 18995 + }, + { + "epoch": 0.05, + "learning_rate": 9.9367357367079e-05, + "loss": 2.4307, + "step": 19000 + }, + { + "epoch": 0.05, + "learning_rate": 9.936702329939595e-05, + "loss": 2.3479, + "step": 19005 + }, + { + "epoch": 0.05, + "learning_rate": 9.936668914409565e-05, + "loss": 2.4379, + "step": 19010 + }, + { + "epoch": 0.05, + "learning_rate": 9.936635490117865e-05, + "loss": 2.3091, + "step": 19015 + }, + { + "epoch": 0.05, + "learning_rate": 9.936602057064554e-05, + "loss": 2.4263, + "step": 19020 + }, + { + "epoch": 0.05, + "learning_rate": 9.936568615249696e-05, + "loss": 2.5074, + "step": 19025 + }, + { + "epoch": 0.05, + "learning_rate": 9.936535164673346e-05, + "loss": 2.4615, + "step": 19030 + }, + { + "epoch": 0.05, + "learning_rate": 9.936501705335563e-05, + "loss": 2.3949, + "step": 19035 + }, + { + "epoch": 0.05, + "learning_rate": 9.93646823723641e-05, + "loss": 2.3662, + "step": 19040 + }, + { + "epoch": 0.05, + "learning_rate": 9.936434760375945e-05, + "loss": 2.3818, + "step": 19045 + }, + { + "epoch": 0.05, + "learning_rate": 9.936401274754226e-05, + "loss": 2.5034, + "step": 19050 + }, + { + "epoch": 0.05, + "learning_rate": 9.936367780371314e-05, + "loss": 2.4383, + "step": 19055 + }, + { + "epoch": 0.05, + "learning_rate": 9.936334277227266e-05, + "loss": 2.3845, + "step": 19060 + }, + { + "epoch": 0.05, + "learning_rate": 9.936300765322146e-05, + "loss": 2.3222, + "step": 19065 + }, + { + "epoch": 0.05, + "learning_rate": 9.936267244656009e-05, + "loss": 2.3915, + "step": 19070 + }, + { + "epoch": 0.05, + "learning_rate": 9.936233715228916e-05, + "loss": 2.3296, + "step": 19075 + }, + { + "epoch": 0.05, + "learning_rate": 9.936200177040929e-05, + "loss": 2.3419, + "step": 19080 + }, + { + "epoch": 0.05, + "learning_rate": 9.936166630092103e-05, + "loss": 2.3701, + "step": 19085 + }, + { + "epoch": 0.05, + "learning_rate": 9.936133074382501e-05, + "loss": 2.3286, + "step": 19090 + }, + { + "epoch": 0.05, + "learning_rate": 9.936099509912181e-05, + "loss": 2.5254, + "step": 19095 + }, + { + "epoch": 0.05, + "learning_rate": 9.936065936681203e-05, + "loss": 2.3507, + "step": 19100 + }, + { + "epoch": 0.05, + "learning_rate": 9.936032354689626e-05, + "loss": 2.4717, + "step": 19105 + }, + { + "epoch": 0.05, + "learning_rate": 9.93599876393751e-05, + "loss": 2.4242, + "step": 19110 + }, + { + "epoch": 0.05, + "learning_rate": 9.935965164424916e-05, + "loss": 2.4696, + "step": 19115 + }, + { + "epoch": 0.05, + "learning_rate": 9.935931556151901e-05, + "loss": 2.3773, + "step": 19120 + }, + { + "epoch": 0.05, + "learning_rate": 9.935897939118526e-05, + "loss": 2.4756, + "step": 19125 + }, + { + "epoch": 0.05, + "learning_rate": 9.935864313324851e-05, + "loss": 2.4229, + "step": 19130 + }, + { + "epoch": 0.05, + "learning_rate": 9.935830678770937e-05, + "loss": 2.4172, + "step": 19135 + }, + { + "epoch": 0.05, + "learning_rate": 9.935797035456842e-05, + "loss": 2.3847, + "step": 19140 + }, + { + "epoch": 0.05, + "learning_rate": 9.935763383382624e-05, + "loss": 2.3651, + "step": 19145 + }, + { + "epoch": 0.05, + "learning_rate": 9.935729722548346e-05, + "loss": 2.4347, + "step": 19150 + }, + { + "epoch": 0.05, + "learning_rate": 9.935696052954065e-05, + "loss": 2.4075, + "step": 19155 + }, + { + "epoch": 0.05, + "learning_rate": 9.935662374599842e-05, + "loss": 2.346, + "step": 19160 + }, + { + "epoch": 0.05, + "learning_rate": 9.935628687485737e-05, + "loss": 2.395, + "step": 19165 + }, + { + "epoch": 0.05, + "learning_rate": 9.935594991611811e-05, + "loss": 2.4314, + "step": 19170 + }, + { + "epoch": 0.05, + "learning_rate": 9.935561286978121e-05, + "loss": 2.4574, + "step": 19175 + }, + { + "epoch": 0.05, + "learning_rate": 9.935527573584728e-05, + "loss": 2.4259, + "step": 19180 + }, + { + "epoch": 0.05, + "learning_rate": 9.935493851431692e-05, + "loss": 2.394, + "step": 19185 + }, + { + "epoch": 0.05, + "learning_rate": 9.935460120519073e-05, + "loss": 2.4896, + "step": 19190 + }, + { + "epoch": 0.05, + "learning_rate": 9.935426380846933e-05, + "loss": 2.4103, + "step": 19195 + }, + { + "epoch": 0.05, + "learning_rate": 9.935392632415327e-05, + "loss": 2.5981, + "step": 19200 + }, + { + "epoch": 0.05, + "learning_rate": 9.935358875224319e-05, + "loss": 2.3989, + "step": 19205 + }, + { + "epoch": 0.05, + "learning_rate": 9.935325109273965e-05, + "loss": 2.3942, + "step": 19210 + }, + { + "epoch": 0.05, + "learning_rate": 9.935291334564329e-05, + "loss": 2.4534, + "step": 19215 + }, + { + "epoch": 0.05, + "learning_rate": 9.935257551095471e-05, + "loss": 2.4487, + "step": 19220 + }, + { + "epoch": 0.05, + "learning_rate": 9.935223758867446e-05, + "loss": 2.3842, + "step": 19225 + }, + { + "epoch": 0.05, + "learning_rate": 9.93518995788032e-05, + "loss": 2.3776, + "step": 19230 + }, + { + "epoch": 0.05, + "learning_rate": 9.935156148134149e-05, + "loss": 2.3643, + "step": 19235 + }, + { + "epoch": 0.05, + "learning_rate": 9.935122329628994e-05, + "loss": 2.3344, + "step": 19240 + }, + { + "epoch": 0.05, + "learning_rate": 9.935088502364916e-05, + "loss": 2.4025, + "step": 19245 + }, + { + "epoch": 0.05, + "learning_rate": 9.935054666341975e-05, + "loss": 2.5025, + "step": 19250 + }, + { + "epoch": 0.05, + "learning_rate": 9.935020821560229e-05, + "loss": 2.4085, + "step": 19255 + }, + { + "epoch": 0.05, + "learning_rate": 9.934986968019739e-05, + "loss": 2.3014, + "step": 19260 + }, + { + "epoch": 0.05, + "learning_rate": 9.934953105720567e-05, + "loss": 2.3582, + "step": 19265 + }, + { + "epoch": 0.05, + "learning_rate": 9.934919234662771e-05, + "loss": 2.4157, + "step": 19270 + }, + { + "epoch": 0.05, + "learning_rate": 9.934885354846411e-05, + "loss": 2.4639, + "step": 19275 + }, + { + "epoch": 0.05, + "learning_rate": 9.934851466271549e-05, + "loss": 2.4146, + "step": 19280 + }, + { + "epoch": 0.05, + "learning_rate": 9.934817568938243e-05, + "loss": 2.3762, + "step": 19285 + }, + { + "epoch": 0.05, + "learning_rate": 9.934783662846554e-05, + "loss": 2.3798, + "step": 19290 + }, + { + "epoch": 0.05, + "learning_rate": 9.934749747996543e-05, + "loss": 2.4962, + "step": 19295 + }, + { + "epoch": 0.05, + "learning_rate": 9.93471582438827e-05, + "loss": 2.4917, + "step": 19300 + }, + { + "epoch": 0.05, + "learning_rate": 9.934681892021794e-05, + "loss": 2.3184, + "step": 19305 + }, + { + "epoch": 0.05, + "learning_rate": 9.934647950897176e-05, + "loss": 2.315, + "step": 19310 + }, + { + "epoch": 0.05, + "learning_rate": 9.934614001014477e-05, + "loss": 2.3702, + "step": 19315 + }, + { + "epoch": 0.05, + "learning_rate": 9.934580042373756e-05, + "loss": 2.4455, + "step": 19320 + }, + { + "epoch": 0.05, + "learning_rate": 9.934546074975074e-05, + "loss": 2.3098, + "step": 19325 + }, + { + "epoch": 0.05, + "learning_rate": 9.93451209881849e-05, + "loss": 2.5048, + "step": 19330 + }, + { + "epoch": 0.05, + "learning_rate": 9.934478113904066e-05, + "loss": 2.5021, + "step": 19335 + }, + { + "epoch": 0.05, + "learning_rate": 9.934444120231862e-05, + "loss": 2.3511, + "step": 19340 + }, + { + "epoch": 0.05, + "learning_rate": 9.934410117801938e-05, + "loss": 2.5518, + "step": 19345 + }, + { + "epoch": 0.05, + "learning_rate": 9.934376106614352e-05, + "loss": 2.3545, + "step": 19350 + }, + { + "epoch": 0.05, + "learning_rate": 9.934342086669169e-05, + "loss": 2.4314, + "step": 19355 + }, + { + "epoch": 0.05, + "learning_rate": 9.934308057966446e-05, + "loss": 2.4468, + "step": 19360 + }, + { + "epoch": 0.05, + "learning_rate": 9.934274020506246e-05, + "loss": 2.5128, + "step": 19365 + }, + { + "epoch": 0.05, + "learning_rate": 9.934239974288627e-05, + "loss": 2.2825, + "step": 19370 + }, + { + "epoch": 0.05, + "learning_rate": 9.934205919313649e-05, + "loss": 2.4339, + "step": 19375 + }, + { + "epoch": 0.05, + "learning_rate": 9.934171855581375e-05, + "loss": 2.4187, + "step": 19380 + }, + { + "epoch": 0.05, + "learning_rate": 9.934137783091865e-05, + "loss": 2.3656, + "step": 19385 + }, + { + "epoch": 0.05, + "learning_rate": 9.934103701845176e-05, + "loss": 2.3657, + "step": 19390 + }, + { + "epoch": 0.05, + "learning_rate": 9.934069611841375e-05, + "loss": 2.3626, + "step": 19395 + }, + { + "epoch": 0.05, + "learning_rate": 9.934035513080516e-05, + "loss": 2.5472, + "step": 19400 + }, + { + "epoch": 0.05, + "learning_rate": 9.934001405562661e-05, + "loss": 2.5348, + "step": 19405 + }, + { + "epoch": 0.05, + "learning_rate": 9.933967289287874e-05, + "loss": 2.4378, + "step": 19410 + }, + { + "epoch": 0.05, + "learning_rate": 9.933933164256213e-05, + "loss": 2.3843, + "step": 19415 + }, + { + "epoch": 0.05, + "learning_rate": 9.933899030467738e-05, + "loss": 2.4056, + "step": 19420 + }, + { + "epoch": 0.05, + "learning_rate": 9.933864887922511e-05, + "loss": 2.4294, + "step": 19425 + }, + { + "epoch": 0.05, + "learning_rate": 9.933830736620592e-05, + "loss": 2.4363, + "step": 19430 + }, + { + "epoch": 0.05, + "learning_rate": 9.933796576562042e-05, + "loss": 2.3676, + "step": 19435 + }, + { + "epoch": 0.05, + "learning_rate": 9.93376240774692e-05, + "loss": 2.3369, + "step": 19440 + }, + { + "epoch": 0.05, + "learning_rate": 9.933728230175287e-05, + "loss": 2.4083, + "step": 19445 + }, + { + "epoch": 0.05, + "learning_rate": 9.933694043847206e-05, + "loss": 2.4538, + "step": 19450 + }, + { + "epoch": 0.05, + "learning_rate": 9.933659848762738e-05, + "loss": 2.4248, + "step": 19455 + }, + { + "epoch": 0.05, + "learning_rate": 9.93362564492194e-05, + "loss": 2.4782, + "step": 19460 + }, + { + "epoch": 0.05, + "learning_rate": 9.933591432324874e-05, + "loss": 2.4906, + "step": 19465 + }, + { + "epoch": 0.05, + "learning_rate": 9.933557210971603e-05, + "loss": 2.4404, + "step": 19470 + }, + { + "epoch": 0.05, + "learning_rate": 9.933522980862185e-05, + "loss": 2.4439, + "step": 19475 + }, + { + "epoch": 0.05, + "learning_rate": 9.933488741996681e-05, + "loss": 2.3882, + "step": 19480 + }, + { + "epoch": 0.05, + "learning_rate": 9.933454494375154e-05, + "loss": 2.452, + "step": 19485 + }, + { + "epoch": 0.05, + "learning_rate": 9.933420237997662e-05, + "loss": 2.3791, + "step": 19490 + }, + { + "epoch": 0.05, + "learning_rate": 9.933385972864268e-05, + "loss": 2.4269, + "step": 19495 + }, + { + "epoch": 0.05, + "learning_rate": 9.933351698975033e-05, + "loss": 2.4224, + "step": 19500 + }, + { + "epoch": 0.05, + "learning_rate": 9.933317416330015e-05, + "loss": 2.4184, + "step": 19505 + }, + { + "epoch": 0.05, + "learning_rate": 9.933283124929278e-05, + "loss": 2.3712, + "step": 19510 + }, + { + "epoch": 0.05, + "learning_rate": 9.93324882477288e-05, + "loss": 2.4342, + "step": 19515 + }, + { + "epoch": 0.05, + "learning_rate": 9.933214515860885e-05, + "loss": 2.52, + "step": 19520 + }, + { + "epoch": 0.05, + "learning_rate": 9.933180198193353e-05, + "loss": 2.4027, + "step": 19525 + }, + { + "epoch": 0.05, + "learning_rate": 9.933145871770342e-05, + "loss": 2.2566, + "step": 19530 + }, + { + "epoch": 0.05, + "learning_rate": 9.933111536591916e-05, + "loss": 2.4106, + "step": 19535 + }, + { + "epoch": 0.05, + "learning_rate": 9.933077192658134e-05, + "loss": 2.3345, + "step": 19540 + }, + { + "epoch": 0.05, + "learning_rate": 9.93304283996906e-05, + "loss": 2.4006, + "step": 19545 + }, + { + "epoch": 0.05, + "learning_rate": 9.933008478524752e-05, + "loss": 2.4226, + "step": 19550 + }, + { + "epoch": 0.05, + "learning_rate": 9.932974108325272e-05, + "loss": 2.5009, + "step": 19555 + }, + { + "epoch": 0.05, + "learning_rate": 9.93293972937068e-05, + "loss": 2.4339, + "step": 19560 + }, + { + "epoch": 0.05, + "learning_rate": 9.932905341661038e-05, + "loss": 2.4253, + "step": 19565 + }, + { + "epoch": 0.05, + "learning_rate": 9.932870945196409e-05, + "loss": 2.4076, + "step": 19570 + }, + { + "epoch": 0.05, + "learning_rate": 9.93283653997685e-05, + "loss": 2.463, + "step": 19575 + }, + { + "epoch": 0.05, + "learning_rate": 9.932802126002427e-05, + "loss": 2.3856, + "step": 19580 + }, + { + "epoch": 0.05, + "learning_rate": 9.932767703273195e-05, + "loss": 2.3991, + "step": 19585 + }, + { + "epoch": 0.05, + "learning_rate": 9.93273327178922e-05, + "loss": 2.4397, + "step": 19590 + }, + { + "epoch": 0.05, + "learning_rate": 9.932698831550561e-05, + "loss": 2.4173, + "step": 19595 + }, + { + "epoch": 0.05, + "learning_rate": 9.93266438255728e-05, + "loss": 2.355, + "step": 19600 + }, + { + "epoch": 0.05, + "learning_rate": 9.932629924809435e-05, + "loss": 2.4573, + "step": 19605 + }, + { + "epoch": 0.05, + "learning_rate": 9.932595458307093e-05, + "loss": 2.5304, + "step": 19610 + }, + { + "epoch": 0.05, + "learning_rate": 9.932560983050311e-05, + "loss": 2.2904, + "step": 19615 + }, + { + "epoch": 0.05, + "learning_rate": 9.932526499039152e-05, + "loss": 2.4407, + "step": 19620 + }, + { + "epoch": 0.05, + "learning_rate": 9.932492006273675e-05, + "loss": 2.3973, + "step": 19625 + }, + { + "epoch": 0.05, + "learning_rate": 9.932457504753943e-05, + "loss": 2.3372, + "step": 19630 + }, + { + "epoch": 0.05, + "learning_rate": 9.932422994480018e-05, + "loss": 2.4738, + "step": 19635 + }, + { + "epoch": 0.05, + "learning_rate": 9.932388475451956e-05, + "loss": 2.3849, + "step": 19640 + }, + { + "epoch": 0.05, + "learning_rate": 9.932353947669827e-05, + "loss": 2.4364, + "step": 19645 + }, + { + "epoch": 0.05, + "learning_rate": 9.932319411133687e-05, + "loss": 2.3943, + "step": 19650 + }, + { + "epoch": 0.05, + "learning_rate": 9.932284865843595e-05, + "loss": 2.3879, + "step": 19655 + }, + { + "epoch": 0.05, + "learning_rate": 9.932250311799618e-05, + "loss": 2.4869, + "step": 19660 + }, + { + "epoch": 0.05, + "learning_rate": 9.932215749001815e-05, + "loss": 2.3205, + "step": 19665 + }, + { + "epoch": 0.05, + "learning_rate": 9.932181177450244e-05, + "loss": 2.322, + "step": 19670 + }, + { + "epoch": 0.05, + "learning_rate": 9.93214659714497e-05, + "loss": 2.4774, + "step": 19675 + }, + { + "epoch": 0.05, + "learning_rate": 9.932112008086056e-05, + "loss": 2.3244, + "step": 19680 + }, + { + "epoch": 0.05, + "learning_rate": 9.93207741027356e-05, + "loss": 2.3187, + "step": 19685 + }, + { + "epoch": 0.05, + "learning_rate": 9.932042803707545e-05, + "loss": 2.4279, + "step": 19690 + }, + { + "epoch": 0.05, + "learning_rate": 9.932008188388069e-05, + "loss": 2.4153, + "step": 19695 + }, + { + "epoch": 0.05, + "learning_rate": 9.931973564315197e-05, + "loss": 2.3708, + "step": 19700 + }, + { + "epoch": 0.05, + "learning_rate": 9.931938931488993e-05, + "loss": 2.4468, + "step": 19705 + }, + { + "epoch": 0.05, + "learning_rate": 9.931904289909513e-05, + "loss": 2.4306, + "step": 19710 + }, + { + "epoch": 0.05, + "learning_rate": 9.931869639576821e-05, + "loss": 2.3818, + "step": 19715 + }, + { + "epoch": 0.05, + "learning_rate": 9.931834980490977e-05, + "loss": 2.394, + "step": 19720 + }, + { + "epoch": 0.05, + "learning_rate": 9.931800312652045e-05, + "loss": 2.4715, + "step": 19725 + }, + { + "epoch": 0.05, + "learning_rate": 9.931765636060084e-05, + "loss": 2.4801, + "step": 19730 + }, + { + "epoch": 0.05, + "learning_rate": 9.931730950715158e-05, + "loss": 2.5016, + "step": 19735 + }, + { + "epoch": 0.05, + "learning_rate": 9.931696256617326e-05, + "loss": 2.2802, + "step": 19740 + }, + { + "epoch": 0.05, + "learning_rate": 9.931661553766654e-05, + "loss": 2.2595, + "step": 19745 + }, + { + "epoch": 0.05, + "learning_rate": 9.931626842163198e-05, + "loss": 2.2038, + "step": 19750 + }, + { + "epoch": 0.05, + "learning_rate": 9.931592121807022e-05, + "loss": 2.3828, + "step": 19755 + }, + { + "epoch": 0.05, + "learning_rate": 9.931557392698187e-05, + "loss": 2.2698, + "step": 19760 + }, + { + "epoch": 0.05, + "learning_rate": 9.931522654836758e-05, + "loss": 2.4323, + "step": 19765 + }, + { + "epoch": 0.05, + "learning_rate": 9.931487908222791e-05, + "loss": 2.4542, + "step": 19770 + }, + { + "epoch": 0.05, + "learning_rate": 9.931453152856353e-05, + "loss": 2.3954, + "step": 19775 + }, + { + "epoch": 0.05, + "learning_rate": 9.931418388737503e-05, + "loss": 2.4959, + "step": 19780 + }, + { + "epoch": 0.05, + "learning_rate": 9.931383615866304e-05, + "loss": 2.2972, + "step": 19785 + }, + { + "epoch": 0.05, + "learning_rate": 9.931348834242816e-05, + "loss": 2.3795, + "step": 19790 + }, + { + "epoch": 0.05, + "learning_rate": 9.931314043867099e-05, + "loss": 2.3395, + "step": 19795 + }, + { + "epoch": 0.05, + "learning_rate": 9.93127924473922e-05, + "loss": 2.4177, + "step": 19800 + }, + { + "epoch": 0.05, + "learning_rate": 9.931244436859237e-05, + "loss": 2.3667, + "step": 19805 + }, + { + "epoch": 0.05, + "learning_rate": 9.931209620227213e-05, + "loss": 2.3732, + "step": 19810 + }, + { + "epoch": 0.05, + "learning_rate": 9.93117479484321e-05, + "loss": 2.3842, + "step": 19815 + }, + { + "epoch": 0.05, + "learning_rate": 9.93113996070729e-05, + "loss": 2.3571, + "step": 19820 + }, + { + "epoch": 0.05, + "learning_rate": 9.931105117819513e-05, + "loss": 2.4143, + "step": 19825 + }, + { + "epoch": 0.05, + "learning_rate": 9.931070266179943e-05, + "loss": 2.3332, + "step": 19830 + }, + { + "epoch": 0.05, + "learning_rate": 9.931035405788639e-05, + "loss": 2.3879, + "step": 19835 + }, + { + "epoch": 0.05, + "learning_rate": 9.931000536645666e-05, + "loss": 2.443, + "step": 19840 + }, + { + "epoch": 0.05, + "learning_rate": 9.930965658751086e-05, + "loss": 2.3054, + "step": 19845 + }, + { + "epoch": 0.05, + "learning_rate": 9.930930772104958e-05, + "loss": 2.3461, + "step": 19850 + }, + { + "epoch": 0.05, + "learning_rate": 9.930895876707346e-05, + "loss": 2.4054, + "step": 19855 + }, + { + "epoch": 0.05, + "learning_rate": 9.930860972558311e-05, + "loss": 2.3699, + "step": 19860 + }, + { + "epoch": 0.05, + "learning_rate": 9.930826059657917e-05, + "loss": 2.3948, + "step": 19865 + }, + { + "epoch": 0.05, + "learning_rate": 9.930791138006221e-05, + "loss": 2.4128, + "step": 19870 + }, + { + "epoch": 0.05, + "learning_rate": 9.930756207603291e-05, + "loss": 2.4012, + "step": 19875 + }, + { + "epoch": 0.05, + "learning_rate": 9.930721268449186e-05, + "loss": 2.3934, + "step": 19880 + }, + { + "epoch": 0.05, + "learning_rate": 9.930686320543966e-05, + "loss": 2.4494, + "step": 19885 + }, + { + "epoch": 0.05, + "learning_rate": 9.9306513638877e-05, + "loss": 2.4052, + "step": 19890 + }, + { + "epoch": 0.05, + "learning_rate": 9.930616398480443e-05, + "loss": 2.4486, + "step": 19895 + }, + { + "epoch": 0.05, + "learning_rate": 9.930581424322258e-05, + "loss": 2.5129, + "step": 19900 + }, + { + "epoch": 0.05, + "learning_rate": 9.930546441413211e-05, + "loss": 2.3661, + "step": 19905 + }, + { + "epoch": 0.05, + "learning_rate": 9.930511449753359e-05, + "loss": 2.2858, + "step": 19910 + }, + { + "epoch": 0.05, + "learning_rate": 9.930476449342769e-05, + "loss": 2.35, + "step": 19915 + }, + { + "epoch": 0.05, + "learning_rate": 9.9304414401815e-05, + "loss": 2.2459, + "step": 19920 + }, + { + "epoch": 0.05, + "learning_rate": 9.930406422269614e-05, + "loss": 2.3407, + "step": 19925 + }, + { + "epoch": 0.05, + "learning_rate": 9.930371395607176e-05, + "loss": 2.3183, + "step": 19930 + }, + { + "epoch": 0.05, + "learning_rate": 9.930336360194245e-05, + "loss": 2.3193, + "step": 19935 + }, + { + "epoch": 0.05, + "learning_rate": 9.930301316030886e-05, + "loss": 2.4396, + "step": 19940 + }, + { + "epoch": 0.05, + "learning_rate": 9.930266263117158e-05, + "loss": 2.4304, + "step": 19945 + }, + { + "epoch": 0.05, + "learning_rate": 9.930231201453126e-05, + "loss": 2.6018, + "step": 19950 + }, + { + "epoch": 0.05, + "learning_rate": 9.930196131038849e-05, + "loss": 2.337, + "step": 19955 + }, + { + "epoch": 0.05, + "learning_rate": 9.930161051874393e-05, + "loss": 2.4205, + "step": 19960 + }, + { + "epoch": 0.05, + "learning_rate": 9.930125963959818e-05, + "loss": 2.3164, + "step": 19965 + }, + { + "epoch": 0.05, + "learning_rate": 9.930090867295188e-05, + "loss": 2.3998, + "step": 19970 + }, + { + "epoch": 0.05, + "learning_rate": 9.930055761880564e-05, + "loss": 2.351, + "step": 19975 + }, + { + "epoch": 0.05, + "learning_rate": 9.930020647716008e-05, + "loss": 2.4399, + "step": 19980 + }, + { + "epoch": 0.05, + "learning_rate": 9.929985524801581e-05, + "loss": 2.4244, + "step": 19985 + }, + { + "epoch": 0.05, + "learning_rate": 9.92995039313735e-05, + "loss": 2.4709, + "step": 19990 + }, + { + "epoch": 0.05, + "learning_rate": 9.929915252723374e-05, + "loss": 2.4771, + "step": 19995 + }, + { + "epoch": 0.05, + "learning_rate": 9.929880103559714e-05, + "loss": 2.4316, + "step": 20000 + }, + { + "epoch": 0.05, + "learning_rate": 9.929844945646435e-05, + "loss": 2.3545, + "step": 20005 + }, + { + "epoch": 0.05, + "learning_rate": 9.929809778983601e-05, + "loss": 2.3807, + "step": 20010 + }, + { + "epoch": 0.05, + "learning_rate": 9.929774603571271e-05, + "loss": 2.4256, + "step": 20015 + }, + { + "epoch": 0.05, + "learning_rate": 9.929739419409507e-05, + "loss": 2.3114, + "step": 20020 + }, + { + "epoch": 0.05, + "learning_rate": 9.929704226498373e-05, + "loss": 2.3944, + "step": 20025 + }, + { + "epoch": 0.05, + "learning_rate": 9.929669024837933e-05, + "loss": 2.3297, + "step": 20030 + }, + { + "epoch": 0.05, + "learning_rate": 9.929633814428246e-05, + "loss": 2.4334, + "step": 20035 + }, + { + "epoch": 0.05, + "learning_rate": 9.929598595269379e-05, + "loss": 2.3695, + "step": 20040 + }, + { + "epoch": 0.05, + "learning_rate": 9.929563367361388e-05, + "loss": 2.3745, + "step": 20045 + }, + { + "epoch": 0.05, + "learning_rate": 9.929528130704342e-05, + "loss": 2.3592, + "step": 20050 + }, + { + "epoch": 0.05, + "learning_rate": 9.9294928852983e-05, + "loss": 2.4016, + "step": 20055 + }, + { + "epoch": 0.05, + "learning_rate": 9.929457631143326e-05, + "loss": 2.3419, + "step": 20060 + }, + { + "epoch": 0.05, + "learning_rate": 9.929422368239483e-05, + "loss": 2.3063, + "step": 20065 + }, + { + "epoch": 0.05, + "learning_rate": 9.92938709658683e-05, + "loss": 2.4001, + "step": 20070 + }, + { + "epoch": 0.05, + "learning_rate": 9.929351816185435e-05, + "loss": 2.3843, + "step": 20075 + }, + { + "epoch": 0.05, + "learning_rate": 9.929316527035357e-05, + "loss": 2.4574, + "step": 20080 + }, + { + "epoch": 0.05, + "learning_rate": 9.929281229136659e-05, + "loss": 2.3523, + "step": 20085 + }, + { + "epoch": 0.05, + "learning_rate": 9.929245922489404e-05, + "loss": 2.4576, + "step": 20090 + }, + { + "epoch": 0.05, + "learning_rate": 9.929210607093655e-05, + "loss": 2.4078, + "step": 20095 + }, + { + "epoch": 0.05, + "learning_rate": 9.929175282949474e-05, + "loss": 2.384, + "step": 20100 + }, + { + "epoch": 0.05, + "learning_rate": 9.929139950056926e-05, + "loss": 2.2969, + "step": 20105 + }, + { + "epoch": 0.05, + "learning_rate": 9.929104608416071e-05, + "loss": 2.5654, + "step": 20110 + }, + { + "epoch": 0.05, + "learning_rate": 9.929069258026971e-05, + "loss": 2.3722, + "step": 20115 + }, + { + "epoch": 0.05, + "learning_rate": 9.929033898889692e-05, + "loss": 2.4103, + "step": 20120 + }, + { + "epoch": 0.05, + "learning_rate": 9.928998531004294e-05, + "loss": 2.409, + "step": 20125 + }, + { + "epoch": 0.05, + "learning_rate": 9.928963154370843e-05, + "loss": 2.391, + "step": 20130 + }, + { + "epoch": 0.05, + "learning_rate": 9.928927768989398e-05, + "loss": 2.4521, + "step": 20135 + }, + { + "epoch": 0.05, + "learning_rate": 9.928892374860024e-05, + "loss": 2.4582, + "step": 20140 + }, + { + "epoch": 0.05, + "learning_rate": 9.928856971982783e-05, + "loss": 2.3449, + "step": 20145 + }, + { + "epoch": 0.05, + "learning_rate": 9.928821560357738e-05, + "loss": 2.4068, + "step": 20150 + }, + { + "epoch": 0.05, + "learning_rate": 9.928786139984951e-05, + "loss": 2.4911, + "step": 20155 + }, + { + "epoch": 0.05, + "learning_rate": 9.928750710864487e-05, + "loss": 2.2703, + "step": 20160 + }, + { + "epoch": 0.05, + "learning_rate": 9.928715272996408e-05, + "loss": 2.5359, + "step": 20165 + }, + { + "epoch": 0.05, + "learning_rate": 9.928679826380775e-05, + "loss": 2.3503, + "step": 20170 + }, + { + "epoch": 0.05, + "learning_rate": 9.928644371017655e-05, + "loss": 2.4202, + "step": 20175 + }, + { + "epoch": 0.05, + "learning_rate": 9.928608906907106e-05, + "loss": 2.4274, + "step": 20180 + }, + { + "epoch": 0.05, + "learning_rate": 9.928573434049195e-05, + "loss": 2.4595, + "step": 20185 + }, + { + "epoch": 0.05, + "learning_rate": 9.928537952443984e-05, + "loss": 2.3859, + "step": 20190 + }, + { + "epoch": 0.05, + "learning_rate": 9.928502462091533e-05, + "loss": 2.3457, + "step": 20195 + }, + { + "epoch": 0.05, + "learning_rate": 9.92846696299191e-05, + "loss": 2.3832, + "step": 20200 + }, + { + "epoch": 0.05, + "learning_rate": 9.928431455145173e-05, + "loss": 2.4064, + "step": 20205 + }, + { + "epoch": 0.05, + "learning_rate": 9.928395938551387e-05, + "loss": 2.3306, + "step": 20210 + }, + { + "epoch": 0.05, + "learning_rate": 9.928360413210616e-05, + "loss": 2.3974, + "step": 20215 + }, + { + "epoch": 0.05, + "learning_rate": 9.928324879122923e-05, + "loss": 2.2759, + "step": 20220 + }, + { + "epoch": 0.05, + "learning_rate": 9.928289336288371e-05, + "loss": 2.4036, + "step": 20225 + }, + { + "epoch": 0.05, + "learning_rate": 9.928253784707023e-05, + "loss": 2.3991, + "step": 20230 + }, + { + "epoch": 0.05, + "learning_rate": 9.92821822437894e-05, + "loss": 2.521, + "step": 20235 + }, + { + "epoch": 0.05, + "learning_rate": 9.928182655304187e-05, + "loss": 2.3759, + "step": 20240 + }, + { + "epoch": 0.05, + "learning_rate": 9.928147077482827e-05, + "loss": 2.4835, + "step": 20245 + }, + { + "epoch": 0.05, + "learning_rate": 9.928111490914923e-05, + "loss": 2.5384, + "step": 20250 + }, + { + "epoch": 0.05, + "learning_rate": 9.928075895600538e-05, + "loss": 2.4639, + "step": 20255 + }, + { + "epoch": 0.05, + "learning_rate": 9.928040291539736e-05, + "loss": 2.4631, + "step": 20260 + }, + { + "epoch": 0.05, + "learning_rate": 9.92800467873258e-05, + "loss": 2.4614, + "step": 20265 + }, + { + "epoch": 0.05, + "learning_rate": 9.927969057179132e-05, + "loss": 2.4612, + "step": 20270 + }, + { + "epoch": 0.05, + "learning_rate": 9.927933426879457e-05, + "loss": 2.3747, + "step": 20275 + }, + { + "epoch": 0.05, + "learning_rate": 9.927897787833616e-05, + "loss": 2.2939, + "step": 20280 + }, + { + "epoch": 0.05, + "learning_rate": 9.927862140041673e-05, + "loss": 2.3911, + "step": 20285 + }, + { + "epoch": 0.05, + "learning_rate": 9.927826483503692e-05, + "loss": 2.4963, + "step": 20290 + }, + { + "epoch": 0.05, + "learning_rate": 9.927790818219738e-05, + "loss": 2.3364, + "step": 20295 + }, + { + "epoch": 0.05, + "learning_rate": 9.92775514418987e-05, + "loss": 2.4765, + "step": 20300 + }, + { + "epoch": 0.05, + "learning_rate": 9.927719461414155e-05, + "loss": 2.6168, + "step": 20305 + }, + { + "epoch": 0.05, + "learning_rate": 9.927683769892655e-05, + "loss": 2.46, + "step": 20310 + }, + { + "epoch": 0.05, + "learning_rate": 9.927648069625433e-05, + "loss": 2.376, + "step": 20315 + }, + { + "epoch": 0.05, + "learning_rate": 9.927612360612553e-05, + "loss": 2.2781, + "step": 20320 + }, + { + "epoch": 0.05, + "learning_rate": 9.927576642854078e-05, + "loss": 2.4303, + "step": 20325 + }, + { + "epoch": 0.05, + "learning_rate": 9.92754091635007e-05, + "loss": 2.5412, + "step": 20330 + }, + { + "epoch": 0.05, + "learning_rate": 9.927505181100595e-05, + "loss": 2.4186, + "step": 20335 + }, + { + "epoch": 0.05, + "learning_rate": 9.927469437105714e-05, + "loss": 2.4706, + "step": 20340 + }, + { + "epoch": 0.05, + "learning_rate": 9.927433684365495e-05, + "loss": 2.3624, + "step": 20345 + }, + { + "epoch": 0.05, + "learning_rate": 9.927397922879996e-05, + "loss": 2.5106, + "step": 20350 + }, + { + "epoch": 0.05, + "learning_rate": 9.927362152649283e-05, + "loss": 2.4488, + "step": 20355 + }, + { + "epoch": 0.05, + "learning_rate": 9.927326373673418e-05, + "loss": 2.401, + "step": 20360 + }, + { + "epoch": 0.05, + "learning_rate": 9.927290585952467e-05, + "loss": 2.3866, + "step": 20365 + }, + { + "epoch": 0.05, + "learning_rate": 9.927254789486492e-05, + "loss": 2.36, + "step": 20370 + }, + { + "epoch": 0.05, + "learning_rate": 9.927218984275556e-05, + "loss": 2.3773, + "step": 20375 + }, + { + "epoch": 0.05, + "learning_rate": 9.927183170319723e-05, + "loss": 2.3937, + "step": 20380 + }, + { + "epoch": 0.05, + "learning_rate": 9.927147347619058e-05, + "loss": 2.409, + "step": 20385 + }, + { + "epoch": 0.05, + "learning_rate": 9.927111516173623e-05, + "loss": 2.4704, + "step": 20390 + }, + { + "epoch": 0.05, + "learning_rate": 9.927075675983481e-05, + "loss": 2.3556, + "step": 20395 + }, + { + "epoch": 0.05, + "learning_rate": 9.927039827048699e-05, + "loss": 2.4034, + "step": 20400 + }, + { + "epoch": 0.05, + "learning_rate": 9.927003969369335e-05, + "loss": 2.4126, + "step": 20405 + }, + { + "epoch": 0.05, + "learning_rate": 9.926968102945458e-05, + "loss": 2.3517, + "step": 20410 + }, + { + "epoch": 0.05, + "learning_rate": 9.92693222777713e-05, + "loss": 2.3139, + "step": 20415 + }, + { + "epoch": 0.05, + "learning_rate": 9.926896343864412e-05, + "loss": 2.397, + "step": 20420 + }, + { + "epoch": 0.05, + "learning_rate": 9.926860451207371e-05, + "loss": 2.3175, + "step": 20425 + }, + { + "epoch": 0.05, + "learning_rate": 9.92682454980607e-05, + "loss": 2.4456, + "step": 20430 + }, + { + "epoch": 0.05, + "learning_rate": 9.926788639660572e-05, + "loss": 2.5055, + "step": 20435 + }, + { + "epoch": 0.05, + "learning_rate": 9.92675272077094e-05, + "loss": 2.3651, + "step": 20440 + }, + { + "epoch": 0.05, + "learning_rate": 9.92671679313724e-05, + "loss": 2.4255, + "step": 20445 + }, + { + "epoch": 0.05, + "learning_rate": 9.926680856759533e-05, + "loss": 2.4983, + "step": 20450 + }, + { + "epoch": 0.05, + "learning_rate": 9.926644911637887e-05, + "loss": 2.4408, + "step": 20455 + }, + { + "epoch": 0.05, + "learning_rate": 9.926608957772361e-05, + "loss": 2.3894, + "step": 20460 + }, + { + "epoch": 0.05, + "learning_rate": 9.926572995163022e-05, + "loss": 2.3816, + "step": 20465 + }, + { + "epoch": 0.05, + "learning_rate": 9.926537023809933e-05, + "loss": 2.4006, + "step": 20470 + }, + { + "epoch": 0.05, + "learning_rate": 9.926501043713155e-05, + "loss": 2.3425, + "step": 20475 + }, + { + "epoch": 0.05, + "learning_rate": 9.926465054872756e-05, + "loss": 2.4785, + "step": 20480 + }, + { + "epoch": 0.05, + "learning_rate": 9.926429057288799e-05, + "loss": 2.4296, + "step": 20485 + }, + { + "epoch": 0.05, + "learning_rate": 9.926393050961347e-05, + "loss": 2.4787, + "step": 20490 + }, + { + "epoch": 0.05, + "learning_rate": 9.926357035890464e-05, + "loss": 2.5422, + "step": 20495 + }, + { + "epoch": 0.05, + "learning_rate": 9.926321012076214e-05, + "loss": 2.3804, + "step": 20500 + }, + { + "epoch": 0.05, + "learning_rate": 9.926284979518661e-05, + "loss": 2.469, + "step": 20505 + }, + { + "epoch": 0.05, + "learning_rate": 9.926248938217867e-05, + "loss": 2.2743, + "step": 20510 + }, + { + "epoch": 0.06, + "learning_rate": 9.926212888173901e-05, + "loss": 2.4257, + "step": 20515 + }, + { + "epoch": 0.06, + "learning_rate": 9.926176829386823e-05, + "loss": 2.2779, + "step": 20520 + }, + { + "epoch": 0.06, + "learning_rate": 9.926140761856698e-05, + "loss": 2.3895, + "step": 20525 + }, + { + "epoch": 0.06, + "learning_rate": 9.926104685583589e-05, + "loss": 2.5135, + "step": 20530 + }, + { + "epoch": 0.06, + "learning_rate": 9.926068600567561e-05, + "loss": 2.4712, + "step": 20535 + }, + { + "epoch": 0.06, + "learning_rate": 9.926032506808678e-05, + "loss": 2.3919, + "step": 20540 + }, + { + "epoch": 0.06, + "learning_rate": 9.925996404307004e-05, + "loss": 2.4558, + "step": 20545 + }, + { + "epoch": 0.06, + "learning_rate": 9.925960293062603e-05, + "loss": 2.3957, + "step": 20550 + }, + { + "epoch": 0.06, + "learning_rate": 9.92592417307554e-05, + "loss": 2.2292, + "step": 20555 + }, + { + "epoch": 0.06, + "learning_rate": 9.925888044345877e-05, + "loss": 2.4311, + "step": 20560 + }, + { + "epoch": 0.06, + "learning_rate": 9.92585190687368e-05, + "loss": 2.3918, + "step": 20565 + }, + { + "epoch": 0.06, + "learning_rate": 9.925815760659014e-05, + "loss": 2.3712, + "step": 20570 + }, + { + "epoch": 0.06, + "learning_rate": 9.925779605701939e-05, + "loss": 2.4746, + "step": 20575 + }, + { + "epoch": 0.06, + "learning_rate": 9.925743442002524e-05, + "loss": 2.4576, + "step": 20580 + }, + { + "epoch": 0.06, + "learning_rate": 9.92570726956083e-05, + "loss": 2.506, + "step": 20585 + }, + { + "epoch": 0.06, + "learning_rate": 9.925671088376922e-05, + "loss": 2.4417, + "step": 20590 + }, + { + "epoch": 0.06, + "learning_rate": 9.925634898450866e-05, + "loss": 2.5234, + "step": 20595 + }, + { + "epoch": 0.06, + "learning_rate": 9.925598699782723e-05, + "loss": 2.3548, + "step": 20600 + }, + { + "epoch": 0.06, + "learning_rate": 9.92556249237256e-05, + "loss": 2.4202, + "step": 20605 + }, + { + "epoch": 0.06, + "learning_rate": 9.925526276220441e-05, + "loss": 2.3817, + "step": 20610 + }, + { + "epoch": 0.06, + "learning_rate": 9.925490051326427e-05, + "loss": 2.3839, + "step": 20615 + }, + { + "epoch": 0.06, + "learning_rate": 9.925453817690587e-05, + "loss": 2.414, + "step": 20620 + }, + { + "epoch": 0.06, + "learning_rate": 9.92541757531298e-05, + "loss": 2.3881, + "step": 20625 + }, + { + "epoch": 0.06, + "learning_rate": 9.925381324193677e-05, + "loss": 2.3954, + "step": 20630 + }, + { + "epoch": 0.06, + "learning_rate": 9.925345064332736e-05, + "loss": 2.361, + "step": 20635 + }, + { + "epoch": 0.06, + "learning_rate": 9.925308795730226e-05, + "loss": 2.3081, + "step": 20640 + }, + { + "epoch": 0.06, + "learning_rate": 9.925272518386209e-05, + "loss": 2.3601, + "step": 20645 + }, + { + "epoch": 0.06, + "learning_rate": 9.92523623230075e-05, + "loss": 2.2406, + "step": 20650 + }, + { + "epoch": 0.06, + "learning_rate": 9.925199937473912e-05, + "loss": 2.234, + "step": 20655 + }, + { + "epoch": 0.06, + "learning_rate": 9.925163633905763e-05, + "loss": 2.4201, + "step": 20660 + }, + { + "epoch": 0.06, + "learning_rate": 9.925127321596364e-05, + "loss": 2.4145, + "step": 20665 + }, + { + "epoch": 0.06, + "learning_rate": 9.925091000545779e-05, + "loss": 2.3459, + "step": 20670 + }, + { + "epoch": 0.06, + "learning_rate": 9.925054670754074e-05, + "loss": 2.3696, + "step": 20675 + }, + { + "epoch": 0.06, + "learning_rate": 9.925018332221316e-05, + "loss": 2.4786, + "step": 20680 + }, + { + "epoch": 0.06, + "learning_rate": 9.924981984947566e-05, + "loss": 2.2985, + "step": 20685 + }, + { + "epoch": 0.06, + "learning_rate": 9.92494562893289e-05, + "loss": 2.3622, + "step": 20690 + }, + { + "epoch": 0.06, + "learning_rate": 9.92490926417735e-05, + "loss": 2.4431, + "step": 20695 + }, + { + "epoch": 0.06, + "learning_rate": 9.924872890681014e-05, + "loss": 2.4266, + "step": 20700 + }, + { + "epoch": 0.06, + "learning_rate": 9.924836508443944e-05, + "loss": 2.4807, + "step": 20705 + }, + { + "epoch": 0.06, + "learning_rate": 9.924800117466208e-05, + "loss": 2.3954, + "step": 20710 + }, + { + "epoch": 0.06, + "learning_rate": 9.924763717747866e-05, + "loss": 2.4907, + "step": 20715 + }, + { + "epoch": 0.06, + "learning_rate": 9.924727309288986e-05, + "loss": 2.3719, + "step": 20720 + }, + { + "epoch": 0.06, + "learning_rate": 9.92469089208963e-05, + "loss": 2.3597, + "step": 20725 + }, + { + "epoch": 0.06, + "learning_rate": 9.924654466149866e-05, + "loss": 2.4444, + "step": 20730 + }, + { + "epoch": 0.06, + "learning_rate": 9.924618031469757e-05, + "loss": 2.2738, + "step": 20735 + }, + { + "epoch": 0.06, + "learning_rate": 9.924581588049366e-05, + "loss": 2.4741, + "step": 20740 + }, + { + "epoch": 0.06, + "learning_rate": 9.924545135888759e-05, + "loss": 2.4662, + "step": 20745 + }, + { + "epoch": 0.06, + "learning_rate": 9.924508674988003e-05, + "loss": 2.4502, + "step": 20750 + }, + { + "epoch": 0.06, + "learning_rate": 9.924472205347158e-05, + "loss": 2.3607, + "step": 20755 + }, + { + "epoch": 0.06, + "learning_rate": 9.924435726966292e-05, + "loss": 2.4336, + "step": 20760 + }, + { + "epoch": 0.06, + "learning_rate": 9.92439923984547e-05, + "loss": 2.4204, + "step": 20765 + }, + { + "epoch": 0.06, + "learning_rate": 9.924362743984754e-05, + "loss": 2.3334, + "step": 20770 + }, + { + "epoch": 0.06, + "learning_rate": 9.924326239384211e-05, + "loss": 2.4082, + "step": 20775 + }, + { + "epoch": 0.06, + "learning_rate": 9.924289726043907e-05, + "loss": 2.3076, + "step": 20780 + }, + { + "epoch": 0.06, + "learning_rate": 9.924253203963905e-05, + "loss": 2.3523, + "step": 20785 + }, + { + "epoch": 0.06, + "learning_rate": 9.924216673144268e-05, + "loss": 2.4333, + "step": 20790 + }, + { + "epoch": 0.06, + "learning_rate": 9.924180133585064e-05, + "loss": 2.4404, + "step": 20795 + }, + { + "epoch": 0.06, + "learning_rate": 9.924143585286356e-05, + "loss": 2.4279, + "step": 20800 + }, + { + "epoch": 0.06, + "learning_rate": 9.92410702824821e-05, + "loss": 2.359, + "step": 20805 + }, + { + "epoch": 0.06, + "learning_rate": 9.92407046247069e-05, + "loss": 2.2904, + "step": 20810 + }, + { + "epoch": 0.06, + "learning_rate": 9.924033887953861e-05, + "loss": 2.4032, + "step": 20815 + }, + { + "epoch": 0.06, + "learning_rate": 9.92399730469779e-05, + "loss": 2.3487, + "step": 20820 + }, + { + "epoch": 0.06, + "learning_rate": 9.923960712702537e-05, + "loss": 2.3643, + "step": 20825 + }, + { + "epoch": 0.06, + "learning_rate": 9.923924111968172e-05, + "loss": 2.3995, + "step": 20830 + }, + { + "epoch": 0.06, + "learning_rate": 9.923887502494759e-05, + "loss": 2.427, + "step": 20835 + }, + { + "epoch": 0.06, + "learning_rate": 9.923850884282359e-05, + "loss": 2.2947, + "step": 20840 + }, + { + "epoch": 0.06, + "learning_rate": 9.923814257331043e-05, + "loss": 2.4008, + "step": 20845 + }, + { + "epoch": 0.06, + "learning_rate": 9.923777621640873e-05, + "loss": 2.4143, + "step": 20850 + }, + { + "epoch": 0.06, + "learning_rate": 9.923740977211912e-05, + "loss": 2.2613, + "step": 20855 + }, + { + "epoch": 0.06, + "learning_rate": 9.923704324044227e-05, + "loss": 2.3564, + "step": 20860 + }, + { + "epoch": 0.06, + "learning_rate": 9.923667662137885e-05, + "loss": 2.4547, + "step": 20865 + }, + { + "epoch": 0.06, + "learning_rate": 9.923630991492948e-05, + "loss": 2.3676, + "step": 20870 + }, + { + "epoch": 0.06, + "learning_rate": 9.923594312109482e-05, + "loss": 2.475, + "step": 20875 + }, + { + "epoch": 0.06, + "learning_rate": 9.923557623987554e-05, + "loss": 2.398, + "step": 20880 + }, + { + "epoch": 0.06, + "learning_rate": 9.923520927127226e-05, + "loss": 2.4338, + "step": 20885 + }, + { + "epoch": 0.06, + "learning_rate": 9.923484221528564e-05, + "loss": 2.3464, + "step": 20890 + }, + { + "epoch": 0.06, + "learning_rate": 9.923447507191634e-05, + "loss": 2.3025, + "step": 20895 + }, + { + "epoch": 0.06, + "learning_rate": 9.923410784116503e-05, + "loss": 2.4581, + "step": 20900 + }, + { + "epoch": 0.06, + "learning_rate": 9.923374052303232e-05, + "loss": 2.3486, + "step": 20905 + }, + { + "epoch": 0.06, + "learning_rate": 9.923337311751889e-05, + "loss": 2.4858, + "step": 20910 + }, + { + "epoch": 0.06, + "learning_rate": 9.923300562462539e-05, + "loss": 2.3862, + "step": 20915 + }, + { + "epoch": 0.06, + "learning_rate": 9.923263804435246e-05, + "loss": 2.5409, + "step": 20920 + }, + { + "epoch": 0.06, + "learning_rate": 9.923227037670075e-05, + "loss": 2.4229, + "step": 20925 + }, + { + "epoch": 0.06, + "learning_rate": 9.923190262167093e-05, + "loss": 2.4032, + "step": 20930 + }, + { + "epoch": 0.06, + "learning_rate": 9.923153477926365e-05, + "loss": 2.3706, + "step": 20935 + }, + { + "epoch": 0.06, + "learning_rate": 9.923116684947956e-05, + "loss": 2.4233, + "step": 20940 + }, + { + "epoch": 0.06, + "learning_rate": 9.92307988323193e-05, + "loss": 2.3677, + "step": 20945 + }, + { + "epoch": 0.06, + "learning_rate": 9.923043072778353e-05, + "loss": 2.3467, + "step": 20950 + }, + { + "epoch": 0.06, + "learning_rate": 9.923006253587292e-05, + "loss": 2.4871, + "step": 20955 + }, + { + "epoch": 0.06, + "learning_rate": 9.92296942565881e-05, + "loss": 2.3195, + "step": 20960 + }, + { + "epoch": 0.06, + "learning_rate": 9.922932588992974e-05, + "loss": 2.353, + "step": 20965 + }, + { + "epoch": 0.06, + "learning_rate": 9.922895743589849e-05, + "loss": 2.5336, + "step": 20970 + }, + { + "epoch": 0.06, + "learning_rate": 9.9228588894495e-05, + "loss": 2.2737, + "step": 20975 + }, + { + "epoch": 0.06, + "learning_rate": 9.922822026571992e-05, + "loss": 2.4351, + "step": 20980 + }, + { + "epoch": 0.06, + "learning_rate": 9.922785154957393e-05, + "loss": 2.3518, + "step": 20985 + }, + { + "epoch": 0.06, + "learning_rate": 9.922748274605764e-05, + "loss": 2.3339, + "step": 20990 + }, + { + "epoch": 0.06, + "learning_rate": 9.922711385517174e-05, + "loss": 2.5374, + "step": 20995 + }, + { + "epoch": 0.06, + "learning_rate": 9.922674487691686e-05, + "loss": 2.4639, + "step": 21000 + }, + { + "epoch": 0.06, + "learning_rate": 9.922637581129368e-05, + "loss": 2.4026, + "step": 21005 + }, + { + "epoch": 0.06, + "learning_rate": 9.922600665830286e-05, + "loss": 2.3707, + "step": 21010 + }, + { + "epoch": 0.06, + "learning_rate": 9.922563741794502e-05, + "loss": 2.4553, + "step": 21015 + }, + { + "epoch": 0.06, + "learning_rate": 9.922526809022084e-05, + "loss": 2.2567, + "step": 21020 + }, + { + "epoch": 0.06, + "learning_rate": 9.922489867513096e-05, + "loss": 2.3812, + "step": 21025 + }, + { + "epoch": 0.06, + "learning_rate": 9.922452917267607e-05, + "loss": 2.5002, + "step": 21030 + }, + { + "epoch": 0.06, + "learning_rate": 9.922415958285677e-05, + "loss": 2.4827, + "step": 21035 + }, + { + "epoch": 0.06, + "learning_rate": 9.922378990567376e-05, + "loss": 2.4009, + "step": 21040 + }, + { + "epoch": 0.06, + "learning_rate": 9.92234201411277e-05, + "loss": 2.2557, + "step": 21045 + }, + { + "epoch": 0.06, + "learning_rate": 9.922305028921921e-05, + "loss": 2.3497, + "step": 21050 + }, + { + "epoch": 0.06, + "learning_rate": 9.922268034994897e-05, + "loss": 2.3588, + "step": 21055 + }, + { + "epoch": 0.06, + "learning_rate": 9.922231032331763e-05, + "loss": 2.39, + "step": 21060 + }, + { + "epoch": 0.06, + "learning_rate": 9.922194020932585e-05, + "loss": 2.3831, + "step": 21065 + }, + { + "epoch": 0.06, + "learning_rate": 9.922157000797428e-05, + "loss": 2.4727, + "step": 21070 + }, + { + "epoch": 0.06, + "learning_rate": 9.92211997192636e-05, + "loss": 2.4531, + "step": 21075 + }, + { + "epoch": 0.06, + "learning_rate": 9.922082934319444e-05, + "loss": 2.4588, + "step": 21080 + }, + { + "epoch": 0.06, + "learning_rate": 9.922045887976747e-05, + "loss": 2.3632, + "step": 21085 + }, + { + "epoch": 0.06, + "learning_rate": 9.922008832898334e-05, + "loss": 2.3319, + "step": 21090 + }, + { + "epoch": 0.06, + "learning_rate": 9.921971769084272e-05, + "loss": 2.3126, + "step": 21095 + }, + { + "epoch": 0.06, + "learning_rate": 9.921934696534625e-05, + "loss": 2.4459, + "step": 21100 + }, + { + "epoch": 0.06, + "learning_rate": 9.921897615249461e-05, + "loss": 2.2738, + "step": 21105 + }, + { + "epoch": 0.06, + "learning_rate": 9.921860525228844e-05, + "loss": 2.4031, + "step": 21110 + }, + { + "epoch": 0.06, + "learning_rate": 9.921823426472841e-05, + "loss": 2.4262, + "step": 21115 + }, + { + "epoch": 0.06, + "learning_rate": 9.921786318981518e-05, + "loss": 2.4512, + "step": 21120 + }, + { + "epoch": 0.06, + "learning_rate": 9.921749202754938e-05, + "loss": 2.3663, + "step": 21125 + }, + { + "epoch": 0.06, + "learning_rate": 9.92171207779317e-05, + "loss": 2.5145, + "step": 21130 + }, + { + "epoch": 0.06, + "learning_rate": 9.92167494409628e-05, + "loss": 2.413, + "step": 21135 + }, + { + "epoch": 0.06, + "learning_rate": 9.921637801664332e-05, + "loss": 2.4459, + "step": 21140 + }, + { + "epoch": 0.06, + "learning_rate": 9.921600650497393e-05, + "loss": 2.4283, + "step": 21145 + }, + { + "epoch": 0.06, + "learning_rate": 9.921563490595529e-05, + "loss": 2.4524, + "step": 21150 + }, + { + "epoch": 0.06, + "learning_rate": 9.921526321958806e-05, + "loss": 2.4417, + "step": 21155 + }, + { + "epoch": 0.06, + "learning_rate": 9.921489144587289e-05, + "loss": 2.3699, + "step": 21160 + }, + { + "epoch": 0.06, + "learning_rate": 9.921451958481045e-05, + "loss": 2.2509, + "step": 21165 + }, + { + "epoch": 0.06, + "learning_rate": 9.92141476364014e-05, + "loss": 2.3528, + "step": 21170 + }, + { + "epoch": 0.06, + "learning_rate": 9.921377560064637e-05, + "loss": 2.2856, + "step": 21175 + }, + { + "epoch": 0.06, + "learning_rate": 9.921340347754608e-05, + "loss": 2.333, + "step": 21180 + }, + { + "epoch": 0.06, + "learning_rate": 9.921303126710115e-05, + "loss": 2.3695, + "step": 21185 + }, + { + "epoch": 0.06, + "learning_rate": 9.921265896931225e-05, + "loss": 2.3174, + "step": 21190 + }, + { + "epoch": 0.06, + "learning_rate": 9.921228658418001e-05, + "loss": 2.3314, + "step": 21195 + }, + { + "epoch": 0.06, + "learning_rate": 9.921191411170514e-05, + "loss": 2.3276, + "step": 21200 + }, + { + "epoch": 0.06, + "learning_rate": 9.921154155188829e-05, + "loss": 2.5684, + "step": 21205 + }, + { + "epoch": 0.06, + "learning_rate": 9.92111689047301e-05, + "loss": 2.3995, + "step": 21210 + }, + { + "epoch": 0.06, + "learning_rate": 9.921079617023125e-05, + "loss": 2.4838, + "step": 21215 + }, + { + "epoch": 0.06, + "learning_rate": 9.921042334839239e-05, + "loss": 2.445, + "step": 21220 + }, + { + "epoch": 0.06, + "learning_rate": 9.921005043921419e-05, + "loss": 2.4214, + "step": 21225 + }, + { + "epoch": 0.06, + "learning_rate": 9.92096774426973e-05, + "loss": 2.2859, + "step": 21230 + }, + { + "epoch": 0.06, + "learning_rate": 9.920930435884239e-05, + "loss": 2.3977, + "step": 21235 + }, + { + "epoch": 0.06, + "learning_rate": 9.920893118765013e-05, + "loss": 2.4821, + "step": 21240 + }, + { + "epoch": 0.06, + "learning_rate": 9.920855792912117e-05, + "loss": 2.347, + "step": 21245 + }, + { + "epoch": 0.06, + "learning_rate": 9.920818458325618e-05, + "loss": 2.4939, + "step": 21250 + }, + { + "epoch": 0.06, + "learning_rate": 9.920781115005582e-05, + "loss": 2.3366, + "step": 21255 + }, + { + "epoch": 0.06, + "learning_rate": 9.920743762952074e-05, + "loss": 2.4748, + "step": 21260 + }, + { + "epoch": 0.06, + "learning_rate": 9.920706402165163e-05, + "loss": 2.4818, + "step": 21265 + }, + { + "epoch": 0.06, + "learning_rate": 9.920669032644913e-05, + "loss": 2.5372, + "step": 21270 + }, + { + "epoch": 0.06, + "learning_rate": 9.920631654391392e-05, + "loss": 2.4395, + "step": 21275 + }, + { + "epoch": 0.06, + "learning_rate": 9.920594267404666e-05, + "loss": 2.4518, + "step": 21280 + }, + { + "epoch": 0.06, + "learning_rate": 9.920556871684801e-05, + "loss": 2.4625, + "step": 21285 + }, + { + "epoch": 0.06, + "learning_rate": 9.920519467231861e-05, + "loss": 2.3971, + "step": 21290 + }, + { + "epoch": 0.06, + "learning_rate": 9.920482054045917e-05, + "loss": 2.453, + "step": 21295 + }, + { + "epoch": 0.06, + "learning_rate": 9.920444632127031e-05, + "loss": 2.3162, + "step": 21300 + }, + { + "epoch": 0.06, + "learning_rate": 9.920407201475273e-05, + "loss": 2.4684, + "step": 21305 + }, + { + "epoch": 0.06, + "learning_rate": 9.920369762090708e-05, + "loss": 2.504, + "step": 21310 + }, + { + "epoch": 0.06, + "learning_rate": 9.920332313973401e-05, + "loss": 2.4527, + "step": 21315 + }, + { + "epoch": 0.06, + "learning_rate": 9.92029485712342e-05, + "loss": 2.4758, + "step": 21320 + }, + { + "epoch": 0.06, + "learning_rate": 9.920257391540832e-05, + "loss": 2.3608, + "step": 21325 + }, + { + "epoch": 0.06, + "learning_rate": 9.920219917225703e-05, + "loss": 2.3036, + "step": 21330 + }, + { + "epoch": 0.06, + "learning_rate": 9.920182434178098e-05, + "loss": 2.4749, + "step": 21335 + }, + { + "epoch": 0.06, + "learning_rate": 9.920144942398086e-05, + "loss": 2.3595, + "step": 21340 + }, + { + "epoch": 0.06, + "learning_rate": 9.920107441885732e-05, + "loss": 2.4565, + "step": 21345 + }, + { + "epoch": 0.06, + "learning_rate": 9.920069932641103e-05, + "loss": 2.4224, + "step": 21350 + }, + { + "epoch": 0.06, + "learning_rate": 9.920032414664266e-05, + "loss": 2.2866, + "step": 21355 + }, + { + "epoch": 0.06, + "learning_rate": 9.919994887955286e-05, + "loss": 2.4072, + "step": 21360 + }, + { + "epoch": 0.06, + "learning_rate": 9.91995735251423e-05, + "loss": 2.2965, + "step": 21365 + }, + { + "epoch": 0.06, + "learning_rate": 9.919919808341168e-05, + "loss": 2.4099, + "step": 21370 + }, + { + "epoch": 0.06, + "learning_rate": 9.919882255436161e-05, + "loss": 2.3574, + "step": 21375 + }, + { + "epoch": 0.06, + "learning_rate": 9.919844693799281e-05, + "loss": 2.4528, + "step": 21380 + }, + { + "epoch": 0.06, + "learning_rate": 9.91980712343059e-05, + "loss": 2.3542, + "step": 21385 + }, + { + "epoch": 0.06, + "learning_rate": 9.919769544330157e-05, + "loss": 2.4792, + "step": 21390 + }, + { + "epoch": 0.06, + "learning_rate": 9.919731956498051e-05, + "loss": 2.3685, + "step": 21395 + }, + { + "epoch": 0.06, + "learning_rate": 9.919694359934335e-05, + "loss": 2.3458, + "step": 21400 + }, + { + "epoch": 0.06, + "learning_rate": 9.919656754639076e-05, + "loss": 2.3365, + "step": 21405 + }, + { + "epoch": 0.06, + "learning_rate": 9.919619140612343e-05, + "loss": 2.3908, + "step": 21410 + }, + { + "epoch": 0.06, + "learning_rate": 9.919581517854202e-05, + "loss": 2.4444, + "step": 21415 + }, + { + "epoch": 0.06, + "learning_rate": 9.919543886364717e-05, + "loss": 2.522, + "step": 21420 + }, + { + "epoch": 0.06, + "learning_rate": 9.919506246143959e-05, + "loss": 2.3136, + "step": 21425 + }, + { + "epoch": 0.06, + "learning_rate": 9.919468597191992e-05, + "loss": 2.2594, + "step": 21430 + }, + { + "epoch": 0.06, + "learning_rate": 9.919430939508885e-05, + "loss": 2.4827, + "step": 21435 + }, + { + "epoch": 0.06, + "learning_rate": 9.919393273094703e-05, + "loss": 2.3935, + "step": 21440 + }, + { + "epoch": 0.06, + "learning_rate": 9.919355597949513e-05, + "loss": 2.5222, + "step": 21445 + }, + { + "epoch": 0.06, + "learning_rate": 9.919317914073383e-05, + "loss": 2.3194, + "step": 21450 + }, + { + "epoch": 0.06, + "learning_rate": 9.919280221466378e-05, + "loss": 2.3971, + "step": 21455 + }, + { + "epoch": 0.06, + "learning_rate": 9.919242520128566e-05, + "loss": 2.3458, + "step": 21460 + }, + { + "epoch": 0.06, + "learning_rate": 9.919204810060016e-05, + "loss": 2.3956, + "step": 21465 + }, + { + "epoch": 0.06, + "learning_rate": 9.91916709126079e-05, + "loss": 2.4637, + "step": 21470 + }, + { + "epoch": 0.06, + "learning_rate": 9.919129363730961e-05, + "loss": 2.536, + "step": 21475 + }, + { + "epoch": 0.06, + "learning_rate": 9.919091627470592e-05, + "loss": 2.3586, + "step": 21480 + }, + { + "epoch": 0.06, + "learning_rate": 9.919053882479749e-05, + "loss": 2.3613, + "step": 21485 + }, + { + "epoch": 0.06, + "learning_rate": 9.919016128758502e-05, + "loss": 2.3604, + "step": 21490 + }, + { + "epoch": 0.06, + "learning_rate": 9.918978366306917e-05, + "loss": 2.5411, + "step": 21495 + }, + { + "epoch": 0.06, + "learning_rate": 9.918940595125059e-05, + "loss": 2.4673, + "step": 21500 + }, + { + "epoch": 0.06, + "learning_rate": 9.918902815212998e-05, + "loss": 2.2146, + "step": 21505 + }, + { + "epoch": 0.06, + "learning_rate": 9.9188650265708e-05, + "loss": 2.4451, + "step": 21510 + }, + { + "epoch": 0.06, + "learning_rate": 9.918827229198533e-05, + "loss": 2.3405, + "step": 21515 + }, + { + "epoch": 0.06, + "learning_rate": 9.918789423096262e-05, + "loss": 2.3241, + "step": 21520 + }, + { + "epoch": 0.06, + "learning_rate": 9.918751608264056e-05, + "loss": 2.4714, + "step": 21525 + }, + { + "epoch": 0.06, + "learning_rate": 9.91871378470198e-05, + "loss": 2.386, + "step": 21530 + }, + { + "epoch": 0.06, + "learning_rate": 9.918675952410102e-05, + "loss": 2.3384, + "step": 21535 + }, + { + "epoch": 0.06, + "learning_rate": 9.918638111388491e-05, + "loss": 2.5089, + "step": 21540 + }, + { + "epoch": 0.06, + "learning_rate": 9.918600261637212e-05, + "loss": 2.5348, + "step": 21545 + }, + { + "epoch": 0.06, + "learning_rate": 9.918562403156333e-05, + "loss": 2.4746, + "step": 21550 + }, + { + "epoch": 0.06, + "learning_rate": 9.918524535945921e-05, + "loss": 2.4144, + "step": 21555 + }, + { + "epoch": 0.06, + "learning_rate": 9.918486660006043e-05, + "loss": 2.2998, + "step": 21560 + }, + { + "epoch": 0.06, + "learning_rate": 9.918448775336766e-05, + "loss": 2.3178, + "step": 21565 + }, + { + "epoch": 0.06, + "learning_rate": 9.91841088193816e-05, + "loss": 2.3462, + "step": 21570 + }, + { + "epoch": 0.06, + "learning_rate": 9.918372979810288e-05, + "loss": 2.3518, + "step": 21575 + }, + { + "epoch": 0.06, + "learning_rate": 9.91833506895322e-05, + "loss": 2.4428, + "step": 21580 + }, + { + "epoch": 0.06, + "learning_rate": 9.918297149367022e-05, + "loss": 2.4786, + "step": 21585 + }, + { + "epoch": 0.06, + "learning_rate": 9.918259221051762e-05, + "loss": 2.3075, + "step": 21590 + }, + { + "epoch": 0.06, + "learning_rate": 9.918221284007507e-05, + "loss": 2.4134, + "step": 21595 + }, + { + "epoch": 0.06, + "learning_rate": 9.918183338234324e-05, + "loss": 2.4539, + "step": 21600 + }, + { + "epoch": 0.06, + "learning_rate": 9.918145383732282e-05, + "loss": 2.3027, + "step": 21605 + }, + { + "epoch": 0.06, + "learning_rate": 9.918107420501446e-05, + "loss": 2.4521, + "step": 21610 + }, + { + "epoch": 0.06, + "learning_rate": 9.918069448541886e-05, + "loss": 2.4049, + "step": 21615 + }, + { + "epoch": 0.06, + "learning_rate": 9.918031467853667e-05, + "loss": 2.4367, + "step": 21620 + }, + { + "epoch": 0.06, + "learning_rate": 9.917993478436858e-05, + "loss": 2.4418, + "step": 21625 + }, + { + "epoch": 0.06, + "learning_rate": 9.917955480291526e-05, + "loss": 2.4163, + "step": 21630 + }, + { + "epoch": 0.06, + "learning_rate": 9.917917473417736e-05, + "loss": 2.4142, + "step": 21635 + }, + { + "epoch": 0.06, + "learning_rate": 9.91787945781556e-05, + "loss": 2.4253, + "step": 21640 + }, + { + "epoch": 0.06, + "learning_rate": 9.917841433485063e-05, + "loss": 2.2906, + "step": 21645 + }, + { + "epoch": 0.06, + "learning_rate": 9.917803400426311e-05, + "loss": 2.4232, + "step": 21650 + }, + { + "epoch": 0.06, + "learning_rate": 9.917765358639375e-05, + "loss": 2.5218, + "step": 21655 + }, + { + "epoch": 0.06, + "learning_rate": 9.917727308124318e-05, + "loss": 2.4109, + "step": 21660 + }, + { + "epoch": 0.06, + "learning_rate": 9.917689248881213e-05, + "loss": 2.3649, + "step": 21665 + }, + { + "epoch": 0.06, + "learning_rate": 9.917651180910123e-05, + "loss": 2.3968, + "step": 21670 + }, + { + "epoch": 0.06, + "learning_rate": 9.917613104211119e-05, + "loss": 2.2747, + "step": 21675 + }, + { + "epoch": 0.06, + "learning_rate": 9.917575018784266e-05, + "loss": 2.5195, + "step": 21680 + }, + { + "epoch": 0.06, + "learning_rate": 9.917536924629632e-05, + "loss": 2.4629, + "step": 21685 + }, + { + "epoch": 0.06, + "learning_rate": 9.917498821747287e-05, + "loss": 2.3058, + "step": 21690 + }, + { + "epoch": 0.06, + "learning_rate": 9.917460710137294e-05, + "loss": 2.466, + "step": 21695 + }, + { + "epoch": 0.06, + "learning_rate": 9.917422589799726e-05, + "loss": 2.3207, + "step": 21700 + }, + { + "epoch": 0.06, + "learning_rate": 9.917384460734646e-05, + "loss": 2.4533, + "step": 21705 + }, + { + "epoch": 0.06, + "learning_rate": 9.917346322942123e-05, + "loss": 2.4476, + "step": 21710 + }, + { + "epoch": 0.06, + "learning_rate": 9.917308176422227e-05, + "loss": 2.4451, + "step": 21715 + }, + { + "epoch": 0.06, + "learning_rate": 9.917270021175024e-05, + "loss": 2.4473, + "step": 21720 + }, + { + "epoch": 0.06, + "learning_rate": 9.917231857200582e-05, + "loss": 2.401, + "step": 21725 + }, + { + "epoch": 0.06, + "learning_rate": 9.917193684498968e-05, + "loss": 2.4293, + "step": 21730 + }, + { + "epoch": 0.06, + "learning_rate": 9.91715550307025e-05, + "loss": 2.4168, + "step": 21735 + }, + { + "epoch": 0.06, + "learning_rate": 9.917117312914496e-05, + "loss": 2.413, + "step": 21740 + }, + { + "epoch": 0.06, + "learning_rate": 9.917079114031775e-05, + "loss": 2.3875, + "step": 21745 + }, + { + "epoch": 0.06, + "learning_rate": 9.917040906422153e-05, + "loss": 2.3977, + "step": 21750 + }, + { + "epoch": 0.06, + "learning_rate": 9.917002690085698e-05, + "loss": 2.3755, + "step": 21755 + }, + { + "epoch": 0.06, + "learning_rate": 9.916964465022477e-05, + "loss": 2.2702, + "step": 21760 + }, + { + "epoch": 0.06, + "learning_rate": 9.916926231232561e-05, + "loss": 2.3624, + "step": 21765 + }, + { + "epoch": 0.06, + "learning_rate": 9.916887988716016e-05, + "loss": 2.4772, + "step": 21770 + }, + { + "epoch": 0.06, + "learning_rate": 9.91684973747291e-05, + "loss": 2.3972, + "step": 21775 + }, + { + "epoch": 0.06, + "learning_rate": 9.916811477503309e-05, + "loss": 2.411, + "step": 21780 + }, + { + "epoch": 0.06, + "learning_rate": 9.916773208807284e-05, + "loss": 2.2566, + "step": 21785 + }, + { + "epoch": 0.06, + "learning_rate": 9.916734931384901e-05, + "loss": 2.3844, + "step": 21790 + }, + { + "epoch": 0.06, + "learning_rate": 9.91669664523623e-05, + "loss": 2.4376, + "step": 21795 + }, + { + "epoch": 0.06, + "learning_rate": 9.916658350361335e-05, + "loss": 2.4515, + "step": 21800 + }, + { + "epoch": 0.06, + "learning_rate": 9.916620046760288e-05, + "loss": 2.4308, + "step": 21805 + }, + { + "epoch": 0.06, + "learning_rate": 9.916581734433154e-05, + "loss": 2.3443, + "step": 21810 + }, + { + "epoch": 0.06, + "learning_rate": 9.916543413380004e-05, + "loss": 2.3964, + "step": 21815 + }, + { + "epoch": 0.06, + "learning_rate": 9.916505083600904e-05, + "loss": 2.3584, + "step": 21820 + }, + { + "epoch": 0.06, + "learning_rate": 9.916466745095922e-05, + "loss": 2.428, + "step": 21825 + }, + { + "epoch": 0.06, + "learning_rate": 9.916428397865128e-05, + "loss": 2.2679, + "step": 21830 + }, + { + "epoch": 0.06, + "learning_rate": 9.916390041908585e-05, + "loss": 2.2005, + "step": 21835 + }, + { + "epoch": 0.06, + "learning_rate": 9.916351677226368e-05, + "loss": 2.3794, + "step": 21840 + }, + { + "epoch": 0.06, + "learning_rate": 9.91631330381854e-05, + "loss": 2.4084, + "step": 21845 + }, + { + "epoch": 0.06, + "learning_rate": 9.916274921685171e-05, + "loss": 2.2898, + "step": 21850 + }, + { + "epoch": 0.06, + "learning_rate": 9.916236530826328e-05, + "loss": 2.3283, + "step": 21855 + }, + { + "epoch": 0.06, + "learning_rate": 9.916198131242083e-05, + "loss": 2.536, + "step": 21860 + }, + { + "epoch": 0.06, + "learning_rate": 9.916159722932499e-05, + "loss": 2.4546, + "step": 21865 + }, + { + "epoch": 0.06, + "learning_rate": 9.916121305897645e-05, + "loss": 2.4407, + "step": 21870 + }, + { + "epoch": 0.06, + "learning_rate": 9.916082880137594e-05, + "loss": 2.3799, + "step": 21875 + }, + { + "epoch": 0.06, + "learning_rate": 9.916044445652408e-05, + "loss": 2.4055, + "step": 21880 + }, + { + "epoch": 0.06, + "learning_rate": 9.91600600244216e-05, + "loss": 2.374, + "step": 21885 + }, + { + "epoch": 0.06, + "learning_rate": 9.915967550506913e-05, + "loss": 2.3538, + "step": 21890 + }, + { + "epoch": 0.06, + "learning_rate": 9.915929089846741e-05, + "loss": 2.348, + "step": 21895 + }, + { + "epoch": 0.06, + "learning_rate": 9.91589062046171e-05, + "loss": 2.4269, + "step": 21900 + }, + { + "epoch": 0.06, + "learning_rate": 9.915852142351887e-05, + "loss": 2.2786, + "step": 21905 + }, + { + "epoch": 0.06, + "learning_rate": 9.915813655517341e-05, + "loss": 2.4619, + "step": 21910 + }, + { + "epoch": 0.06, + "learning_rate": 9.915775159958141e-05, + "loss": 2.3743, + "step": 21915 + }, + { + "epoch": 0.06, + "learning_rate": 9.915736655674355e-05, + "loss": 2.5102, + "step": 21920 + }, + { + "epoch": 0.06, + "learning_rate": 9.915698142666052e-05, + "loss": 2.4028, + "step": 21925 + }, + { + "epoch": 0.06, + "learning_rate": 9.915659620933297e-05, + "loss": 2.3714, + "step": 21930 + }, + { + "epoch": 0.06, + "learning_rate": 9.915621090476164e-05, + "loss": 2.3375, + "step": 21935 + }, + { + "epoch": 0.06, + "learning_rate": 9.915582551294718e-05, + "loss": 2.4107, + "step": 21940 + }, + { + "epoch": 0.06, + "learning_rate": 9.915544003389026e-05, + "loss": 2.4655, + "step": 21945 + }, + { + "epoch": 0.06, + "learning_rate": 9.915505446759159e-05, + "loss": 2.3088, + "step": 21950 + }, + { + "epoch": 0.06, + "learning_rate": 9.915466881405185e-05, + "loss": 2.3208, + "step": 21955 + }, + { + "epoch": 0.06, + "learning_rate": 9.91542830732717e-05, + "loss": 2.3254, + "step": 21960 + }, + { + "epoch": 0.06, + "learning_rate": 9.915389724525186e-05, + "loss": 2.4346, + "step": 21965 + }, + { + "epoch": 0.06, + "learning_rate": 9.9153511329993e-05, + "loss": 2.3487, + "step": 21970 + }, + { + "epoch": 0.06, + "learning_rate": 9.915312532749581e-05, + "loss": 2.3784, + "step": 21975 + }, + { + "epoch": 0.06, + "learning_rate": 9.915273923776096e-05, + "loss": 2.4809, + "step": 21980 + }, + { + "epoch": 0.06, + "learning_rate": 9.915235306078916e-05, + "loss": 2.239, + "step": 21985 + }, + { + "epoch": 0.06, + "learning_rate": 9.915196679658107e-05, + "loss": 2.388, + "step": 21990 + }, + { + "epoch": 0.06, + "learning_rate": 9.915158044513738e-05, + "loss": 2.3925, + "step": 21995 + }, + { + "epoch": 0.06, + "learning_rate": 9.915119400645878e-05, + "loss": 2.4991, + "step": 22000 + }, + { + "epoch": 0.06, + "learning_rate": 9.915080748054597e-05, + "loss": 2.4449, + "step": 22005 + }, + { + "epoch": 0.06, + "learning_rate": 9.915042086739962e-05, + "loss": 2.2948, + "step": 22010 + }, + { + "epoch": 0.06, + "learning_rate": 9.915003416702041e-05, + "loss": 2.4539, + "step": 22015 + }, + { + "epoch": 0.06, + "learning_rate": 9.914964737940904e-05, + "loss": 2.4411, + "step": 22020 + }, + { + "epoch": 0.06, + "learning_rate": 9.914926050456621e-05, + "loss": 2.3514, + "step": 22025 + }, + { + "epoch": 0.06, + "learning_rate": 9.914887354249257e-05, + "loss": 2.3377, + "step": 22030 + }, + { + "epoch": 0.06, + "learning_rate": 9.914848649318881e-05, + "loss": 2.4134, + "step": 22035 + }, + { + "epoch": 0.06, + "learning_rate": 9.914809935665568e-05, + "loss": 2.3621, + "step": 22040 + }, + { + "epoch": 0.06, + "learning_rate": 9.914771213289378e-05, + "loss": 2.4372, + "step": 22045 + }, + { + "epoch": 0.06, + "learning_rate": 9.914732482190385e-05, + "loss": 2.3425, + "step": 22050 + }, + { + "epoch": 0.06, + "learning_rate": 9.914693742368657e-05, + "loss": 2.3996, + "step": 22055 + }, + { + "epoch": 0.06, + "learning_rate": 9.914654993824261e-05, + "loss": 2.3827, + "step": 22060 + }, + { + "epoch": 0.06, + "learning_rate": 9.914616236557268e-05, + "loss": 2.3066, + "step": 22065 + }, + { + "epoch": 0.06, + "learning_rate": 9.914577470567746e-05, + "loss": 2.5102, + "step": 22070 + }, + { + "epoch": 0.06, + "learning_rate": 9.914538695855762e-05, + "loss": 2.4314, + "step": 22075 + }, + { + "epoch": 0.06, + "learning_rate": 9.914499912421388e-05, + "loss": 2.4478, + "step": 22080 + }, + { + "epoch": 0.06, + "learning_rate": 9.914461120264691e-05, + "loss": 2.3265, + "step": 22085 + }, + { + "epoch": 0.06, + "learning_rate": 9.914422319385738e-05, + "loss": 2.4857, + "step": 22090 + }, + { + "epoch": 0.06, + "learning_rate": 9.914383509784603e-05, + "loss": 2.3865, + "step": 22095 + }, + { + "epoch": 0.06, + "learning_rate": 9.914344691461351e-05, + "loss": 2.4877, + "step": 22100 + }, + { + "epoch": 0.06, + "learning_rate": 9.91430586441605e-05, + "loss": 2.3912, + "step": 22105 + }, + { + "epoch": 0.06, + "learning_rate": 9.914267028648771e-05, + "loss": 2.373, + "step": 22110 + }, + { + "epoch": 0.06, + "learning_rate": 9.914228184159585e-05, + "loss": 2.1948, + "step": 22115 + }, + { + "epoch": 0.06, + "learning_rate": 9.914189330948555e-05, + "loss": 2.2775, + "step": 22120 + }, + { + "epoch": 0.06, + "learning_rate": 9.914150469015756e-05, + "loss": 2.3417, + "step": 22125 + }, + { + "epoch": 0.06, + "learning_rate": 9.914111598361253e-05, + "loss": 2.3522, + "step": 22130 + }, + { + "epoch": 0.06, + "learning_rate": 9.914072718985118e-05, + "loss": 2.4842, + "step": 22135 + }, + { + "epoch": 0.06, + "learning_rate": 9.914033830887418e-05, + "loss": 2.4055, + "step": 22140 + }, + { + "epoch": 0.06, + "learning_rate": 9.913994934068221e-05, + "loss": 2.2795, + "step": 22145 + }, + { + "epoch": 0.06, + "learning_rate": 9.913956028527598e-05, + "loss": 2.3957, + "step": 22150 + }, + { + "epoch": 0.06, + "learning_rate": 9.913917114265619e-05, + "loss": 2.5149, + "step": 22155 + }, + { + "epoch": 0.06, + "learning_rate": 9.91387819128235e-05, + "loss": 2.4344, + "step": 22160 + }, + { + "epoch": 0.06, + "learning_rate": 9.913839259577861e-05, + "loss": 2.3573, + "step": 22165 + }, + { + "epoch": 0.06, + "learning_rate": 9.913800319152224e-05, + "loss": 2.4059, + "step": 22170 + }, + { + "epoch": 0.06, + "learning_rate": 9.913761370005504e-05, + "loss": 2.3006, + "step": 22175 + }, + { + "epoch": 0.06, + "learning_rate": 9.913722412137773e-05, + "loss": 2.2571, + "step": 22180 + }, + { + "epoch": 0.06, + "learning_rate": 9.913683445549099e-05, + "loss": 2.397, + "step": 22185 + }, + { + "epoch": 0.06, + "learning_rate": 9.913644470239552e-05, + "loss": 2.3922, + "step": 22190 + }, + { + "epoch": 0.06, + "learning_rate": 9.913605486209198e-05, + "loss": 2.3505, + "step": 22195 + }, + { + "epoch": 0.06, + "learning_rate": 9.91356649345811e-05, + "loss": 2.4446, + "step": 22200 + }, + { + "epoch": 0.06, + "learning_rate": 9.913527491986356e-05, + "loss": 2.3734, + "step": 22205 + }, + { + "epoch": 0.06, + "learning_rate": 9.913488481794007e-05, + "loss": 2.3228, + "step": 22210 + }, + { + "epoch": 0.06, + "learning_rate": 9.913449462881127e-05, + "loss": 2.3533, + "step": 22215 + }, + { + "epoch": 0.06, + "learning_rate": 9.91341043524779e-05, + "loss": 2.3893, + "step": 22220 + }, + { + "epoch": 0.06, + "learning_rate": 9.913371398894064e-05, + "loss": 2.3393, + "step": 22225 + }, + { + "epoch": 0.06, + "learning_rate": 9.913332353820019e-05, + "loss": 2.3467, + "step": 22230 + }, + { + "epoch": 0.06, + "learning_rate": 9.913293300025722e-05, + "loss": 2.3658, + "step": 22235 + }, + { + "epoch": 0.06, + "learning_rate": 9.913254237511245e-05, + "loss": 2.3347, + "step": 22240 + }, + { + "epoch": 0.06, + "learning_rate": 9.913215166276655e-05, + "loss": 2.3663, + "step": 22245 + }, + { + "epoch": 0.06, + "learning_rate": 9.913176086322023e-05, + "loss": 2.3902, + "step": 22250 + }, + { + "epoch": 0.06, + "learning_rate": 9.913136997647416e-05, + "loss": 2.4201, + "step": 22255 + }, + { + "epoch": 0.06, + "learning_rate": 9.913097900252908e-05, + "loss": 2.2548, + "step": 22260 + }, + { + "epoch": 0.06, + "learning_rate": 9.913058794138563e-05, + "loss": 2.3816, + "step": 22265 + }, + { + "epoch": 0.06, + "learning_rate": 9.913019679304454e-05, + "loss": 2.4386, + "step": 22270 + }, + { + "epoch": 0.06, + "learning_rate": 9.91298055575065e-05, + "loss": 2.4507, + "step": 22275 + }, + { + "epoch": 0.06, + "learning_rate": 9.91294142347722e-05, + "loss": 2.4489, + "step": 22280 + }, + { + "epoch": 0.06, + "learning_rate": 9.912902282484231e-05, + "loss": 2.3407, + "step": 22285 + }, + { + "epoch": 0.06, + "learning_rate": 9.912863132771757e-05, + "loss": 2.4158, + "step": 22290 + }, + { + "epoch": 0.06, + "learning_rate": 9.912823974339864e-05, + "loss": 2.4089, + "step": 22295 + }, + { + "epoch": 0.06, + "learning_rate": 9.912784807188624e-05, + "loss": 2.3383, + "step": 22300 + }, + { + "epoch": 0.06, + "learning_rate": 9.912745631318104e-05, + "loss": 2.449, + "step": 22305 + }, + { + "epoch": 0.06, + "learning_rate": 9.912706446728375e-05, + "loss": 2.2877, + "step": 22310 + }, + { + "epoch": 0.06, + "learning_rate": 9.912667253419505e-05, + "loss": 2.4664, + "step": 22315 + }, + { + "epoch": 0.06, + "learning_rate": 9.912628051391567e-05, + "loss": 2.4321, + "step": 22320 + }, + { + "epoch": 0.06, + "learning_rate": 9.912588840644628e-05, + "loss": 2.4902, + "step": 22325 + }, + { + "epoch": 0.06, + "learning_rate": 9.912549621178757e-05, + "loss": 2.3887, + "step": 22330 + }, + { + "epoch": 0.06, + "learning_rate": 9.912510392994025e-05, + "loss": 2.4482, + "step": 22335 + }, + { + "epoch": 0.06, + "learning_rate": 9.912471156090501e-05, + "loss": 2.3591, + "step": 22340 + }, + { + "epoch": 0.06, + "learning_rate": 9.912431910468256e-05, + "loss": 2.2816, + "step": 22345 + }, + { + "epoch": 0.06, + "learning_rate": 9.912392656127358e-05, + "loss": 2.4467, + "step": 22350 + }, + { + "epoch": 0.06, + "learning_rate": 9.912353393067877e-05, + "loss": 2.4528, + "step": 22355 + }, + { + "epoch": 0.06, + "learning_rate": 9.912314121289883e-05, + "loss": 2.4537, + "step": 22360 + }, + { + "epoch": 0.06, + "learning_rate": 9.912274840793445e-05, + "loss": 2.3883, + "step": 22365 + }, + { + "epoch": 0.06, + "learning_rate": 9.912235551578634e-05, + "loss": 2.2388, + "step": 22370 + }, + { + "epoch": 0.06, + "learning_rate": 9.912196253645519e-05, + "loss": 2.3727, + "step": 22375 + }, + { + "epoch": 0.06, + "learning_rate": 9.91215694699417e-05, + "loss": 2.461, + "step": 22380 + }, + { + "epoch": 0.06, + "learning_rate": 9.912117631624656e-05, + "loss": 2.4652, + "step": 22385 + }, + { + "epoch": 0.06, + "learning_rate": 9.912078307537047e-05, + "loss": 2.3738, + "step": 22390 + }, + { + "epoch": 0.06, + "learning_rate": 9.912038974731415e-05, + "loss": 2.4882, + "step": 22395 + }, + { + "epoch": 0.06, + "learning_rate": 9.911999633207825e-05, + "loss": 2.326, + "step": 22400 + }, + { + "epoch": 0.06, + "learning_rate": 9.911960282966353e-05, + "loss": 2.3877, + "step": 22405 + }, + { + "epoch": 0.06, + "learning_rate": 9.911920924007063e-05, + "loss": 2.3649, + "step": 22410 + }, + { + "epoch": 0.06, + "learning_rate": 9.911881556330029e-05, + "loss": 2.4396, + "step": 22415 + }, + { + "epoch": 0.06, + "learning_rate": 9.911842179935317e-05, + "loss": 2.4224, + "step": 22420 + }, + { + "epoch": 0.06, + "learning_rate": 9.911802794823002e-05, + "loss": 2.4095, + "step": 22425 + }, + { + "epoch": 0.06, + "learning_rate": 9.91176340099315e-05, + "loss": 2.3783, + "step": 22430 + }, + { + "epoch": 0.06, + "learning_rate": 9.911723998445832e-05, + "loss": 2.351, + "step": 22435 + }, + { + "epoch": 0.06, + "learning_rate": 9.911684587181118e-05, + "loss": 2.3899, + "step": 22440 + }, + { + "epoch": 0.06, + "learning_rate": 9.911645167199077e-05, + "loss": 2.3822, + "step": 22445 + }, + { + "epoch": 0.06, + "learning_rate": 9.911605738499781e-05, + "loss": 2.439, + "step": 22450 + }, + { + "epoch": 0.06, + "learning_rate": 9.911566301083299e-05, + "loss": 2.3134, + "step": 22455 + }, + { + "epoch": 0.06, + "learning_rate": 9.911526854949701e-05, + "loss": 2.3098, + "step": 22460 + }, + { + "epoch": 0.06, + "learning_rate": 9.911487400099056e-05, + "loss": 2.3129, + "step": 22465 + }, + { + "epoch": 0.06, + "learning_rate": 9.911447936531436e-05, + "loss": 2.2948, + "step": 22470 + }, + { + "epoch": 0.06, + "learning_rate": 9.911408464246907e-05, + "loss": 2.3353, + "step": 22475 + }, + { + "epoch": 0.06, + "learning_rate": 9.911368983245545e-05, + "loss": 2.4309, + "step": 22480 + }, + { + "epoch": 0.06, + "learning_rate": 9.911329493527416e-05, + "loss": 2.5349, + "step": 22485 + }, + { + "epoch": 0.06, + "learning_rate": 9.911289995092592e-05, + "loss": 2.4872, + "step": 22490 + }, + { + "epoch": 0.06, + "learning_rate": 9.91125048794114e-05, + "loss": 2.3979, + "step": 22495 + }, + { + "epoch": 0.06, + "learning_rate": 9.911210972073134e-05, + "loss": 2.2833, + "step": 22500 + }, + { + "epoch": 0.06, + "learning_rate": 9.911171447488643e-05, + "loss": 2.4478, + "step": 22505 + }, + { + "epoch": 0.06, + "learning_rate": 9.911131914187735e-05, + "loss": 2.3915, + "step": 22510 + }, + { + "epoch": 0.06, + "learning_rate": 9.911092372170483e-05, + "loss": 2.5816, + "step": 22515 + }, + { + "epoch": 0.06, + "learning_rate": 9.911052821436956e-05, + "loss": 2.3673, + "step": 22520 + }, + { + "epoch": 0.06, + "learning_rate": 9.911013261987223e-05, + "loss": 2.5325, + "step": 22525 + }, + { + "epoch": 0.06, + "learning_rate": 9.910973693821357e-05, + "loss": 2.3424, + "step": 22530 + }, + { + "epoch": 0.06, + "learning_rate": 9.910934116939425e-05, + "loss": 2.3593, + "step": 22535 + }, + { + "epoch": 0.06, + "learning_rate": 9.910894531341499e-05, + "loss": 2.4267, + "step": 22540 + }, + { + "epoch": 0.06, + "learning_rate": 9.91085493702765e-05, + "loss": 2.4298, + "step": 22545 + }, + { + "epoch": 0.06, + "learning_rate": 9.910815333997947e-05, + "loss": 2.3848, + "step": 22550 + }, + { + "epoch": 0.06, + "learning_rate": 9.910775722252462e-05, + "loss": 2.5902, + "step": 22555 + }, + { + "epoch": 0.06, + "learning_rate": 9.910736101791261e-05, + "loss": 2.3818, + "step": 22560 + }, + { + "epoch": 0.06, + "learning_rate": 9.910696472614421e-05, + "loss": 2.3689, + "step": 22565 + }, + { + "epoch": 0.06, + "learning_rate": 9.910656834722006e-05, + "loss": 2.4154, + "step": 22570 + }, + { + "epoch": 0.06, + "learning_rate": 9.910617188114088e-05, + "loss": 2.3111, + "step": 22575 + }, + { + "epoch": 0.06, + "learning_rate": 9.91057753279074e-05, + "loss": 2.469, + "step": 22580 + }, + { + "epoch": 0.06, + "learning_rate": 9.910537868752031e-05, + "loss": 2.3992, + "step": 22585 + }, + { + "epoch": 0.06, + "learning_rate": 9.910498195998031e-05, + "loss": 2.4154, + "step": 22590 + }, + { + "epoch": 0.06, + "learning_rate": 9.91045851452881e-05, + "loss": 2.3507, + "step": 22595 + }, + { + "epoch": 0.06, + "learning_rate": 9.910418824344438e-05, + "loss": 2.4586, + "step": 22600 + }, + { + "epoch": 0.06, + "learning_rate": 9.910379125444989e-05, + "loss": 2.474, + "step": 22605 + }, + { + "epoch": 0.06, + "learning_rate": 9.91033941783053e-05, + "loss": 2.3097, + "step": 22610 + }, + { + "epoch": 0.06, + "learning_rate": 9.910299701501131e-05, + "loss": 2.4877, + "step": 22615 + }, + { + "epoch": 0.06, + "learning_rate": 9.910259976456863e-05, + "loss": 2.4477, + "step": 22620 + }, + { + "epoch": 0.06, + "learning_rate": 9.910220242697801e-05, + "loss": 2.3421, + "step": 22625 + }, + { + "epoch": 0.06, + "learning_rate": 9.910180500224009e-05, + "loss": 2.3301, + "step": 22630 + }, + { + "epoch": 0.06, + "learning_rate": 9.910140749035561e-05, + "loss": 2.2894, + "step": 22635 + }, + { + "epoch": 0.06, + "learning_rate": 9.910100989132527e-05, + "loss": 2.4569, + "step": 22640 + }, + { + "epoch": 0.06, + "learning_rate": 9.910061220514977e-05, + "loss": 2.294, + "step": 22645 + }, + { + "epoch": 0.06, + "learning_rate": 9.910021443182983e-05, + "loss": 2.3655, + "step": 22650 + }, + { + "epoch": 0.06, + "learning_rate": 9.909981657136613e-05, + "loss": 2.4215, + "step": 22655 + }, + { + "epoch": 0.06, + "learning_rate": 9.909941862375942e-05, + "loss": 2.314, + "step": 22660 + }, + { + "epoch": 0.06, + "learning_rate": 9.909902058901036e-05, + "loss": 2.49, + "step": 22665 + }, + { + "epoch": 0.06, + "learning_rate": 9.909862246711967e-05, + "loss": 2.4417, + "step": 22670 + }, + { + "epoch": 0.06, + "learning_rate": 9.909822425808808e-05, + "loss": 2.3974, + "step": 22675 + }, + { + "epoch": 0.06, + "learning_rate": 9.909782596191627e-05, + "loss": 2.3731, + "step": 22680 + }, + { + "epoch": 0.06, + "learning_rate": 9.909742757860496e-05, + "loss": 2.3297, + "step": 22685 + }, + { + "epoch": 0.06, + "learning_rate": 9.909702910815484e-05, + "loss": 2.453, + "step": 22690 + }, + { + "epoch": 0.06, + "learning_rate": 9.909663055056664e-05, + "loss": 2.4816, + "step": 22695 + }, + { + "epoch": 0.06, + "learning_rate": 9.909623190584105e-05, + "loss": 2.2786, + "step": 22700 + }, + { + "epoch": 0.06, + "learning_rate": 9.90958331739788e-05, + "loss": 2.3803, + "step": 22705 + }, + { + "epoch": 0.06, + "learning_rate": 9.909543435498057e-05, + "loss": 2.464, + "step": 22710 + }, + { + "epoch": 0.06, + "learning_rate": 9.909503544884709e-05, + "loss": 2.4249, + "step": 22715 + }, + { + "epoch": 0.06, + "learning_rate": 9.909463645557906e-05, + "loss": 2.4998, + "step": 22720 + }, + { + "epoch": 0.06, + "learning_rate": 9.909423737517718e-05, + "loss": 2.3922, + "step": 22725 + }, + { + "epoch": 0.06, + "learning_rate": 9.909383820764215e-05, + "loss": 2.456, + "step": 22730 + }, + { + "epoch": 0.06, + "learning_rate": 9.909343895297471e-05, + "loss": 2.4184, + "step": 22735 + }, + { + "epoch": 0.06, + "learning_rate": 9.909303961117555e-05, + "loss": 2.3504, + "step": 22740 + }, + { + "epoch": 0.06, + "learning_rate": 9.909264018224538e-05, + "loss": 2.3269, + "step": 22745 + }, + { + "epoch": 0.06, + "learning_rate": 9.909224066618492e-05, + "loss": 2.4723, + "step": 22750 + }, + { + "epoch": 0.06, + "learning_rate": 9.909184106299485e-05, + "loss": 2.3523, + "step": 22755 + }, + { + "epoch": 0.06, + "learning_rate": 9.90914413726759e-05, + "loss": 2.3391, + "step": 22760 + }, + { + "epoch": 0.06, + "learning_rate": 9.90910415952288e-05, + "loss": 2.4115, + "step": 22765 + }, + { + "epoch": 0.06, + "learning_rate": 9.909064173065421e-05, + "loss": 2.3615, + "step": 22770 + }, + { + "epoch": 0.06, + "learning_rate": 9.909024177895287e-05, + "loss": 2.345, + "step": 22775 + }, + { + "epoch": 0.06, + "learning_rate": 9.90898417401255e-05, + "loss": 2.5123, + "step": 22780 + }, + { + "epoch": 0.06, + "learning_rate": 9.90894416141728e-05, + "loss": 2.4552, + "step": 22785 + }, + { + "epoch": 0.06, + "learning_rate": 9.908904140109546e-05, + "loss": 2.468, + "step": 22790 + }, + { + "epoch": 0.06, + "learning_rate": 9.908864110089421e-05, + "loss": 2.4338, + "step": 22795 + }, + { + "epoch": 0.06, + "learning_rate": 9.908824071356976e-05, + "loss": 2.4243, + "step": 22800 + }, + { + "epoch": 0.06, + "learning_rate": 9.908784023912281e-05, + "loss": 2.3274, + "step": 22805 + }, + { + "epoch": 0.06, + "learning_rate": 9.90874396775541e-05, + "loss": 2.2686, + "step": 22810 + }, + { + "epoch": 0.06, + "learning_rate": 9.90870390288643e-05, + "loss": 2.3235, + "step": 22815 + }, + { + "epoch": 0.06, + "learning_rate": 9.908663829305413e-05, + "loss": 2.3844, + "step": 22820 + }, + { + "epoch": 0.06, + "learning_rate": 9.908623747012433e-05, + "loss": 2.3847, + "step": 22825 + }, + { + "epoch": 0.06, + "learning_rate": 9.90858365600756e-05, + "loss": 2.415, + "step": 22830 + }, + { + "epoch": 0.06, + "learning_rate": 9.908543556290863e-05, + "loss": 2.3407, + "step": 22835 + }, + { + "epoch": 0.06, + "learning_rate": 9.908503447862416e-05, + "loss": 2.402, + "step": 22840 + }, + { + "epoch": 0.06, + "learning_rate": 9.908463330722288e-05, + "loss": 2.4736, + "step": 22845 + }, + { + "epoch": 0.06, + "learning_rate": 9.908423204870552e-05, + "loss": 2.5083, + "step": 22850 + }, + { + "epoch": 0.06, + "learning_rate": 9.908383070307276e-05, + "loss": 2.4858, + "step": 22855 + }, + { + "epoch": 0.06, + "learning_rate": 9.908342927032534e-05, + "loss": 2.3055, + "step": 22860 + }, + { + "epoch": 0.06, + "learning_rate": 9.908302775046398e-05, + "loss": 2.292, + "step": 22865 + }, + { + "epoch": 0.06, + "learning_rate": 9.908262614348938e-05, + "loss": 2.2531, + "step": 22870 + }, + { + "epoch": 0.06, + "learning_rate": 9.908222444940224e-05, + "loss": 2.3329, + "step": 22875 + }, + { + "epoch": 0.06, + "learning_rate": 9.90818226682033e-05, + "loss": 2.3246, + "step": 22880 + }, + { + "epoch": 0.06, + "learning_rate": 9.908142079989326e-05, + "loss": 2.424, + "step": 22885 + }, + { + "epoch": 0.06, + "learning_rate": 9.908101884447281e-05, + "loss": 2.3944, + "step": 22890 + }, + { + "epoch": 0.06, + "learning_rate": 9.908061680194271e-05, + "loss": 2.4544, + "step": 22895 + }, + { + "epoch": 0.06, + "learning_rate": 9.908021467230364e-05, + "loss": 2.4397, + "step": 22900 + }, + { + "epoch": 0.06, + "learning_rate": 9.907981245555631e-05, + "loss": 2.3358, + "step": 22905 + }, + { + "epoch": 0.06, + "learning_rate": 9.907941015170145e-05, + "loss": 2.4035, + "step": 22910 + }, + { + "epoch": 0.06, + "learning_rate": 9.90790077607398e-05, + "loss": 2.3789, + "step": 22915 + }, + { + "epoch": 0.06, + "learning_rate": 9.907860528267201e-05, + "loss": 2.4459, + "step": 22920 + }, + { + "epoch": 0.06, + "learning_rate": 9.907820271749884e-05, + "loss": 2.3939, + "step": 22925 + }, + { + "epoch": 0.06, + "learning_rate": 9.907780006522101e-05, + "loss": 2.3895, + "step": 22930 + }, + { + "epoch": 0.06, + "learning_rate": 9.907739732583918e-05, + "loss": 2.4431, + "step": 22935 + }, + { + "epoch": 0.06, + "learning_rate": 9.907699449935414e-05, + "loss": 2.3227, + "step": 22940 + }, + { + "epoch": 0.06, + "learning_rate": 9.907659158576654e-05, + "loss": 2.3458, + "step": 22945 + }, + { + "epoch": 0.06, + "learning_rate": 9.907618858507714e-05, + "loss": 2.3808, + "step": 22950 + }, + { + "epoch": 0.06, + "learning_rate": 9.907578549728665e-05, + "loss": 2.3929, + "step": 22955 + }, + { + "epoch": 0.06, + "learning_rate": 9.907538232239574e-05, + "loss": 2.3637, + "step": 22960 + }, + { + "epoch": 0.06, + "learning_rate": 9.907497906040517e-05, + "loss": 2.4732, + "step": 22965 + }, + { + "epoch": 0.06, + "learning_rate": 9.907457571131566e-05, + "loss": 2.2371, + "step": 22970 + }, + { + "epoch": 0.06, + "learning_rate": 9.90741722751279e-05, + "loss": 2.5106, + "step": 22975 + }, + { + "epoch": 0.06, + "learning_rate": 9.907376875184262e-05, + "loss": 2.4246, + "step": 22980 + }, + { + "epoch": 0.06, + "learning_rate": 9.907336514146053e-05, + "loss": 2.3396, + "step": 22985 + }, + { + "epoch": 0.06, + "learning_rate": 9.907296144398235e-05, + "loss": 2.3751, + "step": 22990 + }, + { + "epoch": 0.06, + "learning_rate": 9.907255765940879e-05, + "loss": 2.3669, + "step": 22995 + }, + { + "epoch": 0.06, + "learning_rate": 9.907215378774059e-05, + "loss": 2.3467, + "step": 23000 + }, + { + "epoch": 0.06, + "learning_rate": 9.907174982897844e-05, + "loss": 2.2792, + "step": 23005 + }, + { + "epoch": 0.06, + "learning_rate": 9.907134578312304e-05, + "loss": 2.4759, + "step": 23010 + }, + { + "epoch": 0.06, + "learning_rate": 9.907094165017518e-05, + "loss": 2.4041, + "step": 23015 + }, + { + "epoch": 0.06, + "learning_rate": 9.907053743013551e-05, + "loss": 2.3303, + "step": 23020 + }, + { + "epoch": 0.06, + "learning_rate": 9.907013312300475e-05, + "loss": 2.2904, + "step": 23025 + }, + { + "epoch": 0.06, + "learning_rate": 9.906972872878367e-05, + "loss": 2.31, + "step": 23030 + }, + { + "epoch": 0.06, + "learning_rate": 9.906932424747292e-05, + "loss": 2.5214, + "step": 23035 + }, + { + "epoch": 0.06, + "learning_rate": 9.906891967907326e-05, + "loss": 2.2046, + "step": 23040 + }, + { + "epoch": 0.06, + "learning_rate": 9.90685150235854e-05, + "loss": 2.4528, + "step": 23045 + }, + { + "epoch": 0.06, + "learning_rate": 9.906811028101007e-05, + "loss": 2.3958, + "step": 23050 + }, + { + "epoch": 0.06, + "learning_rate": 9.906770545134796e-05, + "loss": 2.47, + "step": 23055 + }, + { + "epoch": 0.06, + "learning_rate": 9.906730053459982e-05, + "loss": 2.434, + "step": 23060 + }, + { + "epoch": 0.06, + "learning_rate": 9.906689553076634e-05, + "loss": 2.3984, + "step": 23065 + }, + { + "epoch": 0.06, + "learning_rate": 9.906649043984826e-05, + "loss": 2.389, + "step": 23070 + }, + { + "epoch": 0.06, + "learning_rate": 9.906608526184628e-05, + "loss": 2.3965, + "step": 23075 + }, + { + "epoch": 0.06, + "learning_rate": 9.906567999676113e-05, + "loss": 2.2819, + "step": 23080 + }, + { + "epoch": 0.06, + "learning_rate": 9.906527464459354e-05, + "loss": 2.4045, + "step": 23085 + }, + { + "epoch": 0.06, + "learning_rate": 9.906486920534421e-05, + "loss": 2.3376, + "step": 23090 + }, + { + "epoch": 0.06, + "learning_rate": 9.906446367901387e-05, + "loss": 2.4308, + "step": 23095 + }, + { + "epoch": 0.06, + "learning_rate": 9.906405806560324e-05, + "loss": 2.3467, + "step": 23100 + }, + { + "epoch": 0.06, + "learning_rate": 9.906365236511303e-05, + "loss": 2.3246, + "step": 23105 + }, + { + "epoch": 0.06, + "learning_rate": 9.906324657754398e-05, + "loss": 2.3806, + "step": 23110 + }, + { + "epoch": 0.06, + "learning_rate": 9.906284070289678e-05, + "loss": 2.218, + "step": 23115 + }, + { + "epoch": 0.06, + "learning_rate": 9.906243474117219e-05, + "loss": 2.5485, + "step": 23120 + }, + { + "epoch": 0.06, + "learning_rate": 9.906202869237089e-05, + "loss": 2.3729, + "step": 23125 + }, + { + "epoch": 0.06, + "learning_rate": 9.906162255649363e-05, + "loss": 2.3391, + "step": 23130 + }, + { + "epoch": 0.06, + "learning_rate": 9.906121633354111e-05, + "loss": 2.3749, + "step": 23135 + }, + { + "epoch": 0.06, + "learning_rate": 9.906081002351406e-05, + "loss": 2.4418, + "step": 23140 + }, + { + "epoch": 0.06, + "learning_rate": 9.906040362641322e-05, + "loss": 2.4451, + "step": 23145 + }, + { + "epoch": 0.06, + "learning_rate": 9.905999714223928e-05, + "loss": 2.369, + "step": 23150 + }, + { + "epoch": 0.06, + "learning_rate": 9.905959057099298e-05, + "loss": 2.4686, + "step": 23155 + }, + { + "epoch": 0.06, + "learning_rate": 9.905918391267503e-05, + "loss": 2.2978, + "step": 23160 + }, + { + "epoch": 0.06, + "learning_rate": 9.905877716728616e-05, + "loss": 2.2757, + "step": 23165 + }, + { + "epoch": 0.06, + "learning_rate": 9.90583703348271e-05, + "loss": 2.3737, + "step": 23170 + }, + { + "epoch": 0.06, + "learning_rate": 9.905796341529856e-05, + "loss": 2.4494, + "step": 23175 + }, + { + "epoch": 0.06, + "learning_rate": 9.905755640870124e-05, + "loss": 2.332, + "step": 23180 + }, + { + "epoch": 0.06, + "learning_rate": 9.905714931503592e-05, + "loss": 2.3837, + "step": 23185 + }, + { + "epoch": 0.06, + "learning_rate": 9.905674213430327e-05, + "loss": 2.3987, + "step": 23190 + }, + { + "epoch": 0.06, + "learning_rate": 9.905633486650404e-05, + "loss": 2.3693, + "step": 23195 + }, + { + "epoch": 0.06, + "learning_rate": 9.905592751163895e-05, + "loss": 2.3246, + "step": 23200 + }, + { + "epoch": 0.06, + "learning_rate": 9.905552006970871e-05, + "loss": 2.3473, + "step": 23205 + }, + { + "epoch": 0.06, + "learning_rate": 9.905511254071405e-05, + "loss": 2.4676, + "step": 23210 + }, + { + "epoch": 0.06, + "learning_rate": 9.90547049246557e-05, + "loss": 2.3819, + "step": 23215 + }, + { + "epoch": 0.06, + "learning_rate": 9.905429722153436e-05, + "loss": 2.3673, + "step": 23220 + }, + { + "epoch": 0.06, + "learning_rate": 9.905388943135078e-05, + "loss": 2.3439, + "step": 23225 + }, + { + "epoch": 0.06, + "learning_rate": 9.905348155410568e-05, + "loss": 2.27, + "step": 23230 + }, + { + "epoch": 0.06, + "learning_rate": 9.905307358979978e-05, + "loss": 2.4386, + "step": 23235 + }, + { + "epoch": 0.06, + "learning_rate": 9.90526655384338e-05, + "loss": 2.466, + "step": 23240 + }, + { + "epoch": 0.06, + "learning_rate": 9.905225740000848e-05, + "loss": 2.3154, + "step": 23245 + }, + { + "epoch": 0.06, + "learning_rate": 9.90518491745245e-05, + "loss": 2.4776, + "step": 23250 + }, + { + "epoch": 0.06, + "learning_rate": 9.905144086198263e-05, + "loss": 2.4491, + "step": 23255 + }, + { + "epoch": 0.06, + "learning_rate": 9.90510324623836e-05, + "loss": 2.4374, + "step": 23260 + }, + { + "epoch": 0.06, + "learning_rate": 9.90506239757281e-05, + "loss": 2.4121, + "step": 23265 + }, + { + "epoch": 0.06, + "learning_rate": 9.905021540201687e-05, + "loss": 2.3118, + "step": 23270 + }, + { + "epoch": 0.06, + "learning_rate": 9.904980674125063e-05, + "loss": 2.2949, + "step": 23275 + }, + { + "epoch": 0.06, + "learning_rate": 9.904939799343013e-05, + "loss": 2.3502, + "step": 23280 + }, + { + "epoch": 0.06, + "learning_rate": 9.904898915855607e-05, + "loss": 2.5275, + "step": 23285 + }, + { + "epoch": 0.06, + "learning_rate": 9.904858023662916e-05, + "loss": 2.4219, + "step": 23290 + }, + { + "epoch": 0.06, + "learning_rate": 9.904817122765017e-05, + "loss": 2.2551, + "step": 23295 + }, + { + "epoch": 0.06, + "learning_rate": 9.904776213161979e-05, + "loss": 2.3345, + "step": 23300 + }, + { + "epoch": 0.06, + "learning_rate": 9.904735294853878e-05, + "loss": 2.3574, + "step": 23305 + }, + { + "epoch": 0.06, + "learning_rate": 9.904694367840784e-05, + "loss": 2.4821, + "step": 23310 + }, + { + "epoch": 0.06, + "learning_rate": 9.904653432122768e-05, + "loss": 2.4475, + "step": 23315 + }, + { + "epoch": 0.06, + "learning_rate": 9.904612487699907e-05, + "loss": 2.3952, + "step": 23320 + }, + { + "epoch": 0.06, + "learning_rate": 9.90457153457227e-05, + "loss": 2.3171, + "step": 23325 + }, + { + "epoch": 0.06, + "learning_rate": 9.904530572739934e-05, + "loss": 2.3477, + "step": 23330 + }, + { + "epoch": 0.06, + "learning_rate": 9.904489602202967e-05, + "loss": 2.3927, + "step": 23335 + }, + { + "epoch": 0.06, + "learning_rate": 9.904448622961444e-05, + "loss": 2.3576, + "step": 23340 + }, + { + "epoch": 0.06, + "learning_rate": 9.904407635015437e-05, + "loss": 2.3351, + "step": 23345 + }, + { + "epoch": 0.06, + "learning_rate": 9.904366638365019e-05, + "loss": 2.3889, + "step": 23350 + }, + { + "epoch": 0.06, + "learning_rate": 9.904325633010263e-05, + "loss": 2.3744, + "step": 23355 + }, + { + "epoch": 0.06, + "learning_rate": 9.904284618951243e-05, + "loss": 2.4722, + "step": 23360 + }, + { + "epoch": 0.06, + "learning_rate": 9.90424359618803e-05, + "loss": 2.2712, + "step": 23365 + }, + { + "epoch": 0.06, + "learning_rate": 9.904202564720697e-05, + "loss": 2.4466, + "step": 23370 + }, + { + "epoch": 0.06, + "learning_rate": 9.904161524549316e-05, + "loss": 2.2673, + "step": 23375 + }, + { + "epoch": 0.06, + "learning_rate": 9.904120475673962e-05, + "loss": 2.3397, + "step": 23380 + }, + { + "epoch": 0.06, + "learning_rate": 9.904079418094707e-05, + "loss": 2.2732, + "step": 23385 + }, + { + "epoch": 0.06, + "learning_rate": 9.904038351811622e-05, + "loss": 2.483, + "step": 23390 + }, + { + "epoch": 0.06, + "learning_rate": 9.903997276824784e-05, + "loss": 2.3527, + "step": 23395 + }, + { + "epoch": 0.06, + "learning_rate": 9.90395619313426e-05, + "loss": 2.355, + "step": 23400 + }, + { + "epoch": 0.06, + "learning_rate": 9.90391510074013e-05, + "loss": 2.3644, + "step": 23405 + }, + { + "epoch": 0.06, + "learning_rate": 9.90387399964246e-05, + "loss": 2.4476, + "step": 23410 + }, + { + "epoch": 0.06, + "learning_rate": 9.903832889841329e-05, + "loss": 2.3937, + "step": 23415 + }, + { + "epoch": 0.06, + "learning_rate": 9.903791771336806e-05, + "loss": 2.4217, + "step": 23420 + }, + { + "epoch": 0.06, + "learning_rate": 9.903750644128967e-05, + "loss": 2.3312, + "step": 23425 + }, + { + "epoch": 0.06, + "learning_rate": 9.90370950821788e-05, + "loss": 2.3948, + "step": 23430 + }, + { + "epoch": 0.06, + "learning_rate": 9.903668363603624e-05, + "loss": 2.3946, + "step": 23435 + }, + { + "epoch": 0.06, + "learning_rate": 9.903627210286268e-05, + "loss": 2.393, + "step": 23440 + }, + { + "epoch": 0.06, + "learning_rate": 9.903586048265886e-05, + "loss": 2.4343, + "step": 23445 + }, + { + "epoch": 0.06, + "learning_rate": 9.903544877542552e-05, + "loss": 2.4604, + "step": 23450 + }, + { + "epoch": 0.06, + "learning_rate": 9.903503698116337e-05, + "loss": 2.4632, + "step": 23455 + }, + { + "epoch": 0.06, + "learning_rate": 9.903462509987316e-05, + "loss": 2.3868, + "step": 23460 + }, + { + "epoch": 0.06, + "learning_rate": 9.903421313155563e-05, + "loss": 2.344, + "step": 23465 + }, + { + "epoch": 0.06, + "learning_rate": 9.903380107621148e-05, + "loss": 2.3076, + "step": 23470 + }, + { + "epoch": 0.06, + "learning_rate": 9.903338893384146e-05, + "loss": 2.294, + "step": 23475 + }, + { + "epoch": 0.06, + "learning_rate": 9.903297670444631e-05, + "loss": 2.3672, + "step": 23480 + }, + { + "epoch": 0.06, + "learning_rate": 9.903256438802675e-05, + "loss": 2.381, + "step": 23485 + }, + { + "epoch": 0.06, + "learning_rate": 9.903215198458351e-05, + "loss": 2.3543, + "step": 23490 + }, + { + "epoch": 0.06, + "learning_rate": 9.903173949411732e-05, + "loss": 2.2525, + "step": 23495 + }, + { + "epoch": 0.06, + "learning_rate": 9.903132691662891e-05, + "loss": 2.3864, + "step": 23500 + }, + { + "epoch": 0.06, + "learning_rate": 9.903091425211903e-05, + "loss": 2.3416, + "step": 23505 + }, + { + "epoch": 0.06, + "learning_rate": 9.90305015005884e-05, + "loss": 2.287, + "step": 23510 + }, + { + "epoch": 0.06, + "learning_rate": 9.903008866203776e-05, + "loss": 2.3912, + "step": 23515 + }, + { + "epoch": 0.06, + "learning_rate": 9.902967573646783e-05, + "loss": 2.3547, + "step": 23520 + }, + { + "epoch": 0.06, + "learning_rate": 9.902926272387936e-05, + "loss": 2.3683, + "step": 23525 + }, + { + "epoch": 0.06, + "learning_rate": 9.902884962427305e-05, + "loss": 2.3864, + "step": 23530 + }, + { + "epoch": 0.06, + "learning_rate": 9.902843643764968e-05, + "loss": 2.2785, + "step": 23535 + }, + { + "epoch": 0.06, + "learning_rate": 9.902802316400994e-05, + "loss": 2.3975, + "step": 23540 + }, + { + "epoch": 0.06, + "learning_rate": 9.90276098033546e-05, + "loss": 2.329, + "step": 23545 + }, + { + "epoch": 0.06, + "learning_rate": 9.902719635568436e-05, + "loss": 2.3383, + "step": 23550 + }, + { + "epoch": 0.06, + "learning_rate": 9.902678282099997e-05, + "loss": 2.3359, + "step": 23555 + }, + { + "epoch": 0.06, + "learning_rate": 9.902636919930218e-05, + "loss": 2.4057, + "step": 23560 + }, + { + "epoch": 0.06, + "learning_rate": 9.90259554905917e-05, + "loss": 2.2681, + "step": 23565 + }, + { + "epoch": 0.06, + "learning_rate": 9.902554169486928e-05, + "loss": 2.3808, + "step": 23570 + }, + { + "epoch": 0.06, + "learning_rate": 9.902512781213564e-05, + "loss": 2.4904, + "step": 23575 + }, + { + "epoch": 0.06, + "learning_rate": 9.902471384239151e-05, + "loss": 2.4174, + "step": 23580 + }, + { + "epoch": 0.06, + "learning_rate": 9.902429978563766e-05, + "loss": 2.3721, + "step": 23585 + }, + { + "epoch": 0.06, + "learning_rate": 9.90238856418748e-05, + "loss": 2.4633, + "step": 23590 + }, + { + "epoch": 0.06, + "learning_rate": 9.902347141110364e-05, + "loss": 2.3747, + "step": 23595 + }, + { + "epoch": 0.06, + "learning_rate": 9.902305709332496e-05, + "loss": 2.2407, + "step": 23600 + }, + { + "epoch": 0.06, + "learning_rate": 9.902264268853947e-05, + "loss": 2.3498, + "step": 23605 + }, + { + "epoch": 0.06, + "learning_rate": 9.902222819674792e-05, + "loss": 2.3141, + "step": 23610 + }, + { + "epoch": 0.06, + "learning_rate": 9.902181361795102e-05, + "loss": 2.5061, + "step": 23615 + }, + { + "epoch": 0.06, + "learning_rate": 9.902139895214954e-05, + "loss": 2.4098, + "step": 23620 + }, + { + "epoch": 0.06, + "learning_rate": 9.90209841993442e-05, + "loss": 2.4185, + "step": 23625 + }, + { + "epoch": 0.06, + "learning_rate": 9.902056935953573e-05, + "loss": 2.3714, + "step": 23630 + }, + { + "epoch": 0.06, + "learning_rate": 9.902015443272488e-05, + "loss": 2.3633, + "step": 23635 + }, + { + "epoch": 0.06, + "learning_rate": 9.901973941891237e-05, + "loss": 2.3729, + "step": 23640 + }, + { + "epoch": 0.06, + "learning_rate": 9.901932431809894e-05, + "loss": 2.4113, + "step": 23645 + }, + { + "epoch": 0.06, + "learning_rate": 9.901890913028533e-05, + "loss": 2.3453, + "step": 23650 + }, + { + "epoch": 0.06, + "learning_rate": 9.90184938554723e-05, + "loss": 2.4516, + "step": 23655 + }, + { + "epoch": 0.06, + "learning_rate": 9.901807849366054e-05, + "loss": 2.3573, + "step": 23660 + }, + { + "epoch": 0.06, + "learning_rate": 9.901766304485083e-05, + "loss": 2.3851, + "step": 23665 + }, + { + "epoch": 0.06, + "learning_rate": 9.901724750904389e-05, + "loss": 2.3898, + "step": 23670 + }, + { + "epoch": 0.06, + "learning_rate": 9.901683188624045e-05, + "loss": 2.3987, + "step": 23675 + }, + { + "epoch": 0.06, + "learning_rate": 9.901641617644125e-05, + "loss": 2.3619, + "step": 23680 + }, + { + "epoch": 0.06, + "learning_rate": 9.901600037964704e-05, + "loss": 2.4053, + "step": 23685 + }, + { + "epoch": 0.06, + "learning_rate": 9.901558449585855e-05, + "loss": 2.4581, + "step": 23690 + }, + { + "epoch": 0.06, + "learning_rate": 9.901516852507653e-05, + "loss": 2.4546, + "step": 23695 + }, + { + "epoch": 0.06, + "learning_rate": 9.90147524673017e-05, + "loss": 2.5228, + "step": 23700 + }, + { + "epoch": 0.06, + "learning_rate": 9.90143363225348e-05, + "loss": 2.3461, + "step": 23705 + }, + { + "epoch": 0.06, + "learning_rate": 9.901392009077657e-05, + "loss": 2.4321, + "step": 23710 + }, + { + "epoch": 0.06, + "learning_rate": 9.901350377202776e-05, + "loss": 2.3575, + "step": 23715 + }, + { + "epoch": 0.06, + "learning_rate": 9.90130873662891e-05, + "loss": 2.2839, + "step": 23720 + }, + { + "epoch": 0.06, + "learning_rate": 9.901267087356134e-05, + "loss": 2.5106, + "step": 23725 + }, + { + "epoch": 0.06, + "learning_rate": 9.90122542938452e-05, + "loss": 2.3689, + "step": 23730 + }, + { + "epoch": 0.06, + "learning_rate": 9.901183762714144e-05, + "loss": 2.3165, + "step": 23735 + }, + { + "epoch": 0.06, + "learning_rate": 9.901142087345077e-05, + "loss": 2.4653, + "step": 23740 + }, + { + "epoch": 0.06, + "learning_rate": 9.901100403277397e-05, + "loss": 2.3673, + "step": 23745 + }, + { + "epoch": 0.06, + "learning_rate": 9.901058710511175e-05, + "loss": 2.4331, + "step": 23750 + }, + { + "epoch": 0.06, + "learning_rate": 9.901017009046485e-05, + "loss": 2.436, + "step": 23755 + }, + { + "epoch": 0.06, + "learning_rate": 9.900975298883403e-05, + "loss": 2.4864, + "step": 23760 + }, + { + "epoch": 0.06, + "learning_rate": 9.900933580022003e-05, + "loss": 2.4007, + "step": 23765 + }, + { + "epoch": 0.06, + "learning_rate": 9.900891852462355e-05, + "loss": 2.4219, + "step": 23770 + }, + { + "epoch": 0.06, + "learning_rate": 9.900850116204539e-05, + "loss": 2.2984, + "step": 23775 + }, + { + "epoch": 0.06, + "learning_rate": 9.900808371248624e-05, + "loss": 2.5128, + "step": 23780 + }, + { + "epoch": 0.06, + "learning_rate": 9.900766617594687e-05, + "loss": 2.2621, + "step": 23785 + }, + { + "epoch": 0.06, + "learning_rate": 9.9007248552428e-05, + "loss": 2.3982, + "step": 23790 + }, + { + "epoch": 0.06, + "learning_rate": 9.90068308419304e-05, + "loss": 2.3191, + "step": 23795 + }, + { + "epoch": 0.06, + "learning_rate": 9.900641304445479e-05, + "loss": 2.2948, + "step": 23800 + }, + { + "epoch": 0.06, + "learning_rate": 9.900599516000191e-05, + "loss": 2.4199, + "step": 23805 + }, + { + "epoch": 0.06, + "learning_rate": 9.900557718857252e-05, + "loss": 2.5658, + "step": 23810 + }, + { + "epoch": 0.06, + "learning_rate": 9.900515913016734e-05, + "loss": 2.4628, + "step": 23815 + }, + { + "epoch": 0.06, + "learning_rate": 9.900474098478714e-05, + "loss": 2.4089, + "step": 23820 + }, + { + "epoch": 0.06, + "learning_rate": 9.900432275243263e-05, + "loss": 2.3854, + "step": 23825 + }, + { + "epoch": 0.06, + "learning_rate": 9.900390443310457e-05, + "loss": 2.5315, + "step": 23830 + }, + { + "epoch": 0.06, + "learning_rate": 9.900348602680368e-05, + "loss": 2.3263, + "step": 23835 + }, + { + "epoch": 0.06, + "learning_rate": 9.900306753353074e-05, + "loss": 2.4096, + "step": 23840 + }, + { + "epoch": 0.06, + "learning_rate": 9.900264895328648e-05, + "loss": 2.3793, + "step": 23845 + }, + { + "epoch": 0.06, + "learning_rate": 9.900223028607162e-05, + "loss": 2.4019, + "step": 23850 + }, + { + "epoch": 0.06, + "learning_rate": 9.900181153188693e-05, + "loss": 2.4249, + "step": 23855 + }, + { + "epoch": 0.06, + "learning_rate": 9.900139269073314e-05, + "loss": 2.3819, + "step": 23860 + }, + { + "epoch": 0.06, + "learning_rate": 9.900097376261101e-05, + "loss": 2.2525, + "step": 23865 + }, + { + "epoch": 0.06, + "learning_rate": 9.900055474752126e-05, + "loss": 2.391, + "step": 23870 + }, + { + "epoch": 0.06, + "learning_rate": 9.900013564546465e-05, + "loss": 2.4226, + "step": 23875 + }, + { + "epoch": 0.06, + "learning_rate": 9.899971645644191e-05, + "loss": 2.3925, + "step": 23880 + }, + { + "epoch": 0.06, + "learning_rate": 9.89992971804538e-05, + "loss": 2.3217, + "step": 23885 + }, + { + "epoch": 0.06, + "learning_rate": 9.899887781750104e-05, + "loss": 2.4477, + "step": 23890 + }, + { + "epoch": 0.06, + "learning_rate": 9.89984583675844e-05, + "loss": 2.3297, + "step": 23895 + }, + { + "epoch": 0.06, + "learning_rate": 9.899803883070462e-05, + "loss": 2.4107, + "step": 23900 + }, + { + "epoch": 0.06, + "learning_rate": 9.899761920686243e-05, + "loss": 2.3698, + "step": 23905 + }, + { + "epoch": 0.06, + "learning_rate": 9.899719949605859e-05, + "loss": 2.3427, + "step": 23910 + }, + { + "epoch": 0.06, + "learning_rate": 9.899677969829384e-05, + "loss": 2.4202, + "step": 23915 + }, + { + "epoch": 0.06, + "learning_rate": 9.899635981356891e-05, + "loss": 2.4542, + "step": 23920 + }, + { + "epoch": 0.06, + "learning_rate": 9.899593984188457e-05, + "loss": 2.2318, + "step": 23925 + }, + { + "epoch": 0.06, + "learning_rate": 9.899551978324158e-05, + "loss": 2.4429, + "step": 23930 + }, + { + "epoch": 0.06, + "learning_rate": 9.899509963764062e-05, + "loss": 2.4783, + "step": 23935 + }, + { + "epoch": 0.06, + "learning_rate": 9.899467940508251e-05, + "loss": 2.4557, + "step": 23940 + }, + { + "epoch": 0.06, + "learning_rate": 9.899425908556794e-05, + "loss": 2.4387, + "step": 23945 + }, + { + "epoch": 0.06, + "learning_rate": 9.899383867909768e-05, + "loss": 2.4107, + "step": 23950 + }, + { + "epoch": 0.06, + "learning_rate": 9.899341818567247e-05, + "loss": 2.437, + "step": 23955 + }, + { + "epoch": 0.06, + "learning_rate": 9.899299760529307e-05, + "loss": 2.3595, + "step": 23960 + }, + { + "epoch": 0.06, + "learning_rate": 9.89925769379602e-05, + "loss": 2.4093, + "step": 23965 + }, + { + "epoch": 0.06, + "learning_rate": 9.899215618367464e-05, + "loss": 2.4616, + "step": 23970 + }, + { + "epoch": 0.06, + "learning_rate": 9.899173534243712e-05, + "loss": 2.4974, + "step": 23975 + }, + { + "epoch": 0.06, + "learning_rate": 9.899131441424838e-05, + "loss": 2.3946, + "step": 23980 + }, + { + "epoch": 0.06, + "learning_rate": 9.899089339910918e-05, + "loss": 2.2979, + "step": 23985 + }, + { + "epoch": 0.06, + "learning_rate": 9.899047229702026e-05, + "loss": 2.2621, + "step": 23990 + }, + { + "epoch": 0.06, + "learning_rate": 9.899005110798236e-05, + "loss": 2.2793, + "step": 23995 + }, + { + "epoch": 0.06, + "learning_rate": 9.898962983199625e-05, + "loss": 2.439, + "step": 24000 + }, + { + "epoch": 0.06, + "learning_rate": 9.898920846906266e-05, + "loss": 2.5537, + "step": 24005 + }, + { + "epoch": 0.06, + "learning_rate": 9.898878701918232e-05, + "loss": 2.4394, + "step": 24010 + }, + { + "epoch": 0.06, + "learning_rate": 9.898836548235603e-05, + "loss": 2.4349, + "step": 24015 + }, + { + "epoch": 0.06, + "learning_rate": 9.89879438585845e-05, + "loss": 2.3885, + "step": 24020 + }, + { + "epoch": 0.06, + "learning_rate": 9.898752214786847e-05, + "loss": 2.4462, + "step": 24025 + }, + { + "epoch": 0.06, + "learning_rate": 9.898710035020871e-05, + "loss": 2.3974, + "step": 24030 + }, + { + "epoch": 0.06, + "learning_rate": 9.898667846560598e-05, + "loss": 2.4397, + "step": 24035 + }, + { + "epoch": 0.06, + "learning_rate": 9.8986256494061e-05, + "loss": 2.3401, + "step": 24040 + }, + { + "epoch": 0.06, + "learning_rate": 9.898583443557453e-05, + "loss": 2.4209, + "step": 24045 + }, + { + "epoch": 0.06, + "learning_rate": 9.898541229014732e-05, + "loss": 2.3993, + "step": 24050 + }, + { + "epoch": 0.06, + "learning_rate": 9.898499005778013e-05, + "loss": 2.4706, + "step": 24055 + }, + { + "epoch": 0.06, + "learning_rate": 9.898456773847368e-05, + "loss": 2.3269, + "step": 24060 + }, + { + "epoch": 0.06, + "learning_rate": 9.898414533222875e-05, + "loss": 2.5113, + "step": 24065 + }, + { + "epoch": 0.06, + "learning_rate": 9.898372283904606e-05, + "loss": 2.3313, + "step": 24070 + }, + { + "epoch": 0.06, + "learning_rate": 9.89833002589264e-05, + "loss": 2.427, + "step": 24075 + }, + { + "epoch": 0.06, + "learning_rate": 9.89828775918705e-05, + "loss": 2.4897, + "step": 24080 + }, + { + "epoch": 0.06, + "learning_rate": 9.89824548378791e-05, + "loss": 2.3881, + "step": 24085 + }, + { + "epoch": 0.06, + "learning_rate": 9.898203199695295e-05, + "loss": 2.3201, + "step": 24090 + }, + { + "epoch": 0.06, + "learning_rate": 9.898160906909282e-05, + "loss": 2.4729, + "step": 24095 + }, + { + "epoch": 0.06, + "learning_rate": 9.898118605429943e-05, + "loss": 2.3449, + "step": 24100 + }, + { + "epoch": 0.06, + "learning_rate": 9.898076295257357e-05, + "loss": 2.3835, + "step": 24105 + }, + { + "epoch": 0.06, + "learning_rate": 9.898033976391597e-05, + "loss": 2.4685, + "step": 24110 + }, + { + "epoch": 0.06, + "learning_rate": 9.897991648832736e-05, + "loss": 2.401, + "step": 24115 + }, + { + "epoch": 0.06, + "learning_rate": 9.897949312580854e-05, + "loss": 2.3162, + "step": 24120 + }, + { + "epoch": 0.06, + "learning_rate": 9.897906967636021e-05, + "loss": 2.4645, + "step": 24125 + }, + { + "epoch": 0.06, + "learning_rate": 9.897864613998318e-05, + "loss": 2.3729, + "step": 24130 + }, + { + "epoch": 0.06, + "learning_rate": 9.897822251667814e-05, + "loss": 2.5051, + "step": 24135 + }, + { + "epoch": 0.06, + "learning_rate": 9.897779880644587e-05, + "loss": 2.493, + "step": 24140 + }, + { + "epoch": 0.06, + "learning_rate": 9.897737500928713e-05, + "loss": 2.351, + "step": 24145 + }, + { + "epoch": 0.06, + "learning_rate": 9.897695112520267e-05, + "loss": 2.4172, + "step": 24150 + }, + { + "epoch": 0.06, + "learning_rate": 9.897652715419323e-05, + "loss": 2.3206, + "step": 24155 + }, + { + "epoch": 0.06, + "learning_rate": 9.897610309625957e-05, + "loss": 2.3006, + "step": 24160 + }, + { + "epoch": 0.06, + "learning_rate": 9.897567895140243e-05, + "loss": 2.351, + "step": 24165 + }, + { + "epoch": 0.06, + "learning_rate": 9.897525471962259e-05, + "loss": 2.3846, + "step": 24170 + }, + { + "epoch": 0.06, + "learning_rate": 9.897483040092077e-05, + "loss": 2.3565, + "step": 24175 + }, + { + "epoch": 0.06, + "learning_rate": 9.897440599529774e-05, + "loss": 2.4035, + "step": 24180 + }, + { + "epoch": 0.06, + "learning_rate": 9.897398150275428e-05, + "loss": 2.3003, + "step": 24185 + }, + { + "epoch": 0.06, + "learning_rate": 9.897355692329108e-05, + "loss": 2.3321, + "step": 24190 + }, + { + "epoch": 0.06, + "learning_rate": 9.897313225690896e-05, + "loss": 2.4374, + "step": 24195 + }, + { + "epoch": 0.06, + "learning_rate": 9.897270750360862e-05, + "loss": 2.4122, + "step": 24200 + }, + { + "epoch": 0.06, + "learning_rate": 9.897228266339086e-05, + "loss": 2.4183, + "step": 24205 + }, + { + "epoch": 0.06, + "learning_rate": 9.897185773625641e-05, + "loss": 2.5064, + "step": 24210 + }, + { + "epoch": 0.06, + "learning_rate": 9.897143272220602e-05, + "loss": 2.3518, + "step": 24215 + }, + { + "epoch": 0.06, + "learning_rate": 9.897100762124046e-05, + "loss": 2.4301, + "step": 24220 + }, + { + "epoch": 0.06, + "learning_rate": 9.897058243336046e-05, + "loss": 2.4754, + "step": 24225 + }, + { + "epoch": 0.06, + "learning_rate": 9.89701571585668e-05, + "loss": 2.403, + "step": 24230 + }, + { + "epoch": 0.06, + "learning_rate": 9.89697317968602e-05, + "loss": 2.3545, + "step": 24235 + }, + { + "epoch": 0.06, + "learning_rate": 9.896930634824146e-05, + "loss": 2.4262, + "step": 24240 + }, + { + "epoch": 0.07, + "learning_rate": 9.896888081271132e-05, + "loss": 2.3687, + "step": 24245 + }, + { + "epoch": 0.07, + "learning_rate": 9.896845519027052e-05, + "loss": 2.5116, + "step": 24250 + }, + { + "epoch": 0.07, + "learning_rate": 9.896802948091981e-05, + "loss": 2.4663, + "step": 24255 + }, + { + "epoch": 0.07, + "learning_rate": 9.896760368465998e-05, + "loss": 2.3304, + "step": 24260 + }, + { + "epoch": 0.07, + "learning_rate": 9.896717780149177e-05, + "loss": 2.3742, + "step": 24265 + }, + { + "epoch": 0.07, + "learning_rate": 9.896675183141592e-05, + "loss": 2.4404, + "step": 24270 + }, + { + "epoch": 0.07, + "learning_rate": 9.89663257744332e-05, + "loss": 2.363, + "step": 24275 + }, + { + "epoch": 0.07, + "learning_rate": 9.896589963054436e-05, + "loss": 2.3249, + "step": 24280 + }, + { + "epoch": 0.07, + "learning_rate": 9.896547339975017e-05, + "loss": 2.4276, + "step": 24285 + }, + { + "epoch": 0.07, + "learning_rate": 9.896504708205136e-05, + "loss": 2.3261, + "step": 24290 + }, + { + "epoch": 0.07, + "learning_rate": 9.896462067744872e-05, + "loss": 2.3321, + "step": 24295 + }, + { + "epoch": 0.07, + "learning_rate": 9.896419418594301e-05, + "loss": 2.3576, + "step": 24300 + }, + { + "epoch": 0.07, + "learning_rate": 9.896376760753493e-05, + "loss": 2.5046, + "step": 24305 + }, + { + "epoch": 0.07, + "learning_rate": 9.896334094222529e-05, + "loss": 2.3477, + "step": 24310 + }, + { + "epoch": 0.07, + "learning_rate": 9.896291419001482e-05, + "loss": 2.4543, + "step": 24315 + }, + { + "epoch": 0.07, + "learning_rate": 9.896248735090431e-05, + "loss": 2.2511, + "step": 24320 + }, + { + "epoch": 0.07, + "learning_rate": 9.896206042489449e-05, + "loss": 2.4888, + "step": 24325 + }, + { + "epoch": 0.07, + "learning_rate": 9.896163341198613e-05, + "loss": 2.4043, + "step": 24330 + }, + { + "epoch": 0.07, + "learning_rate": 9.896120631217995e-05, + "loss": 2.422, + "step": 24335 + }, + { + "epoch": 0.07, + "learning_rate": 9.896077912547677e-05, + "loss": 2.2543, + "step": 24340 + }, + { + "epoch": 0.07, + "learning_rate": 9.89603518518773e-05, + "loss": 2.4614, + "step": 24345 + }, + { + "epoch": 0.07, + "learning_rate": 9.895992449138233e-05, + "loss": 2.3844, + "step": 24350 + }, + { + "epoch": 0.07, + "learning_rate": 9.89594970439926e-05, + "loss": 2.2617, + "step": 24355 + }, + { + "epoch": 0.07, + "learning_rate": 9.895906950970889e-05, + "loss": 2.3863, + "step": 24360 + }, + { + "epoch": 0.07, + "learning_rate": 9.895864188853192e-05, + "loss": 2.1923, + "step": 24365 + }, + { + "epoch": 0.07, + "learning_rate": 9.895821418046249e-05, + "loss": 2.382, + "step": 24370 + }, + { + "epoch": 0.07, + "learning_rate": 9.895778638550132e-05, + "loss": 2.4636, + "step": 24375 + }, + { + "epoch": 0.07, + "learning_rate": 9.89573585036492e-05, + "loss": 2.5402, + "step": 24380 + }, + { + "epoch": 0.07, + "learning_rate": 9.895693053490689e-05, + "loss": 2.2722, + "step": 24385 + }, + { + "epoch": 0.07, + "learning_rate": 9.895650247927514e-05, + "loss": 2.405, + "step": 24390 + }, + { + "epoch": 0.07, + "learning_rate": 9.89560743367547e-05, + "loss": 2.3682, + "step": 24395 + }, + { + "epoch": 0.07, + "learning_rate": 9.895564610734634e-05, + "loss": 2.2819, + "step": 24400 + }, + { + "epoch": 0.07, + "learning_rate": 9.895521779105081e-05, + "loss": 2.4385, + "step": 24405 + }, + { + "epoch": 0.07, + "learning_rate": 9.895478938786889e-05, + "loss": 2.3998, + "step": 24410 + }, + { + "epoch": 0.07, + "learning_rate": 9.895436089780131e-05, + "loss": 2.3269, + "step": 24415 + }, + { + "epoch": 0.07, + "learning_rate": 9.895393232084886e-05, + "loss": 2.4321, + "step": 24420 + }, + { + "epoch": 0.07, + "learning_rate": 9.895350365701231e-05, + "loss": 2.3717, + "step": 24425 + }, + { + "epoch": 0.07, + "learning_rate": 9.895307490629238e-05, + "loss": 2.3811, + "step": 24430 + }, + { + "epoch": 0.07, + "learning_rate": 9.895264606868986e-05, + "loss": 2.5101, + "step": 24435 + }, + { + "epoch": 0.07, + "learning_rate": 9.89522171442055e-05, + "loss": 2.483, + "step": 24440 + }, + { + "epoch": 0.07, + "learning_rate": 9.895178813284005e-05, + "loss": 2.3991, + "step": 24445 + }, + { + "epoch": 0.07, + "learning_rate": 9.89513590345943e-05, + "loss": 2.3846, + "step": 24450 + }, + { + "epoch": 0.07, + "learning_rate": 9.895092984946901e-05, + "loss": 2.3239, + "step": 24455 + }, + { + "epoch": 0.07, + "learning_rate": 9.895050057746492e-05, + "loss": 2.3541, + "step": 24460 + }, + { + "epoch": 0.07, + "learning_rate": 9.895007121858278e-05, + "loss": 2.3731, + "step": 24465 + }, + { + "epoch": 0.07, + "learning_rate": 9.894964177282339e-05, + "loss": 2.3643, + "step": 24470 + }, + { + "epoch": 0.07, + "learning_rate": 9.894921224018749e-05, + "loss": 2.4434, + "step": 24475 + }, + { + "epoch": 0.07, + "learning_rate": 9.894878262067587e-05, + "loss": 2.5061, + "step": 24480 + }, + { + "epoch": 0.07, + "learning_rate": 9.894835291428924e-05, + "loss": 2.3633, + "step": 24485 + }, + { + "epoch": 0.07, + "learning_rate": 9.89479231210284e-05, + "loss": 2.3461, + "step": 24490 + }, + { + "epoch": 0.07, + "learning_rate": 9.894749324089412e-05, + "loss": 2.5233, + "step": 24495 + }, + { + "epoch": 0.07, + "learning_rate": 9.894706327388715e-05, + "loss": 2.2961, + "step": 24500 + }, + { + "epoch": 0.07, + "learning_rate": 9.894663322000823e-05, + "loss": 2.3945, + "step": 24505 + }, + { + "epoch": 0.07, + "learning_rate": 9.894620307925815e-05, + "loss": 2.3924, + "step": 24510 + }, + { + "epoch": 0.07, + "learning_rate": 9.894577285163769e-05, + "loss": 2.3617, + "step": 24515 + }, + { + "epoch": 0.07, + "learning_rate": 9.894534253714755e-05, + "loss": 2.3901, + "step": 24520 + }, + { + "epoch": 0.07, + "learning_rate": 9.894491213578858e-05, + "loss": 2.4889, + "step": 24525 + }, + { + "epoch": 0.07, + "learning_rate": 9.894448164756147e-05, + "loss": 2.4594, + "step": 24530 + }, + { + "epoch": 0.07, + "learning_rate": 9.894405107246702e-05, + "loss": 2.3106, + "step": 24535 + }, + { + "epoch": 0.07, + "learning_rate": 9.8943620410506e-05, + "loss": 2.43, + "step": 24540 + }, + { + "epoch": 0.07, + "learning_rate": 9.894318966167915e-05, + "loss": 2.3456, + "step": 24545 + }, + { + "epoch": 0.07, + "learning_rate": 9.894275882598725e-05, + "loss": 2.4481, + "step": 24550 + }, + { + "epoch": 0.07, + "learning_rate": 9.894232790343105e-05, + "loss": 2.3167, + "step": 24555 + }, + { + "epoch": 0.07, + "learning_rate": 9.894189689401134e-05, + "loss": 2.4283, + "step": 24560 + }, + { + "epoch": 0.07, + "learning_rate": 9.894146579772887e-05, + "loss": 2.4485, + "step": 24565 + }, + { + "epoch": 0.07, + "learning_rate": 9.89410346145844e-05, + "loss": 2.5701, + "step": 24570 + }, + { + "epoch": 0.07, + "learning_rate": 9.894060334457871e-05, + "loss": 2.4093, + "step": 24575 + }, + { + "epoch": 0.07, + "learning_rate": 9.894017198771255e-05, + "loss": 2.4742, + "step": 24580 + }, + { + "epoch": 0.07, + "learning_rate": 9.89397405439867e-05, + "loss": 2.316, + "step": 24585 + }, + { + "epoch": 0.07, + "learning_rate": 9.89393090134019e-05, + "loss": 2.3528, + "step": 24590 + }, + { + "epoch": 0.07, + "learning_rate": 9.893887739595895e-05, + "loss": 2.4229, + "step": 24595 + }, + { + "epoch": 0.07, + "learning_rate": 9.89384456916586e-05, + "loss": 2.3379, + "step": 24600 + }, + { + "epoch": 0.07, + "learning_rate": 9.893801390050162e-05, + "loss": 2.2864, + "step": 24605 + }, + { + "epoch": 0.07, + "learning_rate": 9.893758202248875e-05, + "loss": 2.4099, + "step": 24610 + }, + { + "epoch": 0.07, + "learning_rate": 9.893715005762078e-05, + "loss": 2.2379, + "step": 24615 + }, + { + "epoch": 0.07, + "learning_rate": 9.89367180058985e-05, + "loss": 2.4236, + "step": 24620 + }, + { + "epoch": 0.07, + "learning_rate": 9.893628586732264e-05, + "loss": 2.3518, + "step": 24625 + }, + { + "epoch": 0.07, + "learning_rate": 9.893585364189399e-05, + "loss": 2.3215, + "step": 24630 + }, + { + "epoch": 0.07, + "learning_rate": 9.893542132961329e-05, + "loss": 2.2456, + "step": 24635 + }, + { + "epoch": 0.07, + "learning_rate": 9.893498893048133e-05, + "loss": 2.3336, + "step": 24640 + }, + { + "epoch": 0.07, + "learning_rate": 9.893455644449887e-05, + "loss": 2.3467, + "step": 24645 + }, + { + "epoch": 0.07, + "learning_rate": 9.893412387166668e-05, + "loss": 2.4452, + "step": 24650 + }, + { + "epoch": 0.07, + "learning_rate": 9.893369121198553e-05, + "loss": 2.3149, + "step": 24655 + }, + { + "epoch": 0.07, + "learning_rate": 9.893325846545617e-05, + "loss": 2.3472, + "step": 24660 + }, + { + "epoch": 0.07, + "learning_rate": 9.89328256320794e-05, + "loss": 2.2269, + "step": 24665 + }, + { + "epoch": 0.07, + "learning_rate": 9.893239271185598e-05, + "loss": 2.4867, + "step": 24670 + }, + { + "epoch": 0.07, + "learning_rate": 9.893195970478663e-05, + "loss": 2.2643, + "step": 24675 + }, + { + "epoch": 0.07, + "learning_rate": 9.89315266108722e-05, + "loss": 2.326, + "step": 24680 + }, + { + "epoch": 0.07, + "learning_rate": 9.893109343011338e-05, + "loss": 2.4161, + "step": 24685 + }, + { + "epoch": 0.07, + "learning_rate": 9.8930660162511e-05, + "loss": 2.3257, + "step": 24690 + }, + { + "epoch": 0.07, + "learning_rate": 9.893022680806578e-05, + "loss": 2.3852, + "step": 24695 + }, + { + "epoch": 0.07, + "learning_rate": 9.892979336677855e-05, + "loss": 2.3329, + "step": 24700 + }, + { + "epoch": 0.07, + "learning_rate": 9.892935983865002e-05, + "loss": 2.4075, + "step": 24705 + }, + { + "epoch": 0.07, + "learning_rate": 9.892892622368098e-05, + "loss": 2.326, + "step": 24710 + }, + { + "epoch": 0.07, + "learning_rate": 9.892849252187219e-05, + "loss": 2.3412, + "step": 24715 + }, + { + "epoch": 0.07, + "learning_rate": 9.892805873322446e-05, + "loss": 2.3324, + "step": 24720 + }, + { + "epoch": 0.07, + "learning_rate": 9.89276248577385e-05, + "loss": 2.4745, + "step": 24725 + }, + { + "epoch": 0.07, + "learning_rate": 9.892719089541514e-05, + "loss": 2.4332, + "step": 24730 + }, + { + "epoch": 0.07, + "learning_rate": 9.892675684625511e-05, + "loss": 2.4357, + "step": 24735 + }, + { + "epoch": 0.07, + "learning_rate": 9.892632271025919e-05, + "loss": 2.5156, + "step": 24740 + }, + { + "epoch": 0.07, + "learning_rate": 9.892588848742815e-05, + "loss": 2.3473, + "step": 24745 + }, + { + "epoch": 0.07, + "learning_rate": 9.892545417776276e-05, + "loss": 2.4185, + "step": 24750 + }, + { + "epoch": 0.07, + "learning_rate": 9.89250197812638e-05, + "loss": 2.3676, + "step": 24755 + }, + { + "epoch": 0.07, + "learning_rate": 9.892458529793205e-05, + "loss": 2.4641, + "step": 24760 + }, + { + "epoch": 0.07, + "learning_rate": 9.892415072776824e-05, + "loss": 2.4236, + "step": 24765 + }, + { + "epoch": 0.07, + "learning_rate": 9.892371607077318e-05, + "loss": 2.2757, + "step": 24770 + }, + { + "epoch": 0.07, + "learning_rate": 9.892328132694762e-05, + "loss": 2.3732, + "step": 24775 + }, + { + "epoch": 0.07, + "learning_rate": 9.892284649629236e-05, + "loss": 2.4598, + "step": 24780 + }, + { + "epoch": 0.07, + "learning_rate": 9.892241157880813e-05, + "loss": 2.3706, + "step": 24785 + }, + { + "epoch": 0.07, + "learning_rate": 9.892197657449572e-05, + "loss": 2.3589, + "step": 24790 + }, + { + "epoch": 0.07, + "learning_rate": 9.892154148335591e-05, + "loss": 2.4238, + "step": 24795 + }, + { + "epoch": 0.07, + "learning_rate": 9.892110630538948e-05, + "loss": 2.3, + "step": 24800 + }, + { + "epoch": 0.07, + "learning_rate": 9.892067104059719e-05, + "loss": 2.4446, + "step": 24805 + }, + { + "epoch": 0.07, + "learning_rate": 9.892023568897982e-05, + "loss": 2.4002, + "step": 24810 + }, + { + "epoch": 0.07, + "learning_rate": 9.891980025053811e-05, + "loss": 2.4589, + "step": 24815 + }, + { + "epoch": 0.07, + "learning_rate": 9.891936472527287e-05, + "loss": 2.377, + "step": 24820 + }, + { + "epoch": 0.07, + "learning_rate": 9.891892911318486e-05, + "loss": 2.3787, + "step": 24825 + }, + { + "epoch": 0.07, + "learning_rate": 9.891849341427486e-05, + "loss": 2.3817, + "step": 24830 + }, + { + "epoch": 0.07, + "learning_rate": 9.891805762854362e-05, + "loss": 2.4214, + "step": 24835 + }, + { + "epoch": 0.07, + "learning_rate": 9.891762175599195e-05, + "loss": 2.3078, + "step": 24840 + }, + { + "epoch": 0.07, + "learning_rate": 9.89171857966206e-05, + "loss": 2.3057, + "step": 24845 + }, + { + "epoch": 0.07, + "learning_rate": 9.891674975043035e-05, + "loss": 2.3585, + "step": 24850 + }, + { + "epoch": 0.07, + "learning_rate": 9.891631361742198e-05, + "loss": 2.4252, + "step": 24855 + }, + { + "epoch": 0.07, + "learning_rate": 9.891587739759625e-05, + "loss": 2.4712, + "step": 24860 + }, + { + "epoch": 0.07, + "learning_rate": 9.891544109095392e-05, + "loss": 2.2243, + "step": 24865 + }, + { + "epoch": 0.07, + "learning_rate": 9.891500469749582e-05, + "loss": 2.4045, + "step": 24870 + }, + { + "epoch": 0.07, + "learning_rate": 9.891456821722267e-05, + "loss": 2.3631, + "step": 24875 + }, + { + "epoch": 0.07, + "learning_rate": 9.891413165013526e-05, + "loss": 2.3619, + "step": 24880 + }, + { + "epoch": 0.07, + "learning_rate": 9.891369499623437e-05, + "loss": 2.3857, + "step": 24885 + }, + { + "epoch": 0.07, + "learning_rate": 9.891325825552078e-05, + "loss": 2.2585, + "step": 24890 + }, + { + "epoch": 0.07, + "learning_rate": 9.891282142799526e-05, + "loss": 2.542, + "step": 24895 + }, + { + "epoch": 0.07, + "learning_rate": 9.891238451365859e-05, + "loss": 2.4059, + "step": 24900 + }, + { + "epoch": 0.07, + "learning_rate": 9.891194751251153e-05, + "loss": 2.3682, + "step": 24905 + }, + { + "epoch": 0.07, + "learning_rate": 9.891151042455487e-05, + "loss": 2.2885, + "step": 24910 + }, + { + "epoch": 0.07, + "learning_rate": 9.891107324978937e-05, + "loss": 2.488, + "step": 24915 + }, + { + "epoch": 0.07, + "learning_rate": 9.891063598821584e-05, + "loss": 2.4153, + "step": 24920 + }, + { + "epoch": 0.07, + "learning_rate": 9.891019863983502e-05, + "loss": 2.3742, + "step": 24925 + }, + { + "epoch": 0.07, + "learning_rate": 9.890976120464769e-05, + "loss": 2.3883, + "step": 24930 + }, + { + "epoch": 0.07, + "learning_rate": 9.890932368265464e-05, + "loss": 2.255, + "step": 24935 + }, + { + "epoch": 0.07, + "learning_rate": 9.890888607385664e-05, + "loss": 2.376, + "step": 24940 + }, + { + "epoch": 0.07, + "learning_rate": 9.890844837825448e-05, + "loss": 2.4516, + "step": 24945 + }, + { + "epoch": 0.07, + "learning_rate": 9.890801059584893e-05, + "loss": 2.3627, + "step": 24950 + }, + { + "epoch": 0.07, + "learning_rate": 9.890757272664075e-05, + "loss": 2.2921, + "step": 24955 + }, + { + "epoch": 0.07, + "learning_rate": 9.890713477063074e-05, + "loss": 2.3425, + "step": 24960 + }, + { + "epoch": 0.07, + "learning_rate": 9.890669672781966e-05, + "loss": 2.2403, + "step": 24965 + }, + { + "epoch": 0.07, + "learning_rate": 9.89062585982083e-05, + "loss": 2.3034, + "step": 24970 + }, + { + "epoch": 0.07, + "learning_rate": 9.890582038179742e-05, + "loss": 2.2648, + "step": 24975 + }, + { + "epoch": 0.07, + "learning_rate": 9.890538207858782e-05, + "loss": 2.4659, + "step": 24980 + }, + { + "epoch": 0.07, + "learning_rate": 9.890494368858027e-05, + "loss": 2.3769, + "step": 24985 + }, + { + "epoch": 0.07, + "learning_rate": 9.890450521177555e-05, + "loss": 2.3916, + "step": 24990 + }, + { + "epoch": 0.07, + "learning_rate": 9.890406664817442e-05, + "loss": 2.367, + "step": 24995 + }, + { + "epoch": 0.07, + "learning_rate": 9.890362799777767e-05, + "loss": 2.2988, + "step": 25000 + }, + { + "epoch": 0.07, + "learning_rate": 9.89031892605861e-05, + "loss": 2.2804, + "step": 25005 + }, + { + "epoch": 0.07, + "learning_rate": 9.890275043660044e-05, + "loss": 2.2532, + "step": 25010 + }, + { + "epoch": 0.07, + "learning_rate": 9.890231152582152e-05, + "loss": 2.4236, + "step": 25015 + }, + { + "epoch": 0.07, + "learning_rate": 9.890187252825011e-05, + "loss": 2.4189, + "step": 25020 + }, + { + "epoch": 0.07, + "learning_rate": 9.890143344388696e-05, + "loss": 2.4574, + "step": 25025 + }, + { + "epoch": 0.07, + "learning_rate": 9.890099427273286e-05, + "loss": 2.3832, + "step": 25030 + }, + { + "epoch": 0.07, + "learning_rate": 9.890055501478861e-05, + "loss": 2.329, + "step": 25035 + }, + { + "epoch": 0.07, + "learning_rate": 9.890011567005497e-05, + "loss": 2.4304, + "step": 25040 + }, + { + "epoch": 0.07, + "learning_rate": 9.889967623853272e-05, + "loss": 2.3318, + "step": 25045 + }, + { + "epoch": 0.07, + "learning_rate": 9.889923672022264e-05, + "loss": 2.3673, + "step": 25050 + }, + { + "epoch": 0.07, + "learning_rate": 9.889879711512553e-05, + "loss": 2.3431, + "step": 25055 + }, + { + "epoch": 0.07, + "learning_rate": 9.889835742324214e-05, + "loss": 2.2367, + "step": 25060 + }, + { + "epoch": 0.07, + "learning_rate": 9.889791764457328e-05, + "loss": 2.3344, + "step": 25065 + }, + { + "epoch": 0.07, + "learning_rate": 9.88974777791197e-05, + "loss": 2.3857, + "step": 25070 + }, + { + "epoch": 0.07, + "learning_rate": 9.889703782688221e-05, + "loss": 2.5085, + "step": 25075 + }, + { + "epoch": 0.07, + "learning_rate": 9.889659778786158e-05, + "loss": 2.2335, + "step": 25080 + }, + { + "epoch": 0.07, + "learning_rate": 9.889615766205858e-05, + "loss": 2.4317, + "step": 25085 + }, + { + "epoch": 0.07, + "learning_rate": 9.889571744947398e-05, + "loss": 2.3117, + "step": 25090 + }, + { + "epoch": 0.07, + "learning_rate": 9.889527715010862e-05, + "loss": 2.4529, + "step": 25095 + }, + { + "epoch": 0.07, + "learning_rate": 9.889483676396322e-05, + "loss": 2.5545, + "step": 25100 + }, + { + "epoch": 0.07, + "learning_rate": 9.889439629103858e-05, + "loss": 2.3446, + "step": 25105 + }, + { + "epoch": 0.07, + "learning_rate": 9.889395573133549e-05, + "loss": 2.3729, + "step": 25110 + }, + { + "epoch": 0.07, + "learning_rate": 9.889351508485474e-05, + "loss": 2.3828, + "step": 25115 + }, + { + "epoch": 0.07, + "learning_rate": 9.88930743515971e-05, + "loss": 2.3631, + "step": 25120 + }, + { + "epoch": 0.07, + "learning_rate": 9.889263353156332e-05, + "loss": 2.507, + "step": 25125 + }, + { + "epoch": 0.07, + "learning_rate": 9.889219262475424e-05, + "loss": 2.4274, + "step": 25130 + }, + { + "epoch": 0.07, + "learning_rate": 9.889175163117062e-05, + "loss": 2.3247, + "step": 25135 + }, + { + "epoch": 0.07, + "learning_rate": 9.889131055081323e-05, + "loss": 2.6174, + "step": 25140 + }, + { + "epoch": 0.07, + "learning_rate": 9.889086938368287e-05, + "loss": 2.4147, + "step": 25145 + }, + { + "epoch": 0.07, + "learning_rate": 9.88904281297803e-05, + "loss": 2.4214, + "step": 25150 + }, + { + "epoch": 0.07, + "learning_rate": 9.888998678910633e-05, + "loss": 2.3586, + "step": 25155 + }, + { + "epoch": 0.07, + "learning_rate": 9.888954536166172e-05, + "loss": 2.331, + "step": 25160 + }, + { + "epoch": 0.07, + "learning_rate": 9.888910384744728e-05, + "loss": 2.2806, + "step": 25165 + }, + { + "epoch": 0.07, + "learning_rate": 9.888866224646377e-05, + "loss": 2.4729, + "step": 25170 + }, + { + "epoch": 0.07, + "learning_rate": 9.888822055871198e-05, + "loss": 2.3558, + "step": 25175 + }, + { + "epoch": 0.07, + "learning_rate": 9.888777878419271e-05, + "loss": 2.3748, + "step": 25180 + }, + { + "epoch": 0.07, + "learning_rate": 9.888733692290672e-05, + "loss": 2.4974, + "step": 25185 + }, + { + "epoch": 0.07, + "learning_rate": 9.888689497485482e-05, + "loss": 2.4235, + "step": 25190 + }, + { + "epoch": 0.07, + "learning_rate": 9.888645294003776e-05, + "loss": 2.2875, + "step": 25195 + }, + { + "epoch": 0.07, + "learning_rate": 9.888601081845635e-05, + "loss": 2.473, + "step": 25200 + }, + { + "epoch": 0.07, + "learning_rate": 9.888556861011136e-05, + "loss": 2.4283, + "step": 25205 + }, + { + "epoch": 0.07, + "learning_rate": 9.88851263150036e-05, + "loss": 2.3703, + "step": 25210 + }, + { + "epoch": 0.07, + "learning_rate": 9.888468393313381e-05, + "loss": 2.3397, + "step": 25215 + }, + { + "epoch": 0.07, + "learning_rate": 9.888424146450284e-05, + "loss": 2.3349, + "step": 25220 + }, + { + "epoch": 0.07, + "learning_rate": 9.888379890911141e-05, + "loss": 2.4794, + "step": 25225 + }, + { + "epoch": 0.07, + "learning_rate": 9.888335626696035e-05, + "loss": 2.5126, + "step": 25230 + }, + { + "epoch": 0.07, + "learning_rate": 9.888291353805042e-05, + "loss": 2.3263, + "step": 25235 + }, + { + "epoch": 0.07, + "learning_rate": 9.888247072238241e-05, + "loss": 2.5013, + "step": 25240 + }, + { + "epoch": 0.07, + "learning_rate": 9.888202781995712e-05, + "loss": 2.3788, + "step": 25245 + }, + { + "epoch": 0.07, + "learning_rate": 9.888158483077534e-05, + "loss": 2.3373, + "step": 25250 + }, + { + "epoch": 0.07, + "learning_rate": 9.888114175483783e-05, + "loss": 2.5036, + "step": 25255 + }, + { + "epoch": 0.07, + "learning_rate": 9.88806985921454e-05, + "loss": 2.3512, + "step": 25260 + }, + { + "epoch": 0.07, + "learning_rate": 9.88802553426988e-05, + "loss": 2.3833, + "step": 25265 + }, + { + "epoch": 0.07, + "learning_rate": 9.887981200649888e-05, + "loss": 2.3601, + "step": 25270 + }, + { + "epoch": 0.07, + "learning_rate": 9.887936858354636e-05, + "loss": 2.3636, + "step": 25275 + }, + { + "epoch": 0.07, + "learning_rate": 9.887892507384207e-05, + "loss": 2.4753, + "step": 25280 + }, + { + "epoch": 0.07, + "learning_rate": 9.887848147738678e-05, + "loss": 2.3457, + "step": 25285 + }, + { + "epoch": 0.07, + "learning_rate": 9.887803779418128e-05, + "loss": 2.2386, + "step": 25290 + }, + { + "epoch": 0.07, + "learning_rate": 9.887759402422636e-05, + "loss": 2.4509, + "step": 25295 + }, + { + "epoch": 0.07, + "learning_rate": 9.887715016752281e-05, + "loss": 2.4216, + "step": 25300 + }, + { + "epoch": 0.07, + "learning_rate": 9.887670622407142e-05, + "loss": 2.3848, + "step": 25305 + }, + { + "epoch": 0.07, + "learning_rate": 9.887626219387296e-05, + "loss": 2.4292, + "step": 25310 + }, + { + "epoch": 0.07, + "learning_rate": 9.887581807692823e-05, + "loss": 2.5102, + "step": 25315 + }, + { + "epoch": 0.07, + "learning_rate": 9.887537387323804e-05, + "loss": 2.2805, + "step": 25320 + }, + { + "epoch": 0.07, + "learning_rate": 9.887492958280314e-05, + "loss": 2.4221, + "step": 25325 + }, + { + "epoch": 0.07, + "learning_rate": 9.887448520562432e-05, + "loss": 2.4023, + "step": 25330 + }, + { + "epoch": 0.07, + "learning_rate": 9.887404074170241e-05, + "loss": 2.4413, + "step": 25335 + }, + { + "epoch": 0.07, + "learning_rate": 9.887359619103815e-05, + "loss": 2.3929, + "step": 25340 + }, + { + "epoch": 0.07, + "learning_rate": 9.887315155363237e-05, + "loss": 2.38, + "step": 25345 + }, + { + "epoch": 0.07, + "learning_rate": 9.887270682948584e-05, + "loss": 2.3837, + "step": 25350 + }, + { + "epoch": 0.07, + "learning_rate": 9.887226201859933e-05, + "loss": 2.4138, + "step": 25355 + }, + { + "epoch": 0.07, + "learning_rate": 9.887181712097369e-05, + "loss": 2.4234, + "step": 25360 + }, + { + "epoch": 0.07, + "learning_rate": 9.887137213660964e-05, + "loss": 2.4504, + "step": 25365 + }, + { + "epoch": 0.07, + "learning_rate": 9.8870927065508e-05, + "loss": 2.3705, + "step": 25370 + }, + { + "epoch": 0.07, + "learning_rate": 9.887048190766957e-05, + "loss": 2.4193, + "step": 25375 + }, + { + "epoch": 0.07, + "learning_rate": 9.88700366630951e-05, + "loss": 2.4613, + "step": 25380 + }, + { + "epoch": 0.07, + "learning_rate": 9.886959133178545e-05, + "loss": 2.2534, + "step": 25385 + }, + { + "epoch": 0.07, + "learning_rate": 9.886914591374134e-05, + "loss": 2.388, + "step": 25390 + }, + { + "epoch": 0.07, + "learning_rate": 9.88687004089636e-05, + "loss": 2.3946, + "step": 25395 + }, + { + "epoch": 0.07, + "learning_rate": 9.886825481745303e-05, + "loss": 2.3358, + "step": 25400 + }, + { + "epoch": 0.07, + "learning_rate": 9.886780913921039e-05, + "loss": 2.3997, + "step": 25405 + }, + { + "epoch": 0.07, + "learning_rate": 9.886736337423648e-05, + "loss": 2.4207, + "step": 25410 + }, + { + "epoch": 0.07, + "learning_rate": 9.886691752253207e-05, + "loss": 2.4855, + "step": 25415 + }, + { + "epoch": 0.07, + "learning_rate": 9.886647158409801e-05, + "loss": 2.3843, + "step": 25420 + }, + { + "epoch": 0.07, + "learning_rate": 9.886602555893505e-05, + "loss": 2.3877, + "step": 25425 + }, + { + "epoch": 0.07, + "learning_rate": 9.886557944704397e-05, + "loss": 2.4333, + "step": 25430 + }, + { + "epoch": 0.07, + "learning_rate": 9.88651332484256e-05, + "loss": 2.3824, + "step": 25435 + }, + { + "epoch": 0.07, + "learning_rate": 9.88646869630807e-05, + "loss": 2.4525, + "step": 25440 + }, + { + "epoch": 0.07, + "learning_rate": 9.886424059101008e-05, + "loss": 2.4245, + "step": 25445 + }, + { + "epoch": 0.07, + "learning_rate": 9.886379413221453e-05, + "loss": 2.3863, + "step": 25450 + }, + { + "epoch": 0.07, + "learning_rate": 9.886334758669482e-05, + "loss": 2.4013, + "step": 25455 + }, + { + "epoch": 0.07, + "learning_rate": 9.88629009544518e-05, + "loss": 2.4775, + "step": 25460 + }, + { + "epoch": 0.07, + "learning_rate": 9.886245423548618e-05, + "loss": 2.3344, + "step": 25465 + }, + { + "epoch": 0.07, + "learning_rate": 9.886200742979882e-05, + "loss": 2.5213, + "step": 25470 + }, + { + "epoch": 0.07, + "learning_rate": 9.886156053739048e-05, + "loss": 2.4406, + "step": 25475 + }, + { + "epoch": 0.07, + "learning_rate": 9.886111355826197e-05, + "loss": 2.4757, + "step": 25480 + }, + { + "epoch": 0.07, + "learning_rate": 9.886066649241407e-05, + "loss": 2.465, + "step": 25485 + }, + { + "epoch": 0.07, + "learning_rate": 9.88602193398476e-05, + "loss": 2.4577, + "step": 25490 + }, + { + "epoch": 0.07, + "learning_rate": 9.88597721005633e-05, + "loss": 2.3761, + "step": 25495 + }, + { + "epoch": 0.07, + "learning_rate": 9.885932477456202e-05, + "loss": 2.4558, + "step": 25500 + }, + { + "epoch": 0.07, + "learning_rate": 9.885887736184452e-05, + "loss": 2.3302, + "step": 25505 + }, + { + "epoch": 0.07, + "learning_rate": 9.885842986241161e-05, + "loss": 2.4207, + "step": 25510 + }, + { + "epoch": 0.07, + "learning_rate": 9.885798227626408e-05, + "loss": 2.3263, + "step": 25515 + }, + { + "epoch": 0.07, + "learning_rate": 9.885753460340273e-05, + "loss": 2.5073, + "step": 25520 + }, + { + "epoch": 0.07, + "learning_rate": 9.885708684382834e-05, + "loss": 2.3088, + "step": 25525 + }, + { + "epoch": 0.07, + "learning_rate": 9.885663899754172e-05, + "loss": 2.3677, + "step": 25530 + }, + { + "epoch": 0.07, + "learning_rate": 9.885619106454365e-05, + "loss": 2.3935, + "step": 25535 + }, + { + "epoch": 0.07, + "learning_rate": 9.885574304483493e-05, + "loss": 2.3384, + "step": 25540 + }, + { + "epoch": 0.07, + "learning_rate": 9.885529493841637e-05, + "loss": 2.284, + "step": 25545 + }, + { + "epoch": 0.07, + "learning_rate": 9.885484674528875e-05, + "loss": 2.3304, + "step": 25550 + }, + { + "epoch": 0.07, + "learning_rate": 9.885439846545285e-05, + "loss": 2.362, + "step": 25555 + }, + { + "epoch": 0.07, + "learning_rate": 9.885395009890952e-05, + "loss": 2.3594, + "step": 25560 + }, + { + "epoch": 0.07, + "learning_rate": 9.88535016456595e-05, + "loss": 2.3708, + "step": 25565 + }, + { + "epoch": 0.07, + "learning_rate": 9.88530531057036e-05, + "loss": 2.353, + "step": 25570 + }, + { + "epoch": 0.07, + "learning_rate": 9.885260447904263e-05, + "loss": 2.3968, + "step": 25575 + }, + { + "epoch": 0.07, + "learning_rate": 9.885215576567737e-05, + "loss": 2.2608, + "step": 25580 + }, + { + "epoch": 0.07, + "learning_rate": 9.885170696560863e-05, + "loss": 2.421, + "step": 25585 + }, + { + "epoch": 0.07, + "learning_rate": 9.885125807883721e-05, + "loss": 2.5107, + "step": 25590 + }, + { + "epoch": 0.07, + "learning_rate": 9.885080910536389e-05, + "loss": 2.2747, + "step": 25595 + }, + { + "epoch": 0.07, + "learning_rate": 9.885036004518947e-05, + "loss": 2.4277, + "step": 25600 + }, + { + "epoch": 0.07, + "learning_rate": 9.884991089831475e-05, + "loss": 2.3875, + "step": 25605 + }, + { + "epoch": 0.07, + "learning_rate": 9.884946166474054e-05, + "loss": 2.3827, + "step": 25610 + }, + { + "epoch": 0.07, + "learning_rate": 9.884901234446761e-05, + "loss": 2.4164, + "step": 25615 + }, + { + "epoch": 0.07, + "learning_rate": 9.884856293749679e-05, + "loss": 2.3232, + "step": 25620 + }, + { + "epoch": 0.07, + "learning_rate": 9.884811344382886e-05, + "loss": 2.4657, + "step": 25625 + }, + { + "epoch": 0.07, + "learning_rate": 9.884766386346462e-05, + "loss": 2.3235, + "step": 25630 + }, + { + "epoch": 0.07, + "learning_rate": 9.884721419640486e-05, + "loss": 2.3349, + "step": 25635 + }, + { + "epoch": 0.07, + "learning_rate": 9.884676444265037e-05, + "loss": 2.351, + "step": 25640 + }, + { + "epoch": 0.07, + "learning_rate": 9.884631460220199e-05, + "loss": 2.455, + "step": 25645 + }, + { + "epoch": 0.07, + "learning_rate": 9.884586467506048e-05, + "loss": 2.3912, + "step": 25650 + }, + { + "epoch": 0.07, + "learning_rate": 9.884541466122665e-05, + "loss": 2.1652, + "step": 25655 + }, + { + "epoch": 0.07, + "learning_rate": 9.884496456070131e-05, + "loss": 2.4262, + "step": 25660 + }, + { + "epoch": 0.07, + "learning_rate": 9.884451437348523e-05, + "loss": 2.0949, + "step": 25665 + }, + { + "epoch": 0.07, + "learning_rate": 9.884406409957925e-05, + "loss": 2.4649, + "step": 25670 + }, + { + "epoch": 0.07, + "learning_rate": 9.884361373898412e-05, + "loss": 2.2715, + "step": 25675 + }, + { + "epoch": 0.07, + "learning_rate": 9.884316329170068e-05, + "loss": 2.382, + "step": 25680 + }, + { + "epoch": 0.07, + "learning_rate": 9.884271275772972e-05, + "loss": 2.3579, + "step": 25685 + }, + { + "epoch": 0.07, + "learning_rate": 9.884226213707202e-05, + "loss": 2.4063, + "step": 25690 + }, + { + "epoch": 0.07, + "learning_rate": 9.884181142972839e-05, + "loss": 2.4415, + "step": 25695 + }, + { + "epoch": 0.07, + "learning_rate": 9.884136063569965e-05, + "loss": 2.4302, + "step": 25700 + }, + { + "epoch": 0.07, + "learning_rate": 9.884090975498658e-05, + "loss": 2.3492, + "step": 25705 + }, + { + "epoch": 0.07, + "learning_rate": 9.884045878758999e-05, + "loss": 2.4055, + "step": 25710 + }, + { + "epoch": 0.07, + "learning_rate": 9.884000773351066e-05, + "loss": 2.3355, + "step": 25715 + }, + { + "epoch": 0.07, + "learning_rate": 9.883955659274941e-05, + "loss": 2.4494, + "step": 25720 + }, + { + "epoch": 0.07, + "learning_rate": 9.883910536530702e-05, + "loss": 2.3748, + "step": 25725 + }, + { + "epoch": 0.07, + "learning_rate": 9.883865405118434e-05, + "loss": 2.2918, + "step": 25730 + }, + { + "epoch": 0.07, + "learning_rate": 9.883820265038212e-05, + "loss": 2.4326, + "step": 25735 + }, + { + "epoch": 0.07, + "learning_rate": 9.883775116290117e-05, + "loss": 2.4353, + "step": 25740 + }, + { + "epoch": 0.07, + "learning_rate": 9.883729958874231e-05, + "loss": 2.3957, + "step": 25745 + }, + { + "epoch": 0.07, + "learning_rate": 9.883684792790634e-05, + "loss": 2.3863, + "step": 25750 + }, + { + "epoch": 0.07, + "learning_rate": 9.883639618039405e-05, + "loss": 2.4271, + "step": 25755 + }, + { + "epoch": 0.07, + "learning_rate": 9.883594434620624e-05, + "loss": 2.3698, + "step": 25760 + }, + { + "epoch": 0.07, + "learning_rate": 9.883549242534372e-05, + "loss": 2.374, + "step": 25765 + }, + { + "epoch": 0.07, + "learning_rate": 9.883504041780728e-05, + "loss": 2.4083, + "step": 25770 + }, + { + "epoch": 0.07, + "learning_rate": 9.883458832359773e-05, + "loss": 2.5265, + "step": 25775 + }, + { + "epoch": 0.07, + "learning_rate": 9.883413614271588e-05, + "loss": 2.2341, + "step": 25780 + }, + { + "epoch": 0.07, + "learning_rate": 9.883368387516253e-05, + "loss": 2.4257, + "step": 25785 + }, + { + "epoch": 0.07, + "learning_rate": 9.883323152093848e-05, + "loss": 2.3966, + "step": 25790 + }, + { + "epoch": 0.07, + "learning_rate": 9.883277908004453e-05, + "loss": 2.418, + "step": 25795 + }, + { + "epoch": 0.07, + "learning_rate": 9.883232655248147e-05, + "loss": 2.4402, + "step": 25800 + }, + { + "epoch": 0.07, + "learning_rate": 9.883187393825013e-05, + "loss": 2.3412, + "step": 25805 + }, + { + "epoch": 0.07, + "learning_rate": 9.88314212373513e-05, + "loss": 2.2852, + "step": 25810 + }, + { + "epoch": 0.07, + "learning_rate": 9.883096844978579e-05, + "loss": 2.4177, + "step": 25815 + }, + { + "epoch": 0.07, + "learning_rate": 9.88305155755544e-05, + "loss": 2.3702, + "step": 25820 + }, + { + "epoch": 0.07, + "learning_rate": 9.883006261465793e-05, + "loss": 2.3577, + "step": 25825 + }, + { + "epoch": 0.07, + "learning_rate": 9.882960956709717e-05, + "loss": 2.3342, + "step": 25830 + }, + { + "epoch": 0.07, + "learning_rate": 9.882915643287296e-05, + "loss": 2.3189, + "step": 25835 + }, + { + "epoch": 0.07, + "learning_rate": 9.882870321198607e-05, + "loss": 2.4177, + "step": 25840 + }, + { + "epoch": 0.07, + "learning_rate": 9.882824990443734e-05, + "loss": 2.3295, + "step": 25845 + }, + { + "epoch": 0.07, + "learning_rate": 9.882779651022753e-05, + "loss": 2.3818, + "step": 25850 + }, + { + "epoch": 0.07, + "learning_rate": 9.882734302935748e-05, + "loss": 2.3849, + "step": 25855 + }, + { + "epoch": 0.07, + "learning_rate": 9.882688946182797e-05, + "loss": 2.339, + "step": 25860 + }, + { + "epoch": 0.07, + "learning_rate": 9.882643580763982e-05, + "loss": 2.3753, + "step": 25865 + }, + { + "epoch": 0.07, + "learning_rate": 9.882598206679384e-05, + "loss": 2.4477, + "step": 25870 + }, + { + "epoch": 0.07, + "learning_rate": 9.882552823929083e-05, + "loss": 2.5519, + "step": 25875 + }, + { + "epoch": 0.07, + "learning_rate": 9.88250743251316e-05, + "loss": 2.3976, + "step": 25880 + }, + { + "epoch": 0.07, + "learning_rate": 9.882462032431693e-05, + "loss": 2.2363, + "step": 25885 + }, + { + "epoch": 0.07, + "learning_rate": 9.882416623684765e-05, + "loss": 2.2555, + "step": 25890 + }, + { + "epoch": 0.07, + "learning_rate": 9.882371206272457e-05, + "loss": 2.3153, + "step": 25895 + }, + { + "epoch": 0.07, + "learning_rate": 9.882325780194848e-05, + "loss": 2.3243, + "step": 25900 + }, + { + "epoch": 0.07, + "learning_rate": 9.88228034545202e-05, + "loss": 2.3979, + "step": 25905 + }, + { + "epoch": 0.07, + "learning_rate": 9.882234902044052e-05, + "loss": 2.4729, + "step": 25910 + }, + { + "epoch": 0.07, + "learning_rate": 9.882189449971027e-05, + "loss": 2.4135, + "step": 25915 + }, + { + "epoch": 0.07, + "learning_rate": 9.882143989233023e-05, + "loss": 2.2472, + "step": 25920 + }, + { + "epoch": 0.07, + "learning_rate": 9.882098519830121e-05, + "loss": 2.2189, + "step": 25925 + }, + { + "epoch": 0.07, + "learning_rate": 9.882053041762404e-05, + "loss": 2.3778, + "step": 25930 + }, + { + "epoch": 0.07, + "learning_rate": 9.882007555029953e-05, + "loss": 2.3873, + "step": 25935 + }, + { + "epoch": 0.07, + "learning_rate": 9.881962059632844e-05, + "loss": 2.3209, + "step": 25940 + }, + { + "epoch": 0.07, + "learning_rate": 9.881916555571163e-05, + "loss": 2.4538, + "step": 25945 + }, + { + "epoch": 0.07, + "learning_rate": 9.881871042844988e-05, + "loss": 2.4659, + "step": 25950 + }, + { + "epoch": 0.07, + "learning_rate": 9.8818255214544e-05, + "loss": 2.2967, + "step": 25955 + }, + { + "epoch": 0.07, + "learning_rate": 9.88177999139948e-05, + "loss": 2.3733, + "step": 25960 + }, + { + "epoch": 0.07, + "learning_rate": 9.88173445268031e-05, + "loss": 2.3128, + "step": 25965 + }, + { + "epoch": 0.07, + "learning_rate": 9.88168890529697e-05, + "loss": 2.3012, + "step": 25970 + }, + { + "epoch": 0.07, + "learning_rate": 9.88164334924954e-05, + "loss": 2.4368, + "step": 25975 + }, + { + "epoch": 0.07, + "learning_rate": 9.881597784538101e-05, + "loss": 2.3113, + "step": 25980 + }, + { + "epoch": 0.07, + "learning_rate": 9.881552211162735e-05, + "loss": 2.427, + "step": 25985 + }, + { + "epoch": 0.07, + "learning_rate": 9.881506629123523e-05, + "loss": 2.3586, + "step": 25990 + }, + { + "epoch": 0.07, + "learning_rate": 9.881461038420543e-05, + "loss": 2.3892, + "step": 25995 + }, + { + "epoch": 0.07, + "learning_rate": 9.88141543905388e-05, + "loss": 2.3391, + "step": 26000 + }, + { + "epoch": 0.07, + "learning_rate": 9.881369831023612e-05, + "loss": 2.2443, + "step": 26005 + }, + { + "epoch": 0.07, + "learning_rate": 9.881324214329821e-05, + "loss": 2.4778, + "step": 26010 + }, + { + "epoch": 0.07, + "learning_rate": 9.881278588972588e-05, + "loss": 2.4321, + "step": 26015 + }, + { + "epoch": 0.07, + "learning_rate": 9.881232954951994e-05, + "loss": 2.3424, + "step": 26020 + }, + { + "epoch": 0.07, + "learning_rate": 9.881187312268121e-05, + "loss": 2.3687, + "step": 26025 + }, + { + "epoch": 0.07, + "learning_rate": 9.881141660921047e-05, + "loss": 2.2489, + "step": 26030 + }, + { + "epoch": 0.07, + "learning_rate": 9.881096000910857e-05, + "loss": 2.4344, + "step": 26035 + }, + { + "epoch": 0.07, + "learning_rate": 9.881050332237627e-05, + "loss": 2.3979, + "step": 26040 + }, + { + "epoch": 0.07, + "learning_rate": 9.881004654901444e-05, + "loss": 2.3709, + "step": 26045 + }, + { + "epoch": 0.07, + "learning_rate": 9.880958968902384e-05, + "loss": 2.4214, + "step": 26050 + }, + { + "epoch": 0.07, + "learning_rate": 9.88091327424053e-05, + "loss": 2.3041, + "step": 26055 + }, + { + "epoch": 0.07, + "learning_rate": 9.880867570915965e-05, + "loss": 2.3323, + "step": 26060 + }, + { + "epoch": 0.07, + "learning_rate": 9.880821858928768e-05, + "loss": 2.3669, + "step": 26065 + }, + { + "epoch": 0.07, + "learning_rate": 9.88077613827902e-05, + "loss": 2.4027, + "step": 26070 + }, + { + "epoch": 0.07, + "learning_rate": 9.880730408966803e-05, + "loss": 2.3641, + "step": 26075 + }, + { + "epoch": 0.07, + "learning_rate": 9.880684670992197e-05, + "loss": 2.2864, + "step": 26080 + }, + { + "epoch": 0.07, + "learning_rate": 9.880638924355284e-05, + "loss": 2.4859, + "step": 26085 + }, + { + "epoch": 0.07, + "learning_rate": 9.880593169056147e-05, + "loss": 2.3951, + "step": 26090 + }, + { + "epoch": 0.07, + "learning_rate": 9.880547405094862e-05, + "loss": 2.5591, + "step": 26095 + }, + { + "epoch": 0.07, + "learning_rate": 9.880501632471517e-05, + "loss": 2.4987, + "step": 26100 + }, + { + "epoch": 0.07, + "learning_rate": 9.880455851186188e-05, + "loss": 2.4086, + "step": 26105 + }, + { + "epoch": 0.07, + "learning_rate": 9.880410061238958e-05, + "loss": 2.4221, + "step": 26110 + }, + { + "epoch": 0.07, + "learning_rate": 9.880364262629908e-05, + "loss": 2.4615, + "step": 26115 + }, + { + "epoch": 0.07, + "learning_rate": 9.880318455359122e-05, + "loss": 2.4946, + "step": 26120 + }, + { + "epoch": 0.07, + "learning_rate": 9.880272639426678e-05, + "loss": 2.4905, + "step": 26125 + }, + { + "epoch": 0.07, + "learning_rate": 9.880226814832658e-05, + "loss": 2.3329, + "step": 26130 + }, + { + "epoch": 0.07, + "learning_rate": 9.880180981577143e-05, + "loss": 2.4162, + "step": 26135 + }, + { + "epoch": 0.07, + "learning_rate": 9.880135139660214e-05, + "loss": 2.3651, + "step": 26140 + }, + { + "epoch": 0.07, + "learning_rate": 9.880089289081955e-05, + "loss": 2.4049, + "step": 26145 + }, + { + "epoch": 0.07, + "learning_rate": 9.880043429842446e-05, + "loss": 2.4239, + "step": 26150 + }, + { + "epoch": 0.07, + "learning_rate": 9.879997561941768e-05, + "loss": 2.3031, + "step": 26155 + }, + { + "epoch": 0.07, + "learning_rate": 9.87995168538e-05, + "loss": 2.4126, + "step": 26160 + }, + { + "epoch": 0.07, + "learning_rate": 9.879905800157228e-05, + "loss": 2.5145, + "step": 26165 + }, + { + "epoch": 0.07, + "learning_rate": 9.879859906273532e-05, + "loss": 2.3014, + "step": 26170 + }, + { + "epoch": 0.07, + "learning_rate": 9.879814003728991e-05, + "loss": 2.365, + "step": 26175 + }, + { + "epoch": 0.07, + "learning_rate": 9.879768092523689e-05, + "loss": 2.414, + "step": 26180 + }, + { + "epoch": 0.07, + "learning_rate": 9.879722172657708e-05, + "loss": 2.4048, + "step": 26185 + }, + { + "epoch": 0.07, + "learning_rate": 9.879676244131127e-05, + "loss": 2.4306, + "step": 26190 + }, + { + "epoch": 0.07, + "learning_rate": 9.879630306944028e-05, + "loss": 2.4783, + "step": 26195 + }, + { + "epoch": 0.07, + "learning_rate": 9.879584361096494e-05, + "loss": 2.389, + "step": 26200 + }, + { + "epoch": 0.07, + "learning_rate": 9.879538406588606e-05, + "loss": 2.3925, + "step": 26205 + }, + { + "epoch": 0.07, + "learning_rate": 9.879492443420445e-05, + "loss": 2.5581, + "step": 26210 + }, + { + "epoch": 0.07, + "learning_rate": 9.879446471592093e-05, + "loss": 2.3917, + "step": 26215 + }, + { + "epoch": 0.07, + "learning_rate": 9.879400491103632e-05, + "loss": 2.5019, + "step": 26220 + }, + { + "epoch": 0.07, + "learning_rate": 9.879354501955142e-05, + "loss": 2.4642, + "step": 26225 + }, + { + "epoch": 0.07, + "learning_rate": 9.879308504146707e-05, + "loss": 2.3533, + "step": 26230 + }, + { + "epoch": 0.07, + "learning_rate": 9.879262497678408e-05, + "loss": 2.4543, + "step": 26235 + }, + { + "epoch": 0.07, + "learning_rate": 9.879216482550325e-05, + "loss": 2.442, + "step": 26240 + }, + { + "epoch": 0.07, + "learning_rate": 9.879170458762541e-05, + "loss": 2.3671, + "step": 26245 + }, + { + "epoch": 0.07, + "learning_rate": 9.879124426315137e-05, + "loss": 2.4115, + "step": 26250 + }, + { + "epoch": 0.07, + "learning_rate": 9.879078385208196e-05, + "loss": 2.3563, + "step": 26255 + }, + { + "epoch": 0.07, + "learning_rate": 9.879032335441798e-05, + "loss": 2.3346, + "step": 26260 + }, + { + "epoch": 0.07, + "learning_rate": 9.878986277016025e-05, + "loss": 2.3632, + "step": 26265 + }, + { + "epoch": 0.07, + "learning_rate": 9.87894020993096e-05, + "loss": 2.2863, + "step": 26270 + }, + { + "epoch": 0.07, + "learning_rate": 9.878894134186685e-05, + "loss": 2.5371, + "step": 26275 + }, + { + "epoch": 0.07, + "learning_rate": 9.87884804978328e-05, + "loss": 2.2891, + "step": 26280 + }, + { + "epoch": 0.07, + "learning_rate": 9.878801956720829e-05, + "loss": 2.4692, + "step": 26285 + }, + { + "epoch": 0.07, + "learning_rate": 9.878755854999411e-05, + "loss": 2.3175, + "step": 26290 + }, + { + "epoch": 0.07, + "learning_rate": 9.878709744619109e-05, + "loss": 2.3269, + "step": 26295 + }, + { + "epoch": 0.07, + "learning_rate": 9.878663625580005e-05, + "loss": 2.3121, + "step": 26300 + }, + { + "epoch": 0.07, + "learning_rate": 9.878617497882182e-05, + "loss": 2.3379, + "step": 26305 + }, + { + "epoch": 0.07, + "learning_rate": 9.87857136152572e-05, + "loss": 2.4555, + "step": 26310 + }, + { + "epoch": 0.07, + "learning_rate": 9.878525216510703e-05, + "loss": 2.4054, + "step": 26315 + }, + { + "epoch": 0.07, + "learning_rate": 9.87847906283721e-05, + "loss": 2.3798, + "step": 26320 + }, + { + "epoch": 0.07, + "learning_rate": 9.878432900505326e-05, + "loss": 2.416, + "step": 26325 + }, + { + "epoch": 0.07, + "learning_rate": 9.87838672951513e-05, + "loss": 2.2912, + "step": 26330 + }, + { + "epoch": 0.07, + "learning_rate": 9.878340549866707e-05, + "loss": 2.4268, + "step": 26335 + }, + { + "epoch": 0.07, + "learning_rate": 9.878294361560135e-05, + "loss": 2.3783, + "step": 26340 + }, + { + "epoch": 0.07, + "learning_rate": 9.8782481645955e-05, + "loss": 2.4207, + "step": 26345 + }, + { + "epoch": 0.07, + "learning_rate": 9.878201958972882e-05, + "loss": 2.4228, + "step": 26350 + }, + { + "epoch": 0.07, + "learning_rate": 9.878155744692364e-05, + "loss": 2.481, + "step": 26355 + }, + { + "epoch": 0.07, + "learning_rate": 9.878109521754026e-05, + "loss": 2.4469, + "step": 26360 + }, + { + "epoch": 0.07, + "learning_rate": 9.878063290157953e-05, + "loss": 2.3539, + "step": 26365 + }, + { + "epoch": 0.07, + "learning_rate": 9.878017049904225e-05, + "loss": 2.3982, + "step": 26370 + }, + { + "epoch": 0.07, + "learning_rate": 9.877970800992923e-05, + "loss": 2.4347, + "step": 26375 + }, + { + "epoch": 0.07, + "learning_rate": 9.877924543424133e-05, + "loss": 2.2053, + "step": 26380 + }, + { + "epoch": 0.07, + "learning_rate": 9.877878277197932e-05, + "loss": 2.4025, + "step": 26385 + }, + { + "epoch": 0.07, + "learning_rate": 9.877832002314406e-05, + "loss": 2.3785, + "step": 26390 + }, + { + "epoch": 0.07, + "learning_rate": 9.877785718773638e-05, + "loss": 2.441, + "step": 26395 + }, + { + "epoch": 0.07, + "learning_rate": 9.877739426575705e-05, + "loss": 2.4959, + "step": 26400 + }, + { + "epoch": 0.07, + "learning_rate": 9.877693125720695e-05, + "loss": 2.3954, + "step": 26405 + }, + { + "epoch": 0.07, + "learning_rate": 9.877646816208686e-05, + "loss": 2.4193, + "step": 26410 + }, + { + "epoch": 0.07, + "learning_rate": 9.877600498039761e-05, + "loss": 2.4241, + "step": 26415 + }, + { + "epoch": 0.07, + "learning_rate": 9.877554171214003e-05, + "loss": 2.3533, + "step": 26420 + }, + { + "epoch": 0.07, + "learning_rate": 9.877507835731494e-05, + "loss": 2.4499, + "step": 26425 + }, + { + "epoch": 0.07, + "learning_rate": 9.877461491592317e-05, + "loss": 2.3091, + "step": 26430 + }, + { + "epoch": 0.07, + "learning_rate": 9.877415138796553e-05, + "loss": 2.4081, + "step": 26435 + }, + { + "epoch": 0.07, + "learning_rate": 9.877368777344285e-05, + "loss": 2.3808, + "step": 26440 + }, + { + "epoch": 0.07, + "learning_rate": 9.877322407235594e-05, + "loss": 2.3589, + "step": 26445 + }, + { + "epoch": 0.07, + "learning_rate": 9.877276028470566e-05, + "loss": 2.3794, + "step": 26450 + }, + { + "epoch": 0.07, + "learning_rate": 9.877229641049279e-05, + "loss": 2.3983, + "step": 26455 + }, + { + "epoch": 0.07, + "learning_rate": 9.877183244971816e-05, + "loss": 2.3368, + "step": 26460 + }, + { + "epoch": 0.07, + "learning_rate": 9.877136840238261e-05, + "loss": 2.4904, + "step": 26465 + }, + { + "epoch": 0.07, + "learning_rate": 9.877090426848696e-05, + "loss": 2.3566, + "step": 26470 + }, + { + "epoch": 0.07, + "learning_rate": 9.877044004803205e-05, + "loss": 2.4344, + "step": 26475 + }, + { + "epoch": 0.07, + "learning_rate": 9.876997574101865e-05, + "loss": 2.3321, + "step": 26480 + }, + { + "epoch": 0.07, + "learning_rate": 9.876951134744765e-05, + "loss": 2.5308, + "step": 26485 + }, + { + "epoch": 0.07, + "learning_rate": 9.876904686731983e-05, + "loss": 2.3916, + "step": 26490 + }, + { + "epoch": 0.07, + "learning_rate": 9.876858230063601e-05, + "loss": 2.3959, + "step": 26495 + }, + { + "epoch": 0.07, + "learning_rate": 9.876811764739706e-05, + "loss": 2.278, + "step": 26500 + }, + { + "epoch": 0.07, + "learning_rate": 9.876765290760377e-05, + "loss": 2.3604, + "step": 26505 + }, + { + "epoch": 0.07, + "learning_rate": 9.876718808125696e-05, + "loss": 2.4217, + "step": 26510 + }, + { + "epoch": 0.07, + "learning_rate": 9.876672316835748e-05, + "loss": 2.4375, + "step": 26515 + }, + { + "epoch": 0.07, + "learning_rate": 9.876625816890612e-05, + "loss": 2.3959, + "step": 26520 + }, + { + "epoch": 0.07, + "learning_rate": 9.876579308290375e-05, + "loss": 2.3681, + "step": 26525 + }, + { + "epoch": 0.07, + "learning_rate": 9.876532791035117e-05, + "loss": 2.3602, + "step": 26530 + }, + { + "epoch": 0.07, + "learning_rate": 9.87648626512492e-05, + "loss": 2.3465, + "step": 26535 + }, + { + "epoch": 0.07, + "learning_rate": 9.876439730559868e-05, + "loss": 2.3437, + "step": 26540 + }, + { + "epoch": 0.07, + "learning_rate": 9.876393187340043e-05, + "loss": 2.4019, + "step": 26545 + }, + { + "epoch": 0.07, + "learning_rate": 9.876346635465527e-05, + "loss": 2.4856, + "step": 26550 + }, + { + "epoch": 0.07, + "learning_rate": 9.876300074936402e-05, + "loss": 2.3917, + "step": 26555 + }, + { + "epoch": 0.07, + "learning_rate": 9.876253505752754e-05, + "loss": 2.4039, + "step": 26560 + }, + { + "epoch": 0.07, + "learning_rate": 9.876206927914662e-05, + "loss": 2.332, + "step": 26565 + }, + { + "epoch": 0.07, + "learning_rate": 9.876160341422211e-05, + "loss": 2.2909, + "step": 26570 + }, + { + "epoch": 0.07, + "learning_rate": 9.876113746275483e-05, + "loss": 2.4136, + "step": 26575 + }, + { + "epoch": 0.07, + "learning_rate": 9.87606714247456e-05, + "loss": 2.3214, + "step": 26580 + }, + { + "epoch": 0.07, + "learning_rate": 9.876020530019526e-05, + "loss": 2.4026, + "step": 26585 + }, + { + "epoch": 0.07, + "learning_rate": 9.875973908910461e-05, + "loss": 2.1648, + "step": 26590 + }, + { + "epoch": 0.07, + "learning_rate": 9.875927279147452e-05, + "loss": 2.3663, + "step": 26595 + }, + { + "epoch": 0.07, + "learning_rate": 9.875880640730578e-05, + "loss": 2.4422, + "step": 26600 + }, + { + "epoch": 0.07, + "learning_rate": 9.875833993659925e-05, + "loss": 2.3765, + "step": 26605 + }, + { + "epoch": 0.07, + "learning_rate": 9.875787337935571e-05, + "loss": 2.4223, + "step": 26610 + }, + { + "epoch": 0.07, + "learning_rate": 9.875740673557604e-05, + "loss": 2.4069, + "step": 26615 + }, + { + "epoch": 0.07, + "learning_rate": 9.875694000526104e-05, + "loss": 2.3065, + "step": 26620 + }, + { + "epoch": 0.07, + "learning_rate": 9.875647318841155e-05, + "loss": 2.261, + "step": 26625 + }, + { + "epoch": 0.07, + "learning_rate": 9.87560062850284e-05, + "loss": 2.3668, + "step": 26630 + }, + { + "epoch": 0.07, + "learning_rate": 9.875553929511239e-05, + "loss": 2.379, + "step": 26635 + }, + { + "epoch": 0.07, + "learning_rate": 9.875507221866438e-05, + "loss": 2.4073, + "step": 26640 + }, + { + "epoch": 0.07, + "learning_rate": 9.87546050556852e-05, + "loss": 2.3814, + "step": 26645 + }, + { + "epoch": 0.07, + "learning_rate": 9.875413780617566e-05, + "loss": 2.3199, + "step": 26650 + }, + { + "epoch": 0.07, + "learning_rate": 9.87536704701366e-05, + "loss": 2.4005, + "step": 26655 + }, + { + "epoch": 0.07, + "learning_rate": 9.875320304756885e-05, + "loss": 2.3295, + "step": 26660 + }, + { + "epoch": 0.07, + "learning_rate": 9.875273553847323e-05, + "loss": 2.4269, + "step": 26665 + }, + { + "epoch": 0.07, + "learning_rate": 9.87522679428506e-05, + "loss": 2.3708, + "step": 26670 + }, + { + "epoch": 0.07, + "learning_rate": 9.875180026070174e-05, + "loss": 2.3668, + "step": 26675 + }, + { + "epoch": 0.07, + "learning_rate": 9.875133249202751e-05, + "loss": 2.4264, + "step": 26680 + }, + { + "epoch": 0.07, + "learning_rate": 9.875086463682874e-05, + "loss": 2.4133, + "step": 26685 + }, + { + "epoch": 0.07, + "learning_rate": 9.875039669510628e-05, + "loss": 2.4422, + "step": 26690 + }, + { + "epoch": 0.07, + "learning_rate": 9.874992866686091e-05, + "loss": 2.3896, + "step": 26695 + }, + { + "epoch": 0.07, + "learning_rate": 9.87494605520935e-05, + "loss": 2.352, + "step": 26700 + }, + { + "epoch": 0.07, + "learning_rate": 9.874899235080487e-05, + "loss": 2.3926, + "step": 26705 + }, + { + "epoch": 0.07, + "learning_rate": 9.874852406299584e-05, + "loss": 2.3219, + "step": 26710 + }, + { + "epoch": 0.07, + "learning_rate": 9.874805568866727e-05, + "loss": 2.4338, + "step": 26715 + }, + { + "epoch": 0.07, + "learning_rate": 9.874758722781995e-05, + "loss": 2.3071, + "step": 26720 + }, + { + "epoch": 0.07, + "learning_rate": 9.874711868045477e-05, + "loss": 2.3169, + "step": 26725 + }, + { + "epoch": 0.07, + "learning_rate": 9.874665004657248e-05, + "loss": 2.3901, + "step": 26730 + }, + { + "epoch": 0.07, + "learning_rate": 9.8746181326174e-05, + "loss": 2.389, + "step": 26735 + }, + { + "epoch": 0.07, + "learning_rate": 9.87457125192601e-05, + "loss": 2.4158, + "step": 26740 + }, + { + "epoch": 0.07, + "learning_rate": 9.874524362583162e-05, + "loss": 2.2965, + "step": 26745 + }, + { + "epoch": 0.07, + "learning_rate": 9.874477464588943e-05, + "loss": 2.4138, + "step": 26750 + }, + { + "epoch": 0.07, + "learning_rate": 9.874430557943432e-05, + "loss": 2.3638, + "step": 26755 + }, + { + "epoch": 0.07, + "learning_rate": 9.874383642646714e-05, + "loss": 2.3418, + "step": 26760 + }, + { + "epoch": 0.07, + "learning_rate": 9.874336718698871e-05, + "loss": 2.2724, + "step": 26765 + }, + { + "epoch": 0.07, + "learning_rate": 9.874289786099989e-05, + "loss": 2.3011, + "step": 26770 + }, + { + "epoch": 0.07, + "learning_rate": 9.87424284485015e-05, + "loss": 2.5316, + "step": 26775 + }, + { + "epoch": 0.07, + "learning_rate": 9.874195894949436e-05, + "loss": 2.327, + "step": 26780 + }, + { + "epoch": 0.07, + "learning_rate": 9.87414893639793e-05, + "loss": 2.3946, + "step": 26785 + }, + { + "epoch": 0.07, + "learning_rate": 9.87410196919572e-05, + "loss": 2.4103, + "step": 26790 + }, + { + "epoch": 0.07, + "learning_rate": 9.874054993342885e-05, + "loss": 2.3709, + "step": 26795 + }, + { + "epoch": 0.07, + "learning_rate": 9.874008008839508e-05, + "loss": 2.4276, + "step": 26800 + }, + { + "epoch": 0.07, + "learning_rate": 9.873961015685675e-05, + "loss": 2.3446, + "step": 26805 + }, + { + "epoch": 0.07, + "learning_rate": 9.873914013881467e-05, + "loss": 2.3064, + "step": 26810 + }, + { + "epoch": 0.07, + "learning_rate": 9.873867003426971e-05, + "loss": 2.3843, + "step": 26815 + }, + { + "epoch": 0.07, + "learning_rate": 9.873819984322266e-05, + "loss": 2.4766, + "step": 26820 + }, + { + "epoch": 0.07, + "learning_rate": 9.873772956567438e-05, + "loss": 2.4404, + "step": 26825 + }, + { + "epoch": 0.07, + "learning_rate": 9.873725920162571e-05, + "loss": 2.3637, + "step": 26830 + }, + { + "epoch": 0.07, + "learning_rate": 9.873678875107746e-05, + "loss": 2.4171, + "step": 26835 + }, + { + "epoch": 0.07, + "learning_rate": 9.873631821403048e-05, + "loss": 2.3905, + "step": 26840 + }, + { + "epoch": 0.07, + "learning_rate": 9.873584759048561e-05, + "loss": 2.4108, + "step": 26845 + }, + { + "epoch": 0.07, + "learning_rate": 9.873537688044369e-05, + "loss": 2.3535, + "step": 26850 + }, + { + "epoch": 0.07, + "learning_rate": 9.873490608390553e-05, + "loss": 2.403, + "step": 26855 + }, + { + "epoch": 0.07, + "learning_rate": 9.873443520087198e-05, + "loss": 2.3943, + "step": 26860 + }, + { + "epoch": 0.07, + "learning_rate": 9.873396423134389e-05, + "loss": 2.3612, + "step": 26865 + }, + { + "epoch": 0.07, + "learning_rate": 9.873349317532208e-05, + "loss": 2.4406, + "step": 26870 + }, + { + "epoch": 0.07, + "learning_rate": 9.873302203280739e-05, + "loss": 2.4028, + "step": 26875 + }, + { + "epoch": 0.07, + "learning_rate": 9.873255080380064e-05, + "loss": 2.3852, + "step": 26880 + }, + { + "epoch": 0.07, + "learning_rate": 9.87320794883027e-05, + "loss": 2.5226, + "step": 26885 + }, + { + "epoch": 0.07, + "learning_rate": 9.873160808631438e-05, + "loss": 2.467, + "step": 26890 + }, + { + "epoch": 0.07, + "learning_rate": 9.873113659783652e-05, + "loss": 2.4704, + "step": 26895 + }, + { + "epoch": 0.07, + "learning_rate": 9.873066502286995e-05, + "loss": 2.3348, + "step": 26900 + }, + { + "epoch": 0.07, + "learning_rate": 9.873019336141554e-05, + "loss": 2.2442, + "step": 26905 + }, + { + "epoch": 0.07, + "learning_rate": 9.87297216134741e-05, + "loss": 2.3687, + "step": 26910 + }, + { + "epoch": 0.07, + "learning_rate": 9.872924977904647e-05, + "loss": 2.4186, + "step": 26915 + }, + { + "epoch": 0.07, + "learning_rate": 9.872877785813349e-05, + "loss": 2.2268, + "step": 26920 + }, + { + "epoch": 0.07, + "learning_rate": 9.872830585073601e-05, + "loss": 2.4483, + "step": 26925 + }, + { + "epoch": 0.07, + "learning_rate": 9.872783375685484e-05, + "loss": 2.427, + "step": 26930 + }, + { + "epoch": 0.07, + "learning_rate": 9.872736157649083e-05, + "loss": 2.4028, + "step": 26935 + }, + { + "epoch": 0.07, + "learning_rate": 9.872688930964482e-05, + "loss": 2.3091, + "step": 26940 + }, + { + "epoch": 0.07, + "learning_rate": 9.872641695631766e-05, + "loss": 2.4755, + "step": 26945 + }, + { + "epoch": 0.07, + "learning_rate": 9.872594451651017e-05, + "loss": 2.316, + "step": 26950 + }, + { + "epoch": 0.07, + "learning_rate": 9.87254719902232e-05, + "loss": 2.3455, + "step": 26955 + }, + { + "epoch": 0.07, + "learning_rate": 9.872499937745758e-05, + "loss": 2.3955, + "step": 26960 + }, + { + "epoch": 0.07, + "learning_rate": 9.872452667821417e-05, + "loss": 2.3683, + "step": 26965 + }, + { + "epoch": 0.07, + "learning_rate": 9.872405389249378e-05, + "loss": 2.39, + "step": 26970 + }, + { + "epoch": 0.07, + "learning_rate": 9.872358102029726e-05, + "loss": 2.3117, + "step": 26975 + }, + { + "epoch": 0.07, + "learning_rate": 9.872310806162545e-05, + "loss": 2.448, + "step": 26980 + }, + { + "epoch": 0.07, + "learning_rate": 9.87226350164792e-05, + "loss": 2.3565, + "step": 26985 + }, + { + "epoch": 0.07, + "learning_rate": 9.872216188485933e-05, + "loss": 2.4044, + "step": 26990 + }, + { + "epoch": 0.07, + "learning_rate": 9.87216886667667e-05, + "loss": 2.376, + "step": 26995 + }, + { + "epoch": 0.07, + "learning_rate": 9.872121536220213e-05, + "loss": 2.2432, + "step": 27000 + }, + { + "epoch": 0.07, + "learning_rate": 9.872074197116648e-05, + "loss": 2.3757, + "step": 27005 + }, + { + "epoch": 0.07, + "learning_rate": 9.872026849366056e-05, + "loss": 2.3027, + "step": 27010 + }, + { + "epoch": 0.07, + "learning_rate": 9.871979492968524e-05, + "loss": 2.2446, + "step": 27015 + }, + { + "epoch": 0.07, + "learning_rate": 9.871932127924134e-05, + "loss": 2.2357, + "step": 27020 + }, + { + "epoch": 0.07, + "learning_rate": 9.871884754232973e-05, + "loss": 2.4352, + "step": 27025 + }, + { + "epoch": 0.07, + "learning_rate": 9.871837371895122e-05, + "loss": 2.5012, + "step": 27030 + }, + { + "epoch": 0.07, + "learning_rate": 9.871789980910668e-05, + "loss": 2.3521, + "step": 27035 + }, + { + "epoch": 0.07, + "learning_rate": 9.871742581279693e-05, + "loss": 2.4146, + "step": 27040 + }, + { + "epoch": 0.07, + "learning_rate": 9.87169517300228e-05, + "loss": 2.471, + "step": 27045 + }, + { + "epoch": 0.07, + "learning_rate": 9.871647756078516e-05, + "loss": 2.2491, + "step": 27050 + }, + { + "epoch": 0.07, + "learning_rate": 9.871600330508484e-05, + "loss": 2.386, + "step": 27055 + }, + { + "epoch": 0.07, + "learning_rate": 9.871552896292266e-05, + "loss": 2.4554, + "step": 27060 + }, + { + "epoch": 0.07, + "learning_rate": 9.87150545342995e-05, + "loss": 2.3714, + "step": 27065 + }, + { + "epoch": 0.07, + "learning_rate": 9.871458001921617e-05, + "loss": 2.4498, + "step": 27070 + }, + { + "epoch": 0.07, + "learning_rate": 9.871410541767354e-05, + "loss": 2.3567, + "step": 27075 + }, + { + "epoch": 0.07, + "learning_rate": 9.871363072967243e-05, + "loss": 2.3726, + "step": 27080 + }, + { + "epoch": 0.07, + "learning_rate": 9.871315595521368e-05, + "loss": 2.4091, + "step": 27085 + }, + { + "epoch": 0.07, + "learning_rate": 9.871268109429816e-05, + "loss": 2.4877, + "step": 27090 + }, + { + "epoch": 0.07, + "learning_rate": 9.871220614692672e-05, + "loss": 2.3619, + "step": 27095 + }, + { + "epoch": 0.07, + "learning_rate": 9.871173111310014e-05, + "loss": 2.302, + "step": 27100 + }, + { + "epoch": 0.07, + "learning_rate": 9.87112559928193e-05, + "loss": 2.3986, + "step": 27105 + }, + { + "epoch": 0.07, + "learning_rate": 9.871078078608506e-05, + "loss": 2.3336, + "step": 27110 + }, + { + "epoch": 0.07, + "learning_rate": 9.871030549289825e-05, + "loss": 2.516, + "step": 27115 + }, + { + "epoch": 0.07, + "learning_rate": 9.870983011325971e-05, + "loss": 2.5214, + "step": 27120 + }, + { + "epoch": 0.07, + "learning_rate": 9.870935464717031e-05, + "loss": 2.4546, + "step": 27125 + }, + { + "epoch": 0.07, + "learning_rate": 9.870887909463083e-05, + "loss": 2.4257, + "step": 27130 + }, + { + "epoch": 0.07, + "learning_rate": 9.870840345564219e-05, + "loss": 2.4468, + "step": 27135 + }, + { + "epoch": 0.07, + "learning_rate": 9.870792773020516e-05, + "loss": 2.4611, + "step": 27140 + }, + { + "epoch": 0.07, + "learning_rate": 9.870745191832065e-05, + "loss": 2.2458, + "step": 27145 + }, + { + "epoch": 0.07, + "learning_rate": 9.870697601998947e-05, + "loss": 2.3953, + "step": 27150 + }, + { + "epoch": 0.07, + "learning_rate": 9.870650003521247e-05, + "loss": 2.4527, + "step": 27155 + }, + { + "epoch": 0.07, + "learning_rate": 9.870602396399049e-05, + "loss": 2.4062, + "step": 27160 + }, + { + "epoch": 0.07, + "learning_rate": 9.870554780632439e-05, + "loss": 2.3638, + "step": 27165 + }, + { + "epoch": 0.07, + "learning_rate": 9.870507156221501e-05, + "loss": 2.3474, + "step": 27170 + }, + { + "epoch": 0.07, + "learning_rate": 9.870459523166319e-05, + "loss": 2.4612, + "step": 27175 + }, + { + "epoch": 0.07, + "learning_rate": 9.870411881466978e-05, + "loss": 2.2926, + "step": 27180 + }, + { + "epoch": 0.07, + "learning_rate": 9.87036423112356e-05, + "loss": 2.3848, + "step": 27185 + }, + { + "epoch": 0.07, + "learning_rate": 9.870316572136154e-05, + "loss": 2.3633, + "step": 27190 + }, + { + "epoch": 0.07, + "learning_rate": 9.870268904504843e-05, + "loss": 2.4651, + "step": 27195 + }, + { + "epoch": 0.07, + "learning_rate": 9.87022122822971e-05, + "loss": 2.4265, + "step": 27200 + }, + { + "epoch": 0.07, + "learning_rate": 9.870173543310842e-05, + "loss": 2.4428, + "step": 27205 + }, + { + "epoch": 0.07, + "learning_rate": 9.87012584974832e-05, + "loss": 2.2944, + "step": 27210 + }, + { + "epoch": 0.07, + "learning_rate": 9.870078147542231e-05, + "loss": 2.4198, + "step": 27215 + }, + { + "epoch": 0.07, + "learning_rate": 9.870030436692661e-05, + "loss": 2.4333, + "step": 27220 + }, + { + "epoch": 0.07, + "learning_rate": 9.869982717199693e-05, + "loss": 2.4631, + "step": 27225 + }, + { + "epoch": 0.07, + "learning_rate": 9.869934989063412e-05, + "loss": 2.4647, + "step": 27230 + }, + { + "epoch": 0.07, + "learning_rate": 9.869887252283901e-05, + "loss": 2.4193, + "step": 27235 + }, + { + "epoch": 0.07, + "learning_rate": 9.869839506861248e-05, + "loss": 2.3195, + "step": 27240 + }, + { + "epoch": 0.07, + "learning_rate": 9.869791752795537e-05, + "loss": 2.3319, + "step": 27245 + }, + { + "epoch": 0.07, + "learning_rate": 9.86974399008685e-05, + "loss": 2.4048, + "step": 27250 + }, + { + "epoch": 0.07, + "learning_rate": 9.869696218735273e-05, + "loss": 2.3078, + "step": 27255 + }, + { + "epoch": 0.07, + "learning_rate": 9.869648438740893e-05, + "loss": 2.3478, + "step": 27260 + }, + { + "epoch": 0.07, + "learning_rate": 9.869600650103794e-05, + "loss": 2.3903, + "step": 27265 + }, + { + "epoch": 0.07, + "learning_rate": 9.86955285282406e-05, + "loss": 2.4648, + "step": 27270 + }, + { + "epoch": 0.07, + "learning_rate": 9.869505046901775e-05, + "loss": 2.2536, + "step": 27275 + }, + { + "epoch": 0.07, + "learning_rate": 9.869457232337024e-05, + "loss": 2.3942, + "step": 27280 + }, + { + "epoch": 0.07, + "learning_rate": 9.869409409129892e-05, + "loss": 2.4024, + "step": 27285 + }, + { + "epoch": 0.07, + "learning_rate": 9.869361577280467e-05, + "loss": 2.4691, + "step": 27290 + }, + { + "epoch": 0.07, + "learning_rate": 9.86931373678883e-05, + "loss": 2.2711, + "step": 27295 + }, + { + "epoch": 0.07, + "learning_rate": 9.869265887655069e-05, + "loss": 2.4669, + "step": 27300 + }, + { + "epoch": 0.07, + "learning_rate": 9.869218029879263e-05, + "loss": 2.5487, + "step": 27305 + }, + { + "epoch": 0.07, + "learning_rate": 9.869170163461505e-05, + "loss": 2.293, + "step": 27310 + }, + { + "epoch": 0.07, + "learning_rate": 9.869122288401876e-05, + "loss": 2.431, + "step": 27315 + }, + { + "epoch": 0.07, + "learning_rate": 9.869074404700458e-05, + "loss": 2.4644, + "step": 27320 + }, + { + "epoch": 0.07, + "learning_rate": 9.869026512357342e-05, + "loss": 2.4868, + "step": 27325 + }, + { + "epoch": 0.07, + "learning_rate": 9.868978611372608e-05, + "loss": 2.2775, + "step": 27330 + }, + { + "epoch": 0.07, + "learning_rate": 9.868930701746344e-05, + "loss": 2.4686, + "step": 27335 + }, + { + "epoch": 0.07, + "learning_rate": 9.868882783478634e-05, + "loss": 2.4604, + "step": 27340 + }, + { + "epoch": 0.07, + "learning_rate": 9.868834856569565e-05, + "loss": 2.4226, + "step": 27345 + }, + { + "epoch": 0.07, + "learning_rate": 9.868786921019217e-05, + "loss": 2.3655, + "step": 27350 + }, + { + "epoch": 0.07, + "learning_rate": 9.86873897682768e-05, + "loss": 2.435, + "step": 27355 + }, + { + "epoch": 0.07, + "learning_rate": 9.868691023995039e-05, + "loss": 2.361, + "step": 27360 + }, + { + "epoch": 0.07, + "learning_rate": 9.868643062521375e-05, + "loss": 2.3385, + "step": 27365 + }, + { + "epoch": 0.07, + "learning_rate": 9.868595092406775e-05, + "loss": 2.4418, + "step": 27370 + }, + { + "epoch": 0.07, + "learning_rate": 9.868547113651326e-05, + "loss": 2.4552, + "step": 27375 + }, + { + "epoch": 0.07, + "learning_rate": 9.868499126255112e-05, + "loss": 2.4212, + "step": 27380 + }, + { + "epoch": 0.07, + "learning_rate": 9.868451130218219e-05, + "loss": 2.3703, + "step": 27385 + }, + { + "epoch": 0.07, + "learning_rate": 9.868403125540729e-05, + "loss": 2.3702, + "step": 27390 + }, + { + "epoch": 0.07, + "learning_rate": 9.86835511222273e-05, + "loss": 2.3049, + "step": 27395 + }, + { + "epoch": 0.07, + "learning_rate": 9.868307090264308e-05, + "loss": 2.3944, + "step": 27400 + }, + { + "epoch": 0.07, + "learning_rate": 9.868259059665546e-05, + "loss": 2.296, + "step": 27405 + }, + { + "epoch": 0.07, + "learning_rate": 9.86821102042653e-05, + "loss": 2.3701, + "step": 27410 + }, + { + "epoch": 0.07, + "learning_rate": 9.868162972547345e-05, + "loss": 2.4581, + "step": 27415 + }, + { + "epoch": 0.07, + "learning_rate": 9.868114916028078e-05, + "loss": 2.3803, + "step": 27420 + }, + { + "epoch": 0.07, + "learning_rate": 9.868066850868811e-05, + "loss": 2.4482, + "step": 27425 + }, + { + "epoch": 0.07, + "learning_rate": 9.868018777069633e-05, + "loss": 2.4721, + "step": 27430 + }, + { + "epoch": 0.07, + "learning_rate": 9.867970694630627e-05, + "loss": 2.3328, + "step": 27435 + }, + { + "epoch": 0.07, + "learning_rate": 9.867922603551879e-05, + "loss": 2.3596, + "step": 27440 + }, + { + "epoch": 0.07, + "learning_rate": 9.867874503833474e-05, + "loss": 2.4419, + "step": 27445 + }, + { + "epoch": 0.07, + "learning_rate": 9.867826395475499e-05, + "loss": 2.2254, + "step": 27450 + }, + { + "epoch": 0.07, + "learning_rate": 9.867778278478037e-05, + "loss": 2.3179, + "step": 27455 + }, + { + "epoch": 0.07, + "learning_rate": 9.867730152841174e-05, + "loss": 2.3176, + "step": 27460 + }, + { + "epoch": 0.07, + "learning_rate": 9.867682018564996e-05, + "loss": 2.1977, + "step": 27465 + }, + { + "epoch": 0.07, + "learning_rate": 9.867633875649589e-05, + "loss": 2.3389, + "step": 27470 + }, + { + "epoch": 0.07, + "learning_rate": 9.867585724095037e-05, + "loss": 2.4465, + "step": 27475 + }, + { + "epoch": 0.07, + "learning_rate": 9.867537563901425e-05, + "loss": 2.3419, + "step": 27480 + }, + { + "epoch": 0.07, + "learning_rate": 9.867489395068841e-05, + "loss": 2.3985, + "step": 27485 + }, + { + "epoch": 0.07, + "learning_rate": 9.86744121759737e-05, + "loss": 2.3182, + "step": 27490 + }, + { + "epoch": 0.07, + "learning_rate": 9.867393031487095e-05, + "loss": 2.3318, + "step": 27495 + }, + { + "epoch": 0.07, + "learning_rate": 9.867344836738103e-05, + "loss": 2.5184, + "step": 27500 + }, + { + "epoch": 0.07, + "learning_rate": 9.867296633350483e-05, + "loss": 2.3374, + "step": 27505 + }, + { + "epoch": 0.07, + "learning_rate": 9.867248421324313e-05, + "loss": 2.4409, + "step": 27510 + }, + { + "epoch": 0.07, + "learning_rate": 9.867200200659685e-05, + "loss": 2.3412, + "step": 27515 + }, + { + "epoch": 0.07, + "learning_rate": 9.867151971356682e-05, + "loss": 2.4351, + "step": 27520 + }, + { + "epoch": 0.07, + "learning_rate": 9.867103733415391e-05, + "loss": 2.3372, + "step": 27525 + }, + { + "epoch": 0.07, + "learning_rate": 9.867055486835896e-05, + "loss": 2.4151, + "step": 27530 + }, + { + "epoch": 0.07, + "learning_rate": 9.867007231618283e-05, + "loss": 2.4877, + "step": 27535 + }, + { + "epoch": 0.07, + "learning_rate": 9.866958967762638e-05, + "loss": 2.4344, + "step": 27540 + }, + { + "epoch": 0.07, + "learning_rate": 9.866910695269046e-05, + "loss": 2.3388, + "step": 27545 + }, + { + "epoch": 0.07, + "learning_rate": 9.866862414137594e-05, + "loss": 2.3681, + "step": 27550 + }, + { + "epoch": 0.07, + "learning_rate": 9.866814124368367e-05, + "loss": 2.3465, + "step": 27555 + }, + { + "epoch": 0.07, + "learning_rate": 9.86676582596145e-05, + "loss": 2.3719, + "step": 27560 + }, + { + "epoch": 0.07, + "learning_rate": 9.866717518916928e-05, + "loss": 2.4432, + "step": 27565 + }, + { + "epoch": 0.07, + "learning_rate": 9.86666920323489e-05, + "loss": 2.3873, + "step": 27570 + }, + { + "epoch": 0.07, + "learning_rate": 9.86662087891542e-05, + "loss": 2.303, + "step": 27575 + }, + { + "epoch": 0.07, + "learning_rate": 9.866572545958603e-05, + "loss": 2.2889, + "step": 27580 + }, + { + "epoch": 0.07, + "learning_rate": 9.866524204364525e-05, + "loss": 2.4359, + "step": 27585 + }, + { + "epoch": 0.07, + "learning_rate": 9.866475854133272e-05, + "loss": 2.369, + "step": 27590 + }, + { + "epoch": 0.07, + "learning_rate": 9.866427495264931e-05, + "loss": 2.3323, + "step": 27595 + }, + { + "epoch": 0.07, + "learning_rate": 9.866379127759587e-05, + "loss": 2.5033, + "step": 27600 + }, + { + "epoch": 0.07, + "learning_rate": 9.866330751617325e-05, + "loss": 2.4277, + "step": 27605 + }, + { + "epoch": 0.07, + "learning_rate": 9.86628236683823e-05, + "loss": 2.2758, + "step": 27610 + }, + { + "epoch": 0.07, + "learning_rate": 9.866233973422392e-05, + "loss": 2.4092, + "step": 27615 + }, + { + "epoch": 0.07, + "learning_rate": 9.866185571369892e-05, + "loss": 2.343, + "step": 27620 + }, + { + "epoch": 0.07, + "learning_rate": 9.86613716068082e-05, + "loss": 2.3573, + "step": 27625 + }, + { + "epoch": 0.07, + "learning_rate": 9.866088741355259e-05, + "loss": 2.3547, + "step": 27630 + }, + { + "epoch": 0.07, + "learning_rate": 9.866040313393298e-05, + "loss": 2.3368, + "step": 27635 + }, + { + "epoch": 0.07, + "learning_rate": 9.86599187679502e-05, + "loss": 2.3449, + "step": 27640 + }, + { + "epoch": 0.07, + "learning_rate": 9.86594343156051e-05, + "loss": 2.268, + "step": 27645 + }, + { + "epoch": 0.07, + "learning_rate": 9.865894977689858e-05, + "loss": 2.4262, + "step": 27650 + }, + { + "epoch": 0.07, + "learning_rate": 9.865846515183148e-05, + "loss": 2.4024, + "step": 27655 + }, + { + "epoch": 0.07, + "learning_rate": 9.865798044040465e-05, + "loss": 2.3906, + "step": 27660 + }, + { + "epoch": 0.07, + "learning_rate": 9.865749564261896e-05, + "loss": 2.361, + "step": 27665 + }, + { + "epoch": 0.07, + "learning_rate": 9.865701075847526e-05, + "loss": 2.3608, + "step": 27670 + }, + { + "epoch": 0.07, + "learning_rate": 9.865652578797444e-05, + "loss": 2.5062, + "step": 27675 + }, + { + "epoch": 0.07, + "learning_rate": 9.865604073111734e-05, + "loss": 2.3178, + "step": 27680 + }, + { + "epoch": 0.07, + "learning_rate": 9.865555558790482e-05, + "loss": 2.5139, + "step": 27685 + }, + { + "epoch": 0.07, + "learning_rate": 9.865507035833774e-05, + "loss": 2.317, + "step": 27690 + }, + { + "epoch": 0.07, + "learning_rate": 9.865458504241696e-05, + "loss": 2.4275, + "step": 27695 + }, + { + "epoch": 0.07, + "learning_rate": 9.865409964014337e-05, + "loss": 2.3618, + "step": 27700 + }, + { + "epoch": 0.07, + "learning_rate": 9.865361415151777e-05, + "loss": 2.3817, + "step": 27705 + }, + { + "epoch": 0.07, + "learning_rate": 9.865312857654108e-05, + "loss": 2.3929, + "step": 27710 + }, + { + "epoch": 0.07, + "learning_rate": 9.865264291521414e-05, + "loss": 2.3117, + "step": 27715 + }, + { + "epoch": 0.07, + "learning_rate": 9.865215716753782e-05, + "loss": 2.2182, + "step": 27720 + }, + { + "epoch": 0.07, + "learning_rate": 9.865167133351297e-05, + "loss": 2.3481, + "step": 27725 + }, + { + "epoch": 0.07, + "learning_rate": 9.865118541314045e-05, + "loss": 2.2152, + "step": 27730 + }, + { + "epoch": 0.07, + "learning_rate": 9.865069940642112e-05, + "loss": 2.4556, + "step": 27735 + }, + { + "epoch": 0.07, + "learning_rate": 9.865021331335587e-05, + "loss": 2.407, + "step": 27740 + }, + { + "epoch": 0.07, + "learning_rate": 9.864972713394554e-05, + "loss": 2.3729, + "step": 27745 + }, + { + "epoch": 0.07, + "learning_rate": 9.864924086819099e-05, + "loss": 2.3069, + "step": 27750 + }, + { + "epoch": 0.07, + "learning_rate": 9.86487545160931e-05, + "loss": 2.4177, + "step": 27755 + }, + { + "epoch": 0.07, + "learning_rate": 9.864826807765272e-05, + "loss": 2.3577, + "step": 27760 + }, + { + "epoch": 0.07, + "learning_rate": 9.864778155287073e-05, + "loss": 2.4099, + "step": 27765 + }, + { + "epoch": 0.07, + "learning_rate": 9.864729494174796e-05, + "loss": 2.3488, + "step": 27770 + }, + { + "epoch": 0.07, + "learning_rate": 9.864680824428532e-05, + "loss": 2.3773, + "step": 27775 + }, + { + "epoch": 0.07, + "learning_rate": 9.864632146048363e-05, + "loss": 2.4021, + "step": 27780 + }, + { + "epoch": 0.07, + "learning_rate": 9.864583459034378e-05, + "loss": 2.3609, + "step": 27785 + }, + { + "epoch": 0.07, + "learning_rate": 9.864534763386661e-05, + "loss": 2.3579, + "step": 27790 + }, + { + "epoch": 0.07, + "learning_rate": 9.864486059105302e-05, + "loss": 2.3395, + "step": 27795 + }, + { + "epoch": 0.07, + "learning_rate": 9.864437346190384e-05, + "loss": 2.3855, + "step": 27800 + }, + { + "epoch": 0.07, + "learning_rate": 9.864388624641996e-05, + "loss": 2.371, + "step": 27805 + }, + { + "epoch": 0.07, + "learning_rate": 9.864339894460224e-05, + "loss": 2.3234, + "step": 27810 + }, + { + "epoch": 0.07, + "learning_rate": 9.864291155645153e-05, + "loss": 2.4418, + "step": 27815 + }, + { + "epoch": 0.07, + "learning_rate": 9.864242408196871e-05, + "loss": 2.396, + "step": 27820 + }, + { + "epoch": 0.07, + "learning_rate": 9.864193652115463e-05, + "loss": 2.5454, + "step": 27825 + }, + { + "epoch": 0.07, + "learning_rate": 9.864144887401017e-05, + "loss": 2.4106, + "step": 27830 + }, + { + "epoch": 0.07, + "learning_rate": 9.864096114053618e-05, + "loss": 2.3862, + "step": 27835 + }, + { + "epoch": 0.07, + "learning_rate": 9.864047332073356e-05, + "loss": 2.3959, + "step": 27840 + }, + { + "epoch": 0.07, + "learning_rate": 9.863998541460314e-05, + "loss": 2.1094, + "step": 27845 + }, + { + "epoch": 0.07, + "learning_rate": 9.863949742214579e-05, + "loss": 2.4022, + "step": 27850 + }, + { + "epoch": 0.07, + "learning_rate": 9.863900934336239e-05, + "loss": 2.39, + "step": 27855 + }, + { + "epoch": 0.07, + "learning_rate": 9.863852117825378e-05, + "loss": 2.3414, + "step": 27860 + }, + { + "epoch": 0.07, + "learning_rate": 9.863803292682088e-05, + "loss": 2.506, + "step": 27865 + }, + { + "epoch": 0.07, + "learning_rate": 9.86375445890645e-05, + "loss": 2.4859, + "step": 27870 + }, + { + "epoch": 0.07, + "learning_rate": 9.863705616498554e-05, + "loss": 2.3998, + "step": 27875 + }, + { + "epoch": 0.07, + "learning_rate": 9.863656765458486e-05, + "loss": 2.3938, + "step": 27880 + }, + { + "epoch": 0.07, + "learning_rate": 9.863607905786332e-05, + "loss": 2.2917, + "step": 27885 + }, + { + "epoch": 0.07, + "learning_rate": 9.863559037482178e-05, + "loss": 2.3783, + "step": 27890 + }, + { + "epoch": 0.07, + "learning_rate": 9.863510160546113e-05, + "loss": 2.2851, + "step": 27895 + }, + { + "epoch": 0.07, + "learning_rate": 9.86346127497822e-05, + "loss": 2.4264, + "step": 27900 + }, + { + "epoch": 0.07, + "learning_rate": 9.863412380778591e-05, + "loss": 2.397, + "step": 27905 + }, + { + "epoch": 0.07, + "learning_rate": 9.863363477947309e-05, + "loss": 2.3895, + "step": 27910 + }, + { + "epoch": 0.07, + "learning_rate": 9.863314566484462e-05, + "loss": 2.4197, + "step": 27915 + }, + { + "epoch": 0.07, + "learning_rate": 9.863265646390139e-05, + "loss": 2.4479, + "step": 27920 + }, + { + "epoch": 0.07, + "learning_rate": 9.863216717664422e-05, + "loss": 2.3317, + "step": 27925 + }, + { + "epoch": 0.07, + "learning_rate": 9.8631677803074e-05, + "loss": 2.3301, + "step": 27930 + }, + { + "epoch": 0.07, + "learning_rate": 9.863118834319161e-05, + "loss": 2.4347, + "step": 27935 + }, + { + "epoch": 0.07, + "learning_rate": 9.863069879699792e-05, + "loss": 2.3262, + "step": 27940 + }, + { + "epoch": 0.07, + "learning_rate": 9.863020916449379e-05, + "loss": 2.3866, + "step": 27945 + }, + { + "epoch": 0.07, + "learning_rate": 9.862971944568007e-05, + "loss": 2.4501, + "step": 27950 + }, + { + "epoch": 0.07, + "learning_rate": 9.862922964055765e-05, + "loss": 2.5454, + "step": 27955 + }, + { + "epoch": 0.07, + "learning_rate": 9.862873974912741e-05, + "loss": 2.394, + "step": 27960 + }, + { + "epoch": 0.07, + "learning_rate": 9.862824977139021e-05, + "loss": 2.4937, + "step": 27965 + }, + { + "epoch": 0.07, + "learning_rate": 9.86277597073469e-05, + "loss": 2.3485, + "step": 27970 + }, + { + "epoch": 0.08, + "learning_rate": 9.862726955699838e-05, + "loss": 2.3213, + "step": 27975 + }, + { + "epoch": 0.08, + "learning_rate": 9.86267793203455e-05, + "loss": 2.3419, + "step": 27980 + }, + { + "epoch": 0.08, + "learning_rate": 9.862628899738913e-05, + "loss": 2.385, + "step": 27985 + }, + { + "epoch": 0.08, + "learning_rate": 9.862579858813015e-05, + "loss": 2.2933, + "step": 27990 + }, + { + "epoch": 0.08, + "learning_rate": 9.862530809256943e-05, + "loss": 2.5207, + "step": 27995 + }, + { + "epoch": 0.08, + "learning_rate": 9.862481751070782e-05, + "loss": 2.4057, + "step": 28000 + }, + { + "epoch": 0.08, + "learning_rate": 9.862432684254623e-05, + "loss": 2.4259, + "step": 28005 + }, + { + "epoch": 0.08, + "learning_rate": 9.862383608808549e-05, + "loss": 2.4977, + "step": 28010 + }, + { + "epoch": 0.08, + "learning_rate": 9.86233452473265e-05, + "loss": 2.4448, + "step": 28015 + }, + { + "epoch": 0.08, + "learning_rate": 9.862285432027013e-05, + "loss": 2.3822, + "step": 28020 + }, + { + "epoch": 0.08, + "learning_rate": 9.862236330691723e-05, + "loss": 2.4801, + "step": 28025 + }, + { + "epoch": 0.08, + "learning_rate": 9.862187220726868e-05, + "loss": 2.3741, + "step": 28030 + }, + { + "epoch": 0.08, + "learning_rate": 9.862138102132536e-05, + "loss": 2.4138, + "step": 28035 + }, + { + "epoch": 0.08, + "learning_rate": 9.862088974908814e-05, + "loss": 2.3685, + "step": 28040 + }, + { + "epoch": 0.08, + "learning_rate": 9.86203983905579e-05, + "loss": 2.3549, + "step": 28045 + }, + { + "epoch": 0.08, + "learning_rate": 9.861990694573548e-05, + "loss": 2.4552, + "step": 28050 + }, + { + "epoch": 0.08, + "learning_rate": 9.861941541462178e-05, + "loss": 2.4792, + "step": 28055 + }, + { + "epoch": 0.08, + "learning_rate": 9.861892379721766e-05, + "loss": 2.378, + "step": 28060 + }, + { + "epoch": 0.08, + "learning_rate": 9.861843209352402e-05, + "loss": 2.5066, + "step": 28065 + }, + { + "epoch": 0.08, + "learning_rate": 9.86179403035417e-05, + "loss": 2.4224, + "step": 28070 + }, + { + "epoch": 0.08, + "learning_rate": 9.861744842727156e-05, + "loss": 2.3953, + "step": 28075 + }, + { + "epoch": 0.08, + "learning_rate": 9.861695646471452e-05, + "loss": 2.3549, + "step": 28080 + }, + { + "epoch": 0.08, + "learning_rate": 9.861646441587142e-05, + "loss": 2.3719, + "step": 28085 + }, + { + "epoch": 0.08, + "learning_rate": 9.861597228074316e-05, + "loss": 2.4597, + "step": 28090 + }, + { + "epoch": 0.08, + "learning_rate": 9.861548005933059e-05, + "loss": 2.4348, + "step": 28095 + }, + { + "epoch": 0.08, + "learning_rate": 9.861498775163458e-05, + "loss": 2.4002, + "step": 28100 + }, + { + "epoch": 0.08, + "learning_rate": 9.8614495357656e-05, + "loss": 2.3938, + "step": 28105 + }, + { + "epoch": 0.08, + "learning_rate": 9.861400287739578e-05, + "loss": 2.5002, + "step": 28110 + }, + { + "epoch": 0.08, + "learning_rate": 9.861351031085472e-05, + "loss": 2.379, + "step": 28115 + }, + { + "epoch": 0.08, + "learning_rate": 9.861301765803374e-05, + "loss": 2.4031, + "step": 28120 + }, + { + "epoch": 0.08, + "learning_rate": 9.861252491893369e-05, + "loss": 2.3903, + "step": 28125 + }, + { + "epoch": 0.08, + "learning_rate": 9.861203209355544e-05, + "loss": 2.373, + "step": 28130 + }, + { + "epoch": 0.08, + "learning_rate": 9.861153918189991e-05, + "loss": 2.3591, + "step": 28135 + }, + { + "epoch": 0.08, + "learning_rate": 9.861104618396794e-05, + "loss": 2.4965, + "step": 28140 + }, + { + "epoch": 0.08, + "learning_rate": 9.861055309976039e-05, + "loss": 2.3875, + "step": 28145 + }, + { + "epoch": 0.08, + "learning_rate": 9.861005992927817e-05, + "loss": 2.3563, + "step": 28150 + }, + { + "epoch": 0.08, + "learning_rate": 9.860956667252213e-05, + "loss": 2.4096, + "step": 28155 + }, + { + "epoch": 0.08, + "learning_rate": 9.860907332949316e-05, + "loss": 2.3146, + "step": 28160 + }, + { + "epoch": 0.08, + "learning_rate": 9.860857990019214e-05, + "loss": 2.3417, + "step": 28165 + }, + { + "epoch": 0.08, + "learning_rate": 9.860808638461993e-05, + "loss": 2.3459, + "step": 28170 + }, + { + "epoch": 0.08, + "learning_rate": 9.86075927827774e-05, + "loss": 2.3851, + "step": 28175 + }, + { + "epoch": 0.08, + "learning_rate": 9.860709909466546e-05, + "loss": 2.434, + "step": 28180 + }, + { + "epoch": 0.08, + "learning_rate": 9.860660532028496e-05, + "loss": 2.2856, + "step": 28185 + }, + { + "epoch": 0.08, + "learning_rate": 9.860611145963677e-05, + "loss": 2.4541, + "step": 28190 + }, + { + "epoch": 0.08, + "learning_rate": 9.860561751272177e-05, + "loss": 2.4311, + "step": 28195 + }, + { + "epoch": 0.08, + "learning_rate": 9.860512347954088e-05, + "loss": 2.5552, + "step": 28200 + }, + { + "epoch": 0.08, + "learning_rate": 9.86046293600949e-05, + "loss": 2.4227, + "step": 28205 + }, + { + "epoch": 0.08, + "learning_rate": 9.860413515438477e-05, + "loss": 2.3635, + "step": 28210 + }, + { + "epoch": 0.08, + "learning_rate": 9.860364086241133e-05, + "loss": 2.337, + "step": 28215 + }, + { + "epoch": 0.08, + "learning_rate": 9.860314648417549e-05, + "loss": 2.4689, + "step": 28220 + }, + { + "epoch": 0.08, + "learning_rate": 9.86026520196781e-05, + "loss": 2.371, + "step": 28225 + }, + { + "epoch": 0.08, + "learning_rate": 9.860215746892005e-05, + "loss": 2.5362, + "step": 28230 + }, + { + "epoch": 0.08, + "learning_rate": 9.860166283190223e-05, + "loss": 2.4527, + "step": 28235 + }, + { + "epoch": 0.08, + "learning_rate": 9.860116810862548e-05, + "loss": 2.3733, + "step": 28240 + }, + { + "epoch": 0.08, + "learning_rate": 9.86006732990907e-05, + "loss": 2.3342, + "step": 28245 + }, + { + "epoch": 0.08, + "learning_rate": 9.860017840329878e-05, + "loss": 2.2617, + "step": 28250 + }, + { + "epoch": 0.08, + "learning_rate": 9.859968342125059e-05, + "loss": 2.345, + "step": 28255 + }, + { + "epoch": 0.08, + "learning_rate": 9.8599188352947e-05, + "loss": 2.4048, + "step": 28260 + }, + { + "epoch": 0.08, + "learning_rate": 9.85986931983889e-05, + "loss": 2.3659, + "step": 28265 + }, + { + "epoch": 0.08, + "learning_rate": 9.859819795757716e-05, + "loss": 2.2893, + "step": 28270 + }, + { + "epoch": 0.08, + "learning_rate": 9.859770263051267e-05, + "loss": 2.5235, + "step": 28275 + }, + { + "epoch": 0.08, + "learning_rate": 9.859720721719629e-05, + "loss": 2.3616, + "step": 28280 + }, + { + "epoch": 0.08, + "learning_rate": 9.859671171762891e-05, + "loss": 2.3895, + "step": 28285 + }, + { + "epoch": 0.08, + "learning_rate": 9.859621613181141e-05, + "loss": 2.4328, + "step": 28290 + }, + { + "epoch": 0.08, + "learning_rate": 9.859572045974468e-05, + "loss": 2.3531, + "step": 28295 + }, + { + "epoch": 0.08, + "learning_rate": 9.859522470142958e-05, + "loss": 2.2644, + "step": 28300 + }, + { + "epoch": 0.08, + "learning_rate": 9.8594728856867e-05, + "loss": 2.3947, + "step": 28305 + }, + { + "epoch": 0.08, + "learning_rate": 9.859423292605783e-05, + "loss": 2.3069, + "step": 28310 + }, + { + "epoch": 0.08, + "learning_rate": 9.859373690900293e-05, + "loss": 2.4206, + "step": 28315 + }, + { + "epoch": 0.08, + "learning_rate": 9.85932408057032e-05, + "loss": 2.389, + "step": 28320 + }, + { + "epoch": 0.08, + "learning_rate": 9.859274461615951e-05, + "loss": 2.3736, + "step": 28325 + }, + { + "epoch": 0.08, + "learning_rate": 9.859224834037273e-05, + "loss": 2.2338, + "step": 28330 + }, + { + "epoch": 0.08, + "learning_rate": 9.859175197834376e-05, + "loss": 2.4578, + "step": 28335 + }, + { + "epoch": 0.08, + "learning_rate": 9.859125553007347e-05, + "loss": 2.326, + "step": 28340 + }, + { + "epoch": 0.08, + "learning_rate": 9.859075899556273e-05, + "loss": 2.4544, + "step": 28345 + }, + { + "epoch": 0.08, + "learning_rate": 9.859026237481246e-05, + "loss": 2.449, + "step": 28350 + }, + { + "epoch": 0.08, + "learning_rate": 9.858976566782351e-05, + "loss": 2.3673, + "step": 28355 + }, + { + "epoch": 0.08, + "learning_rate": 9.858926887459677e-05, + "loss": 2.3384, + "step": 28360 + }, + { + "epoch": 0.08, + "learning_rate": 9.858877199513312e-05, + "loss": 2.382, + "step": 28365 + }, + { + "epoch": 0.08, + "learning_rate": 9.858827502943342e-05, + "loss": 2.3557, + "step": 28370 + }, + { + "epoch": 0.08, + "learning_rate": 9.85877779774986e-05, + "loss": 2.395, + "step": 28375 + }, + { + "epoch": 0.08, + "learning_rate": 9.85872808393295e-05, + "loss": 2.3166, + "step": 28380 + }, + { + "epoch": 0.08, + "learning_rate": 9.858678361492702e-05, + "loss": 2.3441, + "step": 28385 + }, + { + "epoch": 0.08, + "learning_rate": 9.858628630429205e-05, + "loss": 2.365, + "step": 28390 + }, + { + "epoch": 0.08, + "learning_rate": 9.858578890742546e-05, + "loss": 2.3848, + "step": 28395 + }, + { + "epoch": 0.08, + "learning_rate": 9.858529142432813e-05, + "loss": 2.3963, + "step": 28400 + }, + { + "epoch": 0.08, + "learning_rate": 9.858479385500097e-05, + "loss": 2.3168, + "step": 28405 + }, + { + "epoch": 0.08, + "learning_rate": 9.858429619944482e-05, + "loss": 2.4015, + "step": 28410 + }, + { + "epoch": 0.08, + "learning_rate": 9.858379845766059e-05, + "loss": 2.4247, + "step": 28415 + }, + { + "epoch": 0.08, + "learning_rate": 9.858330062964916e-05, + "loss": 2.4324, + "step": 28420 + }, + { + "epoch": 0.08, + "learning_rate": 9.858280271541142e-05, + "loss": 2.3196, + "step": 28425 + }, + { + "epoch": 0.08, + "learning_rate": 9.858230471494823e-05, + "loss": 2.4722, + "step": 28430 + }, + { + "epoch": 0.08, + "learning_rate": 9.858180662826051e-05, + "loss": 2.3279, + "step": 28435 + }, + { + "epoch": 0.08, + "learning_rate": 9.858130845534911e-05, + "loss": 2.4463, + "step": 28440 + }, + { + "epoch": 0.08, + "learning_rate": 9.858081019621492e-05, + "loss": 2.3601, + "step": 28445 + }, + { + "epoch": 0.08, + "learning_rate": 9.858031185085885e-05, + "loss": 2.3837, + "step": 28450 + }, + { + "epoch": 0.08, + "learning_rate": 9.857981341928178e-05, + "loss": 2.5022, + "step": 28455 + }, + { + "epoch": 0.08, + "learning_rate": 9.857931490148456e-05, + "loss": 2.3391, + "step": 28460 + }, + { + "epoch": 0.08, + "learning_rate": 9.857881629746809e-05, + "loss": 2.3805, + "step": 28465 + }, + { + "epoch": 0.08, + "learning_rate": 9.857831760723327e-05, + "loss": 2.4499, + "step": 28470 + }, + { + "epoch": 0.08, + "learning_rate": 9.857781883078097e-05, + "loss": 2.3427, + "step": 28475 + }, + { + "epoch": 0.08, + "learning_rate": 9.85773199681121e-05, + "loss": 2.3542, + "step": 28480 + }, + { + "epoch": 0.08, + "learning_rate": 9.857682101922751e-05, + "loss": 2.3442, + "step": 28485 + }, + { + "epoch": 0.08, + "learning_rate": 9.857632198412812e-05, + "loss": 2.3595, + "step": 28490 + }, + { + "epoch": 0.08, + "learning_rate": 9.857582286281478e-05, + "loss": 2.3778, + "step": 28495 + }, + { + "epoch": 0.08, + "learning_rate": 9.857532365528841e-05, + "loss": 2.3868, + "step": 28500 + }, + { + "epoch": 0.08, + "learning_rate": 9.857482436154986e-05, + "loss": 2.4639, + "step": 28505 + }, + { + "epoch": 0.08, + "learning_rate": 9.857432498160006e-05, + "loss": 2.3953, + "step": 28510 + }, + { + "epoch": 0.08, + "learning_rate": 9.857382551543985e-05, + "loss": 2.4454, + "step": 28515 + }, + { + "epoch": 0.08, + "learning_rate": 9.857332596307015e-05, + "loss": 2.428, + "step": 28520 + }, + { + "epoch": 0.08, + "learning_rate": 9.857282632449183e-05, + "loss": 2.293, + "step": 28525 + }, + { + "epoch": 0.08, + "learning_rate": 9.85723265997058e-05, + "loss": 2.2972, + "step": 28530 + }, + { + "epoch": 0.08, + "learning_rate": 9.857182678871291e-05, + "loss": 2.398, + "step": 28535 + }, + { + "epoch": 0.08, + "learning_rate": 9.857132689151408e-05, + "loss": 2.4104, + "step": 28540 + }, + { + "epoch": 0.08, + "learning_rate": 9.857082690811017e-05, + "loss": 2.3759, + "step": 28545 + }, + { + "epoch": 0.08, + "learning_rate": 9.857032683850209e-05, + "loss": 2.2991, + "step": 28550 + }, + { + "epoch": 0.08, + "learning_rate": 9.856982668269071e-05, + "loss": 2.4535, + "step": 28555 + }, + { + "epoch": 0.08, + "learning_rate": 9.856932644067694e-05, + "loss": 2.536, + "step": 28560 + }, + { + "epoch": 0.08, + "learning_rate": 9.856882611246164e-05, + "loss": 2.4682, + "step": 28565 + }, + { + "epoch": 0.08, + "learning_rate": 9.856832569804573e-05, + "loss": 2.4784, + "step": 28570 + }, + { + "epoch": 0.08, + "learning_rate": 9.856782519743006e-05, + "loss": 2.479, + "step": 28575 + }, + { + "epoch": 0.08, + "learning_rate": 9.856732461061555e-05, + "loss": 2.4172, + "step": 28580 + }, + { + "epoch": 0.08, + "learning_rate": 9.856682393760309e-05, + "loss": 2.364, + "step": 28585 + }, + { + "epoch": 0.08, + "learning_rate": 9.856632317839353e-05, + "loss": 2.3591, + "step": 28590 + }, + { + "epoch": 0.08, + "learning_rate": 9.856582233298779e-05, + "loss": 2.407, + "step": 28595 + }, + { + "epoch": 0.08, + "learning_rate": 9.856532140138676e-05, + "loss": 2.3976, + "step": 28600 + }, + { + "epoch": 0.08, + "learning_rate": 9.856482038359131e-05, + "loss": 2.2664, + "step": 28605 + }, + { + "epoch": 0.08, + "learning_rate": 9.856431927960236e-05, + "loss": 2.3756, + "step": 28610 + }, + { + "epoch": 0.08, + "learning_rate": 9.856381808942076e-05, + "loss": 2.4832, + "step": 28615 + }, + { + "epoch": 0.08, + "learning_rate": 9.856331681304745e-05, + "loss": 2.3851, + "step": 28620 + }, + { + "epoch": 0.08, + "learning_rate": 9.856281545048327e-05, + "loss": 2.4511, + "step": 28625 + }, + { + "epoch": 0.08, + "learning_rate": 9.856231400172913e-05, + "loss": 2.3043, + "step": 28630 + }, + { + "epoch": 0.08, + "learning_rate": 9.856181246678591e-05, + "loss": 2.3376, + "step": 28635 + }, + { + "epoch": 0.08, + "learning_rate": 9.856131084565454e-05, + "loss": 2.3419, + "step": 28640 + }, + { + "epoch": 0.08, + "learning_rate": 9.856080913833585e-05, + "loss": 2.3407, + "step": 28645 + }, + { + "epoch": 0.08, + "learning_rate": 9.856030734483077e-05, + "loss": 2.3441, + "step": 28650 + }, + { + "epoch": 0.08, + "learning_rate": 9.855980546514019e-05, + "loss": 2.3311, + "step": 28655 + }, + { + "epoch": 0.08, + "learning_rate": 9.855930349926499e-05, + "loss": 2.377, + "step": 28660 + }, + { + "epoch": 0.08, + "learning_rate": 9.855880144720606e-05, + "loss": 2.3942, + "step": 28665 + }, + { + "epoch": 0.08, + "learning_rate": 9.855829930896429e-05, + "loss": 2.3994, + "step": 28670 + }, + { + "epoch": 0.08, + "learning_rate": 9.855779708454058e-05, + "loss": 2.3088, + "step": 28675 + }, + { + "epoch": 0.08, + "learning_rate": 9.855729477393581e-05, + "loss": 2.4915, + "step": 28680 + }, + { + "epoch": 0.08, + "learning_rate": 9.85567923771509e-05, + "loss": 2.4472, + "step": 28685 + }, + { + "epoch": 0.08, + "learning_rate": 9.85562898941867e-05, + "loss": 2.4129, + "step": 28690 + }, + { + "epoch": 0.08, + "learning_rate": 9.855578732504412e-05, + "loss": 2.3316, + "step": 28695 + }, + { + "epoch": 0.08, + "learning_rate": 9.855528466972405e-05, + "loss": 2.3918, + "step": 28700 + }, + { + "epoch": 0.08, + "learning_rate": 9.85547819282274e-05, + "loss": 2.3064, + "step": 28705 + }, + { + "epoch": 0.08, + "learning_rate": 9.855427910055504e-05, + "loss": 2.3519, + "step": 28710 + }, + { + "epoch": 0.08, + "learning_rate": 9.855377618670787e-05, + "loss": 2.2739, + "step": 28715 + }, + { + "epoch": 0.08, + "learning_rate": 9.855327318668679e-05, + "loss": 2.3112, + "step": 28720 + }, + { + "epoch": 0.08, + "learning_rate": 9.855277010049268e-05, + "loss": 2.2548, + "step": 28725 + }, + { + "epoch": 0.08, + "learning_rate": 9.855226692812644e-05, + "loss": 2.2873, + "step": 28730 + }, + { + "epoch": 0.08, + "learning_rate": 9.855176366958896e-05, + "loss": 2.34, + "step": 28735 + }, + { + "epoch": 0.08, + "learning_rate": 9.855126032488113e-05, + "loss": 2.3672, + "step": 28740 + }, + { + "epoch": 0.08, + "learning_rate": 9.855075689400386e-05, + "loss": 2.3489, + "step": 28745 + }, + { + "epoch": 0.08, + "learning_rate": 9.855025337695803e-05, + "loss": 2.3882, + "step": 28750 + }, + { + "epoch": 0.08, + "learning_rate": 9.854974977374451e-05, + "loss": 2.321, + "step": 28755 + }, + { + "epoch": 0.08, + "learning_rate": 9.854924608436424e-05, + "loss": 2.3164, + "step": 28760 + }, + { + "epoch": 0.08, + "learning_rate": 9.85487423088181e-05, + "loss": 2.3069, + "step": 28765 + }, + { + "epoch": 0.08, + "learning_rate": 9.854823844710697e-05, + "loss": 2.5145, + "step": 28770 + }, + { + "epoch": 0.08, + "learning_rate": 9.854773449923176e-05, + "loss": 2.3322, + "step": 28775 + }, + { + "epoch": 0.08, + "learning_rate": 9.854723046519335e-05, + "loss": 2.5423, + "step": 28780 + }, + { + "epoch": 0.08, + "learning_rate": 9.854672634499262e-05, + "loss": 2.3689, + "step": 28785 + }, + { + "epoch": 0.08, + "learning_rate": 9.85462221386305e-05, + "loss": 2.3449, + "step": 28790 + }, + { + "epoch": 0.08, + "learning_rate": 9.854571784610787e-05, + "loss": 2.3899, + "step": 28795 + }, + { + "epoch": 0.08, + "learning_rate": 9.854521346742563e-05, + "loss": 2.3697, + "step": 28800 + }, + { + "epoch": 0.08, + "learning_rate": 9.854470900258468e-05, + "loss": 2.4779, + "step": 28805 + }, + { + "epoch": 0.08, + "learning_rate": 9.854420445158588e-05, + "loss": 2.4352, + "step": 28810 + }, + { + "epoch": 0.08, + "learning_rate": 9.854369981443017e-05, + "loss": 2.3371, + "step": 28815 + }, + { + "epoch": 0.08, + "learning_rate": 9.854319509111841e-05, + "loss": 2.2933, + "step": 28820 + }, + { + "epoch": 0.08, + "learning_rate": 9.854269028165153e-05, + "loss": 2.3956, + "step": 28825 + }, + { + "epoch": 0.08, + "learning_rate": 9.85421853860304e-05, + "loss": 2.4034, + "step": 28830 + }, + { + "epoch": 0.08, + "learning_rate": 9.854168040425592e-05, + "loss": 2.3193, + "step": 28835 + }, + { + "epoch": 0.08, + "learning_rate": 9.8541175336329e-05, + "loss": 2.4178, + "step": 28840 + }, + { + "epoch": 0.08, + "learning_rate": 9.854067018225052e-05, + "loss": 2.4156, + "step": 28845 + }, + { + "epoch": 0.08, + "learning_rate": 9.854016494202139e-05, + "loss": 2.4092, + "step": 28850 + }, + { + "epoch": 0.08, + "learning_rate": 9.853965961564249e-05, + "loss": 2.4426, + "step": 28855 + }, + { + "epoch": 0.08, + "learning_rate": 9.853915420311474e-05, + "loss": 2.3286, + "step": 28860 + }, + { + "epoch": 0.08, + "learning_rate": 9.853864870443901e-05, + "loss": 2.3603, + "step": 28865 + }, + { + "epoch": 0.08, + "learning_rate": 9.853814311961623e-05, + "loss": 2.3873, + "step": 28870 + }, + { + "epoch": 0.08, + "learning_rate": 9.853763744864726e-05, + "loss": 2.2976, + "step": 28875 + }, + { + "epoch": 0.08, + "learning_rate": 9.853713169153303e-05, + "loss": 2.3691, + "step": 28880 + }, + { + "epoch": 0.08, + "learning_rate": 9.853662584827442e-05, + "loss": 2.4666, + "step": 28885 + }, + { + "epoch": 0.08, + "learning_rate": 9.853611991887231e-05, + "loss": 2.4055, + "step": 28890 + }, + { + "epoch": 0.08, + "learning_rate": 9.853561390332765e-05, + "loss": 2.5536, + "step": 28895 + }, + { + "epoch": 0.08, + "learning_rate": 9.853510780164129e-05, + "loss": 2.4149, + "step": 28900 + }, + { + "epoch": 0.08, + "learning_rate": 9.853460161381416e-05, + "loss": 2.244, + "step": 28905 + }, + { + "epoch": 0.08, + "learning_rate": 9.853409533984712e-05, + "loss": 2.4944, + "step": 28910 + }, + { + "epoch": 0.08, + "learning_rate": 9.853358897974112e-05, + "loss": 2.1401, + "step": 28915 + }, + { + "epoch": 0.08, + "learning_rate": 9.853308253349702e-05, + "loss": 2.3302, + "step": 28920 + }, + { + "epoch": 0.08, + "learning_rate": 9.853257600111573e-05, + "loss": 2.3635, + "step": 28925 + }, + { + "epoch": 0.08, + "learning_rate": 9.853206938259815e-05, + "loss": 2.3568, + "step": 28930 + }, + { + "epoch": 0.08, + "learning_rate": 9.853156267794518e-05, + "loss": 2.3388, + "step": 28935 + }, + { + "epoch": 0.08, + "learning_rate": 9.853105588715772e-05, + "loss": 2.3812, + "step": 28940 + }, + { + "epoch": 0.08, + "learning_rate": 9.853054901023667e-05, + "loss": 2.3556, + "step": 28945 + }, + { + "epoch": 0.08, + "learning_rate": 9.853004204718291e-05, + "loss": 2.3947, + "step": 28950 + }, + { + "epoch": 0.08, + "learning_rate": 9.852953499799735e-05, + "loss": 2.3368, + "step": 28955 + }, + { + "epoch": 0.08, + "learning_rate": 9.852902786268092e-05, + "loss": 2.2333, + "step": 28960 + }, + { + "epoch": 0.08, + "learning_rate": 9.852852064123449e-05, + "loss": 2.4164, + "step": 28965 + }, + { + "epoch": 0.08, + "learning_rate": 9.852801333365898e-05, + "loss": 2.2589, + "step": 28970 + }, + { + "epoch": 0.08, + "learning_rate": 9.852750593995526e-05, + "loss": 2.3612, + "step": 28975 + }, + { + "epoch": 0.08, + "learning_rate": 9.852699846012424e-05, + "loss": 2.4504, + "step": 28980 + }, + { + "epoch": 0.08, + "learning_rate": 9.852649089416684e-05, + "loss": 2.4779, + "step": 28985 + }, + { + "epoch": 0.08, + "learning_rate": 9.852598324208395e-05, + "loss": 2.4394, + "step": 28990 + }, + { + "epoch": 0.08, + "learning_rate": 9.852547550387648e-05, + "loss": 2.3327, + "step": 28995 + }, + { + "epoch": 0.08, + "learning_rate": 9.852496767954529e-05, + "loss": 2.3117, + "step": 29000 + }, + { + "epoch": 0.08, + "learning_rate": 9.852445976909135e-05, + "loss": 2.2704, + "step": 29005 + }, + { + "epoch": 0.08, + "learning_rate": 9.85239517725155e-05, + "loss": 2.4113, + "step": 29010 + }, + { + "epoch": 0.08, + "learning_rate": 9.852344368981867e-05, + "loss": 2.428, + "step": 29015 + }, + { + "epoch": 0.08, + "learning_rate": 9.852293552100176e-05, + "loss": 2.2139, + "step": 29020 + }, + { + "epoch": 0.08, + "learning_rate": 9.852242726606567e-05, + "loss": 2.3744, + "step": 29025 + }, + { + "epoch": 0.08, + "learning_rate": 9.852191892501131e-05, + "loss": 2.4223, + "step": 29030 + }, + { + "epoch": 0.08, + "learning_rate": 9.852141049783956e-05, + "loss": 2.3086, + "step": 29035 + }, + { + "epoch": 0.08, + "learning_rate": 9.852090198455132e-05, + "loss": 2.4235, + "step": 29040 + }, + { + "epoch": 0.08, + "learning_rate": 9.852039338514753e-05, + "loss": 2.395, + "step": 29045 + }, + { + "epoch": 0.08, + "learning_rate": 9.851988469962909e-05, + "loss": 2.2828, + "step": 29050 + }, + { + "epoch": 0.08, + "learning_rate": 9.851937592799684e-05, + "loss": 2.3575, + "step": 29055 + }, + { + "epoch": 0.08, + "learning_rate": 9.851886707025177e-05, + "loss": 2.3011, + "step": 29060 + }, + { + "epoch": 0.08, + "learning_rate": 9.851835812639472e-05, + "loss": 2.2417, + "step": 29065 + }, + { + "epoch": 0.08, + "learning_rate": 9.851784909642661e-05, + "loss": 2.526, + "step": 29070 + }, + { + "epoch": 0.08, + "learning_rate": 9.851733998034835e-05, + "loss": 2.4002, + "step": 29075 + }, + { + "epoch": 0.08, + "learning_rate": 9.851683077816085e-05, + "loss": 2.4431, + "step": 29080 + }, + { + "epoch": 0.08, + "learning_rate": 9.851632148986498e-05, + "loss": 2.4299, + "step": 29085 + }, + { + "epoch": 0.08, + "learning_rate": 9.851581211546169e-05, + "loss": 2.3576, + "step": 29090 + }, + { + "epoch": 0.08, + "learning_rate": 9.851530265495186e-05, + "loss": 2.3488, + "step": 29095 + }, + { + "epoch": 0.08, + "learning_rate": 9.851479310833638e-05, + "loss": 2.352, + "step": 29100 + }, + { + "epoch": 0.08, + "learning_rate": 9.85142834756162e-05, + "loss": 2.3395, + "step": 29105 + }, + { + "epoch": 0.08, + "learning_rate": 9.851377375679216e-05, + "loss": 2.3952, + "step": 29110 + }, + { + "epoch": 0.08, + "learning_rate": 9.851326395186522e-05, + "loss": 2.4069, + "step": 29115 + }, + { + "epoch": 0.08, + "learning_rate": 9.851275406083627e-05, + "loss": 2.2986, + "step": 29120 + }, + { + "epoch": 0.08, + "learning_rate": 9.851224408370619e-05, + "loss": 2.343, + "step": 29125 + }, + { + "epoch": 0.08, + "learning_rate": 9.851173402047592e-05, + "loss": 2.368, + "step": 29130 + }, + { + "epoch": 0.08, + "learning_rate": 9.851122387114636e-05, + "loss": 2.4655, + "step": 29135 + }, + { + "epoch": 0.08, + "learning_rate": 9.851071363571838e-05, + "loss": 2.3869, + "step": 29140 + }, + { + "epoch": 0.08, + "learning_rate": 9.851020331419292e-05, + "loss": 2.4619, + "step": 29145 + }, + { + "epoch": 0.08, + "learning_rate": 9.850969290657088e-05, + "loss": 2.4253, + "step": 29150 + }, + { + "epoch": 0.08, + "learning_rate": 9.850918241285316e-05, + "loss": 2.3768, + "step": 29155 + }, + { + "epoch": 0.08, + "learning_rate": 9.850867183304067e-05, + "loss": 2.3249, + "step": 29160 + }, + { + "epoch": 0.08, + "learning_rate": 9.850816116713431e-05, + "loss": 2.4328, + "step": 29165 + }, + { + "epoch": 0.08, + "learning_rate": 9.8507650415135e-05, + "loss": 2.2322, + "step": 29170 + }, + { + "epoch": 0.08, + "learning_rate": 9.850713957704363e-05, + "loss": 2.5296, + "step": 29175 + }, + { + "epoch": 0.08, + "learning_rate": 9.850662865286113e-05, + "loss": 2.3944, + "step": 29180 + }, + { + "epoch": 0.08, + "learning_rate": 9.850611764258837e-05, + "loss": 2.3671, + "step": 29185 + }, + { + "epoch": 0.08, + "learning_rate": 9.850560654622628e-05, + "loss": 2.3794, + "step": 29190 + }, + { + "epoch": 0.08, + "learning_rate": 9.850509536377576e-05, + "loss": 2.3224, + "step": 29195 + }, + { + "epoch": 0.08, + "learning_rate": 9.850458409523776e-05, + "loss": 2.3293, + "step": 29200 + }, + { + "epoch": 0.08, + "learning_rate": 9.850407274061311e-05, + "loss": 2.4905, + "step": 29205 + }, + { + "epoch": 0.08, + "learning_rate": 9.850356129990277e-05, + "loss": 2.4199, + "step": 29210 + }, + { + "epoch": 0.08, + "learning_rate": 9.850304977310763e-05, + "loss": 2.3524, + "step": 29215 + }, + { + "epoch": 0.08, + "learning_rate": 9.85025381602286e-05, + "loss": 2.4628, + "step": 29220 + }, + { + "epoch": 0.08, + "learning_rate": 9.850202646126661e-05, + "loss": 2.3812, + "step": 29225 + }, + { + "epoch": 0.08, + "learning_rate": 9.850151467622254e-05, + "loss": 2.42, + "step": 29230 + }, + { + "epoch": 0.08, + "learning_rate": 9.850100280509731e-05, + "loss": 2.4446, + "step": 29235 + }, + { + "epoch": 0.08, + "learning_rate": 9.850049084789181e-05, + "loss": 2.4432, + "step": 29240 + }, + { + "epoch": 0.08, + "learning_rate": 9.849997880460698e-05, + "loss": 2.4165, + "step": 29245 + }, + { + "epoch": 0.08, + "learning_rate": 9.849946667524372e-05, + "loss": 2.4498, + "step": 29250 + }, + { + "epoch": 0.08, + "learning_rate": 9.84989544598029e-05, + "loss": 2.4005, + "step": 29255 + }, + { + "epoch": 0.08, + "learning_rate": 9.84984421582855e-05, + "loss": 2.3726, + "step": 29260 + }, + { + "epoch": 0.08, + "learning_rate": 9.849792977069237e-05, + "loss": 2.4065, + "step": 29265 + }, + { + "epoch": 0.08, + "learning_rate": 9.849741729702444e-05, + "loss": 2.3321, + "step": 29270 + }, + { + "epoch": 0.08, + "learning_rate": 9.849690473728261e-05, + "loss": 2.3421, + "step": 29275 + }, + { + "epoch": 0.08, + "learning_rate": 9.849639209146781e-05, + "loss": 2.3983, + "step": 29280 + }, + { + "epoch": 0.08, + "learning_rate": 9.849587935958096e-05, + "loss": 2.429, + "step": 29285 + }, + { + "epoch": 0.08, + "learning_rate": 9.849536654162293e-05, + "loss": 2.3916, + "step": 29290 + }, + { + "epoch": 0.08, + "learning_rate": 9.849485363759463e-05, + "loss": 2.3626, + "step": 29295 + }, + { + "epoch": 0.08, + "learning_rate": 9.849434064749701e-05, + "loss": 2.3743, + "step": 29300 + }, + { + "epoch": 0.08, + "learning_rate": 9.849382757133097e-05, + "loss": 2.3681, + "step": 29305 + }, + { + "epoch": 0.08, + "learning_rate": 9.84933144090974e-05, + "loss": 2.369, + "step": 29310 + }, + { + "epoch": 0.08, + "learning_rate": 9.849280116079721e-05, + "loss": 2.2589, + "step": 29315 + }, + { + "epoch": 0.08, + "learning_rate": 9.849228782643133e-05, + "loss": 2.4516, + "step": 29320 + }, + { + "epoch": 0.08, + "learning_rate": 9.849177440600066e-05, + "loss": 2.3911, + "step": 29325 + }, + { + "epoch": 0.08, + "learning_rate": 9.849126089950613e-05, + "loss": 2.1851, + "step": 29330 + }, + { + "epoch": 0.08, + "learning_rate": 9.849074730694862e-05, + "loss": 2.3311, + "step": 29335 + }, + { + "epoch": 0.08, + "learning_rate": 9.849023362832906e-05, + "loss": 2.3241, + "step": 29340 + }, + { + "epoch": 0.08, + "learning_rate": 9.848971986364837e-05, + "loss": 2.359, + "step": 29345 + }, + { + "epoch": 0.08, + "learning_rate": 9.848920601290743e-05, + "loss": 2.2797, + "step": 29350 + }, + { + "epoch": 0.08, + "learning_rate": 9.84886920761072e-05, + "loss": 2.3545, + "step": 29355 + }, + { + "epoch": 0.08, + "learning_rate": 9.848817805324855e-05, + "loss": 2.3774, + "step": 29360 + }, + { + "epoch": 0.08, + "learning_rate": 9.84876639443324e-05, + "loss": 2.3438, + "step": 29365 + }, + { + "epoch": 0.08, + "learning_rate": 9.848714974935969e-05, + "loss": 2.2963, + "step": 29370 + }, + { + "epoch": 0.08, + "learning_rate": 9.84866354683313e-05, + "loss": 2.3734, + "step": 29375 + }, + { + "epoch": 0.08, + "learning_rate": 9.848612110124813e-05, + "loss": 2.3921, + "step": 29380 + }, + { + "epoch": 0.08, + "learning_rate": 9.848560664811116e-05, + "loss": 2.4374, + "step": 29385 + }, + { + "epoch": 0.08, + "learning_rate": 9.848509210892124e-05, + "loss": 2.3372, + "step": 29390 + }, + { + "epoch": 0.08, + "learning_rate": 9.848457748367931e-05, + "loss": 2.2816, + "step": 29395 + }, + { + "epoch": 0.08, + "learning_rate": 9.848406277238628e-05, + "loss": 2.4036, + "step": 29400 + }, + { + "epoch": 0.08, + "learning_rate": 9.848354797504305e-05, + "loss": 2.3467, + "step": 29405 + }, + { + "epoch": 0.08, + "learning_rate": 9.848303309165055e-05, + "loss": 2.3566, + "step": 29410 + }, + { + "epoch": 0.08, + "learning_rate": 9.84825181222097e-05, + "loss": 2.3753, + "step": 29415 + }, + { + "epoch": 0.08, + "learning_rate": 9.84820030667214e-05, + "loss": 2.405, + "step": 29420 + }, + { + "epoch": 0.08, + "learning_rate": 9.848148792518655e-05, + "loss": 2.4209, + "step": 29425 + }, + { + "epoch": 0.08, + "learning_rate": 9.84809726976061e-05, + "loss": 2.303, + "step": 29430 + }, + { + "epoch": 0.08, + "learning_rate": 9.848045738398093e-05, + "loss": 2.3812, + "step": 29435 + }, + { + "epoch": 0.08, + "learning_rate": 9.847994198431198e-05, + "loss": 2.3485, + "step": 29440 + }, + { + "epoch": 0.08, + "learning_rate": 9.847942649860015e-05, + "loss": 2.4192, + "step": 29445 + }, + { + "epoch": 0.08, + "learning_rate": 9.847891092684635e-05, + "loss": 2.4058, + "step": 29450 + }, + { + "epoch": 0.08, + "learning_rate": 9.847839526905154e-05, + "loss": 2.3801, + "step": 29455 + }, + { + "epoch": 0.08, + "learning_rate": 9.847787952521656e-05, + "loss": 2.2841, + "step": 29460 + }, + { + "epoch": 0.08, + "learning_rate": 9.84773636953424e-05, + "loss": 2.3762, + "step": 29465 + }, + { + "epoch": 0.08, + "learning_rate": 9.847684777942992e-05, + "loss": 2.4444, + "step": 29470 + }, + { + "epoch": 0.08, + "learning_rate": 9.847633177748005e-05, + "loss": 2.3404, + "step": 29475 + }, + { + "epoch": 0.08, + "learning_rate": 9.847581568949373e-05, + "loss": 2.3772, + "step": 29480 + }, + { + "epoch": 0.08, + "learning_rate": 9.847529951547185e-05, + "loss": 2.3333, + "step": 29485 + }, + { + "epoch": 0.08, + "learning_rate": 9.847478325541533e-05, + "loss": 2.3578, + "step": 29490 + }, + { + "epoch": 0.08, + "learning_rate": 9.847426690932509e-05, + "loss": 2.3544, + "step": 29495 + }, + { + "epoch": 0.08, + "learning_rate": 9.847375047720206e-05, + "loss": 2.3076, + "step": 29500 + }, + { + "epoch": 0.08, + "learning_rate": 9.847323395904713e-05, + "loss": 2.3792, + "step": 29505 + }, + { + "epoch": 0.08, + "learning_rate": 9.847271735486124e-05, + "loss": 2.4637, + "step": 29510 + }, + { + "epoch": 0.08, + "learning_rate": 9.847220066464529e-05, + "loss": 2.2513, + "step": 29515 + }, + { + "epoch": 0.08, + "learning_rate": 9.847168388840021e-05, + "loss": 2.3202, + "step": 29520 + }, + { + "epoch": 0.08, + "learning_rate": 9.847116702612691e-05, + "loss": 2.4161, + "step": 29525 + }, + { + "epoch": 0.08, + "learning_rate": 9.847065007782631e-05, + "loss": 2.4063, + "step": 29530 + }, + { + "epoch": 0.08, + "learning_rate": 9.847013304349932e-05, + "loss": 2.3689, + "step": 29535 + }, + { + "epoch": 0.08, + "learning_rate": 9.846961592314687e-05, + "loss": 2.4361, + "step": 29540 + }, + { + "epoch": 0.08, + "learning_rate": 9.846909871676987e-05, + "loss": 2.414, + "step": 29545 + }, + { + "epoch": 0.08, + "learning_rate": 9.846858142436923e-05, + "loss": 2.3437, + "step": 29550 + }, + { + "epoch": 0.08, + "learning_rate": 9.84680640459459e-05, + "loss": 2.4852, + "step": 29555 + }, + { + "epoch": 0.08, + "learning_rate": 9.846754658150075e-05, + "loss": 2.4262, + "step": 29560 + }, + { + "epoch": 0.08, + "learning_rate": 9.846702903103476e-05, + "loss": 2.3485, + "step": 29565 + }, + { + "epoch": 0.08, + "learning_rate": 9.846651139454878e-05, + "loss": 2.518, + "step": 29570 + }, + { + "epoch": 0.08, + "learning_rate": 9.846599367204376e-05, + "loss": 2.3931, + "step": 29575 + }, + { + "epoch": 0.08, + "learning_rate": 9.846547586352064e-05, + "loss": 2.2498, + "step": 29580 + }, + { + "epoch": 0.08, + "learning_rate": 9.846495796898032e-05, + "loss": 2.3804, + "step": 29585 + }, + { + "epoch": 0.08, + "learning_rate": 9.846443998842369e-05, + "loss": 2.3871, + "step": 29590 + }, + { + "epoch": 0.08, + "learning_rate": 9.846392192185174e-05, + "loss": 2.4341, + "step": 29595 + }, + { + "epoch": 0.08, + "learning_rate": 9.846340376926532e-05, + "loss": 2.3645, + "step": 29600 + }, + { + "epoch": 0.08, + "learning_rate": 9.846288553066538e-05, + "loss": 2.3962, + "step": 29605 + }, + { + "epoch": 0.08, + "learning_rate": 9.846236720605284e-05, + "loss": 2.3204, + "step": 29610 + }, + { + "epoch": 0.08, + "learning_rate": 9.846184879542861e-05, + "loss": 2.4025, + "step": 29615 + }, + { + "epoch": 0.08, + "learning_rate": 9.846133029879362e-05, + "loss": 2.4528, + "step": 29620 + }, + { + "epoch": 0.08, + "learning_rate": 9.846081171614881e-05, + "loss": 2.2968, + "step": 29625 + }, + { + "epoch": 0.08, + "learning_rate": 9.846029304749505e-05, + "loss": 2.4585, + "step": 29630 + }, + { + "epoch": 0.08, + "learning_rate": 9.84597742928333e-05, + "loss": 2.346, + "step": 29635 + }, + { + "epoch": 0.08, + "learning_rate": 9.845925545216447e-05, + "loss": 2.3781, + "step": 29640 + }, + { + "epoch": 0.08, + "learning_rate": 9.845873652548947e-05, + "loss": 2.2053, + "step": 29645 + }, + { + "epoch": 0.08, + "learning_rate": 9.845821751280924e-05, + "loss": 2.4366, + "step": 29650 + }, + { + "epoch": 0.08, + "learning_rate": 9.845769841412469e-05, + "loss": 2.3903, + "step": 29655 + }, + { + "epoch": 0.08, + "learning_rate": 9.845717922943674e-05, + "loss": 2.2608, + "step": 29660 + }, + { + "epoch": 0.08, + "learning_rate": 9.845665995874631e-05, + "loss": 2.2978, + "step": 29665 + }, + { + "epoch": 0.08, + "learning_rate": 9.845614060205433e-05, + "loss": 2.3708, + "step": 29670 + }, + { + "epoch": 0.08, + "learning_rate": 9.84556211593617e-05, + "loss": 2.4361, + "step": 29675 + }, + { + "epoch": 0.08, + "learning_rate": 9.84551016306694e-05, + "loss": 2.497, + "step": 29680 + }, + { + "epoch": 0.08, + "learning_rate": 9.845458201597828e-05, + "loss": 2.3844, + "step": 29685 + }, + { + "epoch": 0.08, + "learning_rate": 9.845406231528931e-05, + "loss": 2.4597, + "step": 29690 + }, + { + "epoch": 0.08, + "learning_rate": 9.84535425286034e-05, + "loss": 2.3555, + "step": 29695 + }, + { + "epoch": 0.08, + "learning_rate": 9.845302265592145e-05, + "loss": 2.3681, + "step": 29700 + }, + { + "epoch": 0.08, + "learning_rate": 9.845250269724443e-05, + "loss": 2.4331, + "step": 29705 + }, + { + "epoch": 0.08, + "learning_rate": 9.84519826525732e-05, + "loss": 2.5301, + "step": 29710 + }, + { + "epoch": 0.08, + "learning_rate": 9.845146252190874e-05, + "loss": 2.3633, + "step": 29715 + }, + { + "epoch": 0.08, + "learning_rate": 9.845094230525194e-05, + "loss": 2.3555, + "step": 29720 + }, + { + "epoch": 0.08, + "learning_rate": 9.845042200260374e-05, + "loss": 2.4039, + "step": 29725 + }, + { + "epoch": 0.08, + "learning_rate": 9.844990161396505e-05, + "loss": 2.454, + "step": 29730 + }, + { + "epoch": 0.08, + "learning_rate": 9.844938113933681e-05, + "loss": 2.3914, + "step": 29735 + }, + { + "epoch": 0.08, + "learning_rate": 9.844886057871994e-05, + "loss": 2.3834, + "step": 29740 + }, + { + "epoch": 0.08, + "learning_rate": 9.844833993211534e-05, + "loss": 2.2977, + "step": 29745 + }, + { + "epoch": 0.08, + "learning_rate": 9.844781919952396e-05, + "loss": 2.3644, + "step": 29750 + }, + { + "epoch": 0.08, + "learning_rate": 9.844729838094671e-05, + "loss": 2.2792, + "step": 29755 + }, + { + "epoch": 0.08, + "learning_rate": 9.844677747638453e-05, + "loss": 2.3061, + "step": 29760 + }, + { + "epoch": 0.08, + "learning_rate": 9.844625648583835e-05, + "loss": 2.3261, + "step": 29765 + }, + { + "epoch": 0.08, + "learning_rate": 9.844573540930906e-05, + "loss": 2.3912, + "step": 29770 + }, + { + "epoch": 0.08, + "learning_rate": 9.844521424679761e-05, + "loss": 2.3244, + "step": 29775 + }, + { + "epoch": 0.08, + "learning_rate": 9.844469299830492e-05, + "loss": 2.3987, + "step": 29780 + }, + { + "epoch": 0.08, + "learning_rate": 9.844417166383192e-05, + "loss": 2.366, + "step": 29785 + }, + { + "epoch": 0.08, + "learning_rate": 9.844365024337953e-05, + "loss": 2.3393, + "step": 29790 + }, + { + "epoch": 0.08, + "learning_rate": 9.844312873694868e-05, + "loss": 2.4363, + "step": 29795 + }, + { + "epoch": 0.08, + "learning_rate": 9.844260714454028e-05, + "loss": 2.4965, + "step": 29800 + }, + { + "epoch": 0.08, + "learning_rate": 9.844208546615527e-05, + "loss": 2.4235, + "step": 29805 + }, + { + "epoch": 0.08, + "learning_rate": 9.844156370179458e-05, + "loss": 2.4551, + "step": 29810 + }, + { + "epoch": 0.08, + "learning_rate": 9.844104185145912e-05, + "loss": 2.4756, + "step": 29815 + }, + { + "epoch": 0.08, + "learning_rate": 9.844051991514982e-05, + "loss": 2.4324, + "step": 29820 + }, + { + "epoch": 0.08, + "learning_rate": 9.843999789286763e-05, + "loss": 2.3846, + "step": 29825 + }, + { + "epoch": 0.08, + "learning_rate": 9.843947578461346e-05, + "loss": 2.3594, + "step": 29830 + }, + { + "epoch": 0.08, + "learning_rate": 9.843895359038822e-05, + "loss": 2.2901, + "step": 29835 + }, + { + "epoch": 0.08, + "learning_rate": 9.843843131019285e-05, + "loss": 2.409, + "step": 29840 + }, + { + "epoch": 0.08, + "learning_rate": 9.843790894402829e-05, + "loss": 2.4179, + "step": 29845 + }, + { + "epoch": 0.08, + "learning_rate": 9.843738649189545e-05, + "loss": 2.2962, + "step": 29850 + }, + { + "epoch": 0.08, + "learning_rate": 9.843686395379525e-05, + "loss": 2.3055, + "step": 29855 + }, + { + "epoch": 0.08, + "learning_rate": 9.843634132972866e-05, + "loss": 2.3623, + "step": 29860 + }, + { + "epoch": 0.08, + "learning_rate": 9.843581861969655e-05, + "loss": 2.3431, + "step": 29865 + }, + { + "epoch": 0.08, + "learning_rate": 9.843529582369988e-05, + "loss": 2.4114, + "step": 29870 + }, + { + "epoch": 0.08, + "learning_rate": 9.84347729417396e-05, + "loss": 2.4382, + "step": 29875 + }, + { + "epoch": 0.08, + "learning_rate": 9.843424997381659e-05, + "loss": 2.5575, + "step": 29880 + }, + { + "epoch": 0.08, + "learning_rate": 9.843372691993178e-05, + "loss": 2.527, + "step": 29885 + }, + { + "epoch": 0.08, + "learning_rate": 9.843320378008616e-05, + "loss": 2.4321, + "step": 29890 + }, + { + "epoch": 0.08, + "learning_rate": 9.843268055428058e-05, + "loss": 2.3354, + "step": 29895 + }, + { + "epoch": 0.08, + "learning_rate": 9.843215724251603e-05, + "loss": 2.3317, + "step": 29900 + }, + { + "epoch": 0.08, + "learning_rate": 9.84316338447934e-05, + "loss": 2.3063, + "step": 29905 + }, + { + "epoch": 0.08, + "learning_rate": 9.843111036111365e-05, + "loss": 2.4639, + "step": 29910 + }, + { + "epoch": 0.08, + "learning_rate": 9.843058679147768e-05, + "loss": 2.3499, + "step": 29915 + }, + { + "epoch": 0.08, + "learning_rate": 9.843006313588642e-05, + "loss": 2.4171, + "step": 29920 + }, + { + "epoch": 0.08, + "learning_rate": 9.842953939434083e-05, + "loss": 2.4163, + "step": 29925 + }, + { + "epoch": 0.08, + "learning_rate": 9.84290155668418e-05, + "loss": 2.2954, + "step": 29930 + }, + { + "epoch": 0.08, + "learning_rate": 9.84284916533903e-05, + "loss": 2.2573, + "step": 29935 + }, + { + "epoch": 0.08, + "learning_rate": 9.842796765398722e-05, + "loss": 2.2692, + "step": 29940 + }, + { + "epoch": 0.08, + "learning_rate": 9.842744356863352e-05, + "loss": 2.3399, + "step": 29945 + }, + { + "epoch": 0.08, + "learning_rate": 9.842691939733013e-05, + "loss": 2.2592, + "step": 29950 + }, + { + "epoch": 0.08, + "learning_rate": 9.842639514007796e-05, + "loss": 2.4488, + "step": 29955 + }, + { + "epoch": 0.08, + "learning_rate": 9.842587079687794e-05, + "loss": 2.4487, + "step": 29960 + }, + { + "epoch": 0.08, + "learning_rate": 9.842534636773103e-05, + "loss": 2.3531, + "step": 29965 + }, + { + "epoch": 0.08, + "learning_rate": 9.842482185263813e-05, + "loss": 2.4699, + "step": 29970 + }, + { + "epoch": 0.08, + "learning_rate": 9.842429725160018e-05, + "loss": 2.3858, + "step": 29975 + }, + { + "epoch": 0.08, + "learning_rate": 9.842377256461813e-05, + "loss": 2.3345, + "step": 29980 + }, + { + "epoch": 0.08, + "learning_rate": 9.842324779169289e-05, + "loss": 2.4399, + "step": 29985 + }, + { + "epoch": 0.08, + "learning_rate": 9.842272293282539e-05, + "loss": 2.3577, + "step": 29990 + }, + { + "epoch": 0.08, + "learning_rate": 9.842219798801658e-05, + "loss": 2.3706, + "step": 29995 + }, + { + "epoch": 0.08, + "learning_rate": 9.842167295726737e-05, + "loss": 2.3971, + "step": 30000 + }, + { + "epoch": 0.08, + "learning_rate": 9.84211478405787e-05, + "loss": 2.2774, + "step": 30005 + }, + { + "epoch": 0.08, + "learning_rate": 9.84206226379515e-05, + "loss": 2.4107, + "step": 30010 + }, + { + "epoch": 0.08, + "learning_rate": 9.842009734938671e-05, + "loss": 2.3257, + "step": 30015 + }, + { + "epoch": 0.08, + "learning_rate": 9.841957197488526e-05, + "loss": 2.4827, + "step": 30020 + }, + { + "epoch": 0.08, + "learning_rate": 9.841904651444808e-05, + "loss": 2.291, + "step": 30025 + }, + { + "epoch": 0.08, + "learning_rate": 9.841852096807608e-05, + "loss": 2.373, + "step": 30030 + }, + { + "epoch": 0.08, + "learning_rate": 9.841799533577026e-05, + "loss": 2.3821, + "step": 30035 + }, + { + "epoch": 0.08, + "learning_rate": 9.841746961753148e-05, + "loss": 2.4406, + "step": 30040 + }, + { + "epoch": 0.08, + "learning_rate": 9.84169438133607e-05, + "loss": 2.3586, + "step": 30045 + }, + { + "epoch": 0.08, + "learning_rate": 9.841641792325887e-05, + "loss": 2.298, + "step": 30050 + }, + { + "epoch": 0.08, + "learning_rate": 9.84158919472269e-05, + "loss": 2.3997, + "step": 30055 + }, + { + "epoch": 0.08, + "learning_rate": 9.841536588526574e-05, + "loss": 2.4287, + "step": 30060 + }, + { + "epoch": 0.08, + "learning_rate": 9.841483973737631e-05, + "loss": 2.475, + "step": 30065 + }, + { + "epoch": 0.08, + "learning_rate": 9.841431350355953e-05, + "loss": 2.3701, + "step": 30070 + }, + { + "epoch": 0.08, + "learning_rate": 9.841378718381638e-05, + "loss": 2.3912, + "step": 30075 + }, + { + "epoch": 0.08, + "learning_rate": 9.841326077814777e-05, + "loss": 2.3132, + "step": 30080 + }, + { + "epoch": 0.08, + "learning_rate": 9.841273428655461e-05, + "loss": 2.3434, + "step": 30085 + }, + { + "epoch": 0.08, + "learning_rate": 9.841220770903787e-05, + "loss": 2.3376, + "step": 30090 + }, + { + "epoch": 0.08, + "learning_rate": 9.841168104559846e-05, + "loss": 2.4021, + "step": 30095 + }, + { + "epoch": 0.08, + "learning_rate": 9.841115429623733e-05, + "loss": 2.4277, + "step": 30100 + }, + { + "epoch": 0.08, + "learning_rate": 9.841062746095541e-05, + "loss": 2.3369, + "step": 30105 + }, + { + "epoch": 0.08, + "learning_rate": 9.841010053975363e-05, + "loss": 2.3314, + "step": 30110 + }, + { + "epoch": 0.08, + "learning_rate": 9.840957353263295e-05, + "loss": 2.2873, + "step": 30115 + }, + { + "epoch": 0.08, + "learning_rate": 9.840904643959428e-05, + "loss": 2.2769, + "step": 30120 + }, + { + "epoch": 0.08, + "learning_rate": 9.840851926063854e-05, + "loss": 2.2377, + "step": 30125 + }, + { + "epoch": 0.08, + "learning_rate": 9.840799199576671e-05, + "loss": 2.352, + "step": 30130 + }, + { + "epoch": 0.08, + "learning_rate": 9.84074646449797e-05, + "loss": 2.5103, + "step": 30135 + }, + { + "epoch": 0.08, + "learning_rate": 9.840693720827842e-05, + "loss": 2.4242, + "step": 30140 + }, + { + "epoch": 0.08, + "learning_rate": 9.840640968566387e-05, + "loss": 2.3418, + "step": 30145 + }, + { + "epoch": 0.08, + "learning_rate": 9.840588207713695e-05, + "loss": 2.3626, + "step": 30150 + }, + { + "epoch": 0.08, + "learning_rate": 9.840535438269858e-05, + "loss": 2.3059, + "step": 30155 + }, + { + "epoch": 0.08, + "learning_rate": 9.840482660234971e-05, + "loss": 2.329, + "step": 30160 + }, + { + "epoch": 0.08, + "learning_rate": 9.84042987360913e-05, + "loss": 2.4279, + "step": 30165 + }, + { + "epoch": 0.08, + "learning_rate": 9.840377078392425e-05, + "loss": 2.4671, + "step": 30170 + }, + { + "epoch": 0.08, + "learning_rate": 9.840324274584953e-05, + "loss": 2.3723, + "step": 30175 + }, + { + "epoch": 0.08, + "learning_rate": 9.840271462186805e-05, + "loss": 2.3299, + "step": 30180 + }, + { + "epoch": 0.08, + "learning_rate": 9.840218641198075e-05, + "loss": 2.3062, + "step": 30185 + }, + { + "epoch": 0.08, + "learning_rate": 9.84016581161886e-05, + "loss": 2.3525, + "step": 30190 + }, + { + "epoch": 0.08, + "learning_rate": 9.840112973449251e-05, + "loss": 2.4289, + "step": 30195 + }, + { + "epoch": 0.08, + "learning_rate": 9.84006012668934e-05, + "loss": 2.4156, + "step": 30200 + }, + { + "epoch": 0.08, + "learning_rate": 9.840007271339225e-05, + "loss": 2.4466, + "step": 30205 + }, + { + "epoch": 0.08, + "learning_rate": 9.839954407398998e-05, + "loss": 2.3142, + "step": 30210 + }, + { + "epoch": 0.08, + "learning_rate": 9.839901534868752e-05, + "loss": 2.3597, + "step": 30215 + }, + { + "epoch": 0.08, + "learning_rate": 9.839848653748583e-05, + "loss": 2.391, + "step": 30220 + }, + { + "epoch": 0.08, + "learning_rate": 9.83979576403858e-05, + "loss": 2.2437, + "step": 30225 + }, + { + "epoch": 0.08, + "learning_rate": 9.839742865738842e-05, + "loss": 2.4085, + "step": 30230 + }, + { + "epoch": 0.08, + "learning_rate": 9.839689958849461e-05, + "loss": 2.4871, + "step": 30235 + }, + { + "epoch": 0.08, + "learning_rate": 9.839637043370531e-05, + "loss": 2.4387, + "step": 30240 + }, + { + "epoch": 0.08, + "learning_rate": 9.839584119302146e-05, + "loss": 2.2754, + "step": 30245 + }, + { + "epoch": 0.08, + "learning_rate": 9.8395311866444e-05, + "loss": 2.3698, + "step": 30250 + }, + { + "epoch": 0.08, + "learning_rate": 9.839478245397387e-05, + "loss": 2.5324, + "step": 30255 + }, + { + "epoch": 0.08, + "learning_rate": 9.8394252955612e-05, + "loss": 2.3325, + "step": 30260 + }, + { + "epoch": 0.08, + "learning_rate": 9.839372337135933e-05, + "loss": 2.4364, + "step": 30265 + }, + { + "epoch": 0.08, + "learning_rate": 9.839319370121683e-05, + "loss": 2.4477, + "step": 30270 + }, + { + "epoch": 0.08, + "learning_rate": 9.839266394518539e-05, + "loss": 2.3304, + "step": 30275 + }, + { + "epoch": 0.08, + "learning_rate": 9.839213410326598e-05, + "loss": 2.5105, + "step": 30280 + }, + { + "epoch": 0.08, + "learning_rate": 9.839160417545955e-05, + "loss": 2.3629, + "step": 30285 + }, + { + "epoch": 0.08, + "learning_rate": 9.839107416176703e-05, + "loss": 2.4433, + "step": 30290 + }, + { + "epoch": 0.08, + "learning_rate": 9.839054406218936e-05, + "loss": 2.4256, + "step": 30295 + }, + { + "epoch": 0.08, + "learning_rate": 9.839001387672746e-05, + "loss": 2.4618, + "step": 30300 + }, + { + "epoch": 0.08, + "learning_rate": 9.838948360538231e-05, + "loss": 2.3517, + "step": 30305 + }, + { + "epoch": 0.08, + "learning_rate": 9.838895324815483e-05, + "loss": 2.4635, + "step": 30310 + }, + { + "epoch": 0.08, + "learning_rate": 9.838842280504595e-05, + "loss": 2.4506, + "step": 30315 + }, + { + "epoch": 0.08, + "learning_rate": 9.838789227605662e-05, + "loss": 2.386, + "step": 30320 + }, + { + "epoch": 0.08, + "learning_rate": 9.83873616611878e-05, + "loss": 2.385, + "step": 30325 + }, + { + "epoch": 0.08, + "learning_rate": 9.838683096044042e-05, + "loss": 2.3723, + "step": 30330 + }, + { + "epoch": 0.08, + "learning_rate": 9.838630017381543e-05, + "loss": 2.4678, + "step": 30335 + }, + { + "epoch": 0.08, + "learning_rate": 9.838576930131373e-05, + "loss": 2.4218, + "step": 30340 + }, + { + "epoch": 0.08, + "learning_rate": 9.838523834293632e-05, + "loss": 2.3268, + "step": 30345 + }, + { + "epoch": 0.08, + "learning_rate": 9.838470729868411e-05, + "loss": 2.4475, + "step": 30350 + }, + { + "epoch": 0.08, + "learning_rate": 9.838417616855804e-05, + "loss": 2.2201, + "step": 30355 + }, + { + "epoch": 0.08, + "learning_rate": 9.838364495255907e-05, + "loss": 2.4248, + "step": 30360 + }, + { + "epoch": 0.08, + "learning_rate": 9.838311365068813e-05, + "loss": 2.3542, + "step": 30365 + }, + { + "epoch": 0.08, + "learning_rate": 9.838258226294617e-05, + "loss": 2.4495, + "step": 30370 + }, + { + "epoch": 0.08, + "learning_rate": 9.838205078933413e-05, + "loss": 2.276, + "step": 30375 + }, + { + "epoch": 0.08, + "learning_rate": 9.838151922985295e-05, + "loss": 2.4449, + "step": 30380 + }, + { + "epoch": 0.08, + "learning_rate": 9.838098758450359e-05, + "loss": 2.4613, + "step": 30385 + }, + { + "epoch": 0.08, + "learning_rate": 9.838045585328697e-05, + "loss": 2.5631, + "step": 30390 + }, + { + "epoch": 0.08, + "learning_rate": 9.837992403620404e-05, + "loss": 2.3949, + "step": 30395 + }, + { + "epoch": 0.08, + "learning_rate": 9.837939213325576e-05, + "loss": 2.4941, + "step": 30400 + }, + { + "epoch": 0.08, + "learning_rate": 9.837886014444304e-05, + "loss": 2.4229, + "step": 30405 + }, + { + "epoch": 0.08, + "learning_rate": 9.837832806976685e-05, + "loss": 2.3704, + "step": 30410 + }, + { + "epoch": 0.08, + "learning_rate": 9.837779590922816e-05, + "loss": 2.3869, + "step": 30415 + }, + { + "epoch": 0.08, + "learning_rate": 9.837726366282785e-05, + "loss": 2.2991, + "step": 30420 + }, + { + "epoch": 0.08, + "learning_rate": 9.837673133056693e-05, + "loss": 2.2828, + "step": 30425 + }, + { + "epoch": 0.08, + "learning_rate": 9.83761989124463e-05, + "loss": 2.38, + "step": 30430 + }, + { + "epoch": 0.08, + "learning_rate": 9.837566640846692e-05, + "loss": 2.4003, + "step": 30435 + }, + { + "epoch": 0.08, + "learning_rate": 9.837513381862973e-05, + "loss": 2.3728, + "step": 30440 + }, + { + "epoch": 0.08, + "learning_rate": 9.837460114293569e-05, + "loss": 2.348, + "step": 30445 + }, + { + "epoch": 0.08, + "learning_rate": 9.837406838138572e-05, + "loss": 2.4278, + "step": 30450 + }, + { + "epoch": 0.08, + "learning_rate": 9.837353553398078e-05, + "loss": 2.4915, + "step": 30455 + }, + { + "epoch": 0.08, + "learning_rate": 9.837300260072184e-05, + "loss": 2.3307, + "step": 30460 + }, + { + "epoch": 0.08, + "learning_rate": 9.837246958160982e-05, + "loss": 2.2955, + "step": 30465 + }, + { + "epoch": 0.08, + "learning_rate": 9.837193647664564e-05, + "loss": 2.3509, + "step": 30470 + }, + { + "epoch": 0.08, + "learning_rate": 9.83714032858303e-05, + "loss": 2.3808, + "step": 30475 + }, + { + "epoch": 0.08, + "learning_rate": 9.837087000916472e-05, + "loss": 2.3269, + "step": 30480 + }, + { + "epoch": 0.08, + "learning_rate": 9.837033664664983e-05, + "loss": 2.3538, + "step": 30485 + }, + { + "epoch": 0.08, + "learning_rate": 9.836980319828661e-05, + "loss": 2.4302, + "step": 30490 + }, + { + "epoch": 0.08, + "learning_rate": 9.836926966407597e-05, + "loss": 2.5285, + "step": 30495 + }, + { + "epoch": 0.08, + "learning_rate": 9.836873604401891e-05, + "loss": 2.3287, + "step": 30500 + }, + { + "epoch": 0.08, + "learning_rate": 9.836820233811632e-05, + "loss": 2.4391, + "step": 30505 + }, + { + "epoch": 0.08, + "learning_rate": 9.836766854636919e-05, + "loss": 2.4228, + "step": 30510 + }, + { + "epoch": 0.08, + "learning_rate": 9.836713466877842e-05, + "loss": 2.4578, + "step": 30515 + }, + { + "epoch": 0.08, + "learning_rate": 9.836660070534503e-05, + "loss": 2.4436, + "step": 30520 + }, + { + "epoch": 0.08, + "learning_rate": 9.836606665606989e-05, + "loss": 2.2719, + "step": 30525 + }, + { + "epoch": 0.08, + "learning_rate": 9.8365532520954e-05, + "loss": 2.4273, + "step": 30530 + }, + { + "epoch": 0.08, + "learning_rate": 9.836499829999827e-05, + "loss": 2.4005, + "step": 30535 + }, + { + "epoch": 0.08, + "learning_rate": 9.836446399320368e-05, + "loss": 2.3894, + "step": 30540 + }, + { + "epoch": 0.08, + "learning_rate": 9.836392960057117e-05, + "loss": 2.4204, + "step": 30545 + }, + { + "epoch": 0.08, + "learning_rate": 9.836339512210168e-05, + "loss": 2.3631, + "step": 30550 + }, + { + "epoch": 0.08, + "learning_rate": 9.836286055779618e-05, + "loss": 2.478, + "step": 30555 + }, + { + "epoch": 0.08, + "learning_rate": 9.836232590765558e-05, + "loss": 2.5384, + "step": 30560 + }, + { + "epoch": 0.08, + "learning_rate": 9.836179117168087e-05, + "loss": 2.3497, + "step": 30565 + }, + { + "epoch": 0.08, + "learning_rate": 9.836125634987297e-05, + "loss": 2.3512, + "step": 30570 + }, + { + "epoch": 0.08, + "learning_rate": 9.836072144223283e-05, + "loss": 2.3953, + "step": 30575 + }, + { + "epoch": 0.08, + "learning_rate": 9.836018644876143e-05, + "loss": 2.3914, + "step": 30580 + }, + { + "epoch": 0.08, + "learning_rate": 9.835965136945968e-05, + "loss": 2.4431, + "step": 30585 + }, + { + "epoch": 0.08, + "learning_rate": 9.835911620432856e-05, + "loss": 2.4394, + "step": 30590 + }, + { + "epoch": 0.08, + "learning_rate": 9.835858095336901e-05, + "loss": 2.4159, + "step": 30595 + }, + { + "epoch": 0.08, + "learning_rate": 9.835804561658197e-05, + "loss": 2.4935, + "step": 30600 + }, + { + "epoch": 0.08, + "learning_rate": 9.83575101939684e-05, + "loss": 2.425, + "step": 30605 + }, + { + "epoch": 0.08, + "learning_rate": 9.835697468552926e-05, + "loss": 2.3863, + "step": 30610 + }, + { + "epoch": 0.08, + "learning_rate": 9.835643909126547e-05, + "loss": 2.388, + "step": 30615 + }, + { + "epoch": 0.08, + "learning_rate": 9.835590341117802e-05, + "loss": 2.3942, + "step": 30620 + }, + { + "epoch": 0.08, + "learning_rate": 9.835536764526783e-05, + "loss": 2.3848, + "step": 30625 + }, + { + "epoch": 0.08, + "learning_rate": 9.835483179353585e-05, + "loss": 2.3254, + "step": 30630 + }, + { + "epoch": 0.08, + "learning_rate": 9.835429585598306e-05, + "loss": 2.4165, + "step": 30635 + }, + { + "epoch": 0.08, + "learning_rate": 9.835375983261038e-05, + "loss": 2.3682, + "step": 30640 + }, + { + "epoch": 0.08, + "learning_rate": 9.835322372341878e-05, + "loss": 2.4281, + "step": 30645 + }, + { + "epoch": 0.08, + "learning_rate": 9.83526875284092e-05, + "loss": 2.3779, + "step": 30650 + }, + { + "epoch": 0.08, + "learning_rate": 9.835215124758261e-05, + "loss": 2.4285, + "step": 30655 + }, + { + "epoch": 0.08, + "learning_rate": 9.835161488093994e-05, + "loss": 2.3861, + "step": 30660 + }, + { + "epoch": 0.08, + "learning_rate": 9.835107842848217e-05, + "loss": 2.4054, + "step": 30665 + }, + { + "epoch": 0.08, + "learning_rate": 9.835054189021021e-05, + "loss": 2.3481, + "step": 30670 + }, + { + "epoch": 0.08, + "learning_rate": 9.835000526612504e-05, + "loss": 2.3911, + "step": 30675 + }, + { + "epoch": 0.08, + "learning_rate": 9.834946855622762e-05, + "loss": 2.3546, + "step": 30680 + }, + { + "epoch": 0.08, + "learning_rate": 9.83489317605189e-05, + "loss": 2.3388, + "step": 30685 + }, + { + "epoch": 0.08, + "learning_rate": 9.834839487899979e-05, + "loss": 2.3341, + "step": 30690 + }, + { + "epoch": 0.08, + "learning_rate": 9.83478579116713e-05, + "loss": 2.288, + "step": 30695 + }, + { + "epoch": 0.08, + "learning_rate": 9.834732085853435e-05, + "loss": 2.258, + "step": 30700 + }, + { + "epoch": 0.08, + "learning_rate": 9.834678371958991e-05, + "loss": 2.3749, + "step": 30705 + }, + { + "epoch": 0.08, + "learning_rate": 9.834624649483893e-05, + "loss": 2.5812, + "step": 30710 + }, + { + "epoch": 0.08, + "learning_rate": 9.834570918428234e-05, + "loss": 2.2594, + "step": 30715 + }, + { + "epoch": 0.08, + "learning_rate": 9.834517178792114e-05, + "loss": 2.3262, + "step": 30720 + }, + { + "epoch": 0.08, + "learning_rate": 9.834463430575623e-05, + "loss": 2.4945, + "step": 30725 + }, + { + "epoch": 0.08, + "learning_rate": 9.83440967377886e-05, + "loss": 2.2871, + "step": 30730 + }, + { + "epoch": 0.08, + "learning_rate": 9.83435590840192e-05, + "loss": 2.3728, + "step": 30735 + }, + { + "epoch": 0.08, + "learning_rate": 9.834302134444898e-05, + "loss": 2.4561, + "step": 30740 + }, + { + "epoch": 0.08, + "learning_rate": 9.834248351907889e-05, + "loss": 2.3412, + "step": 30745 + }, + { + "epoch": 0.08, + "learning_rate": 9.834194560790989e-05, + "loss": 2.4677, + "step": 30750 + }, + { + "epoch": 0.08, + "learning_rate": 9.834140761094292e-05, + "loss": 2.4507, + "step": 30755 + }, + { + "epoch": 0.08, + "learning_rate": 9.834086952817898e-05, + "loss": 2.3763, + "step": 30760 + }, + { + "epoch": 0.08, + "learning_rate": 9.834033135961895e-05, + "loss": 2.3554, + "step": 30765 + }, + { + "epoch": 0.08, + "learning_rate": 9.833979310526387e-05, + "loss": 2.3576, + "step": 30770 + }, + { + "epoch": 0.08, + "learning_rate": 9.833925476511463e-05, + "loss": 2.3932, + "step": 30775 + }, + { + "epoch": 0.08, + "learning_rate": 9.83387163391722e-05, + "loss": 2.3341, + "step": 30780 + }, + { + "epoch": 0.08, + "learning_rate": 9.833817782743756e-05, + "loss": 2.3511, + "step": 30785 + }, + { + "epoch": 0.08, + "learning_rate": 9.833763922991164e-05, + "loss": 2.3069, + "step": 30790 + }, + { + "epoch": 0.08, + "learning_rate": 9.833710054659541e-05, + "loss": 2.2747, + "step": 30795 + }, + { + "epoch": 0.08, + "learning_rate": 9.833656177748982e-05, + "loss": 2.3012, + "step": 30800 + }, + { + "epoch": 0.08, + "learning_rate": 9.833602292259584e-05, + "loss": 2.276, + "step": 30805 + }, + { + "epoch": 0.08, + "learning_rate": 9.83354839819144e-05, + "loss": 2.3776, + "step": 30810 + }, + { + "epoch": 0.08, + "learning_rate": 9.833494495544649e-05, + "loss": 2.4959, + "step": 30815 + }, + { + "epoch": 0.08, + "learning_rate": 9.833440584319302e-05, + "loss": 2.2828, + "step": 30820 + }, + { + "epoch": 0.08, + "learning_rate": 9.833386664515498e-05, + "loss": 2.451, + "step": 30825 + }, + { + "epoch": 0.08, + "learning_rate": 9.833332736133334e-05, + "loss": 2.391, + "step": 30830 + }, + { + "epoch": 0.08, + "learning_rate": 9.833278799172903e-05, + "loss": 2.3243, + "step": 30835 + }, + { + "epoch": 0.08, + "learning_rate": 9.8332248536343e-05, + "loss": 2.3686, + "step": 30840 + }, + { + "epoch": 0.08, + "learning_rate": 9.833170899517625e-05, + "loss": 2.3045, + "step": 30845 + }, + { + "epoch": 0.08, + "learning_rate": 9.833116936822969e-05, + "loss": 2.5493, + "step": 30850 + }, + { + "epoch": 0.08, + "learning_rate": 9.83306296555043e-05, + "loss": 2.449, + "step": 30855 + }, + { + "epoch": 0.08, + "learning_rate": 9.833008985700105e-05, + "loss": 2.4554, + "step": 30860 + }, + { + "epoch": 0.08, + "learning_rate": 9.832954997272087e-05, + "loss": 2.5031, + "step": 30865 + }, + { + "epoch": 0.08, + "learning_rate": 9.832901000266474e-05, + "loss": 2.3358, + "step": 30870 + }, + { + "epoch": 0.08, + "learning_rate": 9.832846994683363e-05, + "loss": 2.4543, + "step": 30875 + }, + { + "epoch": 0.08, + "learning_rate": 9.832792980522845e-05, + "loss": 2.3651, + "step": 30880 + }, + { + "epoch": 0.08, + "learning_rate": 9.83273895778502e-05, + "loss": 2.2914, + "step": 30885 + }, + { + "epoch": 0.08, + "learning_rate": 9.832684926469982e-05, + "loss": 2.3521, + "step": 30890 + }, + { + "epoch": 0.08, + "learning_rate": 9.832630886577828e-05, + "loss": 2.4416, + "step": 30895 + }, + { + "epoch": 0.08, + "learning_rate": 9.832576838108653e-05, + "loss": 2.2913, + "step": 30900 + }, + { + "epoch": 0.08, + "learning_rate": 9.832522781062555e-05, + "loss": 2.3119, + "step": 30905 + }, + { + "epoch": 0.08, + "learning_rate": 9.832468715439628e-05, + "loss": 2.3082, + "step": 30910 + }, + { + "epoch": 0.08, + "learning_rate": 9.832414641239968e-05, + "loss": 2.3562, + "step": 30915 + }, + { + "epoch": 0.08, + "learning_rate": 9.83236055846367e-05, + "loss": 2.3306, + "step": 30920 + }, + { + "epoch": 0.08, + "learning_rate": 9.832306467110833e-05, + "loss": 2.3709, + "step": 30925 + }, + { + "epoch": 0.08, + "learning_rate": 9.83225236718155e-05, + "loss": 2.2317, + "step": 30930 + }, + { + "epoch": 0.08, + "learning_rate": 9.83219825867592e-05, + "loss": 2.2604, + "step": 30935 + }, + { + "epoch": 0.08, + "learning_rate": 9.832144141594035e-05, + "loss": 2.2821, + "step": 30940 + }, + { + "epoch": 0.08, + "learning_rate": 9.832090015935995e-05, + "loss": 2.3087, + "step": 30945 + }, + { + "epoch": 0.08, + "learning_rate": 9.832035881701894e-05, + "loss": 2.32, + "step": 30950 + }, + { + "epoch": 0.08, + "learning_rate": 9.831981738891829e-05, + "loss": 2.3731, + "step": 30955 + }, + { + "epoch": 0.08, + "learning_rate": 9.831927587505895e-05, + "loss": 2.4929, + "step": 30960 + }, + { + "epoch": 0.08, + "learning_rate": 9.83187342754419e-05, + "loss": 2.3715, + "step": 30965 + }, + { + "epoch": 0.08, + "learning_rate": 9.831819259006807e-05, + "loss": 2.4901, + "step": 30970 + }, + { + "epoch": 0.08, + "learning_rate": 9.831765081893844e-05, + "loss": 2.4585, + "step": 30975 + }, + { + "epoch": 0.08, + "learning_rate": 9.8317108962054e-05, + "loss": 2.4802, + "step": 30980 + }, + { + "epoch": 0.08, + "learning_rate": 9.831656701941564e-05, + "loss": 2.448, + "step": 30985 + }, + { + "epoch": 0.08, + "learning_rate": 9.83160249910244e-05, + "loss": 2.3377, + "step": 30990 + }, + { + "epoch": 0.08, + "learning_rate": 9.831548287688119e-05, + "loss": 2.427, + "step": 30995 + }, + { + "epoch": 0.08, + "learning_rate": 9.831494067698701e-05, + "loss": 2.3509, + "step": 31000 + }, + { + "epoch": 0.08, + "learning_rate": 9.831439839134278e-05, + "loss": 2.4017, + "step": 31005 + }, + { + "epoch": 0.08, + "learning_rate": 9.83138560199495e-05, + "loss": 2.3316, + "step": 31010 + }, + { + "epoch": 0.08, + "learning_rate": 9.83133135628081e-05, + "loss": 2.2903, + "step": 31015 + }, + { + "epoch": 0.08, + "learning_rate": 9.831277101991956e-05, + "loss": 2.3272, + "step": 31020 + }, + { + "epoch": 0.08, + "learning_rate": 9.831222839128484e-05, + "loss": 2.3881, + "step": 31025 + }, + { + "epoch": 0.08, + "learning_rate": 9.831168567690492e-05, + "loss": 2.3562, + "step": 31030 + }, + { + "epoch": 0.08, + "learning_rate": 9.831114287678073e-05, + "loss": 2.2937, + "step": 31035 + }, + { + "epoch": 0.08, + "learning_rate": 9.831059999091327e-05, + "loss": 2.2818, + "step": 31040 + }, + { + "epoch": 0.08, + "learning_rate": 9.831005701930349e-05, + "loss": 2.4384, + "step": 31045 + }, + { + "epoch": 0.08, + "learning_rate": 9.830951396195234e-05, + "loss": 2.3707, + "step": 31050 + }, + { + "epoch": 0.08, + "learning_rate": 9.830897081886079e-05, + "loss": 2.2838, + "step": 31055 + }, + { + "epoch": 0.08, + "learning_rate": 9.830842759002981e-05, + "loss": 2.506, + "step": 31060 + }, + { + "epoch": 0.08, + "learning_rate": 9.830788427546034e-05, + "loss": 2.3232, + "step": 31065 + }, + { + "epoch": 0.08, + "learning_rate": 9.830734087515339e-05, + "loss": 2.2128, + "step": 31070 + }, + { + "epoch": 0.08, + "learning_rate": 9.830679738910989e-05, + "loss": 2.3492, + "step": 31075 + }, + { + "epoch": 0.08, + "learning_rate": 9.830625381733083e-05, + "loss": 2.2928, + "step": 31080 + }, + { + "epoch": 0.08, + "learning_rate": 9.830571015981713e-05, + "loss": 2.4189, + "step": 31085 + }, + { + "epoch": 0.08, + "learning_rate": 9.830516641656982e-05, + "loss": 2.3326, + "step": 31090 + }, + { + "epoch": 0.08, + "learning_rate": 9.83046225875898e-05, + "loss": 2.3951, + "step": 31095 + }, + { + "epoch": 0.08, + "learning_rate": 9.830407867287808e-05, + "loss": 2.3678, + "step": 31100 + }, + { + "epoch": 0.08, + "learning_rate": 9.830353467243561e-05, + "loss": 2.3508, + "step": 31105 + }, + { + "epoch": 0.08, + "learning_rate": 9.830299058626333e-05, + "loss": 2.4181, + "step": 31110 + }, + { + "epoch": 0.08, + "learning_rate": 9.830244641436226e-05, + "loss": 2.4701, + "step": 31115 + }, + { + "epoch": 0.08, + "learning_rate": 9.830190215673332e-05, + "loss": 2.329, + "step": 31120 + }, + { + "epoch": 0.08, + "learning_rate": 9.83013578133775e-05, + "loss": 2.383, + "step": 31125 + }, + { + "epoch": 0.08, + "learning_rate": 9.830081338429575e-05, + "loss": 2.333, + "step": 31130 + }, + { + "epoch": 0.08, + "learning_rate": 9.830026886948905e-05, + "loss": 2.4334, + "step": 31135 + }, + { + "epoch": 0.08, + "learning_rate": 9.829972426895837e-05, + "loss": 2.395, + "step": 31140 + }, + { + "epoch": 0.08, + "learning_rate": 9.829917958270467e-05, + "loss": 2.4522, + "step": 31145 + }, + { + "epoch": 0.08, + "learning_rate": 9.82986348107289e-05, + "loss": 2.3663, + "step": 31150 + }, + { + "epoch": 0.08, + "learning_rate": 9.829808995303204e-05, + "loss": 2.2708, + "step": 31155 + }, + { + "epoch": 0.08, + "learning_rate": 9.829754500961507e-05, + "loss": 2.432, + "step": 31160 + }, + { + "epoch": 0.08, + "learning_rate": 9.829699998047894e-05, + "loss": 2.4617, + "step": 31165 + }, + { + "epoch": 0.08, + "learning_rate": 9.829645486562463e-05, + "loss": 2.4353, + "step": 31170 + }, + { + "epoch": 0.08, + "learning_rate": 9.829590966505309e-05, + "loss": 2.4967, + "step": 31175 + }, + { + "epoch": 0.08, + "learning_rate": 9.82953643787653e-05, + "loss": 2.5151, + "step": 31180 + }, + { + "epoch": 0.08, + "learning_rate": 9.829481900676222e-05, + "loss": 2.312, + "step": 31185 + }, + { + "epoch": 0.08, + "learning_rate": 9.829427354904483e-05, + "loss": 2.3596, + "step": 31190 + }, + { + "epoch": 0.08, + "learning_rate": 9.829372800561409e-05, + "loss": 2.3328, + "step": 31195 + }, + { + "epoch": 0.08, + "learning_rate": 9.829318237647097e-05, + "loss": 2.4699, + "step": 31200 + }, + { + "epoch": 0.08, + "learning_rate": 9.829263666161644e-05, + "loss": 2.3858, + "step": 31205 + }, + { + "epoch": 0.08, + "learning_rate": 9.829209086105147e-05, + "loss": 2.3188, + "step": 31210 + }, + { + "epoch": 0.08, + "learning_rate": 9.829154497477701e-05, + "loss": 2.3517, + "step": 31215 + }, + { + "epoch": 0.08, + "learning_rate": 9.829099900279407e-05, + "loss": 2.4331, + "step": 31220 + }, + { + "epoch": 0.08, + "learning_rate": 9.829045294510358e-05, + "loss": 2.3365, + "step": 31225 + }, + { + "epoch": 0.08, + "learning_rate": 9.828990680170654e-05, + "loss": 2.3522, + "step": 31230 + }, + { + "epoch": 0.08, + "learning_rate": 9.828936057260387e-05, + "loss": 2.3219, + "step": 31235 + }, + { + "epoch": 0.08, + "learning_rate": 9.82888142577966e-05, + "loss": 2.3306, + "step": 31240 + }, + { + "epoch": 0.08, + "learning_rate": 9.828826785728565e-05, + "loss": 2.2519, + "step": 31245 + }, + { + "epoch": 0.08, + "learning_rate": 9.828772137107201e-05, + "loss": 2.3043, + "step": 31250 + }, + { + "epoch": 0.08, + "learning_rate": 9.828717479915667e-05, + "loss": 2.4015, + "step": 31255 + }, + { + "epoch": 0.08, + "learning_rate": 9.828662814154057e-05, + "loss": 2.357, + "step": 31260 + }, + { + "epoch": 0.08, + "learning_rate": 9.828608139822468e-05, + "loss": 2.3049, + "step": 31265 + }, + { + "epoch": 0.08, + "learning_rate": 9.828553456921001e-05, + "loss": 2.4756, + "step": 31270 + }, + { + "epoch": 0.08, + "learning_rate": 9.828498765449748e-05, + "loss": 2.413, + "step": 31275 + }, + { + "epoch": 0.08, + "learning_rate": 9.828444065408808e-05, + "loss": 2.5032, + "step": 31280 + }, + { + "epoch": 0.08, + "learning_rate": 9.82838935679828e-05, + "loss": 2.3933, + "step": 31285 + }, + { + "epoch": 0.08, + "learning_rate": 9.828334639618258e-05, + "loss": 2.3632, + "step": 31290 + }, + { + "epoch": 0.08, + "learning_rate": 9.82827991386884e-05, + "loss": 2.2604, + "step": 31295 + }, + { + "epoch": 0.08, + "learning_rate": 9.828225179550126e-05, + "loss": 2.4235, + "step": 31300 + }, + { + "epoch": 0.08, + "learning_rate": 9.828170436662209e-05, + "loss": 2.5085, + "step": 31305 + }, + { + "epoch": 0.08, + "learning_rate": 9.828115685205188e-05, + "loss": 2.4931, + "step": 31310 + }, + { + "epoch": 0.08, + "learning_rate": 9.828060925179161e-05, + "loss": 2.3144, + "step": 31315 + }, + { + "epoch": 0.08, + "learning_rate": 9.828006156584224e-05, + "loss": 2.2418, + "step": 31320 + }, + { + "epoch": 0.08, + "learning_rate": 9.827951379420474e-05, + "loss": 2.3214, + "step": 31325 + }, + { + "epoch": 0.08, + "learning_rate": 9.827896593688009e-05, + "loss": 2.3793, + "step": 31330 + }, + { + "epoch": 0.08, + "learning_rate": 9.827841799386926e-05, + "loss": 2.3772, + "step": 31335 + }, + { + "epoch": 0.08, + "learning_rate": 9.827786996517323e-05, + "loss": 2.4034, + "step": 31340 + }, + { + "epoch": 0.08, + "learning_rate": 9.827732185079297e-05, + "loss": 2.3528, + "step": 31345 + }, + { + "epoch": 0.08, + "learning_rate": 9.827677365072942e-05, + "loss": 2.4496, + "step": 31350 + }, + { + "epoch": 0.08, + "learning_rate": 9.827622536498361e-05, + "loss": 2.3902, + "step": 31355 + }, + { + "epoch": 0.08, + "learning_rate": 9.827567699355646e-05, + "loss": 2.4329, + "step": 31360 + }, + { + "epoch": 0.08, + "learning_rate": 9.827512853644898e-05, + "loss": 2.3462, + "step": 31365 + }, + { + "epoch": 0.08, + "learning_rate": 9.827457999366213e-05, + "loss": 2.4624, + "step": 31370 + }, + { + "epoch": 0.08, + "learning_rate": 9.82740313651969e-05, + "loss": 2.2585, + "step": 31375 + }, + { + "epoch": 0.08, + "learning_rate": 9.827348265105423e-05, + "loss": 2.3545, + "step": 31380 + }, + { + "epoch": 0.08, + "learning_rate": 9.827293385123511e-05, + "loss": 2.4451, + "step": 31385 + }, + { + "epoch": 0.08, + "learning_rate": 9.827238496574052e-05, + "loss": 2.4667, + "step": 31390 + }, + { + "epoch": 0.08, + "learning_rate": 9.827183599457144e-05, + "loss": 2.2896, + "step": 31395 + }, + { + "epoch": 0.08, + "learning_rate": 9.827128693772882e-05, + "loss": 2.33, + "step": 31400 + }, + { + "epoch": 0.08, + "learning_rate": 9.827073779521366e-05, + "loss": 2.3866, + "step": 31405 + }, + { + "epoch": 0.08, + "learning_rate": 9.82701885670269e-05, + "loss": 2.2149, + "step": 31410 + }, + { + "epoch": 0.08, + "learning_rate": 9.826963925316957e-05, + "loss": 2.2156, + "step": 31415 + }, + { + "epoch": 0.08, + "learning_rate": 9.826908985364258e-05, + "loss": 2.407, + "step": 31420 + }, + { + "epoch": 0.08, + "learning_rate": 9.826854036844696e-05, + "loss": 2.4897, + "step": 31425 + }, + { + "epoch": 0.08, + "learning_rate": 9.826799079758367e-05, + "loss": 2.2516, + "step": 31430 + }, + { + "epoch": 0.08, + "learning_rate": 9.826744114105366e-05, + "loss": 2.3103, + "step": 31435 + }, + { + "epoch": 0.08, + "learning_rate": 9.826689139885794e-05, + "loss": 2.3747, + "step": 31440 + }, + { + "epoch": 0.08, + "learning_rate": 9.826634157099745e-05, + "loss": 2.4115, + "step": 31445 + }, + { + "epoch": 0.08, + "learning_rate": 9.82657916574732e-05, + "loss": 2.4377, + "step": 31450 + }, + { + "epoch": 0.08, + "learning_rate": 9.826524165828615e-05, + "loss": 2.4583, + "step": 31455 + }, + { + "epoch": 0.08, + "learning_rate": 9.826469157343728e-05, + "loss": 2.3237, + "step": 31460 + }, + { + "epoch": 0.08, + "learning_rate": 9.826414140292755e-05, + "loss": 2.4003, + "step": 31465 + }, + { + "epoch": 0.08, + "learning_rate": 9.826359114675797e-05, + "loss": 2.4149, + "step": 31470 + }, + { + "epoch": 0.08, + "learning_rate": 9.82630408049295e-05, + "loss": 2.2488, + "step": 31475 + }, + { + "epoch": 0.08, + "learning_rate": 9.82624903774431e-05, + "loss": 2.3548, + "step": 31480 + }, + { + "epoch": 0.08, + "learning_rate": 9.826193986429976e-05, + "loss": 2.4327, + "step": 31485 + }, + { + "epoch": 0.08, + "learning_rate": 9.826138926550046e-05, + "loss": 2.3869, + "step": 31490 + }, + { + "epoch": 0.08, + "learning_rate": 9.826083858104618e-05, + "loss": 2.4083, + "step": 31495 + }, + { + "epoch": 0.08, + "learning_rate": 9.82602878109379e-05, + "loss": 2.4087, + "step": 31500 + }, + { + "epoch": 0.08, + "learning_rate": 9.825973695517658e-05, + "loss": 2.3457, + "step": 31505 + }, + { + "epoch": 0.08, + "learning_rate": 9.82591860137632e-05, + "loss": 2.406, + "step": 31510 + }, + { + "epoch": 0.08, + "learning_rate": 9.825863498669877e-05, + "loss": 2.473, + "step": 31515 + }, + { + "epoch": 0.08, + "learning_rate": 9.825808387398421e-05, + "loss": 2.4347, + "step": 31520 + }, + { + "epoch": 0.08, + "learning_rate": 9.825753267562055e-05, + "loss": 2.4298, + "step": 31525 + }, + { + "epoch": 0.08, + "learning_rate": 9.825698139160876e-05, + "loss": 2.343, + "step": 31530 + }, + { + "epoch": 0.08, + "learning_rate": 9.82564300219498e-05, + "loss": 2.3245, + "step": 31535 + }, + { + "epoch": 0.08, + "learning_rate": 9.825587856664465e-05, + "loss": 2.2345, + "step": 31540 + }, + { + "epoch": 0.08, + "learning_rate": 9.82553270256943e-05, + "loss": 2.3544, + "step": 31545 + }, + { + "epoch": 0.08, + "learning_rate": 9.825477539909973e-05, + "loss": 2.3903, + "step": 31550 + }, + { + "epoch": 0.08, + "learning_rate": 9.82542236868619e-05, + "loss": 2.3512, + "step": 31555 + }, + { + "epoch": 0.08, + "learning_rate": 9.825367188898181e-05, + "loss": 2.4361, + "step": 31560 + }, + { + "epoch": 0.08, + "learning_rate": 9.825312000546044e-05, + "loss": 2.4464, + "step": 31565 + }, + { + "epoch": 0.08, + "learning_rate": 9.825256803629875e-05, + "loss": 2.3676, + "step": 31570 + }, + { + "epoch": 0.08, + "learning_rate": 9.825201598149774e-05, + "loss": 2.3612, + "step": 31575 + }, + { + "epoch": 0.08, + "learning_rate": 9.82514638410584e-05, + "loss": 2.4399, + "step": 31580 + }, + { + "epoch": 0.08, + "learning_rate": 9.825091161498165e-05, + "loss": 2.2946, + "step": 31585 + }, + { + "epoch": 0.08, + "learning_rate": 9.825035930326854e-05, + "loss": 2.3482, + "step": 31590 + }, + { + "epoch": 0.08, + "learning_rate": 9.824980690592001e-05, + "loss": 2.4603, + "step": 31595 + }, + { + "epoch": 0.08, + "learning_rate": 9.824925442293704e-05, + "loss": 2.4098, + "step": 31600 + }, + { + "epoch": 0.08, + "learning_rate": 9.824870185432065e-05, + "loss": 2.4227, + "step": 31605 + }, + { + "epoch": 0.08, + "learning_rate": 9.824814920007177e-05, + "loss": 2.347, + "step": 31610 + }, + { + "epoch": 0.08, + "learning_rate": 9.824759646019141e-05, + "loss": 2.4863, + "step": 31615 + }, + { + "epoch": 0.08, + "learning_rate": 9.824704363468055e-05, + "loss": 2.3767, + "step": 31620 + }, + { + "epoch": 0.08, + "learning_rate": 9.824649072354017e-05, + "loss": 2.3272, + "step": 31625 + }, + { + "epoch": 0.08, + "learning_rate": 9.824593772677124e-05, + "loss": 2.437, + "step": 31630 + }, + { + "epoch": 0.08, + "learning_rate": 9.824538464437473e-05, + "loss": 2.2836, + "step": 31635 + }, + { + "epoch": 0.08, + "learning_rate": 9.824483147635168e-05, + "loss": 2.4255, + "step": 31640 + }, + { + "epoch": 0.08, + "learning_rate": 9.824427822270302e-05, + "loss": 2.2305, + "step": 31645 + }, + { + "epoch": 0.08, + "learning_rate": 9.824372488342972e-05, + "loss": 2.2532, + "step": 31650 + }, + { + "epoch": 0.08, + "learning_rate": 9.824317145853281e-05, + "loss": 2.3159, + "step": 31655 + }, + { + "epoch": 0.08, + "learning_rate": 9.824261794801323e-05, + "loss": 2.2781, + "step": 31660 + }, + { + "epoch": 0.08, + "learning_rate": 9.8242064351872e-05, + "loss": 2.3193, + "step": 31665 + }, + { + "epoch": 0.08, + "learning_rate": 9.824151067011008e-05, + "loss": 2.3712, + "step": 31670 + }, + { + "epoch": 0.08, + "learning_rate": 9.824095690272845e-05, + "loss": 2.4556, + "step": 31675 + }, + { + "epoch": 0.08, + "learning_rate": 9.824040304972811e-05, + "loss": 2.362, + "step": 31680 + }, + { + "epoch": 0.08, + "learning_rate": 9.823984911111002e-05, + "loss": 2.3493, + "step": 31685 + }, + { + "epoch": 0.08, + "learning_rate": 9.823929508687517e-05, + "loss": 2.3012, + "step": 31690 + }, + { + "epoch": 0.08, + "learning_rate": 9.823874097702456e-05, + "loss": 2.4252, + "step": 31695 + }, + { + "epoch": 0.08, + "learning_rate": 9.823818678155916e-05, + "loss": 2.3585, + "step": 31700 + }, + { + "epoch": 0.09, + "learning_rate": 9.823763250047994e-05, + "loss": 2.3476, + "step": 31705 + }, + { + "epoch": 0.09, + "learning_rate": 9.823707813378793e-05, + "loss": 2.4129, + "step": 31710 + }, + { + "epoch": 0.09, + "learning_rate": 9.823652368148405e-05, + "loss": 2.4023, + "step": 31715 + }, + { + "epoch": 0.09, + "learning_rate": 9.823596914356933e-05, + "loss": 2.4196, + "step": 31720 + }, + { + "epoch": 0.09, + "learning_rate": 9.823541452004475e-05, + "loss": 2.396, + "step": 31725 + }, + { + "epoch": 0.09, + "learning_rate": 9.823485981091128e-05, + "loss": 2.4136, + "step": 31730 + }, + { + "epoch": 0.09, + "learning_rate": 9.823430501616991e-05, + "loss": 2.3751, + "step": 31735 + }, + { + "epoch": 0.09, + "learning_rate": 9.823375013582163e-05, + "loss": 2.2795, + "step": 31740 + }, + { + "epoch": 0.09, + "learning_rate": 9.823319516986742e-05, + "loss": 2.4119, + "step": 31745 + }, + { + "epoch": 0.09, + "learning_rate": 9.823264011830824e-05, + "loss": 2.3069, + "step": 31750 + }, + { + "epoch": 0.09, + "learning_rate": 9.823208498114512e-05, + "loss": 2.3992, + "step": 31755 + }, + { + "epoch": 0.09, + "learning_rate": 9.823152975837903e-05, + "loss": 2.3897, + "step": 31760 + }, + { + "epoch": 0.09, + "learning_rate": 9.823097445001094e-05, + "loss": 2.4577, + "step": 31765 + }, + { + "epoch": 0.09, + "learning_rate": 9.823041905604185e-05, + "loss": 2.4, + "step": 31770 + }, + { + "epoch": 0.09, + "learning_rate": 9.822986357647275e-05, + "loss": 2.3739, + "step": 31775 + }, + { + "epoch": 0.09, + "learning_rate": 9.82293080113046e-05, + "loss": 2.356, + "step": 31780 + }, + { + "epoch": 0.09, + "learning_rate": 9.822875236053841e-05, + "loss": 2.499, + "step": 31785 + }, + { + "epoch": 0.09, + "learning_rate": 9.822819662417516e-05, + "loss": 2.424, + "step": 31790 + }, + { + "epoch": 0.09, + "learning_rate": 9.822764080221583e-05, + "loss": 2.3587, + "step": 31795 + }, + { + "epoch": 0.09, + "learning_rate": 9.822708489466143e-05, + "loss": 2.4139, + "step": 31800 + }, + { + "epoch": 0.09, + "learning_rate": 9.822652890151291e-05, + "loss": 2.4222, + "step": 31805 + }, + { + "epoch": 0.09, + "learning_rate": 9.822597282277129e-05, + "loss": 2.3716, + "step": 31810 + }, + { + "epoch": 0.09, + "learning_rate": 9.822541665843754e-05, + "loss": 2.4274, + "step": 31815 + }, + { + "epoch": 0.09, + "learning_rate": 9.822486040851264e-05, + "loss": 2.4829, + "step": 31820 + }, + { + "epoch": 0.09, + "learning_rate": 9.82243040729976e-05, + "loss": 2.3482, + "step": 31825 + }, + { + "epoch": 0.09, + "learning_rate": 9.82237476518934e-05, + "loss": 2.4368, + "step": 31830 + }, + { + "epoch": 0.09, + "learning_rate": 9.8223191145201e-05, + "loss": 2.3897, + "step": 31835 + }, + { + "epoch": 0.09, + "learning_rate": 9.822263455292142e-05, + "loss": 2.3835, + "step": 31840 + }, + { + "epoch": 0.09, + "learning_rate": 9.822207787505564e-05, + "loss": 2.486, + "step": 31845 + }, + { + "epoch": 0.09, + "learning_rate": 9.822152111160465e-05, + "loss": 2.3844, + "step": 31850 + }, + { + "epoch": 0.09, + "learning_rate": 9.822096426256943e-05, + "loss": 2.3357, + "step": 31855 + }, + { + "epoch": 0.09, + "learning_rate": 9.822040732795097e-05, + "loss": 2.4314, + "step": 31860 + }, + { + "epoch": 0.09, + "learning_rate": 9.821985030775025e-05, + "loss": 2.3604, + "step": 31865 + }, + { + "epoch": 0.09, + "learning_rate": 9.821929320196828e-05, + "loss": 2.3769, + "step": 31870 + }, + { + "epoch": 0.09, + "learning_rate": 9.821873601060605e-05, + "loss": 2.442, + "step": 31875 + }, + { + "epoch": 0.09, + "learning_rate": 9.821817873366452e-05, + "loss": 2.2075, + "step": 31880 + }, + { + "epoch": 0.09, + "learning_rate": 9.82176213711447e-05, + "loss": 2.3335, + "step": 31885 + }, + { + "epoch": 0.09, + "learning_rate": 9.821706392304757e-05, + "loss": 2.3895, + "step": 31890 + }, + { + "epoch": 0.09, + "learning_rate": 9.821650638937414e-05, + "loss": 2.3374, + "step": 31895 + }, + { + "epoch": 0.09, + "learning_rate": 9.821594877012536e-05, + "loss": 2.4475, + "step": 31900 + }, + { + "epoch": 0.09, + "learning_rate": 9.821539106530226e-05, + "loss": 2.3131, + "step": 31905 + }, + { + "epoch": 0.09, + "learning_rate": 9.821483327490581e-05, + "loss": 2.3372, + "step": 31910 + }, + { + "epoch": 0.09, + "learning_rate": 9.821427539893702e-05, + "loss": 2.2413, + "step": 31915 + }, + { + "epoch": 0.09, + "learning_rate": 9.821371743739685e-05, + "loss": 2.4248, + "step": 31920 + }, + { + "epoch": 0.09, + "learning_rate": 9.82131593902863e-05, + "loss": 2.3014, + "step": 31925 + }, + { + "epoch": 0.09, + "learning_rate": 9.821260125760637e-05, + "loss": 2.4232, + "step": 31930 + }, + { + "epoch": 0.09, + "learning_rate": 9.821204303935803e-05, + "loss": 2.4394, + "step": 31935 + }, + { + "epoch": 0.09, + "learning_rate": 9.82114847355423e-05, + "loss": 2.3836, + "step": 31940 + }, + { + "epoch": 0.09, + "learning_rate": 9.821092634616016e-05, + "loss": 2.3435, + "step": 31945 + }, + { + "epoch": 0.09, + "learning_rate": 9.821036787121259e-05, + "loss": 2.3315, + "step": 31950 + }, + { + "epoch": 0.09, + "learning_rate": 9.820980931070059e-05, + "loss": 2.3118, + "step": 31955 + }, + { + "epoch": 0.09, + "learning_rate": 9.820925066462515e-05, + "loss": 2.2777, + "step": 31960 + }, + { + "epoch": 0.09, + "learning_rate": 9.820869193298726e-05, + "loss": 2.4006, + "step": 31965 + }, + { + "epoch": 0.09, + "learning_rate": 9.820813311578792e-05, + "loss": 2.3444, + "step": 31970 + }, + { + "epoch": 0.09, + "learning_rate": 9.820757421302811e-05, + "loss": 2.4382, + "step": 31975 + }, + { + "epoch": 0.09, + "learning_rate": 9.820701522470883e-05, + "loss": 2.3511, + "step": 31980 + }, + { + "epoch": 0.09, + "learning_rate": 9.820645615083106e-05, + "loss": 2.3835, + "step": 31985 + }, + { + "epoch": 0.09, + "learning_rate": 9.82058969913958e-05, + "loss": 2.4546, + "step": 31990 + }, + { + "epoch": 0.09, + "learning_rate": 9.820533774640406e-05, + "loss": 2.3231, + "step": 31995 + }, + { + "epoch": 0.09, + "learning_rate": 9.820477841585681e-05, + "loss": 2.3826, + "step": 32000 + }, + { + "epoch": 0.09, + "learning_rate": 9.820421899975505e-05, + "loss": 2.4803, + "step": 32005 + }, + { + "epoch": 0.09, + "learning_rate": 9.820365949809976e-05, + "loss": 2.4291, + "step": 32010 + }, + { + "epoch": 0.09, + "learning_rate": 9.820309991089196e-05, + "loss": 2.3649, + "step": 32015 + }, + { + "epoch": 0.09, + "learning_rate": 9.820254023813262e-05, + "loss": 2.2223, + "step": 32020 + }, + { + "epoch": 0.09, + "learning_rate": 9.820198047982274e-05, + "loss": 2.3622, + "step": 32025 + }, + { + "epoch": 0.09, + "learning_rate": 9.820142063596332e-05, + "loss": 2.3589, + "step": 32030 + }, + { + "epoch": 0.09, + "learning_rate": 9.820086070655535e-05, + "loss": 2.3784, + "step": 32035 + }, + { + "epoch": 0.09, + "learning_rate": 9.820030069159981e-05, + "loss": 2.4127, + "step": 32040 + }, + { + "epoch": 0.09, + "learning_rate": 9.819974059109771e-05, + "loss": 2.3947, + "step": 32045 + }, + { + "epoch": 0.09, + "learning_rate": 9.819918040505006e-05, + "loss": 2.3401, + "step": 32050 + }, + { + "epoch": 0.09, + "learning_rate": 9.81986201334578e-05, + "loss": 2.3545, + "step": 32055 + }, + { + "epoch": 0.09, + "learning_rate": 9.819805977632198e-05, + "loss": 2.3665, + "step": 32060 + }, + { + "epoch": 0.09, + "learning_rate": 9.819749933364357e-05, + "loss": 2.4312, + "step": 32065 + }, + { + "epoch": 0.09, + "learning_rate": 9.819693880542357e-05, + "loss": 2.3261, + "step": 32070 + }, + { + "epoch": 0.09, + "learning_rate": 9.819637819166297e-05, + "loss": 2.3756, + "step": 32075 + }, + { + "epoch": 0.09, + "learning_rate": 9.819581749236278e-05, + "loss": 2.3356, + "step": 32080 + }, + { + "epoch": 0.09, + "learning_rate": 9.819525670752396e-05, + "loss": 2.271, + "step": 32085 + }, + { + "epoch": 0.09, + "learning_rate": 9.819469583714756e-05, + "loss": 2.325, + "step": 32090 + }, + { + "epoch": 0.09, + "learning_rate": 9.81941348812345e-05, + "loss": 2.3933, + "step": 32095 + }, + { + "epoch": 0.09, + "learning_rate": 9.819357383978586e-05, + "loss": 2.3754, + "step": 32100 + }, + { + "epoch": 0.09, + "learning_rate": 9.819301271280257e-05, + "loss": 2.376, + "step": 32105 + }, + { + "epoch": 0.09, + "learning_rate": 9.819245150028567e-05, + "loss": 2.3338, + "step": 32110 + }, + { + "epoch": 0.09, + "learning_rate": 9.819189020223611e-05, + "loss": 2.3076, + "step": 32115 + }, + { + "epoch": 0.09, + "learning_rate": 9.819132881865493e-05, + "loss": 2.3962, + "step": 32120 + }, + { + "epoch": 0.09, + "learning_rate": 9.819076734954311e-05, + "loss": 2.301, + "step": 32125 + }, + { + "epoch": 0.09, + "learning_rate": 9.819020579490166e-05, + "loss": 2.492, + "step": 32130 + }, + { + "epoch": 0.09, + "learning_rate": 9.818964415473154e-05, + "loss": 2.3483, + "step": 32135 + }, + { + "epoch": 0.09, + "learning_rate": 9.818908242903378e-05, + "loss": 2.389, + "step": 32140 + }, + { + "epoch": 0.09, + "learning_rate": 9.818852061780936e-05, + "loss": 2.4563, + "step": 32145 + }, + { + "epoch": 0.09, + "learning_rate": 9.818795872105929e-05, + "loss": 2.391, + "step": 32150 + }, + { + "epoch": 0.09, + "learning_rate": 9.818739673878457e-05, + "loss": 2.3898, + "step": 32155 + }, + { + "epoch": 0.09, + "learning_rate": 9.818683467098616e-05, + "loss": 2.3738, + "step": 32160 + }, + { + "epoch": 0.09, + "learning_rate": 9.81862725176651e-05, + "loss": 2.3077, + "step": 32165 + }, + { + "epoch": 0.09, + "learning_rate": 9.818571027882238e-05, + "loss": 2.4003, + "step": 32170 + }, + { + "epoch": 0.09, + "learning_rate": 9.818514795445898e-05, + "loss": 2.4735, + "step": 32175 + }, + { + "epoch": 0.09, + "learning_rate": 9.818458554457593e-05, + "loss": 2.4204, + "step": 32180 + }, + { + "epoch": 0.09, + "learning_rate": 9.818402304917419e-05, + "loss": 2.393, + "step": 32185 + }, + { + "epoch": 0.09, + "learning_rate": 9.818346046825477e-05, + "loss": 2.2739, + "step": 32190 + }, + { + "epoch": 0.09, + "learning_rate": 9.818289780181869e-05, + "loss": 2.4642, + "step": 32195 + }, + { + "epoch": 0.09, + "learning_rate": 9.818233504986694e-05, + "loss": 2.2978, + "step": 32200 + }, + { + "epoch": 0.09, + "learning_rate": 9.818177221240048e-05, + "loss": 2.3595, + "step": 32205 + }, + { + "epoch": 0.09, + "learning_rate": 9.818120928942036e-05, + "loss": 2.4059, + "step": 32210 + }, + { + "epoch": 0.09, + "learning_rate": 9.818064628092756e-05, + "loss": 2.3769, + "step": 32215 + }, + { + "epoch": 0.09, + "learning_rate": 9.818008318692309e-05, + "loss": 2.3697, + "step": 32220 + }, + { + "epoch": 0.09, + "learning_rate": 9.817952000740793e-05, + "loss": 2.5155, + "step": 32225 + }, + { + "epoch": 0.09, + "learning_rate": 9.817895674238308e-05, + "loss": 2.2983, + "step": 32230 + }, + { + "epoch": 0.09, + "learning_rate": 9.817839339184955e-05, + "loss": 2.4855, + "step": 32235 + }, + { + "epoch": 0.09, + "learning_rate": 9.817782995580834e-05, + "loss": 2.4927, + "step": 32240 + }, + { + "epoch": 0.09, + "learning_rate": 9.817726643426044e-05, + "loss": 2.3683, + "step": 32245 + }, + { + "epoch": 0.09, + "learning_rate": 9.817670282720688e-05, + "loss": 2.384, + "step": 32250 + }, + { + "epoch": 0.09, + "learning_rate": 9.817613913464862e-05, + "loss": 2.3773, + "step": 32255 + }, + { + "epoch": 0.09, + "learning_rate": 9.817557535658668e-05, + "loss": 2.3456, + "step": 32260 + }, + { + "epoch": 0.09, + "learning_rate": 9.817501149302206e-05, + "loss": 2.2672, + "step": 32265 + }, + { + "epoch": 0.09, + "learning_rate": 9.817444754395575e-05, + "loss": 2.3597, + "step": 32270 + }, + { + "epoch": 0.09, + "learning_rate": 9.817388350938877e-05, + "loss": 2.3227, + "step": 32275 + }, + { + "epoch": 0.09, + "learning_rate": 9.817331938932212e-05, + "loss": 2.4671, + "step": 32280 + }, + { + "epoch": 0.09, + "learning_rate": 9.817275518375678e-05, + "loss": 2.4655, + "step": 32285 + }, + { + "epoch": 0.09, + "learning_rate": 9.817219089269378e-05, + "loss": 2.3749, + "step": 32290 + }, + { + "epoch": 0.09, + "learning_rate": 9.817162651613409e-05, + "loss": 2.3344, + "step": 32295 + }, + { + "epoch": 0.09, + "learning_rate": 9.817106205407873e-05, + "loss": 2.5011, + "step": 32300 + }, + { + "epoch": 0.09, + "learning_rate": 9.817049750652871e-05, + "loss": 2.469, + "step": 32305 + }, + { + "epoch": 0.09, + "learning_rate": 9.8169932873485e-05, + "loss": 2.3828, + "step": 32310 + }, + { + "epoch": 0.09, + "learning_rate": 9.816936815494864e-05, + "loss": 2.3791, + "step": 32315 + }, + { + "epoch": 0.09, + "learning_rate": 9.816880335092062e-05, + "loss": 2.3476, + "step": 32320 + }, + { + "epoch": 0.09, + "learning_rate": 9.816823846140192e-05, + "loss": 2.5024, + "step": 32325 + }, + { + "epoch": 0.09, + "learning_rate": 9.816767348639358e-05, + "loss": 2.4357, + "step": 32330 + }, + { + "epoch": 0.09, + "learning_rate": 9.816710842589657e-05, + "loss": 2.3863, + "step": 32335 + }, + { + "epoch": 0.09, + "learning_rate": 9.816654327991191e-05, + "loss": 2.4424, + "step": 32340 + }, + { + "epoch": 0.09, + "learning_rate": 9.81659780484406e-05, + "loss": 2.3992, + "step": 32345 + }, + { + "epoch": 0.09, + "learning_rate": 9.816541273148364e-05, + "loss": 2.3731, + "step": 32350 + }, + { + "epoch": 0.09, + "learning_rate": 9.816484732904203e-05, + "loss": 2.3326, + "step": 32355 + }, + { + "epoch": 0.09, + "learning_rate": 9.816428184111679e-05, + "loss": 2.2489, + "step": 32360 + }, + { + "epoch": 0.09, + "learning_rate": 9.816371626770891e-05, + "loss": 2.3505, + "step": 32365 + }, + { + "epoch": 0.09, + "learning_rate": 9.81631506088194e-05, + "loss": 2.3469, + "step": 32370 + }, + { + "epoch": 0.09, + "learning_rate": 9.816258486444926e-05, + "loss": 2.3147, + "step": 32375 + }, + { + "epoch": 0.09, + "learning_rate": 9.81620190345995e-05, + "loss": 2.3128, + "step": 32380 + }, + { + "epoch": 0.09, + "learning_rate": 9.816145311927109e-05, + "loss": 2.3583, + "step": 32385 + }, + { + "epoch": 0.09, + "learning_rate": 9.81608871184651e-05, + "loss": 2.2817, + "step": 32390 + }, + { + "epoch": 0.09, + "learning_rate": 9.816032103218248e-05, + "loss": 2.3831, + "step": 32395 + }, + { + "epoch": 0.09, + "learning_rate": 9.815975486042426e-05, + "loss": 2.3505, + "step": 32400 + }, + { + "epoch": 0.09, + "learning_rate": 9.815918860319142e-05, + "loss": 2.3858, + "step": 32405 + }, + { + "epoch": 0.09, + "learning_rate": 9.8158622260485e-05, + "loss": 2.4306, + "step": 32410 + }, + { + "epoch": 0.09, + "learning_rate": 9.815805583230597e-05, + "loss": 2.4552, + "step": 32415 + }, + { + "epoch": 0.09, + "learning_rate": 9.815748931865537e-05, + "loss": 2.2617, + "step": 32420 + }, + { + "epoch": 0.09, + "learning_rate": 9.815692271953419e-05, + "loss": 2.3656, + "step": 32425 + }, + { + "epoch": 0.09, + "learning_rate": 9.815635603494342e-05, + "loss": 2.4453, + "step": 32430 + }, + { + "epoch": 0.09, + "learning_rate": 9.815578926488409e-05, + "loss": 2.3527, + "step": 32435 + }, + { + "epoch": 0.09, + "learning_rate": 9.815522240935719e-05, + "loss": 2.4465, + "step": 32440 + }, + { + "epoch": 0.09, + "learning_rate": 9.815465546836373e-05, + "loss": 2.3892, + "step": 32445 + }, + { + "epoch": 0.09, + "learning_rate": 9.815408844190473e-05, + "loss": 2.3301, + "step": 32450 + }, + { + "epoch": 0.09, + "learning_rate": 9.815352132998117e-05, + "loss": 2.2864, + "step": 32455 + }, + { + "epoch": 0.09, + "learning_rate": 9.815295413259409e-05, + "loss": 2.4331, + "step": 32460 + }, + { + "epoch": 0.09, + "learning_rate": 9.815238684974446e-05, + "loss": 2.4562, + "step": 32465 + }, + { + "epoch": 0.09, + "learning_rate": 9.81518194814333e-05, + "loss": 2.4398, + "step": 32470 + }, + { + "epoch": 0.09, + "learning_rate": 9.815125202766163e-05, + "loss": 2.3653, + "step": 32475 + }, + { + "epoch": 0.09, + "learning_rate": 9.815068448843045e-05, + "loss": 2.4967, + "step": 32480 + }, + { + "epoch": 0.09, + "learning_rate": 9.815011686374076e-05, + "loss": 2.426, + "step": 32485 + }, + { + "epoch": 0.09, + "learning_rate": 9.814954915359358e-05, + "loss": 2.3261, + "step": 32490 + }, + { + "epoch": 0.09, + "learning_rate": 9.814898135798992e-05, + "loss": 2.2134, + "step": 32495 + }, + { + "epoch": 0.09, + "learning_rate": 9.814841347693075e-05, + "loss": 2.4057, + "step": 32500 + }, + { + "epoch": 0.09, + "learning_rate": 9.814784551041713e-05, + "loss": 2.2999, + "step": 32505 + }, + { + "epoch": 0.09, + "learning_rate": 9.814727745845002e-05, + "loss": 2.348, + "step": 32510 + }, + { + "epoch": 0.09, + "learning_rate": 9.814670932103047e-05, + "loss": 2.3466, + "step": 32515 + }, + { + "epoch": 0.09, + "learning_rate": 9.814614109815947e-05, + "loss": 2.3087, + "step": 32520 + }, + { + "epoch": 0.09, + "learning_rate": 9.814557278983804e-05, + "loss": 2.3214, + "step": 32525 + }, + { + "epoch": 0.09, + "learning_rate": 9.814500439606716e-05, + "loss": 2.3625, + "step": 32530 + }, + { + "epoch": 0.09, + "learning_rate": 9.814443591684788e-05, + "loss": 2.3802, + "step": 32535 + }, + { + "epoch": 0.09, + "learning_rate": 9.814386735218115e-05, + "loss": 2.2907, + "step": 32540 + }, + { + "epoch": 0.09, + "learning_rate": 9.814329870206804e-05, + "loss": 2.4554, + "step": 32545 + }, + { + "epoch": 0.09, + "learning_rate": 9.814272996650953e-05, + "loss": 2.4813, + "step": 32550 + }, + { + "epoch": 0.09, + "learning_rate": 9.814216114550663e-05, + "loss": 2.3728, + "step": 32555 + }, + { + "epoch": 0.09, + "learning_rate": 9.814159223906036e-05, + "loss": 2.3775, + "step": 32560 + }, + { + "epoch": 0.09, + "learning_rate": 9.814102324717172e-05, + "loss": 2.3526, + "step": 32565 + }, + { + "epoch": 0.09, + "learning_rate": 9.814045416984171e-05, + "loss": 2.3147, + "step": 32570 + }, + { + "epoch": 0.09, + "learning_rate": 9.813988500707136e-05, + "loss": 2.329, + "step": 32575 + }, + { + "epoch": 0.09, + "learning_rate": 9.813931575886169e-05, + "loss": 2.3464, + "step": 32580 + }, + { + "epoch": 0.09, + "learning_rate": 9.813874642521367e-05, + "loss": 2.4605, + "step": 32585 + }, + { + "epoch": 0.09, + "learning_rate": 9.813817700612834e-05, + "loss": 2.2421, + "step": 32590 + }, + { + "epoch": 0.09, + "learning_rate": 9.81376075016067e-05, + "loss": 2.3841, + "step": 32595 + }, + { + "epoch": 0.09, + "learning_rate": 9.813703791164977e-05, + "loss": 2.317, + "step": 32600 + }, + { + "epoch": 0.09, + "learning_rate": 9.813646823625856e-05, + "loss": 2.315, + "step": 32605 + }, + { + "epoch": 0.09, + "learning_rate": 9.813589847543408e-05, + "loss": 2.4042, + "step": 32610 + }, + { + "epoch": 0.09, + "learning_rate": 9.813532862917733e-05, + "loss": 2.4361, + "step": 32615 + }, + { + "epoch": 0.09, + "learning_rate": 9.813475869748932e-05, + "loss": 2.307, + "step": 32620 + }, + { + "epoch": 0.09, + "learning_rate": 9.813418868037108e-05, + "loss": 2.4261, + "step": 32625 + }, + { + "epoch": 0.09, + "learning_rate": 9.813361857782361e-05, + "loss": 2.2656, + "step": 32630 + }, + { + "epoch": 0.09, + "learning_rate": 9.813304838984793e-05, + "loss": 2.3819, + "step": 32635 + }, + { + "epoch": 0.09, + "learning_rate": 9.813247811644506e-05, + "loss": 2.3249, + "step": 32640 + }, + { + "epoch": 0.09, + "learning_rate": 9.813190775761598e-05, + "loss": 2.4462, + "step": 32645 + }, + { + "epoch": 0.09, + "learning_rate": 9.813133731336173e-05, + "loss": 2.2666, + "step": 32650 + }, + { + "epoch": 0.09, + "learning_rate": 9.81307667836833e-05, + "loss": 2.3146, + "step": 32655 + }, + { + "epoch": 0.09, + "learning_rate": 9.813019616858173e-05, + "loss": 2.3544, + "step": 32660 + }, + { + "epoch": 0.09, + "learning_rate": 9.812962546805801e-05, + "loss": 2.3756, + "step": 32665 + }, + { + "epoch": 0.09, + "learning_rate": 9.812905468211315e-05, + "loss": 2.3675, + "step": 32670 + }, + { + "epoch": 0.09, + "learning_rate": 9.81284838107482e-05, + "loss": 2.423, + "step": 32675 + }, + { + "epoch": 0.09, + "learning_rate": 9.812791285396414e-05, + "loss": 2.3302, + "step": 32680 + }, + { + "epoch": 0.09, + "learning_rate": 9.812734181176198e-05, + "loss": 2.3018, + "step": 32685 + }, + { + "epoch": 0.09, + "learning_rate": 9.812677068414275e-05, + "loss": 2.402, + "step": 32690 + }, + { + "epoch": 0.09, + "learning_rate": 9.812619947110747e-05, + "loss": 2.3164, + "step": 32695 + }, + { + "epoch": 0.09, + "learning_rate": 9.812562817265712e-05, + "loss": 2.285, + "step": 32700 + }, + { + "epoch": 0.09, + "learning_rate": 9.812505678879276e-05, + "loss": 2.4471, + "step": 32705 + }, + { + "epoch": 0.09, + "learning_rate": 9.812448531951535e-05, + "loss": 2.4698, + "step": 32710 + }, + { + "epoch": 0.09, + "learning_rate": 9.812391376482595e-05, + "loss": 2.1911, + "step": 32715 + }, + { + "epoch": 0.09, + "learning_rate": 9.812334212472557e-05, + "loss": 2.3876, + "step": 32720 + }, + { + "epoch": 0.09, + "learning_rate": 9.81227703992152e-05, + "loss": 2.3982, + "step": 32725 + }, + { + "epoch": 0.09, + "learning_rate": 9.812219858829588e-05, + "loss": 2.4015, + "step": 32730 + }, + { + "epoch": 0.09, + "learning_rate": 9.81216266919686e-05, + "loss": 2.2416, + "step": 32735 + }, + { + "epoch": 0.09, + "learning_rate": 9.812105471023438e-05, + "loss": 2.3388, + "step": 32740 + }, + { + "epoch": 0.09, + "learning_rate": 9.812048264309428e-05, + "loss": 2.3834, + "step": 32745 + }, + { + "epoch": 0.09, + "learning_rate": 9.811991049054923e-05, + "loss": 2.3997, + "step": 32750 + }, + { + "epoch": 0.09, + "learning_rate": 9.811933825260034e-05, + "loss": 2.353, + "step": 32755 + }, + { + "epoch": 0.09, + "learning_rate": 9.811876592924854e-05, + "loss": 2.4441, + "step": 32760 + }, + { + "epoch": 0.09, + "learning_rate": 9.81181935204949e-05, + "loss": 2.395, + "step": 32765 + }, + { + "epoch": 0.09, + "learning_rate": 9.811762102634043e-05, + "loss": 2.4582, + "step": 32770 + }, + { + "epoch": 0.09, + "learning_rate": 9.811704844678613e-05, + "loss": 2.4225, + "step": 32775 + }, + { + "epoch": 0.09, + "learning_rate": 9.811647578183303e-05, + "loss": 2.46, + "step": 32780 + }, + { + "epoch": 0.09, + "learning_rate": 9.811590303148213e-05, + "loss": 2.4733, + "step": 32785 + }, + { + "epoch": 0.09, + "learning_rate": 9.811533019573445e-05, + "loss": 2.3648, + "step": 32790 + }, + { + "epoch": 0.09, + "learning_rate": 9.811475727459103e-05, + "loss": 2.4115, + "step": 32795 + }, + { + "epoch": 0.09, + "learning_rate": 9.811418426805286e-05, + "loss": 2.4732, + "step": 32800 + }, + { + "epoch": 0.09, + "learning_rate": 9.811361117612097e-05, + "loss": 2.3834, + "step": 32805 + }, + { + "epoch": 0.09, + "learning_rate": 9.811303799879637e-05, + "loss": 2.435, + "step": 32810 + }, + { + "epoch": 0.09, + "learning_rate": 9.811246473608009e-05, + "loss": 2.3736, + "step": 32815 + }, + { + "epoch": 0.09, + "learning_rate": 9.811189138797311e-05, + "loss": 2.3545, + "step": 32820 + }, + { + "epoch": 0.09, + "learning_rate": 9.81113179544765e-05, + "loss": 2.4643, + "step": 32825 + }, + { + "epoch": 0.09, + "learning_rate": 9.811074443559128e-05, + "loss": 2.3183, + "step": 32830 + }, + { + "epoch": 0.09, + "learning_rate": 9.811017083131839e-05, + "loss": 2.2639, + "step": 32835 + }, + { + "epoch": 0.09, + "learning_rate": 9.810959714165892e-05, + "loss": 2.3962, + "step": 32840 + }, + { + "epoch": 0.09, + "learning_rate": 9.810902336661388e-05, + "loss": 2.3548, + "step": 32845 + }, + { + "epoch": 0.09, + "learning_rate": 9.810844950618427e-05, + "loss": 2.2773, + "step": 32850 + }, + { + "epoch": 0.09, + "learning_rate": 9.810787556037111e-05, + "loss": 2.299, + "step": 32855 + }, + { + "epoch": 0.09, + "learning_rate": 9.810730152917544e-05, + "loss": 2.4703, + "step": 32860 + }, + { + "epoch": 0.09, + "learning_rate": 9.810672741259824e-05, + "loss": 2.495, + "step": 32865 + }, + { + "epoch": 0.09, + "learning_rate": 9.810615321064056e-05, + "loss": 2.2552, + "step": 32870 + }, + { + "epoch": 0.09, + "learning_rate": 9.81055789233034e-05, + "loss": 2.336, + "step": 32875 + }, + { + "epoch": 0.09, + "learning_rate": 9.810500455058782e-05, + "loss": 2.3081, + "step": 32880 + }, + { + "epoch": 0.09, + "learning_rate": 9.810443009249478e-05, + "loss": 2.3906, + "step": 32885 + }, + { + "epoch": 0.09, + "learning_rate": 9.810385554902535e-05, + "loss": 2.3731, + "step": 32890 + }, + { + "epoch": 0.09, + "learning_rate": 9.81032809201805e-05, + "loss": 2.3443, + "step": 32895 + }, + { + "epoch": 0.09, + "learning_rate": 9.810270620596131e-05, + "loss": 2.3691, + "step": 32900 + }, + { + "epoch": 0.09, + "learning_rate": 9.810213140636874e-05, + "loss": 2.2713, + "step": 32905 + }, + { + "epoch": 0.09, + "learning_rate": 9.810155652140385e-05, + "loss": 2.3741, + "step": 32910 + }, + { + "epoch": 0.09, + "learning_rate": 9.810098155106766e-05, + "loss": 2.2254, + "step": 32915 + }, + { + "epoch": 0.09, + "learning_rate": 9.810040649536116e-05, + "loss": 2.3238, + "step": 32920 + }, + { + "epoch": 0.09, + "learning_rate": 9.80998313542854e-05, + "loss": 2.2251, + "step": 32925 + }, + { + "epoch": 0.09, + "learning_rate": 9.809925612784139e-05, + "loss": 2.3181, + "step": 32930 + }, + { + "epoch": 0.09, + "learning_rate": 9.809868081603015e-05, + "loss": 2.4395, + "step": 32935 + }, + { + "epoch": 0.09, + "learning_rate": 9.80981054188527e-05, + "loss": 2.3525, + "step": 32940 + }, + { + "epoch": 0.09, + "learning_rate": 9.809752993631008e-05, + "loss": 2.3542, + "step": 32945 + }, + { + "epoch": 0.09, + "learning_rate": 9.809695436840327e-05, + "loss": 2.355, + "step": 32950 + }, + { + "epoch": 0.09, + "learning_rate": 9.809637871513333e-05, + "loss": 2.394, + "step": 32955 + }, + { + "epoch": 0.09, + "learning_rate": 9.809580297650128e-05, + "loss": 2.4879, + "step": 32960 + }, + { + "epoch": 0.09, + "learning_rate": 9.80952271525081e-05, + "loss": 2.2932, + "step": 32965 + }, + { + "epoch": 0.09, + "learning_rate": 9.809465124315486e-05, + "loss": 2.2013, + "step": 32970 + }, + { + "epoch": 0.09, + "learning_rate": 9.809407524844257e-05, + "loss": 2.3986, + "step": 32975 + }, + { + "epoch": 0.09, + "learning_rate": 9.809349916837223e-05, + "loss": 2.3974, + "step": 32980 + }, + { + "epoch": 0.09, + "learning_rate": 9.809292300294489e-05, + "loss": 2.3969, + "step": 32985 + }, + { + "epoch": 0.09, + "learning_rate": 9.809234675216157e-05, + "loss": 2.313, + "step": 32990 + }, + { + "epoch": 0.09, + "learning_rate": 9.809177041602326e-05, + "loss": 2.2668, + "step": 32995 + }, + { + "epoch": 0.09, + "learning_rate": 9.809119399453102e-05, + "loss": 2.2387, + "step": 33000 + }, + { + "epoch": 0.09, + "learning_rate": 9.809061748768587e-05, + "loss": 2.3782, + "step": 33005 + }, + { + "epoch": 0.09, + "learning_rate": 9.809004089548881e-05, + "loss": 2.4213, + "step": 33010 + }, + { + "epoch": 0.09, + "learning_rate": 9.808946421794088e-05, + "loss": 2.3494, + "step": 33015 + }, + { + "epoch": 0.09, + "learning_rate": 9.80888874550431e-05, + "loss": 2.4328, + "step": 33020 + }, + { + "epoch": 0.09, + "learning_rate": 9.80883106067965e-05, + "loss": 2.5315, + "step": 33025 + }, + { + "epoch": 0.09, + "learning_rate": 9.80877336732021e-05, + "loss": 2.283, + "step": 33030 + }, + { + "epoch": 0.09, + "learning_rate": 9.80871566542609e-05, + "loss": 2.4191, + "step": 33035 + }, + { + "epoch": 0.09, + "learning_rate": 9.808657954997397e-05, + "loss": 2.4889, + "step": 33040 + }, + { + "epoch": 0.09, + "learning_rate": 9.80860023603423e-05, + "loss": 2.4033, + "step": 33045 + }, + { + "epoch": 0.09, + "learning_rate": 9.808542508536692e-05, + "loss": 2.4849, + "step": 33050 + }, + { + "epoch": 0.09, + "learning_rate": 9.808484772504887e-05, + "loss": 2.4694, + "step": 33055 + }, + { + "epoch": 0.09, + "learning_rate": 9.808427027938915e-05, + "loss": 2.4194, + "step": 33060 + }, + { + "epoch": 0.09, + "learning_rate": 9.808369274838881e-05, + "loss": 2.4505, + "step": 33065 + }, + { + "epoch": 0.09, + "learning_rate": 9.808311513204886e-05, + "loss": 2.3814, + "step": 33070 + }, + { + "epoch": 0.09, + "learning_rate": 9.808253743037032e-05, + "loss": 2.1876, + "step": 33075 + }, + { + "epoch": 0.09, + "learning_rate": 9.808195964335424e-05, + "loss": 2.3729, + "step": 33080 + }, + { + "epoch": 0.09, + "learning_rate": 9.808138177100161e-05, + "loss": 2.3608, + "step": 33085 + }, + { + "epoch": 0.09, + "learning_rate": 9.808080381331349e-05, + "loss": 2.471, + "step": 33090 + }, + { + "epoch": 0.09, + "learning_rate": 9.80802257702909e-05, + "loss": 2.4249, + "step": 33095 + }, + { + "epoch": 0.09, + "learning_rate": 9.807964764193484e-05, + "loss": 2.292, + "step": 33100 + }, + { + "epoch": 0.09, + "learning_rate": 9.807906942824636e-05, + "loss": 2.1674, + "step": 33105 + }, + { + "epoch": 0.09, + "learning_rate": 9.807849112922647e-05, + "loss": 2.3242, + "step": 33110 + }, + { + "epoch": 0.09, + "learning_rate": 9.807791274487621e-05, + "loss": 2.348, + "step": 33115 + }, + { + "epoch": 0.09, + "learning_rate": 9.807733427519661e-05, + "loss": 2.5123, + "step": 33120 + }, + { + "epoch": 0.09, + "learning_rate": 9.807675572018869e-05, + "loss": 2.3368, + "step": 33125 + }, + { + "epoch": 0.09, + "learning_rate": 9.807617707985345e-05, + "loss": 2.3664, + "step": 33130 + }, + { + "epoch": 0.09, + "learning_rate": 9.807559835419196e-05, + "loss": 2.3852, + "step": 33135 + }, + { + "epoch": 0.09, + "learning_rate": 9.807501954320524e-05, + "loss": 2.4346, + "step": 33140 + }, + { + "epoch": 0.09, + "learning_rate": 9.807444064689428e-05, + "loss": 2.3481, + "step": 33145 + }, + { + "epoch": 0.09, + "learning_rate": 9.807386166526016e-05, + "loss": 2.4172, + "step": 33150 + }, + { + "epoch": 0.09, + "learning_rate": 9.807328259830387e-05, + "loss": 2.3877, + "step": 33155 + }, + { + "epoch": 0.09, + "learning_rate": 9.807270344602646e-05, + "loss": 2.2912, + "step": 33160 + }, + { + "epoch": 0.09, + "learning_rate": 9.807212420842893e-05, + "loss": 2.3387, + "step": 33165 + }, + { + "epoch": 0.09, + "learning_rate": 9.807154488551234e-05, + "loss": 2.2814, + "step": 33170 + }, + { + "epoch": 0.09, + "learning_rate": 9.80709654772777e-05, + "loss": 2.5032, + "step": 33175 + }, + { + "epoch": 0.09, + "learning_rate": 9.807038598372605e-05, + "loss": 2.3838, + "step": 33180 + }, + { + "epoch": 0.09, + "learning_rate": 9.806980640485839e-05, + "loss": 2.4358, + "step": 33185 + }, + { + "epoch": 0.09, + "learning_rate": 9.806922674067578e-05, + "loss": 2.494, + "step": 33190 + }, + { + "epoch": 0.09, + "learning_rate": 9.806864699117923e-05, + "loss": 2.4316, + "step": 33195 + }, + { + "epoch": 0.09, + "learning_rate": 9.806806715636979e-05, + "loss": 2.2866, + "step": 33200 + }, + { + "epoch": 0.09, + "learning_rate": 9.806748723624848e-05, + "loss": 2.4296, + "step": 33205 + }, + { + "epoch": 0.09, + "learning_rate": 9.80669072308163e-05, + "loss": 2.425, + "step": 33210 + }, + { + "epoch": 0.09, + "learning_rate": 9.806632714007433e-05, + "loss": 2.328, + "step": 33215 + }, + { + "epoch": 0.09, + "learning_rate": 9.806574696402357e-05, + "loss": 2.2883, + "step": 33220 + }, + { + "epoch": 0.09, + "learning_rate": 9.806516670266505e-05, + "loss": 2.3809, + "step": 33225 + }, + { + "epoch": 0.09, + "learning_rate": 9.80645863559998e-05, + "loss": 2.4033, + "step": 33230 + }, + { + "epoch": 0.09, + "learning_rate": 9.806400592402886e-05, + "loss": 2.4026, + "step": 33235 + }, + { + "epoch": 0.09, + "learning_rate": 9.806342540675324e-05, + "loss": 2.2901, + "step": 33240 + }, + { + "epoch": 0.09, + "learning_rate": 9.8062844804174e-05, + "loss": 2.374, + "step": 33245 + }, + { + "epoch": 0.09, + "learning_rate": 9.806226411629216e-05, + "loss": 2.4859, + "step": 33250 + }, + { + "epoch": 0.09, + "learning_rate": 9.806168334310872e-05, + "loss": 2.3982, + "step": 33255 + }, + { + "epoch": 0.09, + "learning_rate": 9.806110248462476e-05, + "loss": 2.3908, + "step": 33260 + }, + { + "epoch": 0.09, + "learning_rate": 9.806052154084128e-05, + "loss": 2.3354, + "step": 33265 + }, + { + "epoch": 0.09, + "learning_rate": 9.805994051175931e-05, + "loss": 2.3729, + "step": 33270 + }, + { + "epoch": 0.09, + "learning_rate": 9.805935939737989e-05, + "loss": 2.3515, + "step": 33275 + }, + { + "epoch": 0.09, + "learning_rate": 9.805877819770406e-05, + "loss": 2.5205, + "step": 33280 + }, + { + "epoch": 0.09, + "learning_rate": 9.805819691273284e-05, + "loss": 2.3496, + "step": 33285 + }, + { + "epoch": 0.09, + "learning_rate": 9.805761554246726e-05, + "loss": 2.3748, + "step": 33290 + }, + { + "epoch": 0.09, + "learning_rate": 9.805703408690836e-05, + "loss": 2.3404, + "step": 33295 + }, + { + "epoch": 0.09, + "learning_rate": 9.805645254605717e-05, + "loss": 2.4647, + "step": 33300 + }, + { + "epoch": 0.09, + "learning_rate": 9.805587091991471e-05, + "loss": 2.3702, + "step": 33305 + }, + { + "epoch": 0.09, + "learning_rate": 9.805528920848204e-05, + "loss": 2.3247, + "step": 33310 + }, + { + "epoch": 0.09, + "learning_rate": 9.805470741176014e-05, + "loss": 2.3732, + "step": 33315 + }, + { + "epoch": 0.09, + "learning_rate": 9.805412552975011e-05, + "loss": 2.4166, + "step": 33320 + }, + { + "epoch": 0.09, + "learning_rate": 9.805354356245295e-05, + "loss": 2.3587, + "step": 33325 + }, + { + "epoch": 0.09, + "learning_rate": 9.805296150986966e-05, + "loss": 2.4383, + "step": 33330 + }, + { + "epoch": 0.09, + "learning_rate": 9.805237937200133e-05, + "loss": 2.2408, + "step": 33335 + }, + { + "epoch": 0.09, + "learning_rate": 9.805179714884898e-05, + "loss": 2.1116, + "step": 33340 + }, + { + "epoch": 0.09, + "learning_rate": 9.805121484041362e-05, + "loss": 2.4012, + "step": 33345 + }, + { + "epoch": 0.09, + "learning_rate": 9.805063244669629e-05, + "loss": 2.3655, + "step": 33350 + }, + { + "epoch": 0.09, + "learning_rate": 9.805004996769804e-05, + "loss": 2.4396, + "step": 33355 + }, + { + "epoch": 0.09, + "learning_rate": 9.804946740341988e-05, + "loss": 2.3003, + "step": 33360 + }, + { + "epoch": 0.09, + "learning_rate": 9.804888475386287e-05, + "loss": 2.4608, + "step": 33365 + }, + { + "epoch": 0.09, + "learning_rate": 9.804830201902801e-05, + "loss": 2.4171, + "step": 33370 + }, + { + "epoch": 0.09, + "learning_rate": 9.804771919891638e-05, + "loss": 2.3789, + "step": 33375 + }, + { + "epoch": 0.09, + "learning_rate": 9.804713629352899e-05, + "loss": 2.2225, + "step": 33380 + }, + { + "epoch": 0.09, + "learning_rate": 9.804655330286685e-05, + "loss": 2.3442, + "step": 33385 + }, + { + "epoch": 0.09, + "learning_rate": 9.804597022693105e-05, + "loss": 2.375, + "step": 33390 + }, + { + "epoch": 0.09, + "learning_rate": 9.804538706572257e-05, + "loss": 2.3309, + "step": 33395 + }, + { + "epoch": 0.09, + "learning_rate": 9.804480381924248e-05, + "loss": 2.3959, + "step": 33400 + }, + { + "epoch": 0.09, + "learning_rate": 9.80442204874918e-05, + "loss": 2.2896, + "step": 33405 + }, + { + "epoch": 0.09, + "learning_rate": 9.804363707047157e-05, + "loss": 2.3214, + "step": 33410 + }, + { + "epoch": 0.09, + "learning_rate": 9.804305356818282e-05, + "loss": 2.4789, + "step": 33415 + }, + { + "epoch": 0.09, + "learning_rate": 9.804246998062661e-05, + "loss": 2.3825, + "step": 33420 + }, + { + "epoch": 0.09, + "learning_rate": 9.804188630780393e-05, + "loss": 2.4559, + "step": 33425 + }, + { + "epoch": 0.09, + "learning_rate": 9.804130254971586e-05, + "loss": 2.3395, + "step": 33430 + }, + { + "epoch": 0.09, + "learning_rate": 9.804071870636341e-05, + "loss": 2.394, + "step": 33435 + }, + { + "epoch": 0.09, + "learning_rate": 9.804013477774763e-05, + "loss": 2.4224, + "step": 33440 + }, + { + "epoch": 0.09, + "learning_rate": 9.803955076386954e-05, + "loss": 2.3707, + "step": 33445 + }, + { + "epoch": 0.09, + "learning_rate": 9.803896666473021e-05, + "loss": 2.3435, + "step": 33450 + }, + { + "epoch": 0.09, + "learning_rate": 9.803838248033064e-05, + "loss": 2.265, + "step": 33455 + }, + { + "epoch": 0.09, + "learning_rate": 9.803779821067187e-05, + "loss": 2.383, + "step": 33460 + }, + { + "epoch": 0.09, + "learning_rate": 9.803721385575495e-05, + "loss": 2.3525, + "step": 33465 + }, + { + "epoch": 0.09, + "learning_rate": 9.803662941558094e-05, + "loss": 2.2374, + "step": 33470 + }, + { + "epoch": 0.09, + "learning_rate": 9.803604489015082e-05, + "loss": 2.3729, + "step": 33475 + }, + { + "epoch": 0.09, + "learning_rate": 9.803546027946568e-05, + "loss": 2.3152, + "step": 33480 + }, + { + "epoch": 0.09, + "learning_rate": 9.803487558352653e-05, + "loss": 2.46, + "step": 33485 + }, + { + "epoch": 0.09, + "learning_rate": 9.803429080233442e-05, + "loss": 2.3501, + "step": 33490 + }, + { + "epoch": 0.09, + "learning_rate": 9.803370593589038e-05, + "loss": 2.2433, + "step": 33495 + }, + { + "epoch": 0.09, + "learning_rate": 9.803312098419544e-05, + "loss": 2.4494, + "step": 33500 + }, + { + "epoch": 0.09, + "learning_rate": 9.803253594725066e-05, + "loss": 2.3752, + "step": 33505 + }, + { + "epoch": 0.09, + "learning_rate": 9.803195082505707e-05, + "loss": 2.4603, + "step": 33510 + }, + { + "epoch": 0.09, + "learning_rate": 9.80313656176157e-05, + "loss": 2.3529, + "step": 33515 + }, + { + "epoch": 0.09, + "learning_rate": 9.803078032492761e-05, + "loss": 2.3157, + "step": 33520 + }, + { + "epoch": 0.09, + "learning_rate": 9.80301949469938e-05, + "loss": 2.3532, + "step": 33525 + }, + { + "epoch": 0.09, + "learning_rate": 9.802960948381534e-05, + "loss": 2.3579, + "step": 33530 + }, + { + "epoch": 0.09, + "learning_rate": 9.802902393539327e-05, + "loss": 2.3957, + "step": 33535 + }, + { + "epoch": 0.09, + "learning_rate": 9.802843830172861e-05, + "loss": 2.3926, + "step": 33540 + }, + { + "epoch": 0.09, + "learning_rate": 9.80278525828224e-05, + "loss": 2.4837, + "step": 33545 + }, + { + "epoch": 0.09, + "learning_rate": 9.80272667786757e-05, + "loss": 2.457, + "step": 33550 + }, + { + "epoch": 0.09, + "learning_rate": 9.802668088928956e-05, + "loss": 2.3062, + "step": 33555 + }, + { + "epoch": 0.09, + "learning_rate": 9.802609491466498e-05, + "loss": 2.3259, + "step": 33560 + }, + { + "epoch": 0.09, + "learning_rate": 9.802550885480302e-05, + "loss": 2.421, + "step": 33565 + }, + { + "epoch": 0.09, + "learning_rate": 9.802492270970471e-05, + "loss": 2.3093, + "step": 33570 + }, + { + "epoch": 0.09, + "learning_rate": 9.802433647937112e-05, + "loss": 2.4167, + "step": 33575 + }, + { + "epoch": 0.09, + "learning_rate": 9.802375016380325e-05, + "loss": 2.391, + "step": 33580 + }, + { + "epoch": 0.09, + "learning_rate": 9.802316376300218e-05, + "loss": 2.297, + "step": 33585 + }, + { + "epoch": 0.09, + "learning_rate": 9.802257727696891e-05, + "loss": 2.5159, + "step": 33590 + }, + { + "epoch": 0.09, + "learning_rate": 9.802199070570451e-05, + "loss": 2.3786, + "step": 33595 + }, + { + "epoch": 0.09, + "learning_rate": 9.802140404921001e-05, + "loss": 2.3434, + "step": 33600 + }, + { + "epoch": 0.09, + "learning_rate": 9.802081730748647e-05, + "loss": 2.3358, + "step": 33605 + }, + { + "epoch": 0.09, + "learning_rate": 9.80202304805349e-05, + "loss": 2.2862, + "step": 33610 + }, + { + "epoch": 0.09, + "learning_rate": 9.801964356835637e-05, + "loss": 2.3095, + "step": 33615 + }, + { + "epoch": 0.09, + "learning_rate": 9.80190565709519e-05, + "loss": 2.5232, + "step": 33620 + }, + { + "epoch": 0.09, + "learning_rate": 9.801846948832254e-05, + "loss": 2.4374, + "step": 33625 + }, + { + "epoch": 0.09, + "learning_rate": 9.801788232046934e-05, + "loss": 2.3894, + "step": 33630 + }, + { + "epoch": 0.09, + "learning_rate": 9.801729506739332e-05, + "loss": 2.38, + "step": 33635 + }, + { + "epoch": 0.09, + "learning_rate": 9.801670772909555e-05, + "loss": 2.3902, + "step": 33640 + }, + { + "epoch": 0.09, + "learning_rate": 9.801612030557706e-05, + "loss": 2.3229, + "step": 33645 + }, + { + "epoch": 0.09, + "learning_rate": 9.801553279683888e-05, + "loss": 2.4056, + "step": 33650 + }, + { + "epoch": 0.09, + "learning_rate": 9.801494520288208e-05, + "loss": 2.4308, + "step": 33655 + }, + { + "epoch": 0.09, + "learning_rate": 9.801435752370767e-05, + "loss": 2.27, + "step": 33660 + }, + { + "epoch": 0.09, + "learning_rate": 9.801376975931674e-05, + "loss": 2.3702, + "step": 33665 + }, + { + "epoch": 0.09, + "learning_rate": 9.801318190971028e-05, + "loss": 2.4305, + "step": 33670 + }, + { + "epoch": 0.09, + "learning_rate": 9.801259397488936e-05, + "loss": 2.351, + "step": 33675 + }, + { + "epoch": 0.09, + "learning_rate": 9.801200595485501e-05, + "loss": 2.3128, + "step": 33680 + }, + { + "epoch": 0.09, + "learning_rate": 9.80114178496083e-05, + "loss": 2.4182, + "step": 33685 + }, + { + "epoch": 0.09, + "learning_rate": 9.801082965915025e-05, + "loss": 2.3249, + "step": 33690 + }, + { + "epoch": 0.09, + "learning_rate": 9.801024138348192e-05, + "loss": 2.4662, + "step": 33695 + }, + { + "epoch": 0.09, + "learning_rate": 9.800965302260434e-05, + "loss": 2.421, + "step": 33700 + }, + { + "epoch": 0.09, + "learning_rate": 9.800906457651857e-05, + "loss": 2.4048, + "step": 33705 + }, + { + "epoch": 0.09, + "learning_rate": 9.800847604522563e-05, + "loss": 2.3819, + "step": 33710 + }, + { + "epoch": 0.09, + "learning_rate": 9.80078874287266e-05, + "loss": 2.371, + "step": 33715 + }, + { + "epoch": 0.09, + "learning_rate": 9.800729872702247e-05, + "loss": 2.2853, + "step": 33720 + }, + { + "epoch": 0.09, + "learning_rate": 9.800670994011436e-05, + "loss": 2.4097, + "step": 33725 + }, + { + "epoch": 0.09, + "learning_rate": 9.800612106800325e-05, + "loss": 2.4565, + "step": 33730 + }, + { + "epoch": 0.09, + "learning_rate": 9.800553211069021e-05, + "loss": 2.4192, + "step": 33735 + }, + { + "epoch": 0.09, + "learning_rate": 9.800494306817628e-05, + "loss": 2.3538, + "step": 33740 + }, + { + "epoch": 0.09, + "learning_rate": 9.800435394046253e-05, + "loss": 2.2633, + "step": 33745 + }, + { + "epoch": 0.09, + "learning_rate": 9.800376472754995e-05, + "loss": 2.4185, + "step": 33750 + }, + { + "epoch": 0.09, + "learning_rate": 9.800317542943964e-05, + "loss": 2.3246, + "step": 33755 + }, + { + "epoch": 0.09, + "learning_rate": 9.800258604613264e-05, + "loss": 2.3727, + "step": 33760 + }, + { + "epoch": 0.09, + "learning_rate": 9.800199657762996e-05, + "loss": 2.2649, + "step": 33765 + }, + { + "epoch": 0.09, + "learning_rate": 9.800140702393268e-05, + "loss": 2.2963, + "step": 33770 + }, + { + "epoch": 0.09, + "learning_rate": 9.800081738504184e-05, + "loss": 2.351, + "step": 33775 + }, + { + "epoch": 0.09, + "learning_rate": 9.800022766095848e-05, + "loss": 2.413, + "step": 33780 + }, + { + "epoch": 0.09, + "learning_rate": 9.799963785168364e-05, + "loss": 2.3674, + "step": 33785 + }, + { + "epoch": 0.09, + "learning_rate": 9.799904795721838e-05, + "loss": 2.3805, + "step": 33790 + }, + { + "epoch": 0.09, + "learning_rate": 9.799845797756376e-05, + "loss": 2.4206, + "step": 33795 + }, + { + "epoch": 0.09, + "learning_rate": 9.799786791272078e-05, + "loss": 2.4488, + "step": 33800 + }, + { + "epoch": 0.09, + "learning_rate": 9.799727776269054e-05, + "loss": 2.4196, + "step": 33805 + }, + { + "epoch": 0.09, + "learning_rate": 9.799668752747404e-05, + "loss": 2.3469, + "step": 33810 + }, + { + "epoch": 0.09, + "learning_rate": 9.799609720707236e-05, + "loss": 2.3558, + "step": 33815 + }, + { + "epoch": 0.09, + "learning_rate": 9.799550680148655e-05, + "loss": 2.4639, + "step": 33820 + }, + { + "epoch": 0.09, + "learning_rate": 9.799491631071765e-05, + "loss": 2.3154, + "step": 33825 + }, + { + "epoch": 0.09, + "learning_rate": 9.799432573476669e-05, + "loss": 2.3465, + "step": 33830 + }, + { + "epoch": 0.09, + "learning_rate": 9.799373507363474e-05, + "loss": 2.3319, + "step": 33835 + }, + { + "epoch": 0.09, + "learning_rate": 9.799314432732283e-05, + "loss": 2.2891, + "step": 33840 + }, + { + "epoch": 0.09, + "learning_rate": 9.799255349583205e-05, + "loss": 2.4718, + "step": 33845 + }, + { + "epoch": 0.09, + "learning_rate": 9.799196257916338e-05, + "loss": 2.3633, + "step": 33850 + }, + { + "epoch": 0.09, + "learning_rate": 9.799137157731794e-05, + "loss": 2.3877, + "step": 33855 + }, + { + "epoch": 0.09, + "learning_rate": 9.799078049029673e-05, + "loss": 2.4438, + "step": 33860 + }, + { + "epoch": 0.09, + "learning_rate": 9.799018931810082e-05, + "loss": 2.3598, + "step": 33865 + }, + { + "epoch": 0.09, + "learning_rate": 9.798959806073124e-05, + "loss": 2.4049, + "step": 33870 + }, + { + "epoch": 0.09, + "learning_rate": 9.798900671818907e-05, + "loss": 2.4376, + "step": 33875 + }, + { + "epoch": 0.09, + "learning_rate": 9.798841529047536e-05, + "loss": 2.3397, + "step": 33880 + }, + { + "epoch": 0.09, + "learning_rate": 9.798782377759113e-05, + "loss": 2.3025, + "step": 33885 + }, + { + "epoch": 0.09, + "learning_rate": 9.798723217953742e-05, + "loss": 2.3723, + "step": 33890 + }, + { + "epoch": 0.09, + "learning_rate": 9.798664049631533e-05, + "loss": 2.374, + "step": 33895 + }, + { + "epoch": 0.09, + "learning_rate": 9.798604872792588e-05, + "loss": 2.3024, + "step": 33900 + }, + { + "epoch": 0.09, + "learning_rate": 9.79854568743701e-05, + "loss": 2.1894, + "step": 33905 + }, + { + "epoch": 0.09, + "learning_rate": 9.798486493564909e-05, + "loss": 2.3374, + "step": 33910 + }, + { + "epoch": 0.09, + "learning_rate": 9.798427291176387e-05, + "loss": 2.3532, + "step": 33915 + }, + { + "epoch": 0.09, + "learning_rate": 9.79836808027155e-05, + "loss": 2.3846, + "step": 33920 + }, + { + "epoch": 0.09, + "learning_rate": 9.7983088608505e-05, + "loss": 2.2929, + "step": 33925 + }, + { + "epoch": 0.09, + "learning_rate": 9.798249632913347e-05, + "loss": 2.3483, + "step": 33930 + }, + { + "epoch": 0.09, + "learning_rate": 9.798190396460194e-05, + "loss": 2.3219, + "step": 33935 + }, + { + "epoch": 0.09, + "learning_rate": 9.798131151491145e-05, + "loss": 2.4074, + "step": 33940 + }, + { + "epoch": 0.09, + "learning_rate": 9.798071898006306e-05, + "loss": 2.248, + "step": 33945 + }, + { + "epoch": 0.09, + "learning_rate": 9.798012636005783e-05, + "loss": 2.2356, + "step": 33950 + }, + { + "epoch": 0.09, + "learning_rate": 9.797953365489681e-05, + "loss": 2.2952, + "step": 33955 + }, + { + "epoch": 0.09, + "learning_rate": 9.797894086458103e-05, + "loss": 2.3739, + "step": 33960 + }, + { + "epoch": 0.09, + "learning_rate": 9.797834798911157e-05, + "loss": 2.4597, + "step": 33965 + }, + { + "epoch": 0.09, + "learning_rate": 9.797775502848946e-05, + "loss": 2.238, + "step": 33970 + }, + { + "epoch": 0.09, + "learning_rate": 9.797716198271576e-05, + "loss": 2.4153, + "step": 33975 + }, + { + "epoch": 0.09, + "learning_rate": 9.797656885179154e-05, + "loss": 2.3473, + "step": 33980 + }, + { + "epoch": 0.09, + "learning_rate": 9.797597563571783e-05, + "loss": 2.3398, + "step": 33985 + }, + { + "epoch": 0.09, + "learning_rate": 9.79753823344957e-05, + "loss": 2.296, + "step": 33990 + }, + { + "epoch": 0.09, + "learning_rate": 9.79747889481262e-05, + "loss": 2.3903, + "step": 33995 + }, + { + "epoch": 0.09, + "learning_rate": 9.797419547661037e-05, + "loss": 2.4564, + "step": 34000 + }, + { + "epoch": 0.09, + "learning_rate": 9.797360191994926e-05, + "loss": 2.4167, + "step": 34005 + }, + { + "epoch": 0.09, + "learning_rate": 9.797300827814394e-05, + "loss": 2.3739, + "step": 34010 + }, + { + "epoch": 0.09, + "learning_rate": 9.797241455119546e-05, + "loss": 2.4088, + "step": 34015 + }, + { + "epoch": 0.09, + "learning_rate": 9.797182073910487e-05, + "loss": 2.4064, + "step": 34020 + }, + { + "epoch": 0.09, + "learning_rate": 9.797122684187324e-05, + "loss": 2.4592, + "step": 34025 + }, + { + "epoch": 0.09, + "learning_rate": 9.797063285950159e-05, + "loss": 2.3548, + "step": 34030 + }, + { + "epoch": 0.09, + "learning_rate": 9.797003879199101e-05, + "loss": 2.3579, + "step": 34035 + }, + { + "epoch": 0.09, + "learning_rate": 9.796944463934253e-05, + "loss": 2.4094, + "step": 34040 + }, + { + "epoch": 0.09, + "learning_rate": 9.79688504015572e-05, + "loss": 2.3065, + "step": 34045 + }, + { + "epoch": 0.09, + "learning_rate": 9.79682560786361e-05, + "loss": 2.331, + "step": 34050 + }, + { + "epoch": 0.09, + "learning_rate": 9.796766167058027e-05, + "loss": 2.2307, + "step": 34055 + }, + { + "epoch": 0.09, + "learning_rate": 9.796706717739077e-05, + "loss": 2.3141, + "step": 34060 + }, + { + "epoch": 0.09, + "learning_rate": 9.796647259906866e-05, + "loss": 2.3665, + "step": 34065 + }, + { + "epoch": 0.09, + "learning_rate": 9.796587793561497e-05, + "loss": 2.304, + "step": 34070 + }, + { + "epoch": 0.09, + "learning_rate": 9.796528318703078e-05, + "loss": 2.4399, + "step": 34075 + }, + { + "epoch": 0.09, + "learning_rate": 9.796468835331714e-05, + "loss": 2.3582, + "step": 34080 + }, + { + "epoch": 0.09, + "learning_rate": 9.79640934344751e-05, + "loss": 2.3733, + "step": 34085 + }, + { + "epoch": 0.09, + "learning_rate": 9.796349843050571e-05, + "loss": 2.377, + "step": 34090 + }, + { + "epoch": 0.09, + "learning_rate": 9.796290334141008e-05, + "loss": 2.3333, + "step": 34095 + }, + { + "epoch": 0.09, + "learning_rate": 9.796230816718918e-05, + "loss": 2.3628, + "step": 34100 + }, + { + "epoch": 0.09, + "learning_rate": 9.796171290784412e-05, + "loss": 2.3781, + "step": 34105 + }, + { + "epoch": 0.09, + "learning_rate": 9.796111756337594e-05, + "loss": 2.4018, + "step": 34110 + }, + { + "epoch": 0.09, + "learning_rate": 9.79605221337857e-05, + "loss": 2.3593, + "step": 34115 + }, + { + "epoch": 0.09, + "learning_rate": 9.795992661907447e-05, + "loss": 2.5387, + "step": 34120 + }, + { + "epoch": 0.09, + "learning_rate": 9.79593310192433e-05, + "loss": 2.456, + "step": 34125 + }, + { + "epoch": 0.09, + "learning_rate": 9.795873533429322e-05, + "loss": 2.4433, + "step": 34130 + }, + { + "epoch": 0.09, + "learning_rate": 9.795813956422533e-05, + "loss": 2.3683, + "step": 34135 + }, + { + "epoch": 0.09, + "learning_rate": 9.795754370904066e-05, + "loss": 2.443, + "step": 34140 + }, + { + "epoch": 0.09, + "learning_rate": 9.795694776874027e-05, + "loss": 2.2883, + "step": 34145 + }, + { + "epoch": 0.09, + "learning_rate": 9.795635174332523e-05, + "loss": 2.2818, + "step": 34150 + }, + { + "epoch": 0.09, + "learning_rate": 9.795575563279659e-05, + "loss": 2.4003, + "step": 34155 + }, + { + "epoch": 0.09, + "learning_rate": 9.79551594371554e-05, + "loss": 2.373, + "step": 34160 + }, + { + "epoch": 0.09, + "learning_rate": 9.795456315640274e-05, + "loss": 2.389, + "step": 34165 + }, + { + "epoch": 0.09, + "learning_rate": 9.795396679053964e-05, + "loss": 2.2756, + "step": 34170 + }, + { + "epoch": 0.09, + "learning_rate": 9.795337033956717e-05, + "loss": 2.347, + "step": 34175 + }, + { + "epoch": 0.09, + "learning_rate": 9.79527738034864e-05, + "loss": 2.3309, + "step": 34180 + }, + { + "epoch": 0.09, + "learning_rate": 9.795217718229839e-05, + "loss": 2.2932, + "step": 34185 + }, + { + "epoch": 0.09, + "learning_rate": 9.795158047600418e-05, + "loss": 2.3753, + "step": 34190 + }, + { + "epoch": 0.09, + "learning_rate": 9.795098368460484e-05, + "loss": 2.4555, + "step": 34195 + }, + { + "epoch": 0.09, + "learning_rate": 9.795038680810143e-05, + "loss": 2.476, + "step": 34200 + }, + { + "epoch": 0.09, + "learning_rate": 9.794978984649499e-05, + "loss": 2.3676, + "step": 34205 + }, + { + "epoch": 0.09, + "learning_rate": 9.79491927997866e-05, + "loss": 2.3363, + "step": 34210 + }, + { + "epoch": 0.09, + "learning_rate": 9.794859566797734e-05, + "loss": 2.2544, + "step": 34215 + }, + { + "epoch": 0.09, + "learning_rate": 9.794799845106822e-05, + "loss": 2.5413, + "step": 34220 + }, + { + "epoch": 0.09, + "learning_rate": 9.794740114906034e-05, + "loss": 2.3402, + "step": 34225 + }, + { + "epoch": 0.09, + "learning_rate": 9.794680376195474e-05, + "loss": 2.3654, + "step": 34230 + }, + { + "epoch": 0.09, + "learning_rate": 9.794620628975248e-05, + "loss": 2.3497, + "step": 34235 + }, + { + "epoch": 0.09, + "learning_rate": 9.794560873245465e-05, + "loss": 2.3121, + "step": 34240 + }, + { + "epoch": 0.09, + "learning_rate": 9.794501109006227e-05, + "loss": 2.4067, + "step": 34245 + }, + { + "epoch": 0.09, + "learning_rate": 9.794441336257641e-05, + "loss": 2.4079, + "step": 34250 + }, + { + "epoch": 0.09, + "learning_rate": 9.794381554999815e-05, + "loss": 2.3823, + "step": 34255 + }, + { + "epoch": 0.09, + "learning_rate": 9.794321765232853e-05, + "loss": 2.2221, + "step": 34260 + }, + { + "epoch": 0.09, + "learning_rate": 9.794261966956862e-05, + "loss": 2.4409, + "step": 34265 + }, + { + "epoch": 0.09, + "learning_rate": 9.79420216017195e-05, + "loss": 2.3998, + "step": 34270 + }, + { + "epoch": 0.09, + "learning_rate": 9.79414234487822e-05, + "loss": 2.4027, + "step": 34275 + }, + { + "epoch": 0.09, + "learning_rate": 9.794082521075779e-05, + "loss": 2.465, + "step": 34280 + }, + { + "epoch": 0.09, + "learning_rate": 9.794022688764734e-05, + "loss": 2.3968, + "step": 34285 + }, + { + "epoch": 0.09, + "learning_rate": 9.793962847945192e-05, + "loss": 2.2754, + "step": 34290 + }, + { + "epoch": 0.09, + "learning_rate": 9.793902998617257e-05, + "loss": 2.3919, + "step": 34295 + }, + { + "epoch": 0.09, + "learning_rate": 9.793843140781038e-05, + "loss": 2.4088, + "step": 34300 + }, + { + "epoch": 0.09, + "learning_rate": 9.793783274436636e-05, + "loss": 2.3943, + "step": 34305 + }, + { + "epoch": 0.09, + "learning_rate": 9.793723399584165e-05, + "loss": 2.2849, + "step": 34310 + }, + { + "epoch": 0.09, + "learning_rate": 9.793663516223724e-05, + "loss": 2.2475, + "step": 34315 + }, + { + "epoch": 0.09, + "learning_rate": 9.793603624355422e-05, + "loss": 2.301, + "step": 34320 + }, + { + "epoch": 0.09, + "learning_rate": 9.793543723979368e-05, + "loss": 2.3309, + "step": 34325 + }, + { + "epoch": 0.09, + "learning_rate": 9.793483815095665e-05, + "loss": 2.344, + "step": 34330 + }, + { + "epoch": 0.09, + "learning_rate": 9.793423897704421e-05, + "loss": 2.3889, + "step": 34335 + }, + { + "epoch": 0.09, + "learning_rate": 9.79336397180574e-05, + "loss": 2.4662, + "step": 34340 + }, + { + "epoch": 0.09, + "learning_rate": 9.79330403739973e-05, + "loss": 2.3972, + "step": 34345 + }, + { + "epoch": 0.09, + "learning_rate": 9.7932440944865e-05, + "loss": 2.2908, + "step": 34350 + }, + { + "epoch": 0.09, + "learning_rate": 9.793184143066151e-05, + "loss": 2.2545, + "step": 34355 + }, + { + "epoch": 0.09, + "learning_rate": 9.793124183138793e-05, + "loss": 2.3474, + "step": 34360 + }, + { + "epoch": 0.09, + "learning_rate": 9.793064214704531e-05, + "loss": 2.3655, + "step": 34365 + }, + { + "epoch": 0.09, + "learning_rate": 9.793004237763472e-05, + "loss": 2.2845, + "step": 34370 + }, + { + "epoch": 0.09, + "learning_rate": 9.792944252315723e-05, + "loss": 2.433, + "step": 34375 + }, + { + "epoch": 0.09, + "learning_rate": 9.792884258361391e-05, + "loss": 2.3742, + "step": 34380 + }, + { + "epoch": 0.09, + "learning_rate": 9.79282425590058e-05, + "loss": 2.3647, + "step": 34385 + }, + { + "epoch": 0.09, + "learning_rate": 9.792764244933399e-05, + "loss": 2.4314, + "step": 34390 + }, + { + "epoch": 0.09, + "learning_rate": 9.792704225459952e-05, + "loss": 2.4093, + "step": 34395 + }, + { + "epoch": 0.09, + "learning_rate": 9.792644197480347e-05, + "loss": 2.4203, + "step": 34400 + }, + { + "epoch": 0.09, + "learning_rate": 9.792584160994692e-05, + "loss": 2.3028, + "step": 34405 + }, + { + "epoch": 0.09, + "learning_rate": 9.79252411600309e-05, + "loss": 2.3609, + "step": 34410 + }, + { + "epoch": 0.09, + "learning_rate": 9.792464062505651e-05, + "loss": 2.243, + "step": 34415 + }, + { + "epoch": 0.09, + "learning_rate": 9.792404000502477e-05, + "loss": 2.3984, + "step": 34420 + }, + { + "epoch": 0.09, + "learning_rate": 9.792343929993682e-05, + "loss": 2.3666, + "step": 34425 + }, + { + "epoch": 0.09, + "learning_rate": 9.792283850979366e-05, + "loss": 2.2989, + "step": 34430 + }, + { + "epoch": 0.09, + "learning_rate": 9.792223763459638e-05, + "loss": 2.4209, + "step": 34435 + }, + { + "epoch": 0.09, + "learning_rate": 9.792163667434606e-05, + "loss": 2.3492, + "step": 34440 + }, + { + "epoch": 0.09, + "learning_rate": 9.792103562904374e-05, + "loss": 2.3634, + "step": 34445 + }, + { + "epoch": 0.09, + "learning_rate": 9.792043449869051e-05, + "loss": 2.4425, + "step": 34450 + }, + { + "epoch": 0.09, + "learning_rate": 9.791983328328742e-05, + "loss": 2.3115, + "step": 34455 + }, + { + "epoch": 0.09, + "learning_rate": 9.791923198283555e-05, + "loss": 2.3379, + "step": 34460 + }, + { + "epoch": 0.09, + "learning_rate": 9.791863059733595e-05, + "loss": 2.3333, + "step": 34465 + }, + { + "epoch": 0.09, + "learning_rate": 9.79180291267897e-05, + "loss": 2.411, + "step": 34470 + }, + { + "epoch": 0.09, + "learning_rate": 9.791742757119787e-05, + "loss": 2.4703, + "step": 34475 + }, + { + "epoch": 0.09, + "learning_rate": 9.791682593056154e-05, + "loss": 2.3118, + "step": 34480 + }, + { + "epoch": 0.09, + "learning_rate": 9.791622420488174e-05, + "loss": 2.375, + "step": 34485 + }, + { + "epoch": 0.09, + "learning_rate": 9.791562239415954e-05, + "loss": 2.4167, + "step": 34490 + }, + { + "epoch": 0.09, + "learning_rate": 9.791502049839606e-05, + "loss": 2.3799, + "step": 34495 + }, + { + "epoch": 0.09, + "learning_rate": 9.791441851759232e-05, + "loss": 2.2588, + "step": 34500 + }, + { + "epoch": 0.09, + "learning_rate": 9.791381645174942e-05, + "loss": 2.3925, + "step": 34505 + }, + { + "epoch": 0.09, + "learning_rate": 9.791321430086839e-05, + "loss": 2.3176, + "step": 34510 + }, + { + "epoch": 0.09, + "learning_rate": 9.791261206495032e-05, + "loss": 2.5058, + "step": 34515 + }, + { + "epoch": 0.09, + "learning_rate": 9.791200974399631e-05, + "loss": 2.3766, + "step": 34520 + }, + { + "epoch": 0.09, + "learning_rate": 9.791140733800738e-05, + "loss": 2.2879, + "step": 34525 + }, + { + "epoch": 0.09, + "learning_rate": 9.791080484698462e-05, + "loss": 2.4644, + "step": 34530 + }, + { + "epoch": 0.09, + "learning_rate": 9.791020227092908e-05, + "loss": 2.2838, + "step": 34535 + }, + { + "epoch": 0.09, + "learning_rate": 9.790959960984187e-05, + "loss": 2.4027, + "step": 34540 + }, + { + "epoch": 0.09, + "learning_rate": 9.790899686372404e-05, + "loss": 2.2972, + "step": 34545 + }, + { + "epoch": 0.09, + "learning_rate": 9.790839403257664e-05, + "loss": 2.4892, + "step": 34550 + }, + { + "epoch": 0.09, + "learning_rate": 9.790779111640077e-05, + "loss": 2.3674, + "step": 34555 + }, + { + "epoch": 0.09, + "learning_rate": 9.790718811519747e-05, + "loss": 2.3331, + "step": 34560 + }, + { + "epoch": 0.09, + "learning_rate": 9.790658502896784e-05, + "loss": 2.3683, + "step": 34565 + }, + { + "epoch": 0.09, + "learning_rate": 9.790598185771292e-05, + "loss": 2.4006, + "step": 34570 + }, + { + "epoch": 0.09, + "learning_rate": 9.79053786014338e-05, + "loss": 2.4971, + "step": 34575 + }, + { + "epoch": 0.09, + "learning_rate": 9.790477526013157e-05, + "loss": 2.3845, + "step": 34580 + }, + { + "epoch": 0.09, + "learning_rate": 9.790417183380726e-05, + "loss": 2.4268, + "step": 34585 + }, + { + "epoch": 0.09, + "learning_rate": 9.790356832246197e-05, + "loss": 2.3856, + "step": 34590 + }, + { + "epoch": 0.09, + "learning_rate": 9.790296472609675e-05, + "loss": 2.3908, + "step": 34595 + }, + { + "epoch": 0.09, + "learning_rate": 9.79023610447127e-05, + "loss": 2.3796, + "step": 34600 + }, + { + "epoch": 0.09, + "learning_rate": 9.790175727831086e-05, + "loss": 2.2493, + "step": 34605 + }, + { + "epoch": 0.09, + "learning_rate": 9.79011534268923e-05, + "loss": 2.2388, + "step": 34610 + }, + { + "epoch": 0.09, + "learning_rate": 9.790054949045812e-05, + "loss": 2.4712, + "step": 34615 + }, + { + "epoch": 0.09, + "learning_rate": 9.789994546900939e-05, + "loss": 2.4648, + "step": 34620 + }, + { + "epoch": 0.09, + "learning_rate": 9.789934136254718e-05, + "loss": 2.345, + "step": 34625 + }, + { + "epoch": 0.09, + "learning_rate": 9.789873717107252e-05, + "loss": 2.4425, + "step": 34630 + }, + { + "epoch": 0.09, + "learning_rate": 9.789813289458653e-05, + "loss": 2.3472, + "step": 34635 + }, + { + "epoch": 0.09, + "learning_rate": 9.789752853309026e-05, + "loss": 2.3665, + "step": 34640 + }, + { + "epoch": 0.09, + "learning_rate": 9.789692408658481e-05, + "loss": 2.3855, + "step": 34645 + }, + { + "epoch": 0.09, + "learning_rate": 9.789631955507122e-05, + "loss": 2.3542, + "step": 34650 + }, + { + "epoch": 0.09, + "learning_rate": 9.789571493855058e-05, + "loss": 2.3552, + "step": 34655 + }, + { + "epoch": 0.09, + "learning_rate": 9.789511023702396e-05, + "loss": 2.4362, + "step": 34660 + }, + { + "epoch": 0.09, + "learning_rate": 9.789450545049242e-05, + "loss": 2.3823, + "step": 34665 + }, + { + "epoch": 0.09, + "learning_rate": 9.789390057895706e-05, + "loss": 2.4315, + "step": 34670 + }, + { + "epoch": 0.09, + "learning_rate": 9.789329562241894e-05, + "loss": 2.4059, + "step": 34675 + }, + { + "epoch": 0.09, + "learning_rate": 9.789269058087912e-05, + "loss": 2.4141, + "step": 34680 + }, + { + "epoch": 0.09, + "learning_rate": 9.78920854543387e-05, + "loss": 2.3841, + "step": 34685 + }, + { + "epoch": 0.09, + "learning_rate": 9.789148024279872e-05, + "loss": 2.4283, + "step": 34690 + }, + { + "epoch": 0.09, + "learning_rate": 9.78908749462603e-05, + "loss": 2.3631, + "step": 34695 + }, + { + "epoch": 0.09, + "learning_rate": 9.789026956472448e-05, + "loss": 2.3988, + "step": 34700 + }, + { + "epoch": 0.09, + "learning_rate": 9.788966409819233e-05, + "loss": 2.4378, + "step": 34705 + }, + { + "epoch": 0.09, + "learning_rate": 9.788905854666496e-05, + "loss": 2.3676, + "step": 34710 + }, + { + "epoch": 0.09, + "learning_rate": 9.788845291014341e-05, + "loss": 2.4075, + "step": 34715 + }, + { + "epoch": 0.09, + "learning_rate": 9.788784718862877e-05, + "loss": 2.4331, + "step": 34720 + }, + { + "epoch": 0.09, + "learning_rate": 9.788724138212211e-05, + "loss": 2.3539, + "step": 34725 + }, + { + "epoch": 0.09, + "learning_rate": 9.788663549062452e-05, + "loss": 2.3299, + "step": 34730 + }, + { + "epoch": 0.09, + "learning_rate": 9.788602951413706e-05, + "loss": 2.3588, + "step": 34735 + }, + { + "epoch": 0.09, + "learning_rate": 9.788542345266081e-05, + "loss": 2.3705, + "step": 34740 + }, + { + "epoch": 0.09, + "learning_rate": 9.788481730619683e-05, + "loss": 2.3535, + "step": 34745 + }, + { + "epoch": 0.09, + "learning_rate": 9.788421107474624e-05, + "loss": 2.3969, + "step": 34750 + }, + { + "epoch": 0.09, + "learning_rate": 9.788360475831005e-05, + "loss": 2.4627, + "step": 34755 + }, + { + "epoch": 0.09, + "learning_rate": 9.788299835688938e-05, + "loss": 2.2584, + "step": 34760 + }, + { + "epoch": 0.09, + "learning_rate": 9.788239187048532e-05, + "loss": 2.3454, + "step": 34765 + }, + { + "epoch": 0.09, + "learning_rate": 9.788178529909892e-05, + "loss": 2.3475, + "step": 34770 + }, + { + "epoch": 0.09, + "learning_rate": 9.788117864273126e-05, + "loss": 2.2953, + "step": 34775 + }, + { + "epoch": 0.09, + "learning_rate": 9.78805719013834e-05, + "loss": 2.3197, + "step": 34780 + }, + { + "epoch": 0.09, + "learning_rate": 9.787996507505645e-05, + "loss": 2.3858, + "step": 34785 + }, + { + "epoch": 0.09, + "learning_rate": 9.787935816375147e-05, + "loss": 2.4161, + "step": 34790 + }, + { + "epoch": 0.09, + "learning_rate": 9.787875116746953e-05, + "loss": 2.3431, + "step": 34795 + }, + { + "epoch": 0.09, + "learning_rate": 9.787814408621174e-05, + "loss": 2.3724, + "step": 34800 + }, + { + "epoch": 0.09, + "learning_rate": 9.787753691997914e-05, + "loss": 2.3187, + "step": 34805 + }, + { + "epoch": 0.09, + "learning_rate": 9.787692966877281e-05, + "loss": 2.2366, + "step": 34810 + }, + { + "epoch": 0.09, + "learning_rate": 9.787632233259387e-05, + "loss": 2.5568, + "step": 34815 + }, + { + "epoch": 0.09, + "learning_rate": 9.787571491144334e-05, + "loss": 2.3215, + "step": 34820 + }, + { + "epoch": 0.09, + "learning_rate": 9.787510740532233e-05, + "loss": 2.334, + "step": 34825 + }, + { + "epoch": 0.09, + "learning_rate": 9.787449981423192e-05, + "loss": 2.4184, + "step": 34830 + }, + { + "epoch": 0.09, + "learning_rate": 9.787389213817318e-05, + "loss": 2.3731, + "step": 34835 + }, + { + "epoch": 0.09, + "learning_rate": 9.78732843771472e-05, + "loss": 2.3509, + "step": 34840 + }, + { + "epoch": 0.09, + "learning_rate": 9.787267653115503e-05, + "loss": 2.4394, + "step": 34845 + }, + { + "epoch": 0.09, + "learning_rate": 9.787206860019778e-05, + "loss": 2.3596, + "step": 34850 + }, + { + "epoch": 0.09, + "learning_rate": 9.787146058427651e-05, + "loss": 2.4604, + "step": 34855 + }, + { + "epoch": 0.09, + "learning_rate": 9.787085248339232e-05, + "loss": 2.3192, + "step": 34860 + }, + { + "epoch": 0.09, + "learning_rate": 9.787024429754626e-05, + "loss": 2.3635, + "step": 34865 + }, + { + "epoch": 0.09, + "learning_rate": 9.786963602673945e-05, + "loss": 2.2249, + "step": 34870 + }, + { + "epoch": 0.09, + "learning_rate": 9.786902767097292e-05, + "loss": 2.3737, + "step": 34875 + }, + { + "epoch": 0.09, + "learning_rate": 9.786841923024777e-05, + "loss": 2.3222, + "step": 34880 + }, + { + "epoch": 0.09, + "learning_rate": 9.78678107045651e-05, + "loss": 2.3701, + "step": 34885 + }, + { + "epoch": 0.09, + "learning_rate": 9.786720209392597e-05, + "loss": 2.3218, + "step": 34890 + }, + { + "epoch": 0.09, + "learning_rate": 9.786659339833145e-05, + "loss": 2.2478, + "step": 34895 + }, + { + "epoch": 0.09, + "learning_rate": 9.786598461778265e-05, + "loss": 2.4121, + "step": 34900 + }, + { + "epoch": 0.09, + "learning_rate": 9.786537575228063e-05, + "loss": 2.4794, + "step": 34905 + }, + { + "epoch": 0.09, + "learning_rate": 9.786476680182649e-05, + "loss": 2.399, + "step": 34910 + }, + { + "epoch": 0.09, + "learning_rate": 9.786415776642128e-05, + "loss": 2.2882, + "step": 34915 + }, + { + "epoch": 0.09, + "learning_rate": 9.78635486460661e-05, + "loss": 2.3221, + "step": 34920 + }, + { + "epoch": 0.09, + "learning_rate": 9.786293944076202e-05, + "loss": 2.4792, + "step": 34925 + }, + { + "epoch": 0.09, + "learning_rate": 9.786233015051014e-05, + "loss": 2.3345, + "step": 34930 + }, + { + "epoch": 0.09, + "learning_rate": 9.786172077531152e-05, + "loss": 2.3932, + "step": 34935 + }, + { + "epoch": 0.09, + "learning_rate": 9.786111131516727e-05, + "loss": 2.3604, + "step": 34940 + }, + { + "epoch": 0.09, + "learning_rate": 9.786050177007844e-05, + "loss": 2.418, + "step": 34945 + }, + { + "epoch": 0.09, + "learning_rate": 9.785989214004614e-05, + "loss": 2.3892, + "step": 34950 + }, + { + "epoch": 0.09, + "learning_rate": 9.785928242507143e-05, + "loss": 2.3394, + "step": 34955 + }, + { + "epoch": 0.09, + "learning_rate": 9.78586726251554e-05, + "loss": 2.2381, + "step": 34960 + }, + { + "epoch": 0.09, + "learning_rate": 9.785806274029913e-05, + "loss": 2.2831, + "step": 34965 + }, + { + "epoch": 0.09, + "learning_rate": 9.785745277050372e-05, + "loss": 2.4317, + "step": 34970 + }, + { + "epoch": 0.09, + "learning_rate": 9.78568427157702e-05, + "loss": 2.1514, + "step": 34975 + }, + { + "epoch": 0.09, + "learning_rate": 9.785623257609973e-05, + "loss": 2.3764, + "step": 34980 + }, + { + "epoch": 0.09, + "learning_rate": 9.785562235149333e-05, + "loss": 2.441, + "step": 34985 + }, + { + "epoch": 0.09, + "learning_rate": 9.785501204195212e-05, + "loss": 2.3792, + "step": 34990 + }, + { + "epoch": 0.09, + "learning_rate": 9.785440164747716e-05, + "loss": 2.3742, + "step": 34995 + }, + { + "epoch": 0.09, + "learning_rate": 9.785379116806955e-05, + "loss": 2.4611, + "step": 35000 + }, + { + "epoch": 0.09, + "learning_rate": 9.785318060373035e-05, + "loss": 2.4081, + "step": 35005 + }, + { + "epoch": 0.09, + "learning_rate": 9.785256995446069e-05, + "loss": 2.3393, + "step": 35010 + }, + { + "epoch": 0.09, + "learning_rate": 9.785195922026159e-05, + "loss": 2.3711, + "step": 35015 + }, + { + "epoch": 0.09, + "learning_rate": 9.785134840113418e-05, + "loss": 2.4473, + "step": 35020 + }, + { + "epoch": 0.09, + "learning_rate": 9.785073749707954e-05, + "loss": 2.3487, + "step": 35025 + }, + { + "epoch": 0.09, + "learning_rate": 9.785012650809872e-05, + "loss": 2.3492, + "step": 35030 + }, + { + "epoch": 0.09, + "learning_rate": 9.784951543419286e-05, + "loss": 2.3103, + "step": 35035 + }, + { + "epoch": 0.09, + "learning_rate": 9.7848904275363e-05, + "loss": 2.3547, + "step": 35040 + }, + { + "epoch": 0.09, + "learning_rate": 9.784829303161025e-05, + "loss": 2.3592, + "step": 35045 + }, + { + "epoch": 0.09, + "learning_rate": 9.784768170293568e-05, + "loss": 2.3788, + "step": 35050 + }, + { + "epoch": 0.09, + "learning_rate": 9.784707028934037e-05, + "loss": 2.3452, + "step": 35055 + }, + { + "epoch": 0.09, + "learning_rate": 9.784645879082542e-05, + "loss": 2.3796, + "step": 35060 + }, + { + "epoch": 0.09, + "learning_rate": 9.784584720739191e-05, + "loss": 2.3196, + "step": 35065 + }, + { + "epoch": 0.09, + "learning_rate": 9.784523553904093e-05, + "loss": 2.3142, + "step": 35070 + }, + { + "epoch": 0.09, + "learning_rate": 9.784462378577354e-05, + "loss": 2.4035, + "step": 35075 + }, + { + "epoch": 0.09, + "learning_rate": 9.784401194759087e-05, + "loss": 2.2935, + "step": 35080 + }, + { + "epoch": 0.09, + "learning_rate": 9.784340002449398e-05, + "loss": 2.4539, + "step": 35085 + }, + { + "epoch": 0.09, + "learning_rate": 9.784278801648395e-05, + "loss": 2.4173, + "step": 35090 + }, + { + "epoch": 0.09, + "learning_rate": 9.784217592356188e-05, + "loss": 2.3388, + "step": 35095 + }, + { + "epoch": 0.09, + "learning_rate": 9.784156374572885e-05, + "loss": 2.3639, + "step": 35100 + }, + { + "epoch": 0.09, + "learning_rate": 9.784095148298595e-05, + "loss": 2.3187, + "step": 35105 + }, + { + "epoch": 0.09, + "learning_rate": 9.784033913533426e-05, + "loss": 2.2903, + "step": 35110 + }, + { + "epoch": 0.09, + "learning_rate": 9.783972670277486e-05, + "loss": 2.4509, + "step": 35115 + }, + { + "epoch": 0.09, + "learning_rate": 9.783911418530887e-05, + "loss": 2.2837, + "step": 35120 + }, + { + "epoch": 0.09, + "learning_rate": 9.783850158293734e-05, + "loss": 2.4618, + "step": 35125 + }, + { + "epoch": 0.09, + "learning_rate": 9.783788889566136e-05, + "loss": 2.3125, + "step": 35130 + }, + { + "epoch": 0.09, + "learning_rate": 9.783727612348206e-05, + "loss": 2.425, + "step": 35135 + }, + { + "epoch": 0.09, + "learning_rate": 9.783666326640049e-05, + "loss": 2.211, + "step": 35140 + }, + { + "epoch": 0.09, + "learning_rate": 9.783605032441773e-05, + "loss": 2.3869, + "step": 35145 + }, + { + "epoch": 0.09, + "learning_rate": 9.78354372975349e-05, + "loss": 2.3878, + "step": 35150 + }, + { + "epoch": 0.09, + "learning_rate": 9.783482418575305e-05, + "loss": 2.3682, + "step": 35155 + }, + { + "epoch": 0.09, + "learning_rate": 9.783421098907331e-05, + "loss": 2.3626, + "step": 35160 + }, + { + "epoch": 0.09, + "learning_rate": 9.783359770749672e-05, + "loss": 2.429, + "step": 35165 + }, + { + "epoch": 0.09, + "learning_rate": 9.783298434102442e-05, + "loss": 2.5226, + "step": 35170 + }, + { + "epoch": 0.09, + "learning_rate": 9.783237088965747e-05, + "loss": 2.4353, + "step": 35175 + }, + { + "epoch": 0.09, + "learning_rate": 9.783175735339696e-05, + "loss": 2.2844, + "step": 35180 + }, + { + "epoch": 0.09, + "learning_rate": 9.783114373224398e-05, + "loss": 2.383, + "step": 35185 + }, + { + "epoch": 0.09, + "learning_rate": 9.783053002619962e-05, + "loss": 2.3418, + "step": 35190 + }, + { + "epoch": 0.09, + "learning_rate": 9.782991623526499e-05, + "loss": 2.2684, + "step": 35195 + }, + { + "epoch": 0.09, + "learning_rate": 9.782930235944113e-05, + "loss": 2.3547, + "step": 35200 + }, + { + "epoch": 0.09, + "learning_rate": 9.782868839872917e-05, + "loss": 2.3868, + "step": 35205 + }, + { + "epoch": 0.09, + "learning_rate": 9.782807435313019e-05, + "loss": 2.3526, + "step": 35210 + }, + { + "epoch": 0.09, + "learning_rate": 9.782746022264528e-05, + "loss": 2.3183, + "step": 35215 + }, + { + "epoch": 0.09, + "learning_rate": 9.782684600727553e-05, + "loss": 2.3734, + "step": 35220 + }, + { + "epoch": 0.09, + "learning_rate": 9.782623170702203e-05, + "loss": 2.1607, + "step": 35225 + }, + { + "epoch": 0.09, + "learning_rate": 9.782561732188586e-05, + "loss": 2.4981, + "step": 35230 + }, + { + "epoch": 0.09, + "learning_rate": 9.782500285186813e-05, + "loss": 2.4116, + "step": 35235 + }, + { + "epoch": 0.09, + "learning_rate": 9.782438829696991e-05, + "loss": 2.4017, + "step": 35240 + }, + { + "epoch": 0.09, + "learning_rate": 9.782377365719231e-05, + "loss": 2.2683, + "step": 35245 + }, + { + "epoch": 0.09, + "learning_rate": 9.78231589325364e-05, + "loss": 2.3797, + "step": 35250 + }, + { + "epoch": 0.09, + "learning_rate": 9.782254412300329e-05, + "loss": 2.4468, + "step": 35255 + }, + { + "epoch": 0.09, + "learning_rate": 9.782192922859407e-05, + "loss": 2.3517, + "step": 35260 + }, + { + "epoch": 0.09, + "learning_rate": 9.782131424930981e-05, + "loss": 2.3665, + "step": 35265 + }, + { + "epoch": 0.09, + "learning_rate": 9.782069918515164e-05, + "loss": 2.4435, + "step": 35270 + }, + { + "epoch": 0.09, + "learning_rate": 9.78200840361206e-05, + "loss": 2.3604, + "step": 35275 + }, + { + "epoch": 0.09, + "learning_rate": 9.781946880221782e-05, + "loss": 2.3295, + "step": 35280 + }, + { + "epoch": 0.09, + "learning_rate": 9.781885348344439e-05, + "loss": 2.4523, + "step": 35285 + }, + { + "epoch": 0.09, + "learning_rate": 9.781823807980139e-05, + "loss": 2.3532, + "step": 35290 + }, + { + "epoch": 0.09, + "learning_rate": 9.781762259128991e-05, + "loss": 2.3216, + "step": 35295 + }, + { + "epoch": 0.09, + "learning_rate": 9.781700701791106e-05, + "loss": 2.3899, + "step": 35300 + }, + { + "epoch": 0.09, + "learning_rate": 9.781639135966592e-05, + "loss": 2.3114, + "step": 35305 + }, + { + "epoch": 0.09, + "learning_rate": 9.781577561655557e-05, + "loss": 2.3482, + "step": 35310 + }, + { + "epoch": 0.09, + "learning_rate": 9.781515978858114e-05, + "loss": 2.5081, + "step": 35315 + }, + { + "epoch": 0.09, + "learning_rate": 9.781454387574367e-05, + "loss": 2.3413, + "step": 35320 + }, + { + "epoch": 0.09, + "learning_rate": 9.78139278780443e-05, + "loss": 2.332, + "step": 35325 + }, + { + "epoch": 0.09, + "learning_rate": 9.781331179548409e-05, + "loss": 2.3128, + "step": 35330 + }, + { + "epoch": 0.09, + "learning_rate": 9.781269562806416e-05, + "loss": 2.3119, + "step": 35335 + }, + { + "epoch": 0.09, + "learning_rate": 9.78120793757856e-05, + "loss": 2.4656, + "step": 35340 + }, + { + "epoch": 0.09, + "learning_rate": 9.78114630386495e-05, + "loss": 2.2898, + "step": 35345 + }, + { + "epoch": 0.09, + "learning_rate": 9.781084661665693e-05, + "loss": 2.4831, + "step": 35350 + }, + { + "epoch": 0.09, + "learning_rate": 9.781023010980903e-05, + "loss": 2.4393, + "step": 35355 + }, + { + "epoch": 0.09, + "learning_rate": 9.780961351810684e-05, + "loss": 2.3521, + "step": 35360 + }, + { + "epoch": 0.09, + "learning_rate": 9.780899684155152e-05, + "loss": 2.3513, + "step": 35365 + }, + { + "epoch": 0.09, + "learning_rate": 9.78083800801441e-05, + "loss": 2.3545, + "step": 35370 + }, + { + "epoch": 0.09, + "learning_rate": 9.780776323388571e-05, + "loss": 2.4133, + "step": 35375 + }, + { + "epoch": 0.09, + "learning_rate": 9.780714630277744e-05, + "loss": 2.5706, + "step": 35380 + }, + { + "epoch": 0.09, + "learning_rate": 9.780652928682036e-05, + "loss": 2.437, + "step": 35385 + }, + { + "epoch": 0.09, + "learning_rate": 9.780591218601562e-05, + "loss": 2.3785, + "step": 35390 + }, + { + "epoch": 0.09, + "learning_rate": 9.780529500036425e-05, + "loss": 2.3604, + "step": 35395 + }, + { + "epoch": 0.09, + "learning_rate": 9.780467772986741e-05, + "loss": 2.3513, + "step": 35400 + }, + { + "epoch": 0.09, + "learning_rate": 9.780406037452614e-05, + "loss": 2.344, + "step": 35405 + }, + { + "epoch": 0.09, + "learning_rate": 9.780344293434159e-05, + "loss": 2.5036, + "step": 35410 + }, + { + "epoch": 0.09, + "learning_rate": 9.780282540931477e-05, + "loss": 2.3187, + "step": 35415 + }, + { + "epoch": 0.09, + "learning_rate": 9.780220779944688e-05, + "loss": 2.4144, + "step": 35420 + }, + { + "epoch": 0.09, + "learning_rate": 9.780159010473895e-05, + "loss": 2.3352, + "step": 35425 + }, + { + "epoch": 0.09, + "learning_rate": 9.78009723251921e-05, + "loss": 2.2838, + "step": 35430 + }, + { + "epoch": 0.1, + "learning_rate": 9.780035446080739e-05, + "loss": 2.5635, + "step": 35435 + }, + { + "epoch": 0.1, + "learning_rate": 9.779973651158598e-05, + "loss": 2.4107, + "step": 35440 + }, + { + "epoch": 0.1, + "learning_rate": 9.779911847752891e-05, + "loss": 2.3368, + "step": 35445 + }, + { + "epoch": 0.1, + "learning_rate": 9.779850035863732e-05, + "loss": 2.3725, + "step": 35450 + }, + { + "epoch": 0.1, + "learning_rate": 9.779788215491228e-05, + "loss": 2.367, + "step": 35455 + }, + { + "epoch": 0.1, + "learning_rate": 9.77972638663549e-05, + "loss": 2.437, + "step": 35460 + }, + { + "epoch": 0.1, + "learning_rate": 9.779664549296626e-05, + "loss": 2.3188, + "step": 35465 + }, + { + "epoch": 0.1, + "learning_rate": 9.779602703474747e-05, + "loss": 2.3349, + "step": 35470 + }, + { + "epoch": 0.1, + "learning_rate": 9.779540849169963e-05, + "loss": 2.2347, + "step": 35475 + }, + { + "epoch": 0.1, + "learning_rate": 9.779478986382385e-05, + "loss": 2.4194, + "step": 35480 + }, + { + "epoch": 0.1, + "learning_rate": 9.779417115112118e-05, + "loss": 2.471, + "step": 35485 + }, + { + "epoch": 0.1, + "learning_rate": 9.779355235359277e-05, + "loss": 2.3925, + "step": 35490 + }, + { + "epoch": 0.1, + "learning_rate": 9.77929334712397e-05, + "loss": 2.4235, + "step": 35495 + }, + { + "epoch": 0.1, + "learning_rate": 9.779231450406305e-05, + "loss": 2.4815, + "step": 35500 + }, + { + "epoch": 0.1, + "learning_rate": 9.779169545206396e-05, + "loss": 2.4694, + "step": 35505 + }, + { + "epoch": 0.1, + "learning_rate": 9.779107631524348e-05, + "loss": 2.282, + "step": 35510 + }, + { + "epoch": 0.1, + "learning_rate": 9.779045709360275e-05, + "loss": 2.3994, + "step": 35515 + }, + { + "epoch": 0.1, + "learning_rate": 9.778983778714286e-05, + "loss": 2.3845, + "step": 35520 + }, + { + "epoch": 0.1, + "learning_rate": 9.778921839586488e-05, + "loss": 2.3143, + "step": 35525 + }, + { + "epoch": 0.1, + "learning_rate": 9.778859891976995e-05, + "loss": 2.2786, + "step": 35530 + }, + { + "epoch": 0.1, + "learning_rate": 9.778797935885912e-05, + "loss": 2.3957, + "step": 35535 + }, + { + "epoch": 0.1, + "learning_rate": 9.778735971313355e-05, + "loss": 2.2986, + "step": 35540 + }, + { + "epoch": 0.1, + "learning_rate": 9.778673998259431e-05, + "loss": 2.3606, + "step": 35545 + }, + { + "epoch": 0.1, + "learning_rate": 9.778612016724249e-05, + "loss": 2.3403, + "step": 35550 + }, + { + "epoch": 0.1, + "learning_rate": 9.778550026707919e-05, + "loss": 2.3688, + "step": 35555 + }, + { + "epoch": 0.1, + "learning_rate": 9.778488028210553e-05, + "loss": 2.4817, + "step": 35560 + }, + { + "epoch": 0.1, + "learning_rate": 9.77842602123226e-05, + "loss": 2.5515, + "step": 35565 + }, + { + "epoch": 0.1, + "learning_rate": 9.77836400577315e-05, + "loss": 2.4589, + "step": 35570 + }, + { + "epoch": 0.1, + "learning_rate": 9.778301981833333e-05, + "loss": 2.4934, + "step": 35575 + }, + { + "epoch": 0.1, + "learning_rate": 9.778239949412919e-05, + "loss": 2.4144, + "step": 35580 + }, + { + "epoch": 0.1, + "learning_rate": 9.778177908512018e-05, + "loss": 2.4364, + "step": 35585 + }, + { + "epoch": 0.1, + "learning_rate": 9.778115859130741e-05, + "loss": 2.3835, + "step": 35590 + }, + { + "epoch": 0.1, + "learning_rate": 9.778053801269198e-05, + "loss": 2.3293, + "step": 35595 + }, + { + "epoch": 0.1, + "learning_rate": 9.777991734927498e-05, + "loss": 2.3545, + "step": 35600 + }, + { + "epoch": 0.1, + "learning_rate": 9.777929660105752e-05, + "loss": 2.3044, + "step": 35605 + }, + { + "epoch": 0.1, + "learning_rate": 9.77786757680407e-05, + "loss": 2.3864, + "step": 35610 + }, + { + "epoch": 0.1, + "learning_rate": 9.777805485022562e-05, + "loss": 2.2958, + "step": 35615 + }, + { + "epoch": 0.1, + "learning_rate": 9.777743384761338e-05, + "loss": 2.3325, + "step": 35620 + }, + { + "epoch": 0.1, + "learning_rate": 9.777681276020509e-05, + "loss": 2.4, + "step": 35625 + }, + { + "epoch": 0.1, + "learning_rate": 9.777619158800186e-05, + "loss": 2.3289, + "step": 35630 + }, + { + "epoch": 0.1, + "learning_rate": 9.777557033100476e-05, + "loss": 2.311, + "step": 35635 + }, + { + "epoch": 0.1, + "learning_rate": 9.777494898921493e-05, + "loss": 2.3723, + "step": 35640 + }, + { + "epoch": 0.1, + "learning_rate": 9.777432756263345e-05, + "loss": 2.4337, + "step": 35645 + }, + { + "epoch": 0.1, + "learning_rate": 9.777370605126143e-05, + "loss": 2.4019, + "step": 35650 + }, + { + "epoch": 0.1, + "learning_rate": 9.777308445509996e-05, + "loss": 2.4151, + "step": 35655 + }, + { + "epoch": 0.1, + "learning_rate": 9.777246277415018e-05, + "loss": 2.326, + "step": 35660 + }, + { + "epoch": 0.1, + "learning_rate": 9.777184100841315e-05, + "loss": 2.3066, + "step": 35665 + }, + { + "epoch": 0.1, + "learning_rate": 9.777121915789e-05, + "loss": 2.3194, + "step": 35670 + }, + { + "epoch": 0.1, + "learning_rate": 9.777059722258182e-05, + "loss": 2.4021, + "step": 35675 + }, + { + "epoch": 0.1, + "learning_rate": 9.776997520248973e-05, + "loss": 2.4273, + "step": 35680 + }, + { + "epoch": 0.1, + "learning_rate": 9.776935309761482e-05, + "loss": 2.2723, + "step": 35685 + }, + { + "epoch": 0.1, + "learning_rate": 9.77687309079582e-05, + "loss": 2.2665, + "step": 35690 + }, + { + "epoch": 0.1, + "learning_rate": 9.776810863352095e-05, + "loss": 2.3515, + "step": 35695 + }, + { + "epoch": 0.1, + "learning_rate": 9.776748627430422e-05, + "loss": 2.3308, + "step": 35700 + }, + { + "epoch": 0.1, + "learning_rate": 9.776686383030908e-05, + "loss": 2.3102, + "step": 35705 + }, + { + "epoch": 0.1, + "learning_rate": 9.776624130153665e-05, + "loss": 2.3462, + "step": 35710 + }, + { + "epoch": 0.1, + "learning_rate": 9.776561868798803e-05, + "loss": 2.2972, + "step": 35715 + }, + { + "epoch": 0.1, + "learning_rate": 9.776499598966433e-05, + "loss": 2.3407, + "step": 35720 + }, + { + "epoch": 0.1, + "learning_rate": 9.776437320656666e-05, + "loss": 2.4195, + "step": 35725 + }, + { + "epoch": 0.1, + "learning_rate": 9.776375033869609e-05, + "loss": 2.3101, + "step": 35730 + }, + { + "epoch": 0.1, + "learning_rate": 9.776312738605378e-05, + "loss": 2.432, + "step": 35735 + }, + { + "epoch": 0.1, + "learning_rate": 9.776250434864078e-05, + "loss": 2.3134, + "step": 35740 + }, + { + "epoch": 0.1, + "learning_rate": 9.776188122645823e-05, + "loss": 2.3893, + "step": 35745 + }, + { + "epoch": 0.1, + "learning_rate": 9.776125801950725e-05, + "loss": 2.3618, + "step": 35750 + }, + { + "epoch": 0.1, + "learning_rate": 9.776063472778891e-05, + "loss": 2.3421, + "step": 35755 + }, + { + "epoch": 0.1, + "learning_rate": 9.776001135130434e-05, + "loss": 2.3922, + "step": 35760 + }, + { + "epoch": 0.1, + "learning_rate": 9.775938789005463e-05, + "loss": 2.2769, + "step": 35765 + }, + { + "epoch": 0.1, + "learning_rate": 9.77587643440409e-05, + "loss": 2.39, + "step": 35770 + }, + { + "epoch": 0.1, + "learning_rate": 9.775814071326424e-05, + "loss": 2.337, + "step": 35775 + }, + { + "epoch": 0.1, + "learning_rate": 9.775751699772578e-05, + "loss": 2.3524, + "step": 35780 + }, + { + "epoch": 0.1, + "learning_rate": 9.77568931974266e-05, + "loss": 2.3698, + "step": 35785 + }, + { + "epoch": 0.1, + "learning_rate": 9.775626931236783e-05, + "loss": 2.3536, + "step": 35790 + }, + { + "epoch": 0.1, + "learning_rate": 9.775564534255057e-05, + "loss": 2.3174, + "step": 35795 + }, + { + "epoch": 0.1, + "learning_rate": 9.775502128797593e-05, + "loss": 2.3959, + "step": 35800 + }, + { + "epoch": 0.1, + "learning_rate": 9.775439714864501e-05, + "loss": 2.4051, + "step": 35805 + }, + { + "epoch": 0.1, + "learning_rate": 9.775377292455892e-05, + "loss": 2.3521, + "step": 35810 + }, + { + "epoch": 0.1, + "learning_rate": 9.775314861571877e-05, + "loss": 2.4173, + "step": 35815 + }, + { + "epoch": 0.1, + "learning_rate": 9.775252422212566e-05, + "loss": 2.4111, + "step": 35820 + }, + { + "epoch": 0.1, + "learning_rate": 9.775189974378072e-05, + "loss": 2.416, + "step": 35825 + }, + { + "epoch": 0.1, + "learning_rate": 9.775127518068505e-05, + "loss": 2.3822, + "step": 35830 + }, + { + "epoch": 0.1, + "learning_rate": 9.775065053283973e-05, + "loss": 2.3199, + "step": 35835 + }, + { + "epoch": 0.1, + "learning_rate": 9.775002580024591e-05, + "loss": 2.3821, + "step": 35840 + }, + { + "epoch": 0.1, + "learning_rate": 9.774940098290468e-05, + "loss": 2.2709, + "step": 35845 + }, + { + "epoch": 0.1, + "learning_rate": 9.774877608081714e-05, + "loss": 2.3769, + "step": 35850 + }, + { + "epoch": 0.1, + "learning_rate": 9.774815109398441e-05, + "loss": 2.283, + "step": 35855 + }, + { + "epoch": 0.1, + "learning_rate": 9.77475260224076e-05, + "loss": 2.3159, + "step": 35860 + }, + { + "epoch": 0.1, + "learning_rate": 9.774690086608782e-05, + "loss": 2.2845, + "step": 35865 + }, + { + "epoch": 0.1, + "learning_rate": 9.774627562502616e-05, + "loss": 2.4684, + "step": 35870 + }, + { + "epoch": 0.1, + "learning_rate": 9.774565029922377e-05, + "loss": 2.3986, + "step": 35875 + }, + { + "epoch": 0.1, + "learning_rate": 9.774502488868173e-05, + "loss": 2.2935, + "step": 35880 + }, + { + "epoch": 0.1, + "learning_rate": 9.774439939340114e-05, + "loss": 2.3478, + "step": 35885 + }, + { + "epoch": 0.1, + "learning_rate": 9.774377381338314e-05, + "loss": 2.3602, + "step": 35890 + }, + { + "epoch": 0.1, + "learning_rate": 9.774314814862882e-05, + "loss": 2.3476, + "step": 35895 + }, + { + "epoch": 0.1, + "learning_rate": 9.77425223991393e-05, + "loss": 2.4453, + "step": 35900 + }, + { + "epoch": 0.1, + "learning_rate": 9.774189656491569e-05, + "loss": 2.1822, + "step": 35905 + }, + { + "epoch": 0.1, + "learning_rate": 9.77412706459591e-05, + "loss": 2.3554, + "step": 35910 + }, + { + "epoch": 0.1, + "learning_rate": 9.774064464227064e-05, + "loss": 2.4025, + "step": 35915 + }, + { + "epoch": 0.1, + "learning_rate": 9.77400185538514e-05, + "loss": 2.4149, + "step": 35920 + }, + { + "epoch": 0.1, + "learning_rate": 9.773939238070254e-05, + "loss": 2.3902, + "step": 35925 + }, + { + "epoch": 0.1, + "learning_rate": 9.773876612282512e-05, + "loss": 2.3489, + "step": 35930 + }, + { + "epoch": 0.1, + "learning_rate": 9.773813978022028e-05, + "loss": 2.3908, + "step": 35935 + }, + { + "epoch": 0.1, + "learning_rate": 9.773751335288913e-05, + "loss": 2.439, + "step": 35940 + }, + { + "epoch": 0.1, + "learning_rate": 9.773688684083277e-05, + "loss": 2.229, + "step": 35945 + }, + { + "epoch": 0.1, + "learning_rate": 9.773626024405234e-05, + "loss": 2.2623, + "step": 35950 + }, + { + "epoch": 0.1, + "learning_rate": 9.773563356254891e-05, + "loss": 2.3504, + "step": 35955 + }, + { + "epoch": 0.1, + "learning_rate": 9.773500679632361e-05, + "loss": 2.3221, + "step": 35960 + }, + { + "epoch": 0.1, + "learning_rate": 9.773437994537758e-05, + "loss": 2.2429, + "step": 35965 + }, + { + "epoch": 0.1, + "learning_rate": 9.77337530097119e-05, + "loss": 2.3784, + "step": 35970 + }, + { + "epoch": 0.1, + "learning_rate": 9.773312598932768e-05, + "loss": 2.4067, + "step": 35975 + }, + { + "epoch": 0.1, + "learning_rate": 9.773249888422604e-05, + "loss": 2.2951, + "step": 35980 + }, + { + "epoch": 0.1, + "learning_rate": 9.773187169440812e-05, + "loss": 2.3401, + "step": 35985 + }, + { + "epoch": 0.1, + "learning_rate": 9.7731244419875e-05, + "loss": 2.2891, + "step": 35990 + }, + { + "epoch": 0.1, + "learning_rate": 9.77306170606278e-05, + "loss": 2.303, + "step": 35995 + }, + { + "epoch": 0.1, + "learning_rate": 9.772998961666764e-05, + "loss": 2.4122, + "step": 36000 + }, + { + "epoch": 0.1, + "learning_rate": 9.772936208799564e-05, + "loss": 2.4553, + "step": 36005 + }, + { + "epoch": 0.1, + "learning_rate": 9.772873447461288e-05, + "loss": 2.4013, + "step": 36010 + }, + { + "epoch": 0.1, + "learning_rate": 9.772810677652051e-05, + "loss": 2.3869, + "step": 36015 + }, + { + "epoch": 0.1, + "learning_rate": 9.772747899371964e-05, + "loss": 2.4086, + "step": 36020 + }, + { + "epoch": 0.1, + "learning_rate": 9.772685112621139e-05, + "loss": 2.3086, + "step": 36025 + }, + { + "epoch": 0.1, + "learning_rate": 9.772622317399683e-05, + "loss": 2.3421, + "step": 36030 + }, + { + "epoch": 0.1, + "learning_rate": 9.772559513707712e-05, + "loss": 2.4232, + "step": 36035 + }, + { + "epoch": 0.1, + "learning_rate": 9.772496701545335e-05, + "loss": 2.4238, + "step": 36040 + }, + { + "epoch": 0.1, + "learning_rate": 9.772433880912666e-05, + "loss": 2.3397, + "step": 36045 + }, + { + "epoch": 0.1, + "learning_rate": 9.772371051809815e-05, + "loss": 2.3978, + "step": 36050 + }, + { + "epoch": 0.1, + "learning_rate": 9.772308214236892e-05, + "loss": 2.3261, + "step": 36055 + }, + { + "epoch": 0.1, + "learning_rate": 9.77224536819401e-05, + "loss": 2.2485, + "step": 36060 + }, + { + "epoch": 0.1, + "learning_rate": 9.772182513681283e-05, + "loss": 2.356, + "step": 36065 + }, + { + "epoch": 0.1, + "learning_rate": 9.77211965069882e-05, + "loss": 2.4955, + "step": 36070 + }, + { + "epoch": 0.1, + "learning_rate": 9.77205677924673e-05, + "loss": 2.3483, + "step": 36075 + }, + { + "epoch": 0.1, + "learning_rate": 9.771993899325128e-05, + "loss": 2.4906, + "step": 36080 + }, + { + "epoch": 0.1, + "learning_rate": 9.771931010934126e-05, + "loss": 2.2538, + "step": 36085 + }, + { + "epoch": 0.1, + "learning_rate": 9.771868114073834e-05, + "loss": 2.2751, + "step": 36090 + }, + { + "epoch": 0.1, + "learning_rate": 9.771805208744364e-05, + "loss": 2.35, + "step": 36095 + }, + { + "epoch": 0.1, + "learning_rate": 9.771742294945827e-05, + "loss": 2.3443, + "step": 36100 + }, + { + "epoch": 0.1, + "learning_rate": 9.771679372678337e-05, + "loss": 2.3756, + "step": 36105 + }, + { + "epoch": 0.1, + "learning_rate": 9.771616441942003e-05, + "loss": 2.3761, + "step": 36110 + }, + { + "epoch": 0.1, + "learning_rate": 9.771553502736938e-05, + "loss": 2.3837, + "step": 36115 + }, + { + "epoch": 0.1, + "learning_rate": 9.771490555063253e-05, + "loss": 2.3764, + "step": 36120 + }, + { + "epoch": 0.1, + "learning_rate": 9.771427598921062e-05, + "loss": 2.3243, + "step": 36125 + }, + { + "epoch": 0.1, + "learning_rate": 9.771364634310472e-05, + "loss": 2.4538, + "step": 36130 + }, + { + "epoch": 0.1, + "learning_rate": 9.7713016612316e-05, + "loss": 2.3583, + "step": 36135 + }, + { + "epoch": 0.1, + "learning_rate": 9.771238679684556e-05, + "loss": 2.2275, + "step": 36140 + }, + { + "epoch": 0.1, + "learning_rate": 9.771175689669449e-05, + "loss": 2.2817, + "step": 36145 + }, + { + "epoch": 0.1, + "learning_rate": 9.771112691186395e-05, + "loss": 2.3511, + "step": 36150 + }, + { + "epoch": 0.1, + "learning_rate": 9.771049684235503e-05, + "loss": 2.4888, + "step": 36155 + }, + { + "epoch": 0.1, + "learning_rate": 9.770986668816885e-05, + "loss": 2.425, + "step": 36160 + }, + { + "epoch": 0.1, + "learning_rate": 9.770923644930654e-05, + "loss": 2.3217, + "step": 36165 + }, + { + "epoch": 0.1, + "learning_rate": 9.770860612576923e-05, + "loss": 2.3572, + "step": 36170 + }, + { + "epoch": 0.1, + "learning_rate": 9.7707975717558e-05, + "loss": 2.3285, + "step": 36175 + }, + { + "epoch": 0.1, + "learning_rate": 9.7707345224674e-05, + "loss": 2.2925, + "step": 36180 + }, + { + "epoch": 0.1, + "learning_rate": 9.770671464711834e-05, + "loss": 2.3166, + "step": 36185 + }, + { + "epoch": 0.1, + "learning_rate": 9.770608398489214e-05, + "loss": 2.3882, + "step": 36190 + }, + { + "epoch": 0.1, + "learning_rate": 9.770545323799652e-05, + "loss": 2.3616, + "step": 36195 + }, + { + "epoch": 0.1, + "learning_rate": 9.77048224064326e-05, + "loss": 2.5085, + "step": 36200 + }, + { + "epoch": 0.1, + "learning_rate": 9.770419149020149e-05, + "loss": 2.4042, + "step": 36205 + }, + { + "epoch": 0.1, + "learning_rate": 9.770356048930432e-05, + "loss": 2.4683, + "step": 36210 + }, + { + "epoch": 0.1, + "learning_rate": 9.770292940374222e-05, + "loss": 2.4066, + "step": 36215 + }, + { + "epoch": 0.1, + "learning_rate": 9.770229823351629e-05, + "loss": 2.4704, + "step": 36220 + }, + { + "epoch": 0.1, + "learning_rate": 9.770166697862763e-05, + "loss": 2.2418, + "step": 36225 + }, + { + "epoch": 0.1, + "learning_rate": 9.770103563907741e-05, + "loss": 2.4405, + "step": 36230 + }, + { + "epoch": 0.1, + "learning_rate": 9.770040421486674e-05, + "loss": 2.3934, + "step": 36235 + }, + { + "epoch": 0.1, + "learning_rate": 9.769977270599672e-05, + "loss": 2.2997, + "step": 36240 + }, + { + "epoch": 0.1, + "learning_rate": 9.76991411124685e-05, + "loss": 2.4037, + "step": 36245 + }, + { + "epoch": 0.1, + "learning_rate": 9.769850943428315e-05, + "loss": 2.2973, + "step": 36250 + }, + { + "epoch": 0.1, + "learning_rate": 9.769787767144184e-05, + "loss": 2.3555, + "step": 36255 + }, + { + "epoch": 0.1, + "learning_rate": 9.769724582394567e-05, + "loss": 2.2796, + "step": 36260 + }, + { + "epoch": 0.1, + "learning_rate": 9.769661389179576e-05, + "loss": 2.4351, + "step": 36265 + }, + { + "epoch": 0.1, + "learning_rate": 9.769598187499325e-05, + "loss": 2.3836, + "step": 36270 + }, + { + "epoch": 0.1, + "learning_rate": 9.769534977353923e-05, + "loss": 2.394, + "step": 36275 + }, + { + "epoch": 0.1, + "learning_rate": 9.769471758743484e-05, + "loss": 2.3218, + "step": 36280 + }, + { + "epoch": 0.1, + "learning_rate": 9.769408531668121e-05, + "loss": 2.4122, + "step": 36285 + }, + { + "epoch": 0.1, + "learning_rate": 9.769345296127946e-05, + "loss": 2.3056, + "step": 36290 + }, + { + "epoch": 0.1, + "learning_rate": 9.76928205212307e-05, + "loss": 2.2929, + "step": 36295 + }, + { + "epoch": 0.1, + "learning_rate": 9.769218799653605e-05, + "loss": 2.4277, + "step": 36300 + }, + { + "epoch": 0.1, + "learning_rate": 9.769155538719666e-05, + "loss": 2.3938, + "step": 36305 + }, + { + "epoch": 0.1, + "learning_rate": 9.769092269321363e-05, + "loss": 2.37, + "step": 36310 + }, + { + "epoch": 0.1, + "learning_rate": 9.769028991458809e-05, + "loss": 2.374, + "step": 36315 + }, + { + "epoch": 0.1, + "learning_rate": 9.768965705132114e-05, + "loss": 2.5616, + "step": 36320 + }, + { + "epoch": 0.1, + "learning_rate": 9.768902410341395e-05, + "loss": 2.4191, + "step": 36325 + }, + { + "epoch": 0.1, + "learning_rate": 9.768839107086759e-05, + "loss": 2.3441, + "step": 36330 + }, + { + "epoch": 0.1, + "learning_rate": 9.768775795368323e-05, + "loss": 2.3585, + "step": 36335 + }, + { + "epoch": 0.1, + "learning_rate": 9.768712475186197e-05, + "loss": 2.4388, + "step": 36340 + }, + { + "epoch": 0.1, + "learning_rate": 9.768649146540494e-05, + "loss": 2.3798, + "step": 36345 + }, + { + "epoch": 0.1, + "learning_rate": 9.768585809431325e-05, + "loss": 2.3852, + "step": 36350 + }, + { + "epoch": 0.1, + "learning_rate": 9.768522463858804e-05, + "loss": 2.3187, + "step": 36355 + }, + { + "epoch": 0.1, + "learning_rate": 9.768459109823044e-05, + "loss": 2.4189, + "step": 36360 + }, + { + "epoch": 0.1, + "learning_rate": 9.768395747324155e-05, + "loss": 2.37, + "step": 36365 + }, + { + "epoch": 0.1, + "learning_rate": 9.768332376362253e-05, + "loss": 2.3454, + "step": 36370 + }, + { + "epoch": 0.1, + "learning_rate": 9.768268996937446e-05, + "loss": 2.35, + "step": 36375 + }, + { + "epoch": 0.1, + "learning_rate": 9.76820560904985e-05, + "loss": 2.4532, + "step": 36380 + }, + { + "epoch": 0.1, + "learning_rate": 9.768142212699577e-05, + "loss": 2.3214, + "step": 36385 + }, + { + "epoch": 0.1, + "learning_rate": 9.768078807886738e-05, + "loss": 2.2954, + "step": 36390 + }, + { + "epoch": 0.1, + "learning_rate": 9.768015394611447e-05, + "loss": 2.4174, + "step": 36395 + }, + { + "epoch": 0.1, + "learning_rate": 9.767951972873817e-05, + "loss": 2.3583, + "step": 36400 + }, + { + "epoch": 0.1, + "learning_rate": 9.767888542673957e-05, + "loss": 2.445, + "step": 36405 + }, + { + "epoch": 0.1, + "learning_rate": 9.767825104011984e-05, + "loss": 2.4601, + "step": 36410 + }, + { + "epoch": 0.1, + "learning_rate": 9.767761656888008e-05, + "loss": 2.3126, + "step": 36415 + }, + { + "epoch": 0.1, + "learning_rate": 9.767698201302143e-05, + "loss": 2.4557, + "step": 36420 + }, + { + "epoch": 0.1, + "learning_rate": 9.7676347372545e-05, + "loss": 2.3614, + "step": 36425 + }, + { + "epoch": 0.1, + "learning_rate": 9.767571264745192e-05, + "loss": 2.1857, + "step": 36430 + }, + { + "epoch": 0.1, + "learning_rate": 9.767507783774335e-05, + "loss": 2.2797, + "step": 36435 + }, + { + "epoch": 0.1, + "learning_rate": 9.767444294342036e-05, + "loss": 2.2911, + "step": 36440 + }, + { + "epoch": 0.1, + "learning_rate": 9.767380796448411e-05, + "loss": 2.4117, + "step": 36445 + }, + { + "epoch": 0.1, + "learning_rate": 9.767317290093574e-05, + "loss": 2.5128, + "step": 36450 + }, + { + "epoch": 0.1, + "learning_rate": 9.767253775277633e-05, + "loss": 2.5078, + "step": 36455 + }, + { + "epoch": 0.1, + "learning_rate": 9.767190252000706e-05, + "loss": 2.4692, + "step": 36460 + }, + { + "epoch": 0.1, + "learning_rate": 9.767126720262905e-05, + "loss": 2.3519, + "step": 36465 + }, + { + "epoch": 0.1, + "learning_rate": 9.767063180064337e-05, + "loss": 2.4714, + "step": 36470 + }, + { + "epoch": 0.1, + "learning_rate": 9.766999631405123e-05, + "loss": 2.3727, + "step": 36475 + }, + { + "epoch": 0.1, + "learning_rate": 9.766936074285368e-05, + "loss": 2.3745, + "step": 36480 + }, + { + "epoch": 0.1, + "learning_rate": 9.766872508705191e-05, + "loss": 2.3071, + "step": 36485 + }, + { + "epoch": 0.1, + "learning_rate": 9.766808934664702e-05, + "loss": 2.4579, + "step": 36490 + }, + { + "epoch": 0.1, + "learning_rate": 9.766745352164015e-05, + "loss": 2.3591, + "step": 36495 + }, + { + "epoch": 0.1, + "learning_rate": 9.766681761203242e-05, + "loss": 2.3571, + "step": 36500 + }, + { + "epoch": 0.1, + "learning_rate": 9.766618161782494e-05, + "loss": 2.3085, + "step": 36505 + }, + { + "epoch": 0.1, + "learning_rate": 9.766554553901887e-05, + "loss": 2.3055, + "step": 36510 + }, + { + "epoch": 0.1, + "learning_rate": 9.766490937561533e-05, + "loss": 2.343, + "step": 36515 + }, + { + "epoch": 0.1, + "learning_rate": 9.766427312761545e-05, + "loss": 2.3466, + "step": 36520 + }, + { + "epoch": 0.1, + "learning_rate": 9.766363679502035e-05, + "loss": 2.3193, + "step": 36525 + }, + { + "epoch": 0.1, + "learning_rate": 9.766300037783117e-05, + "loss": 2.2577, + "step": 36530 + }, + { + "epoch": 0.1, + "learning_rate": 9.766236387604902e-05, + "loss": 2.2301, + "step": 36535 + }, + { + "epoch": 0.1, + "learning_rate": 9.766172728967504e-05, + "loss": 2.4296, + "step": 36540 + }, + { + "epoch": 0.1, + "learning_rate": 9.766109061871038e-05, + "loss": 2.3908, + "step": 36545 + }, + { + "epoch": 0.1, + "learning_rate": 9.766045386315615e-05, + "loss": 2.2778, + "step": 36550 + }, + { + "epoch": 0.1, + "learning_rate": 9.765981702301347e-05, + "loss": 2.3586, + "step": 36555 + }, + { + "epoch": 0.1, + "learning_rate": 9.76591800982835e-05, + "loss": 2.316, + "step": 36560 + }, + { + "epoch": 0.1, + "learning_rate": 9.765854308896736e-05, + "loss": 2.3436, + "step": 36565 + }, + { + "epoch": 0.1, + "learning_rate": 9.765790599506616e-05, + "loss": 2.5454, + "step": 36570 + }, + { + "epoch": 0.1, + "learning_rate": 9.765726881658106e-05, + "loss": 2.3865, + "step": 36575 + }, + { + "epoch": 0.1, + "learning_rate": 9.765663155351317e-05, + "loss": 2.3595, + "step": 36580 + }, + { + "epoch": 0.1, + "learning_rate": 9.765599420586362e-05, + "loss": 2.3172, + "step": 36585 + }, + { + "epoch": 0.1, + "learning_rate": 9.765535677363356e-05, + "loss": 2.3551, + "step": 36590 + }, + { + "epoch": 0.1, + "learning_rate": 9.76547192568241e-05, + "loss": 2.4674, + "step": 36595 + }, + { + "epoch": 0.1, + "learning_rate": 9.76540816554364e-05, + "loss": 2.3729, + "step": 36600 + }, + { + "epoch": 0.1, + "learning_rate": 9.765344396947155e-05, + "loss": 2.3697, + "step": 36605 + }, + { + "epoch": 0.1, + "learning_rate": 9.765280619893071e-05, + "loss": 2.3415, + "step": 36610 + }, + { + "epoch": 0.1, + "learning_rate": 9.765216834381501e-05, + "loss": 2.4132, + "step": 36615 + }, + { + "epoch": 0.1, + "learning_rate": 9.765153040412559e-05, + "loss": 2.4939, + "step": 36620 + }, + { + "epoch": 0.1, + "learning_rate": 9.765089237986356e-05, + "loss": 2.4114, + "step": 36625 + }, + { + "epoch": 0.1, + "learning_rate": 9.765025427103006e-05, + "loss": 2.317, + "step": 36630 + }, + { + "epoch": 0.1, + "learning_rate": 9.764961607762624e-05, + "loss": 2.1173, + "step": 36635 + }, + { + "epoch": 0.1, + "learning_rate": 9.76489777996532e-05, + "loss": 2.4754, + "step": 36640 + }, + { + "epoch": 0.1, + "learning_rate": 9.76483394371121e-05, + "loss": 2.3623, + "step": 36645 + }, + { + "epoch": 0.1, + "learning_rate": 9.764770099000406e-05, + "loss": 2.2872, + "step": 36650 + }, + { + "epoch": 0.1, + "learning_rate": 9.764706245833024e-05, + "loss": 2.3719, + "step": 36655 + }, + { + "epoch": 0.1, + "learning_rate": 9.764642384209173e-05, + "loss": 2.4494, + "step": 36660 + }, + { + "epoch": 0.1, + "learning_rate": 9.764578514128969e-05, + "loss": 2.4511, + "step": 36665 + }, + { + "epoch": 0.1, + "learning_rate": 9.764514635592525e-05, + "loss": 2.3972, + "step": 36670 + }, + { + "epoch": 0.1, + "learning_rate": 9.764450748599954e-05, + "loss": 2.2725, + "step": 36675 + }, + { + "epoch": 0.1, + "learning_rate": 9.764386853151369e-05, + "loss": 2.4026, + "step": 36680 + }, + { + "epoch": 0.1, + "learning_rate": 9.764322949246885e-05, + "loss": 2.2753, + "step": 36685 + }, + { + "epoch": 0.1, + "learning_rate": 9.764259036886614e-05, + "loss": 2.3791, + "step": 36690 + }, + { + "epoch": 0.1, + "learning_rate": 9.764195116070668e-05, + "loss": 2.351, + "step": 36695 + }, + { + "epoch": 0.1, + "learning_rate": 9.764131186799166e-05, + "loss": 2.4374, + "step": 36700 + }, + { + "epoch": 0.1, + "learning_rate": 9.764067249072215e-05, + "loss": 2.3897, + "step": 36705 + }, + { + "epoch": 0.1, + "learning_rate": 9.764003302889932e-05, + "loss": 2.4072, + "step": 36710 + }, + { + "epoch": 0.1, + "learning_rate": 9.763939348252429e-05, + "loss": 2.3224, + "step": 36715 + }, + { + "epoch": 0.1, + "learning_rate": 9.763875385159821e-05, + "loss": 2.3343, + "step": 36720 + }, + { + "epoch": 0.1, + "learning_rate": 9.76381141361222e-05, + "loss": 2.353, + "step": 36725 + }, + { + "epoch": 0.1, + "learning_rate": 9.76374743360974e-05, + "loss": 2.337, + "step": 36730 + }, + { + "epoch": 0.1, + "learning_rate": 9.763683445152497e-05, + "loss": 2.4543, + "step": 36735 + }, + { + "epoch": 0.1, + "learning_rate": 9.763619448240601e-05, + "loss": 2.3849, + "step": 36740 + }, + { + "epoch": 0.1, + "learning_rate": 9.763555442874166e-05, + "loss": 2.4835, + "step": 36745 + }, + { + "epoch": 0.1, + "learning_rate": 9.763491429053309e-05, + "loss": 2.5529, + "step": 36750 + }, + { + "epoch": 0.1, + "learning_rate": 9.763427406778139e-05, + "loss": 2.3212, + "step": 36755 + }, + { + "epoch": 0.1, + "learning_rate": 9.763363376048773e-05, + "loss": 2.4145, + "step": 36760 + }, + { + "epoch": 0.1, + "learning_rate": 9.763299336865322e-05, + "loss": 2.357, + "step": 36765 + }, + { + "epoch": 0.1, + "learning_rate": 9.763235289227903e-05, + "loss": 2.3748, + "step": 36770 + }, + { + "epoch": 0.1, + "learning_rate": 9.763171233136626e-05, + "loss": 2.3804, + "step": 36775 + }, + { + "epoch": 0.1, + "learning_rate": 9.763107168591607e-05, + "loss": 2.2639, + "step": 36780 + }, + { + "epoch": 0.1, + "learning_rate": 9.763043095592959e-05, + "loss": 2.3502, + "step": 36785 + }, + { + "epoch": 0.1, + "learning_rate": 9.762979014140797e-05, + "loss": 2.3973, + "step": 36790 + }, + { + "epoch": 0.1, + "learning_rate": 9.762914924235233e-05, + "loss": 2.5186, + "step": 36795 + }, + { + "epoch": 0.1, + "learning_rate": 9.762850825876382e-05, + "loss": 2.3391, + "step": 36800 + }, + { + "epoch": 0.1, + "learning_rate": 9.762786719064356e-05, + "loss": 2.4074, + "step": 36805 + }, + { + "epoch": 0.1, + "learning_rate": 9.76272260379927e-05, + "loss": 2.4035, + "step": 36810 + }, + { + "epoch": 0.1, + "learning_rate": 9.762658480081238e-05, + "loss": 2.2602, + "step": 36815 + }, + { + "epoch": 0.1, + "learning_rate": 9.762594347910374e-05, + "loss": 2.3823, + "step": 36820 + }, + { + "epoch": 0.1, + "learning_rate": 9.762530207286791e-05, + "loss": 2.377, + "step": 36825 + }, + { + "epoch": 0.1, + "learning_rate": 9.762466058210602e-05, + "loss": 2.2811, + "step": 36830 + }, + { + "epoch": 0.1, + "learning_rate": 9.762401900681923e-05, + "loss": 2.4305, + "step": 36835 + }, + { + "epoch": 0.1, + "learning_rate": 9.762337734700867e-05, + "loss": 2.3694, + "step": 36840 + }, + { + "epoch": 0.1, + "learning_rate": 9.762273560267547e-05, + "loss": 2.3259, + "step": 36845 + }, + { + "epoch": 0.1, + "learning_rate": 9.762209377382079e-05, + "loss": 2.3982, + "step": 36850 + }, + { + "epoch": 0.1, + "learning_rate": 9.762145186044575e-05, + "loss": 2.3232, + "step": 36855 + }, + { + "epoch": 0.1, + "learning_rate": 9.76208098625515e-05, + "loss": 2.282, + "step": 36860 + }, + { + "epoch": 0.1, + "learning_rate": 9.762016778013916e-05, + "loss": 2.3827, + "step": 36865 + }, + { + "epoch": 0.1, + "learning_rate": 9.76195256132099e-05, + "loss": 2.4196, + "step": 36870 + }, + { + "epoch": 0.1, + "learning_rate": 9.761888336176484e-05, + "loss": 2.2927, + "step": 36875 + }, + { + "epoch": 0.1, + "learning_rate": 9.761824102580511e-05, + "loss": 2.382, + "step": 36880 + }, + { + "epoch": 0.1, + "learning_rate": 9.761759860533189e-05, + "loss": 2.3762, + "step": 36885 + }, + { + "epoch": 0.1, + "learning_rate": 9.761695610034627e-05, + "loss": 2.4991, + "step": 36890 + }, + { + "epoch": 0.1, + "learning_rate": 9.761631351084941e-05, + "loss": 2.5029, + "step": 36895 + }, + { + "epoch": 0.1, + "learning_rate": 9.761567083684248e-05, + "loss": 2.379, + "step": 36900 + }, + { + "epoch": 0.1, + "learning_rate": 9.761502807832658e-05, + "loss": 2.288, + "step": 36905 + }, + { + "epoch": 0.1, + "learning_rate": 9.761438523530286e-05, + "loss": 2.3899, + "step": 36910 + }, + { + "epoch": 0.1, + "learning_rate": 9.761374230777248e-05, + "loss": 2.3667, + "step": 36915 + }, + { + "epoch": 0.1, + "learning_rate": 9.761309929573657e-05, + "loss": 2.4304, + "step": 36920 + }, + { + "epoch": 0.1, + "learning_rate": 9.761245619919626e-05, + "loss": 2.281, + "step": 36925 + }, + { + "epoch": 0.1, + "learning_rate": 9.76118130181527e-05, + "loss": 2.355, + "step": 36930 + }, + { + "epoch": 0.1, + "learning_rate": 9.761116975260703e-05, + "loss": 2.3842, + "step": 36935 + }, + { + "epoch": 0.1, + "learning_rate": 9.761052640256041e-05, + "loss": 2.378, + "step": 36940 + }, + { + "epoch": 0.1, + "learning_rate": 9.760988296801394e-05, + "loss": 2.3277, + "step": 36945 + }, + { + "epoch": 0.1, + "learning_rate": 9.76092394489688e-05, + "loss": 2.4369, + "step": 36950 + }, + { + "epoch": 0.1, + "learning_rate": 9.760859584542613e-05, + "loss": 2.4536, + "step": 36955 + }, + { + "epoch": 0.1, + "learning_rate": 9.760795215738705e-05, + "loss": 2.4553, + "step": 36960 + }, + { + "epoch": 0.1, + "learning_rate": 9.760730838485273e-05, + "loss": 2.3888, + "step": 36965 + }, + { + "epoch": 0.1, + "learning_rate": 9.760666452782427e-05, + "loss": 2.4045, + "step": 36970 + }, + { + "epoch": 0.1, + "learning_rate": 9.760602058630287e-05, + "loss": 2.4194, + "step": 36975 + }, + { + "epoch": 0.1, + "learning_rate": 9.760537656028963e-05, + "loss": 2.3281, + "step": 36980 + }, + { + "epoch": 0.1, + "learning_rate": 9.76047324497857e-05, + "loss": 2.377, + "step": 36985 + }, + { + "epoch": 0.1, + "learning_rate": 9.760408825479224e-05, + "loss": 2.3103, + "step": 36990 + }, + { + "epoch": 0.1, + "learning_rate": 9.760344397531037e-05, + "loss": 2.4442, + "step": 36995 + }, + { + "epoch": 0.1, + "learning_rate": 9.760279961134124e-05, + "loss": 2.4474, + "step": 37000 + }, + { + "epoch": 0.1, + "learning_rate": 9.760215516288602e-05, + "loss": 2.3978, + "step": 37005 + }, + { + "epoch": 0.1, + "learning_rate": 9.760151062994583e-05, + "loss": 2.2877, + "step": 37010 + }, + { + "epoch": 0.1, + "learning_rate": 9.760086601252181e-05, + "loss": 2.3424, + "step": 37015 + }, + { + "epoch": 0.1, + "learning_rate": 9.76002213106151e-05, + "loss": 2.3691, + "step": 37020 + }, + { + "epoch": 0.1, + "learning_rate": 9.759957652422688e-05, + "loss": 2.3978, + "step": 37025 + }, + { + "epoch": 0.1, + "learning_rate": 9.759893165335824e-05, + "loss": 2.3029, + "step": 37030 + }, + { + "epoch": 0.1, + "learning_rate": 9.75982866980104e-05, + "loss": 2.4603, + "step": 37035 + }, + { + "epoch": 0.1, + "learning_rate": 9.759764165818441e-05, + "loss": 2.3209, + "step": 37040 + }, + { + "epoch": 0.1, + "learning_rate": 9.759699653388149e-05, + "loss": 2.4333, + "step": 37045 + }, + { + "epoch": 0.1, + "learning_rate": 9.759635132510275e-05, + "loss": 2.4299, + "step": 37050 + }, + { + "epoch": 0.1, + "learning_rate": 9.759570603184935e-05, + "loss": 2.3797, + "step": 37055 + }, + { + "epoch": 0.1, + "learning_rate": 9.759506065412243e-05, + "loss": 2.2898, + "step": 37060 + }, + { + "epoch": 0.1, + "learning_rate": 9.759441519192312e-05, + "loss": 2.3864, + "step": 37065 + }, + { + "epoch": 0.1, + "learning_rate": 9.75937696452526e-05, + "loss": 2.3584, + "step": 37070 + }, + { + "epoch": 0.1, + "learning_rate": 9.759312401411198e-05, + "loss": 2.4032, + "step": 37075 + }, + { + "epoch": 0.1, + "learning_rate": 9.759247829850242e-05, + "loss": 2.3226, + "step": 37080 + }, + { + "epoch": 0.1, + "learning_rate": 9.759183249842509e-05, + "loss": 2.417, + "step": 37085 + }, + { + "epoch": 0.1, + "learning_rate": 9.75911866138811e-05, + "loss": 2.323, + "step": 37090 + }, + { + "epoch": 0.1, + "learning_rate": 9.759054064487161e-05, + "loss": 2.2975, + "step": 37095 + }, + { + "epoch": 0.1, + "learning_rate": 9.758989459139776e-05, + "loss": 2.4981, + "step": 37100 + }, + { + "epoch": 0.1, + "learning_rate": 9.758924845346072e-05, + "loss": 2.4531, + "step": 37105 + }, + { + "epoch": 0.1, + "learning_rate": 9.75886022310616e-05, + "loss": 2.2945, + "step": 37110 + }, + { + "epoch": 0.1, + "learning_rate": 9.758795592420158e-05, + "loss": 2.4491, + "step": 37115 + }, + { + "epoch": 0.1, + "learning_rate": 9.758730953288179e-05, + "loss": 2.3367, + "step": 37120 + }, + { + "epoch": 0.1, + "learning_rate": 9.758666305710337e-05, + "loss": 2.447, + "step": 37125 + }, + { + "epoch": 0.1, + "learning_rate": 9.75860164968675e-05, + "loss": 2.5252, + "step": 37130 + }, + { + "epoch": 0.1, + "learning_rate": 9.75853698521753e-05, + "loss": 2.3075, + "step": 37135 + }, + { + "epoch": 0.1, + "learning_rate": 9.758472312302791e-05, + "loss": 2.5197, + "step": 37140 + }, + { + "epoch": 0.1, + "learning_rate": 9.75840763094265e-05, + "loss": 2.426, + "step": 37145 + }, + { + "epoch": 0.1, + "learning_rate": 9.758342941137221e-05, + "loss": 2.4378, + "step": 37150 + }, + { + "epoch": 0.1, + "learning_rate": 9.75827824288662e-05, + "loss": 2.4295, + "step": 37155 + }, + { + "epoch": 0.1, + "learning_rate": 9.75821353619096e-05, + "loss": 2.3844, + "step": 37160 + }, + { + "epoch": 0.1, + "learning_rate": 9.758148821050356e-05, + "loss": 2.3126, + "step": 37165 + }, + { + "epoch": 0.1, + "learning_rate": 9.758084097464924e-05, + "loss": 2.3186, + "step": 37170 + }, + { + "epoch": 0.1, + "learning_rate": 9.758019365434776e-05, + "loss": 2.2461, + "step": 37175 + }, + { + "epoch": 0.1, + "learning_rate": 9.757954624960032e-05, + "loss": 2.3481, + "step": 37180 + }, + { + "epoch": 0.1, + "learning_rate": 9.757889876040802e-05, + "loss": 2.4545, + "step": 37185 + }, + { + "epoch": 0.1, + "learning_rate": 9.757825118677205e-05, + "loss": 2.2602, + "step": 37190 + }, + { + "epoch": 0.1, + "learning_rate": 9.757760352869352e-05, + "loss": 2.4054, + "step": 37195 + }, + { + "epoch": 0.1, + "learning_rate": 9.757695578617358e-05, + "loss": 2.4326, + "step": 37200 + }, + { + "epoch": 0.1, + "learning_rate": 9.757630795921343e-05, + "loss": 2.4223, + "step": 37205 + }, + { + "epoch": 0.1, + "learning_rate": 9.757566004781418e-05, + "loss": 2.4133, + "step": 37210 + }, + { + "epoch": 0.1, + "learning_rate": 9.757501205197698e-05, + "loss": 2.2774, + "step": 37215 + }, + { + "epoch": 0.1, + "learning_rate": 9.757436397170298e-05, + "loss": 2.3529, + "step": 37220 + }, + { + "epoch": 0.1, + "learning_rate": 9.757371580699335e-05, + "loss": 2.3108, + "step": 37225 + }, + { + "epoch": 0.1, + "learning_rate": 9.757306755784923e-05, + "loss": 2.4017, + "step": 37230 + }, + { + "epoch": 0.1, + "learning_rate": 9.757241922427176e-05, + "loss": 2.3626, + "step": 37235 + }, + { + "epoch": 0.1, + "learning_rate": 9.757177080626211e-05, + "loss": 2.4159, + "step": 37240 + }, + { + "epoch": 0.1, + "learning_rate": 9.75711223038214e-05, + "loss": 2.4441, + "step": 37245 + }, + { + "epoch": 0.1, + "learning_rate": 9.757047371695081e-05, + "loss": 2.434, + "step": 37250 + }, + { + "epoch": 0.1, + "learning_rate": 9.756982504565148e-05, + "loss": 2.3351, + "step": 37255 + }, + { + "epoch": 0.1, + "learning_rate": 9.756917628992458e-05, + "loss": 2.3425, + "step": 37260 + }, + { + "epoch": 0.1, + "learning_rate": 9.756852744977124e-05, + "loss": 2.3786, + "step": 37265 + }, + { + "epoch": 0.1, + "learning_rate": 9.75678785251926e-05, + "loss": 2.3955, + "step": 37270 + }, + { + "epoch": 0.1, + "learning_rate": 9.756722951618984e-05, + "loss": 2.4487, + "step": 37275 + }, + { + "epoch": 0.1, + "learning_rate": 9.756658042276409e-05, + "loss": 2.3597, + "step": 37280 + }, + { + "epoch": 0.1, + "learning_rate": 9.756593124491651e-05, + "loss": 2.3001, + "step": 37285 + }, + { + "epoch": 0.1, + "learning_rate": 9.756528198264827e-05, + "loss": 2.4058, + "step": 37290 + }, + { + "epoch": 0.1, + "learning_rate": 9.75646326359605e-05, + "loss": 2.4109, + "step": 37295 + }, + { + "epoch": 0.1, + "learning_rate": 9.756398320485435e-05, + "loss": 2.2686, + "step": 37300 + }, + { + "epoch": 0.1, + "learning_rate": 9.756333368933098e-05, + "loss": 2.3596, + "step": 37305 + }, + { + "epoch": 0.1, + "learning_rate": 9.756268408939155e-05, + "loss": 2.288, + "step": 37310 + }, + { + "epoch": 0.1, + "learning_rate": 9.756203440503722e-05, + "loss": 2.4913, + "step": 37315 + }, + { + "epoch": 0.1, + "learning_rate": 9.756138463626912e-05, + "loss": 2.3669, + "step": 37320 + }, + { + "epoch": 0.1, + "learning_rate": 9.75607347830884e-05, + "loss": 2.3089, + "step": 37325 + }, + { + "epoch": 0.1, + "learning_rate": 9.756008484549624e-05, + "loss": 2.3869, + "step": 37330 + }, + { + "epoch": 0.1, + "learning_rate": 9.755943482349377e-05, + "loss": 2.3172, + "step": 37335 + }, + { + "epoch": 0.1, + "learning_rate": 9.755878471708216e-05, + "loss": 2.402, + "step": 37340 + }, + { + "epoch": 0.1, + "learning_rate": 9.755813452626257e-05, + "loss": 2.3135, + "step": 37345 + }, + { + "epoch": 0.1, + "learning_rate": 9.755748425103612e-05, + "loss": 2.3724, + "step": 37350 + }, + { + "epoch": 0.1, + "learning_rate": 9.7556833891404e-05, + "loss": 2.2577, + "step": 37355 + }, + { + "epoch": 0.1, + "learning_rate": 9.755618344736735e-05, + "loss": 2.4102, + "step": 37360 + }, + { + "epoch": 0.1, + "learning_rate": 9.755553291892731e-05, + "loss": 2.3556, + "step": 37365 + }, + { + "epoch": 0.1, + "learning_rate": 9.755488230608507e-05, + "loss": 2.4168, + "step": 37370 + }, + { + "epoch": 0.1, + "learning_rate": 9.755423160884175e-05, + "loss": 2.3401, + "step": 37375 + }, + { + "epoch": 0.1, + "learning_rate": 9.755358082719852e-05, + "loss": 2.3475, + "step": 37380 + }, + { + "epoch": 0.1, + "learning_rate": 9.755292996115652e-05, + "loss": 2.3542, + "step": 37385 + }, + { + "epoch": 0.1, + "learning_rate": 9.755227901071695e-05, + "loss": 2.3031, + "step": 37390 + }, + { + "epoch": 0.1, + "learning_rate": 9.75516279758809e-05, + "loss": 2.449, + "step": 37395 + }, + { + "epoch": 0.1, + "learning_rate": 9.755097685664958e-05, + "loss": 2.4117, + "step": 37400 + }, + { + "epoch": 0.1, + "learning_rate": 9.755032565302413e-05, + "loss": 2.4061, + "step": 37405 + }, + { + "epoch": 0.1, + "learning_rate": 9.754967436500569e-05, + "loss": 2.3356, + "step": 37410 + }, + { + "epoch": 0.1, + "learning_rate": 9.754902299259543e-05, + "loss": 2.3576, + "step": 37415 + }, + { + "epoch": 0.1, + "learning_rate": 9.75483715357945e-05, + "loss": 2.4381, + "step": 37420 + }, + { + "epoch": 0.1, + "learning_rate": 9.754771999460406e-05, + "loss": 2.494, + "step": 37425 + }, + { + "epoch": 0.1, + "learning_rate": 9.754706836902528e-05, + "loss": 2.2908, + "step": 37430 + }, + { + "epoch": 0.1, + "learning_rate": 9.754641665905928e-05, + "loss": 2.4761, + "step": 37435 + }, + { + "epoch": 0.1, + "learning_rate": 9.754576486470724e-05, + "loss": 2.3972, + "step": 37440 + }, + { + "epoch": 0.1, + "learning_rate": 9.754511298597033e-05, + "loss": 2.3971, + "step": 37445 + }, + { + "epoch": 0.1, + "learning_rate": 9.754446102284968e-05, + "loss": 2.4397, + "step": 37450 + }, + { + "epoch": 0.1, + "learning_rate": 9.754380897534648e-05, + "loss": 2.2916, + "step": 37455 + }, + { + "epoch": 0.1, + "learning_rate": 9.754315684346184e-05, + "loss": 2.4419, + "step": 37460 + }, + { + "epoch": 0.1, + "learning_rate": 9.754250462719696e-05, + "loss": 2.2446, + "step": 37465 + }, + { + "epoch": 0.1, + "learning_rate": 9.754185232655297e-05, + "loss": 2.3521, + "step": 37470 + }, + { + "epoch": 0.1, + "learning_rate": 9.754119994153103e-05, + "loss": 2.3788, + "step": 37475 + }, + { + "epoch": 0.1, + "learning_rate": 9.754054747213234e-05, + "loss": 2.275, + "step": 37480 + }, + { + "epoch": 0.1, + "learning_rate": 9.7539894918358e-05, + "loss": 2.425, + "step": 37485 + }, + { + "epoch": 0.1, + "learning_rate": 9.75392422802092e-05, + "loss": 2.4088, + "step": 37490 + }, + { + "epoch": 0.1, + "learning_rate": 9.753858955768711e-05, + "loss": 2.3454, + "step": 37495 + }, + { + "epoch": 0.1, + "learning_rate": 9.753793675079285e-05, + "loss": 2.2947, + "step": 37500 + }, + { + "epoch": 0.1, + "learning_rate": 9.753728385952761e-05, + "loss": 2.2641, + "step": 37505 + }, + { + "epoch": 0.1, + "learning_rate": 9.753663088389253e-05, + "loss": 2.3171, + "step": 37510 + }, + { + "epoch": 0.1, + "learning_rate": 9.753597782388877e-05, + "loss": 2.2305, + "step": 37515 + }, + { + "epoch": 0.1, + "learning_rate": 9.75353246795175e-05, + "loss": 2.4434, + "step": 37520 + }, + { + "epoch": 0.1, + "learning_rate": 9.753467145077988e-05, + "loss": 2.3985, + "step": 37525 + }, + { + "epoch": 0.1, + "learning_rate": 9.753401813767706e-05, + "loss": 2.3617, + "step": 37530 + }, + { + "epoch": 0.1, + "learning_rate": 9.753336474021019e-05, + "loss": 2.4467, + "step": 37535 + }, + { + "epoch": 0.1, + "learning_rate": 9.753271125838047e-05, + "loss": 2.3936, + "step": 37540 + }, + { + "epoch": 0.1, + "learning_rate": 9.753205769218903e-05, + "loss": 2.3501, + "step": 37545 + }, + { + "epoch": 0.1, + "learning_rate": 9.753140404163702e-05, + "loss": 2.4798, + "step": 37550 + }, + { + "epoch": 0.1, + "learning_rate": 9.753075030672563e-05, + "loss": 2.2524, + "step": 37555 + }, + { + "epoch": 0.1, + "learning_rate": 9.753009648745597e-05, + "loss": 2.2863, + "step": 37560 + }, + { + "epoch": 0.1, + "learning_rate": 9.752944258382927e-05, + "loss": 2.3094, + "step": 37565 + }, + { + "epoch": 0.1, + "learning_rate": 9.752878859584664e-05, + "loss": 2.384, + "step": 37570 + }, + { + "epoch": 0.1, + "learning_rate": 9.752813452350925e-05, + "loss": 2.4674, + "step": 37575 + }, + { + "epoch": 0.1, + "learning_rate": 9.752748036681827e-05, + "loss": 2.3975, + "step": 37580 + }, + { + "epoch": 0.1, + "learning_rate": 9.752682612577486e-05, + "loss": 2.4288, + "step": 37585 + }, + { + "epoch": 0.1, + "learning_rate": 9.752617180038017e-05, + "loss": 2.3184, + "step": 37590 + }, + { + "epoch": 0.1, + "learning_rate": 9.752551739063538e-05, + "loss": 2.2842, + "step": 37595 + }, + { + "epoch": 0.1, + "learning_rate": 9.752486289654164e-05, + "loss": 2.2755, + "step": 37600 + }, + { + "epoch": 0.1, + "learning_rate": 9.752420831810011e-05, + "loss": 2.3662, + "step": 37605 + }, + { + "epoch": 0.1, + "learning_rate": 9.752355365531195e-05, + "loss": 2.2742, + "step": 37610 + }, + { + "epoch": 0.1, + "learning_rate": 9.752289890817833e-05, + "loss": 2.3926, + "step": 37615 + }, + { + "epoch": 0.1, + "learning_rate": 9.752224407670041e-05, + "loss": 2.3105, + "step": 37620 + }, + { + "epoch": 0.1, + "learning_rate": 9.752158916087935e-05, + "loss": 2.3947, + "step": 37625 + }, + { + "epoch": 0.1, + "learning_rate": 9.75209341607163e-05, + "loss": 2.2808, + "step": 37630 + }, + { + "epoch": 0.1, + "learning_rate": 9.752027907621244e-05, + "loss": 2.3886, + "step": 37635 + }, + { + "epoch": 0.1, + "learning_rate": 9.751962390736895e-05, + "loss": 2.2719, + "step": 37640 + }, + { + "epoch": 0.1, + "learning_rate": 9.751896865418694e-05, + "loss": 2.3557, + "step": 37645 + }, + { + "epoch": 0.1, + "learning_rate": 9.751831331666762e-05, + "loss": 2.3008, + "step": 37650 + }, + { + "epoch": 0.1, + "learning_rate": 9.751765789481214e-05, + "loss": 2.3455, + "step": 37655 + }, + { + "epoch": 0.1, + "learning_rate": 9.751700238862164e-05, + "loss": 2.4013, + "step": 37660 + }, + { + "epoch": 0.1, + "learning_rate": 9.751634679809734e-05, + "loss": 2.3398, + "step": 37665 + }, + { + "epoch": 0.1, + "learning_rate": 9.751569112324033e-05, + "loss": 2.4788, + "step": 37670 + }, + { + "epoch": 0.1, + "learning_rate": 9.751503536405185e-05, + "loss": 2.3809, + "step": 37675 + }, + { + "epoch": 0.1, + "learning_rate": 9.751437952053299e-05, + "loss": 2.3441, + "step": 37680 + }, + { + "epoch": 0.1, + "learning_rate": 9.751372359268496e-05, + "loss": 2.2093, + "step": 37685 + }, + { + "epoch": 0.1, + "learning_rate": 9.751306758050891e-05, + "loss": 2.4303, + "step": 37690 + }, + { + "epoch": 0.1, + "learning_rate": 9.7512411484006e-05, + "loss": 2.3663, + "step": 37695 + }, + { + "epoch": 0.1, + "learning_rate": 9.751175530317742e-05, + "loss": 2.4401, + "step": 37700 + }, + { + "epoch": 0.1, + "learning_rate": 9.75110990380243e-05, + "loss": 2.3514, + "step": 37705 + }, + { + "epoch": 0.1, + "learning_rate": 9.751044268854783e-05, + "loss": 2.4667, + "step": 37710 + }, + { + "epoch": 0.1, + "learning_rate": 9.750978625474916e-05, + "loss": 2.3638, + "step": 37715 + }, + { + "epoch": 0.1, + "learning_rate": 9.750912973662947e-05, + "loss": 2.3816, + "step": 37720 + }, + { + "epoch": 0.1, + "learning_rate": 9.750847313418991e-05, + "loss": 2.3118, + "step": 37725 + }, + { + "epoch": 0.1, + "learning_rate": 9.750781644743165e-05, + "loss": 2.3838, + "step": 37730 + }, + { + "epoch": 0.1, + "learning_rate": 9.750715967635585e-05, + "loss": 2.3014, + "step": 37735 + }, + { + "epoch": 0.1, + "learning_rate": 9.75065028209637e-05, + "loss": 2.4844, + "step": 37740 + }, + { + "epoch": 0.1, + "learning_rate": 9.750584588125632e-05, + "loss": 2.5473, + "step": 37745 + }, + { + "epoch": 0.1, + "learning_rate": 9.750518885723494e-05, + "loss": 2.4236, + "step": 37750 + }, + { + "epoch": 0.1, + "learning_rate": 9.750453174890066e-05, + "loss": 2.3943, + "step": 37755 + }, + { + "epoch": 0.1, + "learning_rate": 9.750387455625469e-05, + "loss": 2.4728, + "step": 37760 + }, + { + "epoch": 0.1, + "learning_rate": 9.750321727929818e-05, + "loss": 2.3247, + "step": 37765 + }, + { + "epoch": 0.1, + "learning_rate": 9.750255991803231e-05, + "loss": 2.3036, + "step": 37770 + }, + { + "epoch": 0.1, + "learning_rate": 9.750190247245823e-05, + "loss": 2.41, + "step": 37775 + }, + { + "epoch": 0.1, + "learning_rate": 9.750124494257711e-05, + "loss": 2.3726, + "step": 37780 + }, + { + "epoch": 0.1, + "learning_rate": 9.750058732839012e-05, + "loss": 2.3015, + "step": 37785 + }, + { + "epoch": 0.1, + "learning_rate": 9.749992962989842e-05, + "loss": 2.3323, + "step": 37790 + }, + { + "epoch": 0.1, + "learning_rate": 9.749927184710319e-05, + "loss": 2.3527, + "step": 37795 + }, + { + "epoch": 0.1, + "learning_rate": 9.74986139800056e-05, + "loss": 2.2805, + "step": 37800 + }, + { + "epoch": 0.1, + "learning_rate": 9.74979560286068e-05, + "loss": 2.338, + "step": 37805 + }, + { + "epoch": 0.1, + "learning_rate": 9.749729799290797e-05, + "loss": 2.4714, + "step": 37810 + }, + { + "epoch": 0.1, + "learning_rate": 9.749663987291028e-05, + "loss": 2.515, + "step": 37815 + }, + { + "epoch": 0.1, + "learning_rate": 9.749598166861489e-05, + "loss": 2.4582, + "step": 37820 + }, + { + "epoch": 0.1, + "learning_rate": 9.749532338002297e-05, + "loss": 2.4887, + "step": 37825 + }, + { + "epoch": 0.1, + "learning_rate": 9.749466500713569e-05, + "loss": 2.4111, + "step": 37830 + }, + { + "epoch": 0.1, + "learning_rate": 9.749400654995422e-05, + "loss": 2.3664, + "step": 37835 + }, + { + "epoch": 0.1, + "learning_rate": 9.749334800847971e-05, + "loss": 2.3859, + "step": 37840 + }, + { + "epoch": 0.1, + "learning_rate": 9.749268938271337e-05, + "loss": 2.3879, + "step": 37845 + }, + { + "epoch": 0.1, + "learning_rate": 9.749203067265633e-05, + "loss": 2.3262, + "step": 37850 + }, + { + "epoch": 0.1, + "learning_rate": 9.74913718783098e-05, + "loss": 2.4083, + "step": 37855 + }, + { + "epoch": 0.1, + "learning_rate": 9.749071299967489e-05, + "loss": 2.3509, + "step": 37860 + }, + { + "epoch": 0.1, + "learning_rate": 9.74900540367528e-05, + "loss": 2.3476, + "step": 37865 + }, + { + "epoch": 0.1, + "learning_rate": 9.748939498954472e-05, + "loss": 2.4143, + "step": 37870 + }, + { + "epoch": 0.1, + "learning_rate": 9.74887358580518e-05, + "loss": 2.2402, + "step": 37875 + }, + { + "epoch": 0.1, + "learning_rate": 9.748807664227521e-05, + "loss": 2.4421, + "step": 37880 + }, + { + "epoch": 0.1, + "learning_rate": 9.748741734221611e-05, + "loss": 2.2666, + "step": 37885 + }, + { + "epoch": 0.1, + "learning_rate": 9.74867579578757e-05, + "loss": 2.3361, + "step": 37890 + }, + { + "epoch": 0.1, + "learning_rate": 9.748609848925512e-05, + "loss": 2.4552, + "step": 37895 + }, + { + "epoch": 0.1, + "learning_rate": 9.748543893635556e-05, + "loss": 2.3552, + "step": 37900 + }, + { + "epoch": 0.1, + "learning_rate": 9.748477929917816e-05, + "loss": 2.4517, + "step": 37905 + }, + { + "epoch": 0.1, + "learning_rate": 9.748411957772414e-05, + "loss": 2.4212, + "step": 37910 + }, + { + "epoch": 0.1, + "learning_rate": 9.748345977199463e-05, + "loss": 2.3793, + "step": 37915 + }, + { + "epoch": 0.1, + "learning_rate": 9.748279988199083e-05, + "loss": 2.481, + "step": 37920 + }, + { + "epoch": 0.1, + "learning_rate": 9.748213990771387e-05, + "loss": 2.385, + "step": 37925 + }, + { + "epoch": 0.1, + "learning_rate": 9.748147984916497e-05, + "loss": 2.3657, + "step": 37930 + }, + { + "epoch": 0.1, + "learning_rate": 9.748081970634529e-05, + "loss": 2.3285, + "step": 37935 + }, + { + "epoch": 0.1, + "learning_rate": 9.748015947925597e-05, + "loss": 2.3068, + "step": 37940 + }, + { + "epoch": 0.1, + "learning_rate": 9.74794991678982e-05, + "loss": 2.3396, + "step": 37945 + }, + { + "epoch": 0.1, + "learning_rate": 9.747883877227314e-05, + "loss": 2.3782, + "step": 37950 + }, + { + "epoch": 0.1, + "learning_rate": 9.747817829238202e-05, + "loss": 2.3167, + "step": 37955 + }, + { + "epoch": 0.1, + "learning_rate": 9.747751772822595e-05, + "loss": 2.3628, + "step": 37960 + }, + { + "epoch": 0.1, + "learning_rate": 9.747685707980612e-05, + "loss": 2.4236, + "step": 37965 + }, + { + "epoch": 0.1, + "learning_rate": 9.74761963471237e-05, + "loss": 2.2582, + "step": 37970 + }, + { + "epoch": 0.1, + "learning_rate": 9.747553553017988e-05, + "loss": 2.2783, + "step": 37975 + }, + { + "epoch": 0.1, + "learning_rate": 9.74748746289758e-05, + "loss": 2.2971, + "step": 37980 + }, + { + "epoch": 0.1, + "learning_rate": 9.747421364351265e-05, + "loss": 2.404, + "step": 37985 + }, + { + "epoch": 0.1, + "learning_rate": 9.747355257379162e-05, + "loss": 2.4429, + "step": 37990 + }, + { + "epoch": 0.1, + "learning_rate": 9.747289141981386e-05, + "loss": 2.3083, + "step": 37995 + }, + { + "epoch": 0.1, + "learning_rate": 9.747223018158056e-05, + "loss": 2.3763, + "step": 38000 + }, + { + "epoch": 0.1, + "learning_rate": 9.747156885909289e-05, + "loss": 2.4523, + "step": 38005 + }, + { + "epoch": 0.1, + "learning_rate": 9.747090745235202e-05, + "loss": 2.4768, + "step": 38010 + }, + { + "epoch": 0.1, + "learning_rate": 9.747024596135911e-05, + "loss": 2.4968, + "step": 38015 + }, + { + "epoch": 0.1, + "learning_rate": 9.746958438611536e-05, + "loss": 2.3373, + "step": 38020 + }, + { + "epoch": 0.1, + "learning_rate": 9.746892272662191e-05, + "loss": 2.3818, + "step": 38025 + }, + { + "epoch": 0.1, + "learning_rate": 9.746826098287998e-05, + "loss": 2.3436, + "step": 38030 + }, + { + "epoch": 0.1, + "learning_rate": 9.746759915489071e-05, + "loss": 2.351, + "step": 38035 + }, + { + "epoch": 0.1, + "learning_rate": 9.746693724265529e-05, + "loss": 2.2754, + "step": 38040 + }, + { + "epoch": 0.1, + "learning_rate": 9.746627524617488e-05, + "loss": 2.2824, + "step": 38045 + }, + { + "epoch": 0.1, + "learning_rate": 9.746561316545068e-05, + "loss": 2.2744, + "step": 38050 + }, + { + "epoch": 0.1, + "learning_rate": 9.746495100048383e-05, + "loss": 2.3362, + "step": 38055 + }, + { + "epoch": 0.1, + "learning_rate": 9.746428875127554e-05, + "loss": 2.4859, + "step": 38060 + }, + { + "epoch": 0.1, + "learning_rate": 9.746362641782696e-05, + "loss": 2.305, + "step": 38065 + }, + { + "epoch": 0.1, + "learning_rate": 9.746296400013929e-05, + "loss": 2.3372, + "step": 38070 + }, + { + "epoch": 0.1, + "learning_rate": 9.746230149821368e-05, + "loss": 2.3339, + "step": 38075 + }, + { + "epoch": 0.1, + "learning_rate": 9.746163891205132e-05, + "loss": 2.4276, + "step": 38080 + }, + { + "epoch": 0.1, + "learning_rate": 9.746097624165339e-05, + "loss": 2.3599, + "step": 38085 + }, + { + "epoch": 0.1, + "learning_rate": 9.746031348702104e-05, + "loss": 2.2868, + "step": 38090 + }, + { + "epoch": 0.1, + "learning_rate": 9.745965064815549e-05, + "loss": 2.3739, + "step": 38095 + }, + { + "epoch": 0.1, + "learning_rate": 9.745898772505789e-05, + "loss": 2.3279, + "step": 38100 + }, + { + "epoch": 0.1, + "learning_rate": 9.74583247177294e-05, + "loss": 2.417, + "step": 38105 + }, + { + "epoch": 0.1, + "learning_rate": 9.745766162617122e-05, + "loss": 2.4176, + "step": 38110 + }, + { + "epoch": 0.1, + "learning_rate": 9.745699845038453e-05, + "loss": 2.3063, + "step": 38115 + }, + { + "epoch": 0.1, + "learning_rate": 9.745633519037049e-05, + "loss": 2.3839, + "step": 38120 + }, + { + "epoch": 0.1, + "learning_rate": 9.74556718461303e-05, + "loss": 2.3429, + "step": 38125 + }, + { + "epoch": 0.1, + "learning_rate": 9.745500841766512e-05, + "loss": 2.4886, + "step": 38130 + }, + { + "epoch": 0.1, + "learning_rate": 9.745434490497613e-05, + "loss": 2.4032, + "step": 38135 + }, + { + "epoch": 0.1, + "learning_rate": 9.745368130806451e-05, + "loss": 2.3231, + "step": 38140 + }, + { + "epoch": 0.1, + "learning_rate": 9.745301762693143e-05, + "loss": 2.4154, + "step": 38145 + }, + { + "epoch": 0.1, + "learning_rate": 9.745235386157808e-05, + "loss": 2.456, + "step": 38150 + }, + { + "epoch": 0.1, + "learning_rate": 9.745169001200562e-05, + "loss": 2.3884, + "step": 38155 + }, + { + "epoch": 0.1, + "learning_rate": 9.745102607821526e-05, + "loss": 2.3956, + "step": 38160 + }, + { + "epoch": 0.1, + "learning_rate": 9.745036206020815e-05, + "loss": 2.2924, + "step": 38165 + }, + { + "epoch": 0.1, + "learning_rate": 9.744969795798547e-05, + "loss": 2.4663, + "step": 38170 + }, + { + "epoch": 0.1, + "learning_rate": 9.744903377154842e-05, + "loss": 2.3649, + "step": 38175 + }, + { + "epoch": 0.1, + "learning_rate": 9.744836950089815e-05, + "loss": 2.364, + "step": 38180 + }, + { + "epoch": 0.1, + "learning_rate": 9.744770514603587e-05, + "loss": 2.2779, + "step": 38185 + }, + { + "epoch": 0.1, + "learning_rate": 9.744704070696273e-05, + "loss": 2.3207, + "step": 38190 + }, + { + "epoch": 0.1, + "learning_rate": 9.744637618367994e-05, + "loss": 2.3931, + "step": 38195 + }, + { + "epoch": 0.1, + "learning_rate": 9.744571157618865e-05, + "loss": 2.3121, + "step": 38200 + }, + { + "epoch": 0.1, + "learning_rate": 9.744504688449005e-05, + "loss": 2.3913, + "step": 38205 + }, + { + "epoch": 0.1, + "learning_rate": 9.744438210858533e-05, + "loss": 2.4506, + "step": 38210 + }, + { + "epoch": 0.1, + "learning_rate": 9.744371724847565e-05, + "loss": 2.3435, + "step": 38215 + }, + { + "epoch": 0.1, + "learning_rate": 9.744305230416219e-05, + "loss": 2.3434, + "step": 38220 + }, + { + "epoch": 0.1, + "learning_rate": 9.744238727564617e-05, + "loss": 2.3342, + "step": 38225 + }, + { + "epoch": 0.1, + "learning_rate": 9.744172216292873e-05, + "loss": 2.4525, + "step": 38230 + }, + { + "epoch": 0.1, + "learning_rate": 9.744105696601105e-05, + "loss": 2.3847, + "step": 38235 + }, + { + "epoch": 0.1, + "learning_rate": 9.744039168489433e-05, + "loss": 2.3484, + "step": 38240 + }, + { + "epoch": 0.1, + "learning_rate": 9.743972631957974e-05, + "loss": 2.4101, + "step": 38245 + }, + { + "epoch": 0.1, + "learning_rate": 9.74390608700685e-05, + "loss": 2.3977, + "step": 38250 + }, + { + "epoch": 0.1, + "learning_rate": 9.74383953363617e-05, + "loss": 2.3511, + "step": 38255 + }, + { + "epoch": 0.1, + "learning_rate": 9.74377297184606e-05, + "loss": 2.3466, + "step": 38260 + }, + { + "epoch": 0.1, + "learning_rate": 9.743706401636637e-05, + "loss": 2.2358, + "step": 38265 + }, + { + "epoch": 0.1, + "learning_rate": 9.743639823008016e-05, + "loss": 2.3911, + "step": 38270 + }, + { + "epoch": 0.1, + "learning_rate": 9.743573235960317e-05, + "loss": 2.2525, + "step": 38275 + }, + { + "epoch": 0.1, + "learning_rate": 9.743506640493661e-05, + "loss": 2.3252, + "step": 38280 + }, + { + "epoch": 0.1, + "learning_rate": 9.74344003660816e-05, + "loss": 2.3426, + "step": 38285 + }, + { + "epoch": 0.1, + "learning_rate": 9.743373424303938e-05, + "loss": 2.2957, + "step": 38290 + }, + { + "epoch": 0.1, + "learning_rate": 9.743306803581109e-05, + "loss": 2.4036, + "step": 38295 + }, + { + "epoch": 0.1, + "learning_rate": 9.743240174439795e-05, + "loss": 2.4132, + "step": 38300 + }, + { + "epoch": 0.1, + "learning_rate": 9.743173536880111e-05, + "loss": 2.2772, + "step": 38305 + }, + { + "epoch": 0.1, + "learning_rate": 9.743106890902178e-05, + "loss": 2.4171, + "step": 38310 + }, + { + "epoch": 0.1, + "learning_rate": 9.743040236506112e-05, + "loss": 2.5278, + "step": 38315 + }, + { + "epoch": 0.1, + "learning_rate": 9.742973573692033e-05, + "loss": 2.3673, + "step": 38320 + }, + { + "epoch": 0.1, + "learning_rate": 9.742906902460057e-05, + "loss": 2.3383, + "step": 38325 + }, + { + "epoch": 0.1, + "learning_rate": 9.742840222810305e-05, + "loss": 2.4085, + "step": 38330 + }, + { + "epoch": 0.1, + "learning_rate": 9.742773534742895e-05, + "loss": 2.3128, + "step": 38335 + }, + { + "epoch": 0.1, + "learning_rate": 9.742706838257943e-05, + "loss": 2.2704, + "step": 38340 + }, + { + "epoch": 0.1, + "learning_rate": 9.742640133355569e-05, + "loss": 2.3937, + "step": 38345 + }, + { + "epoch": 0.1, + "learning_rate": 9.742573420035893e-05, + "loss": 2.41, + "step": 38350 + }, + { + "epoch": 0.1, + "learning_rate": 9.74250669829903e-05, + "loss": 2.3969, + "step": 38355 + }, + { + "epoch": 0.1, + "learning_rate": 9.7424399681451e-05, + "loss": 2.3018, + "step": 38360 + }, + { + "epoch": 0.1, + "learning_rate": 9.742373229574224e-05, + "loss": 2.5241, + "step": 38365 + }, + { + "epoch": 0.1, + "learning_rate": 9.742306482586517e-05, + "loss": 2.3358, + "step": 38370 + }, + { + "epoch": 0.1, + "learning_rate": 9.742239727182097e-05, + "loss": 2.2944, + "step": 38375 + }, + { + "epoch": 0.1, + "learning_rate": 9.742172963361084e-05, + "loss": 2.3724, + "step": 38380 + }, + { + "epoch": 0.1, + "learning_rate": 9.742106191123599e-05, + "loss": 2.1795, + "step": 38385 + }, + { + "epoch": 0.1, + "learning_rate": 9.742039410469757e-05, + "loss": 2.312, + "step": 38390 + }, + { + "epoch": 0.1, + "learning_rate": 9.741972621399676e-05, + "loss": 2.3995, + "step": 38395 + }, + { + "epoch": 0.1, + "learning_rate": 9.741905823913477e-05, + "loss": 2.4347, + "step": 38400 + }, + { + "epoch": 0.1, + "learning_rate": 9.741839018011278e-05, + "loss": 2.3, + "step": 38405 + }, + { + "epoch": 0.1, + "learning_rate": 9.741772203693197e-05, + "loss": 2.4274, + "step": 38410 + }, + { + "epoch": 0.1, + "learning_rate": 9.741705380959352e-05, + "loss": 2.4067, + "step": 38415 + }, + { + "epoch": 0.1, + "learning_rate": 9.741638549809864e-05, + "loss": 2.3441, + "step": 38420 + }, + { + "epoch": 0.1, + "learning_rate": 9.741571710244848e-05, + "loss": 2.3145, + "step": 38425 + }, + { + "epoch": 0.1, + "learning_rate": 9.741504862264426e-05, + "loss": 2.3518, + "step": 38430 + }, + { + "epoch": 0.1, + "learning_rate": 9.741438005868714e-05, + "loss": 2.4705, + "step": 38435 + }, + { + "epoch": 0.1, + "learning_rate": 9.741371141057834e-05, + "loss": 2.3769, + "step": 38440 + }, + { + "epoch": 0.1, + "learning_rate": 9.7413042678319e-05, + "loss": 2.2036, + "step": 38445 + }, + { + "epoch": 0.1, + "learning_rate": 9.741237386191035e-05, + "loss": 2.3056, + "step": 38450 + }, + { + "epoch": 0.1, + "learning_rate": 9.741170496135355e-05, + "loss": 2.4542, + "step": 38455 + }, + { + "epoch": 0.1, + "learning_rate": 9.741103597664981e-05, + "loss": 2.467, + "step": 38460 + }, + { + "epoch": 0.1, + "learning_rate": 9.74103669078003e-05, + "loss": 2.3964, + "step": 38465 + }, + { + "epoch": 0.1, + "learning_rate": 9.740969775480619e-05, + "loss": 2.3285, + "step": 38470 + }, + { + "epoch": 0.1, + "learning_rate": 9.740902851766871e-05, + "loss": 2.2899, + "step": 38475 + }, + { + "epoch": 0.1, + "learning_rate": 9.740835919638902e-05, + "loss": 2.269, + "step": 38480 + }, + { + "epoch": 0.1, + "learning_rate": 9.740768979096832e-05, + "loss": 2.2893, + "step": 38485 + }, + { + "epoch": 0.1, + "learning_rate": 9.74070203014078e-05, + "loss": 2.3586, + "step": 38490 + }, + { + "epoch": 0.1, + "learning_rate": 9.740635072770863e-05, + "loss": 2.4316, + "step": 38495 + }, + { + "epoch": 0.1, + "learning_rate": 9.740568106987202e-05, + "loss": 2.3753, + "step": 38500 + }, + { + "epoch": 0.1, + "learning_rate": 9.740501132789913e-05, + "loss": 2.365, + "step": 38505 + }, + { + "epoch": 0.1, + "learning_rate": 9.740434150179118e-05, + "loss": 2.3598, + "step": 38510 + }, + { + "epoch": 0.1, + "learning_rate": 9.740367159154934e-05, + "loss": 2.1705, + "step": 38515 + }, + { + "epoch": 0.1, + "learning_rate": 9.740300159717481e-05, + "loss": 2.4852, + "step": 38520 + }, + { + "epoch": 0.1, + "learning_rate": 9.740233151866876e-05, + "loss": 2.2309, + "step": 38525 + }, + { + "epoch": 0.1, + "learning_rate": 9.740166135603241e-05, + "loss": 2.2746, + "step": 38530 + }, + { + "epoch": 0.1, + "learning_rate": 9.740099110926693e-05, + "loss": 2.3422, + "step": 38535 + }, + { + "epoch": 0.1, + "learning_rate": 9.74003207783735e-05, + "loss": 2.4009, + "step": 38540 + }, + { + "epoch": 0.1, + "learning_rate": 9.739965036335333e-05, + "loss": 2.3671, + "step": 38545 + }, + { + "epoch": 0.1, + "learning_rate": 9.73989798642076e-05, + "loss": 2.4033, + "step": 38550 + }, + { + "epoch": 0.1, + "learning_rate": 9.739830928093751e-05, + "loss": 2.3523, + "step": 38555 + }, + { + "epoch": 0.1, + "learning_rate": 9.739763861354422e-05, + "loss": 2.2519, + "step": 38560 + }, + { + "epoch": 0.1, + "learning_rate": 9.739696786202897e-05, + "loss": 2.398, + "step": 38565 + }, + { + "epoch": 0.1, + "learning_rate": 9.73962970263929e-05, + "loss": 2.5205, + "step": 38570 + }, + { + "epoch": 0.1, + "learning_rate": 9.739562610663724e-05, + "loss": 2.3918, + "step": 38575 + }, + { + "epoch": 0.1, + "learning_rate": 9.739495510276315e-05, + "loss": 2.3573, + "step": 38580 + }, + { + "epoch": 0.1, + "learning_rate": 9.739428401477185e-05, + "loss": 2.354, + "step": 38585 + }, + { + "epoch": 0.1, + "learning_rate": 9.739361284266448e-05, + "loss": 2.381, + "step": 38590 + }, + { + "epoch": 0.1, + "learning_rate": 9.73929415864423e-05, + "loss": 2.406, + "step": 38595 + }, + { + "epoch": 0.1, + "learning_rate": 9.739227024610646e-05, + "loss": 2.3476, + "step": 38600 + }, + { + "epoch": 0.1, + "learning_rate": 9.739159882165816e-05, + "loss": 2.4274, + "step": 38605 + }, + { + "epoch": 0.1, + "learning_rate": 9.739092731309858e-05, + "loss": 2.3119, + "step": 38610 + }, + { + "epoch": 0.1, + "learning_rate": 9.739025572042895e-05, + "loss": 2.4298, + "step": 38615 + }, + { + "epoch": 0.1, + "learning_rate": 9.738958404365042e-05, + "loss": 2.3505, + "step": 38620 + }, + { + "epoch": 0.1, + "learning_rate": 9.73889122827642e-05, + "loss": 2.3904, + "step": 38625 + }, + { + "epoch": 0.1, + "learning_rate": 9.738824043777147e-05, + "loss": 2.3991, + "step": 38630 + }, + { + "epoch": 0.1, + "learning_rate": 9.738756850867343e-05, + "loss": 2.4359, + "step": 38635 + }, + { + "epoch": 0.1, + "learning_rate": 9.738689649547129e-05, + "loss": 2.3672, + "step": 38640 + }, + { + "epoch": 0.1, + "learning_rate": 9.738622439816622e-05, + "loss": 2.3004, + "step": 38645 + }, + { + "epoch": 0.1, + "learning_rate": 9.738555221675942e-05, + "loss": 2.3027, + "step": 38650 + }, + { + "epoch": 0.1, + "learning_rate": 9.738487995125208e-05, + "loss": 2.3879, + "step": 38655 + }, + { + "epoch": 0.1, + "learning_rate": 9.738420760164539e-05, + "loss": 2.3385, + "step": 38660 + }, + { + "epoch": 0.1, + "learning_rate": 9.738353516794056e-05, + "loss": 2.5025, + "step": 38665 + }, + { + "epoch": 0.1, + "learning_rate": 9.738286265013878e-05, + "loss": 2.4029, + "step": 38670 + }, + { + "epoch": 0.1, + "learning_rate": 9.738219004824123e-05, + "loss": 2.5345, + "step": 38675 + }, + { + "epoch": 0.1, + "learning_rate": 9.73815173622491e-05, + "loss": 2.3545, + "step": 38680 + }, + { + "epoch": 0.1, + "learning_rate": 9.738084459216361e-05, + "loss": 2.5006, + "step": 38685 + }, + { + "epoch": 0.1, + "learning_rate": 9.738017173798593e-05, + "loss": 2.3205, + "step": 38690 + }, + { + "epoch": 0.1, + "learning_rate": 9.737949879971726e-05, + "loss": 2.4087, + "step": 38695 + }, + { + "epoch": 0.1, + "learning_rate": 9.737882577735881e-05, + "loss": 2.3321, + "step": 38700 + }, + { + "epoch": 0.1, + "learning_rate": 9.737815267091175e-05, + "loss": 2.3436, + "step": 38705 + }, + { + "epoch": 0.1, + "learning_rate": 9.737747948037729e-05, + "loss": 2.3469, + "step": 38710 + }, + { + "epoch": 0.1, + "learning_rate": 9.737680620575661e-05, + "loss": 2.4362, + "step": 38715 + }, + { + "epoch": 0.1, + "learning_rate": 9.737613284705093e-05, + "loss": 2.3676, + "step": 38720 + }, + { + "epoch": 0.1, + "learning_rate": 9.737545940426143e-05, + "loss": 2.3542, + "step": 38725 + }, + { + "epoch": 0.1, + "learning_rate": 9.737478587738928e-05, + "loss": 2.3955, + "step": 38730 + }, + { + "epoch": 0.1, + "learning_rate": 9.737411226643574e-05, + "loss": 2.3717, + "step": 38735 + }, + { + "epoch": 0.1, + "learning_rate": 9.737343857140194e-05, + "loss": 2.1425, + "step": 38740 + }, + { + "epoch": 0.1, + "learning_rate": 9.737276479228911e-05, + "loss": 2.3939, + "step": 38745 + }, + { + "epoch": 0.1, + "learning_rate": 9.737209092909844e-05, + "loss": 2.3282, + "step": 38750 + }, + { + "epoch": 0.1, + "learning_rate": 9.737141698183112e-05, + "loss": 2.3548, + "step": 38755 + }, + { + "epoch": 0.1, + "learning_rate": 9.737074295048836e-05, + "loss": 2.3478, + "step": 38760 + }, + { + "epoch": 0.1, + "learning_rate": 9.737006883507133e-05, + "loss": 2.4962, + "step": 38765 + }, + { + "epoch": 0.1, + "learning_rate": 9.736939463558126e-05, + "loss": 2.4725, + "step": 38770 + }, + { + "epoch": 0.1, + "learning_rate": 9.736872035201932e-05, + "loss": 2.3507, + "step": 38775 + }, + { + "epoch": 0.1, + "learning_rate": 9.736804598438672e-05, + "loss": 2.4408, + "step": 38780 + }, + { + "epoch": 0.1, + "learning_rate": 9.736737153268465e-05, + "loss": 2.3422, + "step": 38785 + }, + { + "epoch": 0.1, + "learning_rate": 9.736669699691432e-05, + "loss": 2.4702, + "step": 38790 + }, + { + "epoch": 0.1, + "learning_rate": 9.73660223770769e-05, + "loss": 2.3714, + "step": 38795 + }, + { + "epoch": 0.1, + "learning_rate": 9.736534767317362e-05, + "loss": 2.3143, + "step": 38800 + }, + { + "epoch": 0.1, + "learning_rate": 9.736467288520566e-05, + "loss": 2.3959, + "step": 38805 + }, + { + "epoch": 0.1, + "learning_rate": 9.736399801317421e-05, + "loss": 2.3688, + "step": 38810 + }, + { + "epoch": 0.1, + "learning_rate": 9.736332305708049e-05, + "loss": 2.5148, + "step": 38815 + }, + { + "epoch": 0.1, + "learning_rate": 9.736264801692569e-05, + "loss": 2.3134, + "step": 38820 + }, + { + "epoch": 0.1, + "learning_rate": 9.736197289271098e-05, + "loss": 2.4286, + "step": 38825 + }, + { + "epoch": 0.1, + "learning_rate": 9.73612976844376e-05, + "loss": 2.3983, + "step": 38830 + }, + { + "epoch": 0.1, + "learning_rate": 9.736062239210673e-05, + "loss": 2.2572, + "step": 38835 + }, + { + "epoch": 0.1, + "learning_rate": 9.735994701571956e-05, + "loss": 2.2873, + "step": 38840 + }, + { + "epoch": 0.1, + "learning_rate": 9.735927155527732e-05, + "loss": 2.3606, + "step": 38845 + }, + { + "epoch": 0.1, + "learning_rate": 9.735859601078117e-05, + "loss": 2.4633, + "step": 38850 + }, + { + "epoch": 0.1, + "learning_rate": 9.735792038223231e-05, + "loss": 2.3195, + "step": 38855 + }, + { + "epoch": 0.1, + "learning_rate": 9.735724466963198e-05, + "loss": 2.3462, + "step": 38860 + }, + { + "epoch": 0.1, + "learning_rate": 9.735656887298135e-05, + "loss": 2.3172, + "step": 38865 + }, + { + "epoch": 0.1, + "learning_rate": 9.735589299228161e-05, + "loss": 2.252, + "step": 38870 + }, + { + "epoch": 0.1, + "learning_rate": 9.735521702753397e-05, + "loss": 2.3043, + "step": 38875 + }, + { + "epoch": 0.1, + "learning_rate": 9.735454097873965e-05, + "loss": 2.3493, + "step": 38880 + }, + { + "epoch": 0.1, + "learning_rate": 9.735386484589981e-05, + "loss": 2.364, + "step": 38885 + }, + { + "epoch": 0.1, + "learning_rate": 9.73531886290157e-05, + "loss": 2.2648, + "step": 38890 + }, + { + "epoch": 0.1, + "learning_rate": 9.735251232808847e-05, + "loss": 2.3241, + "step": 38895 + }, + { + "epoch": 0.1, + "learning_rate": 9.735183594311936e-05, + "loss": 2.4687, + "step": 38900 + }, + { + "epoch": 0.1, + "learning_rate": 9.735115947410953e-05, + "loss": 2.3583, + "step": 38905 + }, + { + "epoch": 0.1, + "learning_rate": 9.735048292106022e-05, + "loss": 2.4724, + "step": 38910 + }, + { + "epoch": 0.1, + "learning_rate": 9.734980628397262e-05, + "loss": 2.4048, + "step": 38915 + }, + { + "epoch": 0.1, + "learning_rate": 9.734912956284792e-05, + "loss": 2.2754, + "step": 38920 + }, + { + "epoch": 0.1, + "learning_rate": 9.734845275768732e-05, + "loss": 2.5024, + "step": 38925 + }, + { + "epoch": 0.1, + "learning_rate": 9.734777586849201e-05, + "loss": 2.3638, + "step": 38930 + }, + { + "epoch": 0.1, + "learning_rate": 9.734709889526325e-05, + "loss": 2.4373, + "step": 38935 + }, + { + "epoch": 0.1, + "learning_rate": 9.734642183800218e-05, + "loss": 2.3123, + "step": 38940 + }, + { + "epoch": 0.1, + "learning_rate": 9.734574469671003e-05, + "loss": 2.2966, + "step": 38945 + }, + { + "epoch": 0.1, + "learning_rate": 9.734506747138798e-05, + "loss": 2.3137, + "step": 38950 + }, + { + "epoch": 0.1, + "learning_rate": 9.734439016203725e-05, + "loss": 2.3519, + "step": 38955 + }, + { + "epoch": 0.1, + "learning_rate": 9.734371276865904e-05, + "loss": 2.3037, + "step": 38960 + }, + { + "epoch": 0.1, + "learning_rate": 9.734303529125454e-05, + "loss": 2.4504, + "step": 38965 + }, + { + "epoch": 0.1, + "learning_rate": 9.734235772982499e-05, + "loss": 2.5047, + "step": 38970 + }, + { + "epoch": 0.1, + "learning_rate": 9.734168008437155e-05, + "loss": 2.3478, + "step": 38975 + }, + { + "epoch": 0.1, + "learning_rate": 9.734100235489544e-05, + "loss": 2.2531, + "step": 38980 + }, + { + "epoch": 0.1, + "learning_rate": 9.734032454139785e-05, + "loss": 2.4206, + "step": 38985 + }, + { + "epoch": 0.1, + "learning_rate": 9.733964664388e-05, + "loss": 2.3578, + "step": 38990 + }, + { + "epoch": 0.1, + "learning_rate": 9.73389686623431e-05, + "loss": 2.3947, + "step": 38995 + }, + { + "epoch": 0.1, + "learning_rate": 9.733829059678831e-05, + "loss": 2.4514, + "step": 39000 + }, + { + "epoch": 0.1, + "learning_rate": 9.73376124472169e-05, + "loss": 2.3438, + "step": 39005 + }, + { + "epoch": 0.1, + "learning_rate": 9.733693421363002e-05, + "loss": 2.3942, + "step": 39010 + }, + { + "epoch": 0.1, + "learning_rate": 9.733625589602889e-05, + "loss": 2.2841, + "step": 39015 + }, + { + "epoch": 0.1, + "learning_rate": 9.73355774944147e-05, + "loss": 2.4226, + "step": 39020 + }, + { + "epoch": 0.1, + "learning_rate": 9.733489900878869e-05, + "loss": 2.3244, + "step": 39025 + }, + { + "epoch": 0.1, + "learning_rate": 9.733422043915204e-05, + "loss": 2.2595, + "step": 39030 + }, + { + "epoch": 0.1, + "learning_rate": 9.733354178550596e-05, + "loss": 2.4618, + "step": 39035 + }, + { + "epoch": 0.1, + "learning_rate": 9.733286304785165e-05, + "loss": 2.302, + "step": 39040 + }, + { + "epoch": 0.1, + "learning_rate": 9.73321842261903e-05, + "loss": 2.3075, + "step": 39045 + }, + { + "epoch": 0.1, + "learning_rate": 9.733150532052316e-05, + "loss": 2.4105, + "step": 39050 + }, + { + "epoch": 0.1, + "learning_rate": 9.733082633085138e-05, + "loss": 2.2652, + "step": 39055 + }, + { + "epoch": 0.1, + "learning_rate": 9.73301472571762e-05, + "loss": 2.348, + "step": 39060 + }, + { + "epoch": 0.1, + "learning_rate": 9.732946809949882e-05, + "loss": 2.2996, + "step": 39065 + }, + { + "epoch": 0.1, + "learning_rate": 9.732878885782044e-05, + "loss": 2.3286, + "step": 39070 + }, + { + "epoch": 0.1, + "learning_rate": 9.732810953214227e-05, + "loss": 2.4087, + "step": 39075 + }, + { + "epoch": 0.1, + "learning_rate": 9.73274301224655e-05, + "loss": 2.3778, + "step": 39080 + }, + { + "epoch": 0.1, + "learning_rate": 9.732675062879138e-05, + "loss": 2.3281, + "step": 39085 + }, + { + "epoch": 0.1, + "learning_rate": 9.732607105112106e-05, + "loss": 2.3418, + "step": 39090 + }, + { + "epoch": 0.1, + "learning_rate": 9.732539138945576e-05, + "loss": 2.4684, + "step": 39095 + }, + { + "epoch": 0.1, + "learning_rate": 9.732471164379673e-05, + "loss": 2.3083, + "step": 39100 + }, + { + "epoch": 0.1, + "learning_rate": 9.732403181414511e-05, + "loss": 2.3276, + "step": 39105 + }, + { + "epoch": 0.1, + "learning_rate": 9.732335190050216e-05, + "loss": 2.4486, + "step": 39110 + }, + { + "epoch": 0.1, + "learning_rate": 9.732267190286906e-05, + "loss": 2.3476, + "step": 39115 + }, + { + "epoch": 0.1, + "learning_rate": 9.732199182124704e-05, + "loss": 2.4669, + "step": 39120 + }, + { + "epoch": 0.1, + "learning_rate": 9.732131165563728e-05, + "loss": 2.3922, + "step": 39125 + }, + { + "epoch": 0.1, + "learning_rate": 9.7320631406041e-05, + "loss": 2.3489, + "step": 39130 + }, + { + "epoch": 0.1, + "learning_rate": 9.731995107245939e-05, + "loss": 2.2243, + "step": 39135 + }, + { + "epoch": 0.1, + "learning_rate": 9.731927065489369e-05, + "loss": 2.3494, + "step": 39140 + }, + { + "epoch": 0.1, + "learning_rate": 9.731859015334508e-05, + "loss": 2.4119, + "step": 39145 + }, + { + "epoch": 0.1, + "learning_rate": 9.731790956781479e-05, + "loss": 2.4149, + "step": 39150 + }, + { + "epoch": 0.1, + "learning_rate": 9.731722889830401e-05, + "loss": 2.421, + "step": 39155 + }, + { + "epoch": 0.1, + "learning_rate": 9.731654814481395e-05, + "loss": 2.4216, + "step": 39160 + }, + { + "epoch": 0.11, + "learning_rate": 9.731586730734583e-05, + "loss": 2.405, + "step": 39165 + }, + { + "epoch": 0.11, + "learning_rate": 9.731518638590085e-05, + "loss": 2.3074, + "step": 39170 + }, + { + "epoch": 0.11, + "learning_rate": 9.731450538048022e-05, + "loss": 2.429, + "step": 39175 + }, + { + "epoch": 0.11, + "learning_rate": 9.731382429108513e-05, + "loss": 2.3266, + "step": 39180 + }, + { + "epoch": 0.11, + "learning_rate": 9.731314311771684e-05, + "loss": 2.3573, + "step": 39185 + }, + { + "epoch": 0.11, + "learning_rate": 9.73124618603765e-05, + "loss": 2.33, + "step": 39190 + }, + { + "epoch": 0.11, + "learning_rate": 9.731178051906536e-05, + "loss": 2.3115, + "step": 39195 + }, + { + "epoch": 0.11, + "learning_rate": 9.73110990937846e-05, + "loss": 2.2854, + "step": 39200 + }, + { + "epoch": 0.11, + "learning_rate": 9.731041758453545e-05, + "loss": 2.3213, + "step": 39205 + }, + { + "epoch": 0.11, + "learning_rate": 9.73097359913191e-05, + "loss": 2.5004, + "step": 39210 + }, + { + "epoch": 0.11, + "learning_rate": 9.730905431413679e-05, + "loss": 2.3673, + "step": 39215 + }, + { + "epoch": 0.11, + "learning_rate": 9.73083725529897e-05, + "loss": 2.2836, + "step": 39220 + }, + { + "epoch": 0.11, + "learning_rate": 9.730769070787906e-05, + "loss": 2.4069, + "step": 39225 + }, + { + "epoch": 0.11, + "learning_rate": 9.730700877880607e-05, + "loss": 2.4677, + "step": 39230 + }, + { + "epoch": 0.11, + "learning_rate": 9.730632676577196e-05, + "loss": 2.4385, + "step": 39235 + }, + { + "epoch": 0.11, + "learning_rate": 9.73056446687779e-05, + "loss": 2.3048, + "step": 39240 + }, + { + "epoch": 0.11, + "learning_rate": 9.730496248782513e-05, + "loss": 2.3253, + "step": 39245 + }, + { + "epoch": 0.11, + "learning_rate": 9.730428022291486e-05, + "loss": 2.4129, + "step": 39250 + }, + { + "epoch": 0.11, + "learning_rate": 9.73035978740483e-05, + "loss": 2.3974, + "step": 39255 + }, + { + "epoch": 0.11, + "learning_rate": 9.730291544122665e-05, + "loss": 2.4096, + "step": 39260 + }, + { + "epoch": 0.11, + "learning_rate": 9.730223292445113e-05, + "loss": 2.4459, + "step": 39265 + }, + { + "epoch": 0.11, + "learning_rate": 9.730155032372295e-05, + "loss": 2.3859, + "step": 39270 + }, + { + "epoch": 0.11, + "learning_rate": 9.730086763904331e-05, + "loss": 2.376, + "step": 39275 + }, + { + "epoch": 0.11, + "learning_rate": 9.730018487041345e-05, + "loss": 2.4105, + "step": 39280 + }, + { + "epoch": 0.11, + "learning_rate": 9.729950201783455e-05, + "loss": 2.284, + "step": 39285 + }, + { + "epoch": 0.11, + "learning_rate": 9.729881908130784e-05, + "loss": 2.4516, + "step": 39290 + }, + { + "epoch": 0.11, + "learning_rate": 9.729813606083453e-05, + "loss": 2.4115, + "step": 39295 + }, + { + "epoch": 0.11, + "learning_rate": 9.729745295641584e-05, + "loss": 2.2722, + "step": 39300 + }, + { + "epoch": 0.11, + "learning_rate": 9.729676976805295e-05, + "loss": 2.3203, + "step": 39305 + }, + { + "epoch": 0.11, + "learning_rate": 9.729608649574711e-05, + "loss": 2.3234, + "step": 39310 + }, + { + "epoch": 0.11, + "learning_rate": 9.729540313949952e-05, + "loss": 2.4279, + "step": 39315 + }, + { + "epoch": 0.11, + "learning_rate": 9.729471969931138e-05, + "loss": 2.4929, + "step": 39320 + }, + { + "epoch": 0.11, + "learning_rate": 9.729403617518394e-05, + "loss": 2.3518, + "step": 39325 + }, + { + "epoch": 0.11, + "learning_rate": 9.729335256711835e-05, + "loss": 2.4293, + "step": 39330 + }, + { + "epoch": 0.11, + "learning_rate": 9.729266887511587e-05, + "loss": 2.3686, + "step": 39335 + }, + { + "epoch": 0.11, + "learning_rate": 9.729198509917771e-05, + "loss": 2.2912, + "step": 39340 + }, + { + "epoch": 0.11, + "learning_rate": 9.72913012393051e-05, + "loss": 2.1717, + "step": 39345 + }, + { + "epoch": 0.11, + "learning_rate": 9.72906172954992e-05, + "loss": 2.2758, + "step": 39350 + }, + { + "epoch": 0.11, + "learning_rate": 9.728993326776127e-05, + "loss": 2.4296, + "step": 39355 + }, + { + "epoch": 0.11, + "learning_rate": 9.72892491560925e-05, + "loss": 2.2897, + "step": 39360 + }, + { + "epoch": 0.11, + "learning_rate": 9.728856496049413e-05, + "loss": 2.3682, + "step": 39365 + }, + { + "epoch": 0.11, + "learning_rate": 9.728788068096734e-05, + "loss": 2.3215, + "step": 39370 + }, + { + "epoch": 0.11, + "learning_rate": 9.728719631751336e-05, + "loss": 2.363, + "step": 39375 + }, + { + "epoch": 0.11, + "learning_rate": 9.728651187013341e-05, + "loss": 2.4182, + "step": 39380 + }, + { + "epoch": 0.11, + "learning_rate": 9.72858273388287e-05, + "loss": 2.2277, + "step": 39385 + }, + { + "epoch": 0.11, + "learning_rate": 9.728514272360045e-05, + "loss": 2.4193, + "step": 39390 + }, + { + "epoch": 0.11, + "learning_rate": 9.728445802444988e-05, + "loss": 2.3977, + "step": 39395 + }, + { + "epoch": 0.11, + "learning_rate": 9.728377324137819e-05, + "loss": 2.3119, + "step": 39400 + }, + { + "epoch": 0.11, + "learning_rate": 9.72830883743866e-05, + "loss": 2.4442, + "step": 39405 + }, + { + "epoch": 0.11, + "learning_rate": 9.728240342347631e-05, + "loss": 2.4293, + "step": 39410 + }, + { + "epoch": 0.11, + "learning_rate": 9.728171838864858e-05, + "loss": 2.437, + "step": 39415 + }, + { + "epoch": 0.11, + "learning_rate": 9.728103326990459e-05, + "loss": 2.407, + "step": 39420 + }, + { + "epoch": 0.11, + "learning_rate": 9.728034806724555e-05, + "loss": 2.3013, + "step": 39425 + }, + { + "epoch": 0.11, + "learning_rate": 9.727966278067271e-05, + "loss": 2.3918, + "step": 39430 + }, + { + "epoch": 0.11, + "learning_rate": 9.727897741018725e-05, + "loss": 2.5037, + "step": 39435 + }, + { + "epoch": 0.11, + "learning_rate": 9.727829195579043e-05, + "loss": 2.273, + "step": 39440 + }, + { + "epoch": 0.11, + "learning_rate": 9.727760641748342e-05, + "loss": 2.3738, + "step": 39445 + }, + { + "epoch": 0.11, + "learning_rate": 9.727692079526746e-05, + "loss": 2.3636, + "step": 39450 + }, + { + "epoch": 0.11, + "learning_rate": 9.727623508914376e-05, + "loss": 2.3675, + "step": 39455 + }, + { + "epoch": 0.11, + "learning_rate": 9.727554929911355e-05, + "loss": 2.4005, + "step": 39460 + }, + { + "epoch": 0.11, + "learning_rate": 9.727486342517804e-05, + "loss": 2.4045, + "step": 39465 + }, + { + "epoch": 0.11, + "learning_rate": 9.727417746733842e-05, + "loss": 2.5091, + "step": 39470 + }, + { + "epoch": 0.11, + "learning_rate": 9.727349142559596e-05, + "loss": 2.3668, + "step": 39475 + }, + { + "epoch": 0.11, + "learning_rate": 9.727280529995184e-05, + "loss": 2.3767, + "step": 39480 + }, + { + "epoch": 0.11, + "learning_rate": 9.72721190904073e-05, + "loss": 2.4027, + "step": 39485 + }, + { + "epoch": 0.11, + "learning_rate": 9.727143279696352e-05, + "loss": 2.2501, + "step": 39490 + }, + { + "epoch": 0.11, + "learning_rate": 9.727074641962176e-05, + "loss": 2.4418, + "step": 39495 + }, + { + "epoch": 0.11, + "learning_rate": 9.727005995838322e-05, + "loss": 2.497, + "step": 39500 + }, + { + "epoch": 0.11, + "learning_rate": 9.726937341324913e-05, + "loss": 2.355, + "step": 39505 + }, + { + "epoch": 0.11, + "learning_rate": 9.726868678422067e-05, + "loss": 2.371, + "step": 39510 + }, + { + "epoch": 0.11, + "learning_rate": 9.726800007129912e-05, + "loss": 2.37, + "step": 39515 + }, + { + "epoch": 0.11, + "learning_rate": 9.726731327448565e-05, + "loss": 2.2264, + "step": 39520 + }, + { + "epoch": 0.11, + "learning_rate": 9.72666263937815e-05, + "loss": 2.435, + "step": 39525 + }, + { + "epoch": 0.11, + "learning_rate": 9.726593942918788e-05, + "loss": 2.3383, + "step": 39530 + }, + { + "epoch": 0.11, + "learning_rate": 9.726525238070601e-05, + "loss": 2.4431, + "step": 39535 + }, + { + "epoch": 0.11, + "learning_rate": 9.726456524833712e-05, + "loss": 2.2697, + "step": 39540 + }, + { + "epoch": 0.11, + "learning_rate": 9.726387803208242e-05, + "loss": 2.3826, + "step": 39545 + }, + { + "epoch": 0.11, + "learning_rate": 9.726319073194313e-05, + "loss": 2.3459, + "step": 39550 + }, + { + "epoch": 0.11, + "learning_rate": 9.726250334792049e-05, + "loss": 2.308, + "step": 39555 + }, + { + "epoch": 0.11, + "learning_rate": 9.726181588001568e-05, + "loss": 2.3237, + "step": 39560 + }, + { + "epoch": 0.11, + "learning_rate": 9.726112832822995e-05, + "loss": 2.3341, + "step": 39565 + }, + { + "epoch": 0.11, + "learning_rate": 9.72604406925645e-05, + "loss": 2.4034, + "step": 39570 + }, + { + "epoch": 0.11, + "learning_rate": 9.725975297302058e-05, + "loss": 2.3836, + "step": 39575 + }, + { + "epoch": 0.11, + "learning_rate": 9.725906516959938e-05, + "loss": 2.348, + "step": 39580 + }, + { + "epoch": 0.11, + "learning_rate": 9.725837728230214e-05, + "loss": 2.3972, + "step": 39585 + }, + { + "epoch": 0.11, + "learning_rate": 9.725768931113009e-05, + "loss": 2.2388, + "step": 39590 + }, + { + "epoch": 0.11, + "learning_rate": 9.725700125608441e-05, + "loss": 2.3487, + "step": 39595 + }, + { + "epoch": 0.11, + "learning_rate": 9.725631311716637e-05, + "loss": 2.3518, + "step": 39600 + }, + { + "epoch": 0.11, + "learning_rate": 9.725562489437715e-05, + "loss": 2.3043, + "step": 39605 + }, + { + "epoch": 0.11, + "learning_rate": 9.7254936587718e-05, + "loss": 2.2385, + "step": 39610 + }, + { + "epoch": 0.11, + "learning_rate": 9.725424819719011e-05, + "loss": 2.2463, + "step": 39615 + }, + { + "epoch": 0.11, + "learning_rate": 9.725355972279475e-05, + "loss": 2.3706, + "step": 39620 + }, + { + "epoch": 0.11, + "learning_rate": 9.725287116453311e-05, + "loss": 2.3207, + "step": 39625 + }, + { + "epoch": 0.11, + "learning_rate": 9.72521825224064e-05, + "loss": 2.3301, + "step": 39630 + }, + { + "epoch": 0.11, + "learning_rate": 9.725149379641589e-05, + "loss": 2.3224, + "step": 39635 + }, + { + "epoch": 0.11, + "learning_rate": 9.725080498656275e-05, + "loss": 2.3831, + "step": 39640 + }, + { + "epoch": 0.11, + "learning_rate": 9.725011609284822e-05, + "loss": 2.2589, + "step": 39645 + }, + { + "epoch": 0.11, + "learning_rate": 9.724942711527355e-05, + "loss": 2.3144, + "step": 39650 + }, + { + "epoch": 0.11, + "learning_rate": 9.724873805383991e-05, + "loss": 2.5022, + "step": 39655 + }, + { + "epoch": 0.11, + "learning_rate": 9.724804890854857e-05, + "loss": 2.4912, + "step": 39660 + }, + { + "epoch": 0.11, + "learning_rate": 9.724735967940075e-05, + "loss": 2.4212, + "step": 39665 + }, + { + "epoch": 0.11, + "learning_rate": 9.724667036639763e-05, + "loss": 2.3182, + "step": 39670 + }, + { + "epoch": 0.11, + "learning_rate": 9.724598096954048e-05, + "loss": 2.295, + "step": 39675 + }, + { + "epoch": 0.11, + "learning_rate": 9.72452914888305e-05, + "loss": 2.3026, + "step": 39680 + }, + { + "epoch": 0.11, + "learning_rate": 9.724460192426892e-05, + "loss": 2.4549, + "step": 39685 + }, + { + "epoch": 0.11, + "learning_rate": 9.724391227585697e-05, + "loss": 2.2738, + "step": 39690 + }, + { + "epoch": 0.11, + "learning_rate": 9.724322254359586e-05, + "loss": 2.4051, + "step": 39695 + }, + { + "epoch": 0.11, + "learning_rate": 9.724253272748683e-05, + "loss": 2.4311, + "step": 39700 + }, + { + "epoch": 0.11, + "learning_rate": 9.724184282753109e-05, + "loss": 2.3394, + "step": 39705 + }, + { + "epoch": 0.11, + "learning_rate": 9.724115284372987e-05, + "loss": 2.5365, + "step": 39710 + }, + { + "epoch": 0.11, + "learning_rate": 9.724046277608439e-05, + "loss": 2.4147, + "step": 39715 + }, + { + "epoch": 0.11, + "learning_rate": 9.723977262459589e-05, + "loss": 2.3371, + "step": 39720 + }, + { + "epoch": 0.11, + "learning_rate": 9.723908238926558e-05, + "loss": 2.336, + "step": 39725 + }, + { + "epoch": 0.11, + "learning_rate": 9.723839207009469e-05, + "loss": 2.3692, + "step": 39730 + }, + { + "epoch": 0.11, + "learning_rate": 9.723770166708445e-05, + "loss": 2.3554, + "step": 39735 + }, + { + "epoch": 0.11, + "learning_rate": 9.723701118023608e-05, + "loss": 2.47, + "step": 39740 + }, + { + "epoch": 0.11, + "learning_rate": 9.723632060955079e-05, + "loss": 2.4257, + "step": 39745 + }, + { + "epoch": 0.11, + "learning_rate": 9.723562995502985e-05, + "loss": 2.3955, + "step": 39750 + }, + { + "epoch": 0.11, + "learning_rate": 9.723493921667443e-05, + "loss": 2.3261, + "step": 39755 + }, + { + "epoch": 0.11, + "learning_rate": 9.723424839448578e-05, + "loss": 2.5155, + "step": 39760 + }, + { + "epoch": 0.11, + "learning_rate": 9.723355748846515e-05, + "loss": 2.4501, + "step": 39765 + }, + { + "epoch": 0.11, + "learning_rate": 9.723286649861374e-05, + "loss": 2.3989, + "step": 39770 + }, + { + "epoch": 0.11, + "learning_rate": 9.723217542493277e-05, + "loss": 2.3743, + "step": 39775 + }, + { + "epoch": 0.11, + "learning_rate": 9.72314842674235e-05, + "loss": 2.427, + "step": 39780 + }, + { + "epoch": 0.11, + "learning_rate": 9.72307930260871e-05, + "loss": 2.3146, + "step": 39785 + }, + { + "epoch": 0.11, + "learning_rate": 9.723010170092487e-05, + "loss": 2.2777, + "step": 39790 + }, + { + "epoch": 0.11, + "learning_rate": 9.722941029193798e-05, + "loss": 2.4544, + "step": 39795 + }, + { + "epoch": 0.11, + "learning_rate": 9.722871879912767e-05, + "loss": 2.3021, + "step": 39800 + }, + { + "epoch": 0.11, + "learning_rate": 9.72280272224952e-05, + "loss": 2.3168, + "step": 39805 + }, + { + "epoch": 0.11, + "learning_rate": 9.722733556204175e-05, + "loss": 2.3389, + "step": 39810 + }, + { + "epoch": 0.11, + "learning_rate": 9.722664381776855e-05, + "loss": 2.2857, + "step": 39815 + }, + { + "epoch": 0.11, + "learning_rate": 9.722595198967687e-05, + "loss": 2.3563, + "step": 39820 + }, + { + "epoch": 0.11, + "learning_rate": 9.722526007776792e-05, + "loss": 2.342, + "step": 39825 + }, + { + "epoch": 0.11, + "learning_rate": 9.722456808204292e-05, + "loss": 2.2438, + "step": 39830 + }, + { + "epoch": 0.11, + "learning_rate": 9.72238760025031e-05, + "loss": 2.4056, + "step": 39835 + }, + { + "epoch": 0.11, + "learning_rate": 9.722318383914966e-05, + "loss": 2.4626, + "step": 39840 + }, + { + "epoch": 0.11, + "learning_rate": 9.722249159198387e-05, + "loss": 2.4023, + "step": 39845 + }, + { + "epoch": 0.11, + "learning_rate": 9.722179926100696e-05, + "loss": 2.244, + "step": 39850 + }, + { + "epoch": 0.11, + "learning_rate": 9.722110684622014e-05, + "loss": 2.3466, + "step": 39855 + }, + { + "epoch": 0.11, + "learning_rate": 9.722041434762463e-05, + "loss": 2.4027, + "step": 39860 + }, + { + "epoch": 0.11, + "learning_rate": 9.721972176522169e-05, + "loss": 2.395, + "step": 39865 + }, + { + "epoch": 0.11, + "learning_rate": 9.721902909901253e-05, + "loss": 2.4321, + "step": 39870 + }, + { + "epoch": 0.11, + "learning_rate": 9.721833634899837e-05, + "loss": 2.3761, + "step": 39875 + }, + { + "epoch": 0.11, + "learning_rate": 9.721764351518045e-05, + "loss": 2.265, + "step": 39880 + }, + { + "epoch": 0.11, + "learning_rate": 9.721695059756e-05, + "loss": 2.3969, + "step": 39885 + }, + { + "epoch": 0.11, + "learning_rate": 9.721625759613825e-05, + "loss": 2.2698, + "step": 39890 + }, + { + "epoch": 0.11, + "learning_rate": 9.721556451091643e-05, + "loss": 2.3052, + "step": 39895 + }, + { + "epoch": 0.11, + "learning_rate": 9.721487134189578e-05, + "loss": 2.3138, + "step": 39900 + }, + { + "epoch": 0.11, + "learning_rate": 9.721417808907751e-05, + "loss": 2.3962, + "step": 39905 + }, + { + "epoch": 0.11, + "learning_rate": 9.721348475246285e-05, + "loss": 2.528, + "step": 39910 + }, + { + "epoch": 0.11, + "learning_rate": 9.721279133205306e-05, + "loss": 2.4287, + "step": 39915 + }, + { + "epoch": 0.11, + "learning_rate": 9.721209782784934e-05, + "loss": 2.3089, + "step": 39920 + }, + { + "epoch": 0.11, + "learning_rate": 9.721140423985294e-05, + "loss": 2.3995, + "step": 39925 + }, + { + "epoch": 0.11, + "learning_rate": 9.721071056806508e-05, + "loss": 2.1767, + "step": 39930 + }, + { + "epoch": 0.11, + "learning_rate": 9.721001681248698e-05, + "loss": 2.4451, + "step": 39935 + }, + { + "epoch": 0.11, + "learning_rate": 9.720932297311991e-05, + "loss": 2.4271, + "step": 39940 + }, + { + "epoch": 0.11, + "learning_rate": 9.720862904996507e-05, + "loss": 2.2254, + "step": 39945 + }, + { + "epoch": 0.11, + "learning_rate": 9.720793504302369e-05, + "loss": 2.3405, + "step": 39950 + }, + { + "epoch": 0.11, + "learning_rate": 9.720724095229701e-05, + "loss": 2.2309, + "step": 39955 + }, + { + "epoch": 0.11, + "learning_rate": 9.720654677778627e-05, + "loss": 2.3976, + "step": 39960 + }, + { + "epoch": 0.11, + "learning_rate": 9.72058525194927e-05, + "loss": 2.4428, + "step": 39965 + }, + { + "epoch": 0.11, + "learning_rate": 9.720515817741751e-05, + "loss": 2.3674, + "step": 39970 + }, + { + "epoch": 0.11, + "learning_rate": 9.720446375156196e-05, + "loss": 2.4054, + "step": 39975 + }, + { + "epoch": 0.11, + "learning_rate": 9.720376924192726e-05, + "loss": 2.3644, + "step": 39980 + }, + { + "epoch": 0.11, + "learning_rate": 9.720307464851468e-05, + "loss": 2.4418, + "step": 39985 + }, + { + "epoch": 0.11, + "learning_rate": 9.72023799713254e-05, + "loss": 2.4314, + "step": 39990 + }, + { + "epoch": 0.11, + "learning_rate": 9.72016852103607e-05, + "loss": 2.3178, + "step": 39995 + }, + { + "epoch": 0.11, + "learning_rate": 9.720099036562178e-05, + "loss": 2.3289, + "step": 40000 + }, + { + "epoch": 0.11, + "learning_rate": 9.720029543710988e-05, + "loss": 2.4942, + "step": 40005 + }, + { + "epoch": 0.11, + "learning_rate": 9.719960042482625e-05, + "loss": 2.2999, + "step": 40010 + }, + { + "epoch": 0.11, + "learning_rate": 9.719890532877211e-05, + "loss": 2.3376, + "step": 40015 + }, + { + "epoch": 0.11, + "learning_rate": 9.71982101489487e-05, + "loss": 2.3425, + "step": 40020 + }, + { + "epoch": 0.11, + "learning_rate": 9.719751488535725e-05, + "loss": 2.386, + "step": 40025 + }, + { + "epoch": 0.11, + "learning_rate": 9.719681953799898e-05, + "loss": 2.3755, + "step": 40030 + }, + { + "epoch": 0.11, + "learning_rate": 9.719612410687515e-05, + "loss": 2.4056, + "step": 40035 + }, + { + "epoch": 0.11, + "learning_rate": 9.719542859198698e-05, + "loss": 2.337, + "step": 40040 + }, + { + "epoch": 0.11, + "learning_rate": 9.719473299333572e-05, + "loss": 2.4664, + "step": 40045 + }, + { + "epoch": 0.11, + "learning_rate": 9.719403731092259e-05, + "loss": 2.3078, + "step": 40050 + }, + { + "epoch": 0.11, + "learning_rate": 9.719334154474881e-05, + "loss": 2.4355, + "step": 40055 + }, + { + "epoch": 0.11, + "learning_rate": 9.719264569481564e-05, + "loss": 2.4097, + "step": 40060 + }, + { + "epoch": 0.11, + "learning_rate": 9.719194976112429e-05, + "loss": 2.3344, + "step": 40065 + }, + { + "epoch": 0.11, + "learning_rate": 9.719125374367604e-05, + "loss": 2.4952, + "step": 40070 + }, + { + "epoch": 0.11, + "learning_rate": 9.719055764247208e-05, + "loss": 2.2419, + "step": 40075 + }, + { + "epoch": 0.11, + "learning_rate": 9.718986145751366e-05, + "loss": 2.5508, + "step": 40080 + }, + { + "epoch": 0.11, + "learning_rate": 9.718916518880202e-05, + "loss": 2.504, + "step": 40085 + }, + { + "epoch": 0.11, + "learning_rate": 9.718846883633841e-05, + "loss": 2.3456, + "step": 40090 + }, + { + "epoch": 0.11, + "learning_rate": 9.718777240012402e-05, + "loss": 2.3852, + "step": 40095 + }, + { + "epoch": 0.11, + "learning_rate": 9.718707588016014e-05, + "loss": 2.2872, + "step": 40100 + }, + { + "epoch": 0.11, + "learning_rate": 9.718637927644796e-05, + "loss": 2.293, + "step": 40105 + }, + { + "epoch": 0.11, + "learning_rate": 9.718568258898876e-05, + "loss": 2.3987, + "step": 40110 + }, + { + "epoch": 0.11, + "learning_rate": 9.718498581778375e-05, + "loss": 2.344, + "step": 40115 + }, + { + "epoch": 0.11, + "learning_rate": 9.718428896283415e-05, + "loss": 2.4227, + "step": 40120 + }, + { + "epoch": 0.11, + "learning_rate": 9.718359202414124e-05, + "loss": 2.3508, + "step": 40125 + }, + { + "epoch": 0.11, + "learning_rate": 9.718289500170623e-05, + "loss": 2.4491, + "step": 40130 + }, + { + "epoch": 0.11, + "learning_rate": 9.718219789553036e-05, + "loss": 2.3182, + "step": 40135 + }, + { + "epoch": 0.11, + "learning_rate": 9.718150070561487e-05, + "loss": 2.1895, + "step": 40140 + }, + { + "epoch": 0.11, + "learning_rate": 9.7180803431961e-05, + "loss": 2.4239, + "step": 40145 + }, + { + "epoch": 0.11, + "learning_rate": 9.718010607456997e-05, + "loss": 2.3591, + "step": 40150 + }, + { + "epoch": 0.11, + "learning_rate": 9.717940863344305e-05, + "loss": 2.3554, + "step": 40155 + }, + { + "epoch": 0.11, + "learning_rate": 9.717871110858145e-05, + "loss": 2.4027, + "step": 40160 + }, + { + "epoch": 0.11, + "learning_rate": 9.717801349998642e-05, + "loss": 2.3544, + "step": 40165 + }, + { + "epoch": 0.11, + "learning_rate": 9.717731580765919e-05, + "loss": 2.2857, + "step": 40170 + }, + { + "epoch": 0.11, + "learning_rate": 9.717661803160101e-05, + "loss": 2.4096, + "step": 40175 + }, + { + "epoch": 0.11, + "learning_rate": 9.717592017181311e-05, + "loss": 2.3565, + "step": 40180 + }, + { + "epoch": 0.11, + "learning_rate": 9.717522222829674e-05, + "loss": 2.4324, + "step": 40185 + }, + { + "epoch": 0.11, + "learning_rate": 9.717452420105312e-05, + "loss": 2.3059, + "step": 40190 + }, + { + "epoch": 0.11, + "learning_rate": 9.717382609008349e-05, + "loss": 2.3397, + "step": 40195 + }, + { + "epoch": 0.11, + "learning_rate": 9.717312789538912e-05, + "loss": 2.3917, + "step": 40200 + }, + { + "epoch": 0.11, + "learning_rate": 9.717242961697121e-05, + "loss": 2.2662, + "step": 40205 + }, + { + "epoch": 0.11, + "learning_rate": 9.717173125483102e-05, + "loss": 2.3733, + "step": 40210 + }, + { + "epoch": 0.11, + "learning_rate": 9.717103280896978e-05, + "loss": 2.3663, + "step": 40215 + }, + { + "epoch": 0.11, + "learning_rate": 9.717033427938875e-05, + "loss": 2.3397, + "step": 40220 + }, + { + "epoch": 0.11, + "learning_rate": 9.716963566608915e-05, + "loss": 2.3479, + "step": 40225 + }, + { + "epoch": 0.11, + "learning_rate": 9.716893696907221e-05, + "loss": 2.4106, + "step": 40230 + }, + { + "epoch": 0.11, + "learning_rate": 9.71682381883392e-05, + "loss": 2.285, + "step": 40235 + }, + { + "epoch": 0.11, + "learning_rate": 9.716753932389133e-05, + "loss": 2.4067, + "step": 40240 + }, + { + "epoch": 0.11, + "learning_rate": 9.716684037572987e-05, + "loss": 2.3888, + "step": 40245 + }, + { + "epoch": 0.11, + "learning_rate": 9.716614134385604e-05, + "loss": 2.4275, + "step": 40250 + }, + { + "epoch": 0.11, + "learning_rate": 9.716544222827109e-05, + "loss": 2.3207, + "step": 40255 + }, + { + "epoch": 0.11, + "learning_rate": 9.716474302897624e-05, + "loss": 2.3899, + "step": 40260 + }, + { + "epoch": 0.11, + "learning_rate": 9.716404374597278e-05, + "loss": 2.3919, + "step": 40265 + }, + { + "epoch": 0.11, + "learning_rate": 9.716334437926189e-05, + "loss": 2.3637, + "step": 40270 + }, + { + "epoch": 0.11, + "learning_rate": 9.716264492884486e-05, + "loss": 2.154, + "step": 40275 + }, + { + "epoch": 0.11, + "learning_rate": 9.71619453947229e-05, + "loss": 2.4218, + "step": 40280 + }, + { + "epoch": 0.11, + "learning_rate": 9.716124577689727e-05, + "loss": 2.3237, + "step": 40285 + }, + { + "epoch": 0.11, + "learning_rate": 9.716054607536921e-05, + "loss": 2.3473, + "step": 40290 + }, + { + "epoch": 0.11, + "learning_rate": 9.715984629013995e-05, + "loss": 2.2757, + "step": 40295 + }, + { + "epoch": 0.11, + "learning_rate": 9.715914642121074e-05, + "loss": 2.3401, + "step": 40300 + }, + { + "epoch": 0.11, + "learning_rate": 9.715844646858281e-05, + "loss": 2.2443, + "step": 40305 + }, + { + "epoch": 0.11, + "learning_rate": 9.715774643225744e-05, + "loss": 2.3662, + "step": 40310 + }, + { + "epoch": 0.11, + "learning_rate": 9.715704631223582e-05, + "loss": 2.315, + "step": 40315 + }, + { + "epoch": 0.11, + "learning_rate": 9.715634610851922e-05, + "loss": 2.3747, + "step": 40320 + }, + { + "epoch": 0.11, + "learning_rate": 9.71556458211089e-05, + "loss": 2.3631, + "step": 40325 + }, + { + "epoch": 0.11, + "learning_rate": 9.715494545000607e-05, + "loss": 2.4622, + "step": 40330 + }, + { + "epoch": 0.11, + "learning_rate": 9.7154244995212e-05, + "loss": 2.3286, + "step": 40335 + }, + { + "epoch": 0.11, + "learning_rate": 9.715354445672791e-05, + "loss": 2.3369, + "step": 40340 + }, + { + "epoch": 0.11, + "learning_rate": 9.715284383455505e-05, + "loss": 2.4598, + "step": 40345 + }, + { + "epoch": 0.11, + "learning_rate": 9.715214312869467e-05, + "loss": 2.3471, + "step": 40350 + }, + { + "epoch": 0.11, + "learning_rate": 9.715144233914801e-05, + "loss": 2.2578, + "step": 40355 + }, + { + "epoch": 0.11, + "learning_rate": 9.715074146591632e-05, + "loss": 2.3788, + "step": 40360 + }, + { + "epoch": 0.11, + "learning_rate": 9.715004050900083e-05, + "loss": 2.334, + "step": 40365 + }, + { + "epoch": 0.11, + "learning_rate": 9.714933946840281e-05, + "loss": 2.401, + "step": 40370 + }, + { + "epoch": 0.11, + "learning_rate": 9.714863834412346e-05, + "loss": 2.4128, + "step": 40375 + }, + { + "epoch": 0.11, + "learning_rate": 9.714793713616408e-05, + "loss": 2.3079, + "step": 40380 + }, + { + "epoch": 0.11, + "learning_rate": 9.714723584452587e-05, + "loss": 2.2658, + "step": 40385 + }, + { + "epoch": 0.11, + "learning_rate": 9.71465344692101e-05, + "loss": 2.2517, + "step": 40390 + }, + { + "epoch": 0.11, + "learning_rate": 9.7145833010218e-05, + "loss": 2.3656, + "step": 40395 + }, + { + "epoch": 0.11, + "learning_rate": 9.714513146755081e-05, + "loss": 2.4066, + "step": 40400 + }, + { + "epoch": 0.11, + "learning_rate": 9.714442984120979e-05, + "loss": 2.3807, + "step": 40405 + }, + { + "epoch": 0.11, + "learning_rate": 9.71437281311962e-05, + "loss": 2.4011, + "step": 40410 + }, + { + "epoch": 0.11, + "learning_rate": 9.714302633751122e-05, + "loss": 2.4083, + "step": 40415 + }, + { + "epoch": 0.11, + "learning_rate": 9.714232446015618e-05, + "loss": 2.3009, + "step": 40420 + }, + { + "epoch": 0.11, + "learning_rate": 9.714162249913228e-05, + "loss": 2.3185, + "step": 40425 + }, + { + "epoch": 0.11, + "learning_rate": 9.714092045444077e-05, + "loss": 2.3426, + "step": 40430 + }, + { + "epoch": 0.11, + "learning_rate": 9.714021832608289e-05, + "loss": 2.2668, + "step": 40435 + }, + { + "epoch": 0.11, + "learning_rate": 9.713951611405992e-05, + "loss": 2.2879, + "step": 40440 + }, + { + "epoch": 0.11, + "learning_rate": 9.713881381837305e-05, + "loss": 2.4705, + "step": 40445 + }, + { + "epoch": 0.11, + "learning_rate": 9.713811143902357e-05, + "loss": 2.4086, + "step": 40450 + }, + { + "epoch": 0.11, + "learning_rate": 9.713740897601273e-05, + "loss": 2.4074, + "step": 40455 + }, + { + "epoch": 0.11, + "learning_rate": 9.713670642934174e-05, + "loss": 2.3616, + "step": 40460 + }, + { + "epoch": 0.11, + "learning_rate": 9.713600379901187e-05, + "loss": 2.3991, + "step": 40465 + }, + { + "epoch": 0.11, + "learning_rate": 9.713530108502438e-05, + "loss": 2.3479, + "step": 40470 + }, + { + "epoch": 0.11, + "learning_rate": 9.71345982873805e-05, + "loss": 2.4043, + "step": 40475 + }, + { + "epoch": 0.11, + "learning_rate": 9.713389540608147e-05, + "loss": 2.1718, + "step": 40480 + }, + { + "epoch": 0.11, + "learning_rate": 9.713319244112854e-05, + "loss": 2.435, + "step": 40485 + }, + { + "epoch": 0.11, + "learning_rate": 9.713248939252297e-05, + "loss": 2.5254, + "step": 40490 + }, + { + "epoch": 0.11, + "learning_rate": 9.713178626026602e-05, + "loss": 2.4904, + "step": 40495 + }, + { + "epoch": 0.11, + "learning_rate": 9.71310830443589e-05, + "loss": 2.3444, + "step": 40500 + }, + { + "epoch": 0.11, + "learning_rate": 9.713037974480288e-05, + "loss": 2.2606, + "step": 40505 + }, + { + "epoch": 0.11, + "learning_rate": 9.712967636159923e-05, + "loss": 2.302, + "step": 40510 + }, + { + "epoch": 0.11, + "learning_rate": 9.712897289474913e-05, + "loss": 2.4354, + "step": 40515 + }, + { + "epoch": 0.11, + "learning_rate": 9.712826934425391e-05, + "loss": 2.2884, + "step": 40520 + }, + { + "epoch": 0.11, + "learning_rate": 9.712756571011479e-05, + "loss": 2.3065, + "step": 40525 + }, + { + "epoch": 0.11, + "learning_rate": 9.712686199233299e-05, + "loss": 2.4215, + "step": 40530 + }, + { + "epoch": 0.11, + "learning_rate": 9.712615819090977e-05, + "loss": 2.3851, + "step": 40535 + }, + { + "epoch": 0.11, + "learning_rate": 9.71254543058464e-05, + "loss": 2.4878, + "step": 40540 + }, + { + "epoch": 0.11, + "learning_rate": 9.712475033714413e-05, + "loss": 2.364, + "step": 40545 + }, + { + "epoch": 0.11, + "learning_rate": 9.71240462848042e-05, + "loss": 2.447, + "step": 40550 + }, + { + "epoch": 0.11, + "learning_rate": 9.712334214882784e-05, + "loss": 2.2771, + "step": 40555 + }, + { + "epoch": 0.11, + "learning_rate": 9.712263792921632e-05, + "loss": 2.3597, + "step": 40560 + }, + { + "epoch": 0.11, + "learning_rate": 9.71219336259709e-05, + "loss": 2.3897, + "step": 40565 + }, + { + "epoch": 0.11, + "learning_rate": 9.71212292390928e-05, + "loss": 2.3055, + "step": 40570 + }, + { + "epoch": 0.11, + "learning_rate": 9.71205247685833e-05, + "loss": 2.4107, + "step": 40575 + }, + { + "epoch": 0.11, + "learning_rate": 9.711982021444364e-05, + "loss": 2.3562, + "step": 40580 + }, + { + "epoch": 0.11, + "learning_rate": 9.711911557667505e-05, + "loss": 2.4083, + "step": 40585 + }, + { + "epoch": 0.11, + "learning_rate": 9.711841085527882e-05, + "loss": 2.5099, + "step": 40590 + }, + { + "epoch": 0.11, + "learning_rate": 9.711770605025617e-05, + "loss": 2.3268, + "step": 40595 + }, + { + "epoch": 0.11, + "learning_rate": 9.711700116160835e-05, + "loss": 2.3757, + "step": 40600 + }, + { + "epoch": 0.11, + "learning_rate": 9.711629618933663e-05, + "loss": 2.2882, + "step": 40605 + }, + { + "epoch": 0.11, + "learning_rate": 9.711559113344225e-05, + "loss": 2.4424, + "step": 40610 + }, + { + "epoch": 0.11, + "learning_rate": 9.711488599392646e-05, + "loss": 2.2789, + "step": 40615 + }, + { + "epoch": 0.11, + "learning_rate": 9.711418077079053e-05, + "loss": 2.4383, + "step": 40620 + }, + { + "epoch": 0.11, + "learning_rate": 9.711347546403569e-05, + "loss": 2.4055, + "step": 40625 + }, + { + "epoch": 0.11, + "learning_rate": 9.711277007366319e-05, + "loss": 2.3761, + "step": 40630 + }, + { + "epoch": 0.11, + "learning_rate": 9.71120645996743e-05, + "loss": 2.3619, + "step": 40635 + }, + { + "epoch": 0.11, + "learning_rate": 9.711135904207026e-05, + "loss": 2.3942, + "step": 40640 + }, + { + "epoch": 0.11, + "learning_rate": 9.711065340085233e-05, + "loss": 2.3691, + "step": 40645 + }, + { + "epoch": 0.11, + "learning_rate": 9.710994767602176e-05, + "loss": 2.4663, + "step": 40650 + }, + { + "epoch": 0.11, + "learning_rate": 9.710924186757976e-05, + "loss": 2.3401, + "step": 40655 + }, + { + "epoch": 0.11, + "learning_rate": 9.710853597552767e-05, + "loss": 2.3483, + "step": 40660 + }, + { + "epoch": 0.11, + "learning_rate": 9.710782999986668e-05, + "loss": 2.3718, + "step": 40665 + }, + { + "epoch": 0.11, + "learning_rate": 9.710712394059806e-05, + "loss": 2.3725, + "step": 40670 + }, + { + "epoch": 0.11, + "learning_rate": 9.710641779772306e-05, + "loss": 2.4178, + "step": 40675 + }, + { + "epoch": 0.11, + "learning_rate": 9.710571157124293e-05, + "loss": 2.3504, + "step": 40680 + }, + { + "epoch": 0.11, + "learning_rate": 9.710500526115893e-05, + "loss": 2.4804, + "step": 40685 + }, + { + "epoch": 0.11, + "learning_rate": 9.710429886747233e-05, + "loss": 2.2767, + "step": 40690 + }, + { + "epoch": 0.11, + "learning_rate": 9.710359239018433e-05, + "loss": 2.3388, + "step": 40695 + }, + { + "epoch": 0.11, + "learning_rate": 9.710288582929625e-05, + "loss": 2.2406, + "step": 40700 + }, + { + "epoch": 0.11, + "learning_rate": 9.710217918480929e-05, + "loss": 2.2288, + "step": 40705 + }, + { + "epoch": 0.11, + "learning_rate": 9.710147245672473e-05, + "loss": 2.2785, + "step": 40710 + }, + { + "epoch": 0.11, + "learning_rate": 9.710076564504384e-05, + "loss": 2.4413, + "step": 40715 + }, + { + "epoch": 0.11, + "learning_rate": 9.710005874976786e-05, + "loss": 2.2765, + "step": 40720 + }, + { + "epoch": 0.11, + "learning_rate": 9.709935177089801e-05, + "loss": 2.3839, + "step": 40725 + }, + { + "epoch": 0.11, + "learning_rate": 9.70986447084356e-05, + "loss": 2.4064, + "step": 40730 + }, + { + "epoch": 0.11, + "learning_rate": 9.709793756238185e-05, + "loss": 2.368, + "step": 40735 + }, + { + "epoch": 0.11, + "learning_rate": 9.709723033273803e-05, + "loss": 2.3496, + "step": 40740 + }, + { + "epoch": 0.11, + "learning_rate": 9.70965230195054e-05, + "loss": 2.3686, + "step": 40745 + }, + { + "epoch": 0.11, + "learning_rate": 9.70958156226852e-05, + "loss": 2.3957, + "step": 40750 + }, + { + "epoch": 0.11, + "learning_rate": 9.709510814227868e-05, + "loss": 2.2933, + "step": 40755 + }, + { + "epoch": 0.11, + "learning_rate": 9.709440057828711e-05, + "loss": 2.4821, + "step": 40760 + }, + { + "epoch": 0.11, + "learning_rate": 9.709369293071175e-05, + "loss": 2.3368, + "step": 40765 + }, + { + "epoch": 0.11, + "learning_rate": 9.709298519955385e-05, + "loss": 2.3767, + "step": 40770 + }, + { + "epoch": 0.11, + "learning_rate": 9.709227738481466e-05, + "loss": 2.3057, + "step": 40775 + }, + { + "epoch": 0.11, + "learning_rate": 9.709156948649545e-05, + "loss": 2.2205, + "step": 40780 + }, + { + "epoch": 0.11, + "learning_rate": 9.709086150459746e-05, + "loss": 2.4503, + "step": 40785 + }, + { + "epoch": 0.11, + "learning_rate": 9.709015343912194e-05, + "loss": 2.3211, + "step": 40790 + }, + { + "epoch": 0.11, + "learning_rate": 9.708944529007018e-05, + "loss": 2.4522, + "step": 40795 + }, + { + "epoch": 0.11, + "learning_rate": 9.708873705744342e-05, + "loss": 2.3718, + "step": 40800 + }, + { + "epoch": 0.11, + "learning_rate": 9.70880287412429e-05, + "loss": 2.3605, + "step": 40805 + }, + { + "epoch": 0.11, + "learning_rate": 9.70873203414699e-05, + "loss": 2.3718, + "step": 40810 + }, + { + "epoch": 0.11, + "learning_rate": 9.708661185812567e-05, + "loss": 2.427, + "step": 40815 + }, + { + "epoch": 0.11, + "learning_rate": 9.708590329121147e-05, + "loss": 2.3847, + "step": 40820 + }, + { + "epoch": 0.11, + "learning_rate": 9.708519464072855e-05, + "loss": 2.2095, + "step": 40825 + }, + { + "epoch": 0.11, + "learning_rate": 9.708448590667817e-05, + "loss": 2.3153, + "step": 40830 + }, + { + "epoch": 0.11, + "learning_rate": 9.708377708906159e-05, + "loss": 2.342, + "step": 40835 + }, + { + "epoch": 0.11, + "learning_rate": 9.708306818788007e-05, + "loss": 2.448, + "step": 40840 + }, + { + "epoch": 0.11, + "learning_rate": 9.708235920313486e-05, + "loss": 2.4348, + "step": 40845 + }, + { + "epoch": 0.11, + "learning_rate": 9.708165013482724e-05, + "loss": 2.24, + "step": 40850 + }, + { + "epoch": 0.11, + "learning_rate": 9.708094098295844e-05, + "loss": 2.3258, + "step": 40855 + }, + { + "epoch": 0.11, + "learning_rate": 9.708023174752973e-05, + "loss": 2.2715, + "step": 40860 + }, + { + "epoch": 0.11, + "learning_rate": 9.707952242854237e-05, + "loss": 2.4291, + "step": 40865 + }, + { + "epoch": 0.11, + "learning_rate": 9.707881302599762e-05, + "loss": 2.4309, + "step": 40870 + }, + { + "epoch": 0.11, + "learning_rate": 9.707810353989674e-05, + "loss": 2.4548, + "step": 40875 + }, + { + "epoch": 0.11, + "learning_rate": 9.7077393970241e-05, + "loss": 2.4531, + "step": 40880 + }, + { + "epoch": 0.11, + "learning_rate": 9.707668431703163e-05, + "loss": 2.3909, + "step": 40885 + }, + { + "epoch": 0.11, + "learning_rate": 9.707597458026991e-05, + "loss": 2.394, + "step": 40890 + }, + { + "epoch": 0.11, + "learning_rate": 9.70752647599571e-05, + "loss": 2.3897, + "step": 40895 + }, + { + "epoch": 0.11, + "learning_rate": 9.707455485609445e-05, + "loss": 2.3187, + "step": 40900 + }, + { + "epoch": 0.11, + "learning_rate": 9.707384486868322e-05, + "loss": 2.1897, + "step": 40905 + }, + { + "epoch": 0.11, + "learning_rate": 9.707313479772468e-05, + "loss": 2.3501, + "step": 40910 + }, + { + "epoch": 0.11, + "learning_rate": 9.707242464322008e-05, + "loss": 2.2568, + "step": 40915 + }, + { + "epoch": 0.11, + "learning_rate": 9.707171440517072e-05, + "loss": 2.3413, + "step": 40920 + }, + { + "epoch": 0.11, + "learning_rate": 9.707100408357779e-05, + "loss": 2.3134, + "step": 40925 + }, + { + "epoch": 0.11, + "learning_rate": 9.70702936784426e-05, + "loss": 2.317, + "step": 40930 + }, + { + "epoch": 0.11, + "learning_rate": 9.706958318976639e-05, + "loss": 2.3361, + "step": 40935 + }, + { + "epoch": 0.11, + "learning_rate": 9.706887261755043e-05, + "loss": 2.2952, + "step": 40940 + }, + { + "epoch": 0.11, + "learning_rate": 9.706816196179598e-05, + "loss": 2.4005, + "step": 40945 + }, + { + "epoch": 0.11, + "learning_rate": 9.706745122250432e-05, + "loss": 2.3353, + "step": 40950 + }, + { + "epoch": 0.11, + "learning_rate": 9.706674039967668e-05, + "loss": 2.3009, + "step": 40955 + }, + { + "epoch": 0.11, + "learning_rate": 9.706602949331433e-05, + "loss": 2.3424, + "step": 40960 + }, + { + "epoch": 0.11, + "learning_rate": 9.706531850341855e-05, + "loss": 2.3382, + "step": 40965 + }, + { + "epoch": 0.11, + "learning_rate": 9.706460742999057e-05, + "loss": 2.332, + "step": 40970 + }, + { + "epoch": 0.11, + "learning_rate": 9.706389627303169e-05, + "loss": 2.2635, + "step": 40975 + }, + { + "epoch": 0.11, + "learning_rate": 9.706318503254315e-05, + "loss": 2.2736, + "step": 40980 + }, + { + "epoch": 0.11, + "learning_rate": 9.70624737085262e-05, + "loss": 2.3744, + "step": 40985 + }, + { + "epoch": 0.11, + "learning_rate": 9.706176230098211e-05, + "loss": 2.4321, + "step": 40990 + }, + { + "epoch": 0.11, + "learning_rate": 9.706105080991218e-05, + "loss": 2.3111, + "step": 40995 + }, + { + "epoch": 0.11, + "learning_rate": 9.706033923531763e-05, + "loss": 2.3571, + "step": 41000 + }, + { + "epoch": 0.11, + "learning_rate": 9.705962757719973e-05, + "loss": 2.3032, + "step": 41005 + }, + { + "epoch": 0.11, + "learning_rate": 9.705891583555975e-05, + "loss": 2.3908, + "step": 41010 + }, + { + "epoch": 0.11, + "learning_rate": 9.705820401039895e-05, + "loss": 2.3655, + "step": 41015 + }, + { + "epoch": 0.11, + "learning_rate": 9.70574921017186e-05, + "loss": 2.3544, + "step": 41020 + }, + { + "epoch": 0.11, + "learning_rate": 9.705678010951996e-05, + "loss": 2.3904, + "step": 41025 + }, + { + "epoch": 0.11, + "learning_rate": 9.70560680338043e-05, + "loss": 2.3288, + "step": 41030 + }, + { + "epoch": 0.11, + "learning_rate": 9.705535587457286e-05, + "loss": 2.3222, + "step": 41035 + }, + { + "epoch": 0.11, + "learning_rate": 9.705464363182692e-05, + "loss": 2.26, + "step": 41040 + }, + { + "epoch": 0.11, + "learning_rate": 9.705393130556776e-05, + "loss": 2.4208, + "step": 41045 + }, + { + "epoch": 0.11, + "learning_rate": 9.705321889579661e-05, + "loss": 2.296, + "step": 41050 + }, + { + "epoch": 0.11, + "learning_rate": 9.705250640251476e-05, + "loss": 2.2638, + "step": 41055 + }, + { + "epoch": 0.11, + "learning_rate": 9.705179382572347e-05, + "loss": 2.4382, + "step": 41060 + }, + { + "epoch": 0.11, + "learning_rate": 9.705108116542401e-05, + "loss": 2.3222, + "step": 41065 + }, + { + "epoch": 0.11, + "learning_rate": 9.705036842161762e-05, + "loss": 2.295, + "step": 41070 + }, + { + "epoch": 0.11, + "learning_rate": 9.70496555943056e-05, + "loss": 2.2723, + "step": 41075 + }, + { + "epoch": 0.11, + "learning_rate": 9.704894268348918e-05, + "loss": 2.3058, + "step": 41080 + }, + { + "epoch": 0.11, + "learning_rate": 9.704822968916964e-05, + "loss": 2.3042, + "step": 41085 + }, + { + "epoch": 0.11, + "learning_rate": 9.704751661134827e-05, + "loss": 2.4083, + "step": 41090 + }, + { + "epoch": 0.11, + "learning_rate": 9.70468034500263e-05, + "loss": 2.2842, + "step": 41095 + }, + { + "epoch": 0.11, + "learning_rate": 9.7046090205205e-05, + "loss": 2.449, + "step": 41100 + }, + { + "epoch": 0.11, + "learning_rate": 9.704537687688566e-05, + "loss": 2.3818, + "step": 41105 + }, + { + "epoch": 0.11, + "learning_rate": 9.704466346506953e-05, + "loss": 2.2848, + "step": 41110 + }, + { + "epoch": 0.11, + "learning_rate": 9.704394996975787e-05, + "loss": 2.4058, + "step": 41115 + }, + { + "epoch": 0.11, + "learning_rate": 9.704323639095195e-05, + "loss": 2.2304, + "step": 41120 + }, + { + "epoch": 0.11, + "learning_rate": 9.704252272865304e-05, + "loss": 2.3758, + "step": 41125 + }, + { + "epoch": 0.11, + "learning_rate": 9.704180898286241e-05, + "loss": 2.455, + "step": 41130 + }, + { + "epoch": 0.11, + "learning_rate": 9.704109515358133e-05, + "loss": 2.3971, + "step": 41135 + }, + { + "epoch": 0.11, + "learning_rate": 9.704038124081106e-05, + "loss": 2.3646, + "step": 41140 + }, + { + "epoch": 0.11, + "learning_rate": 9.703966724455287e-05, + "loss": 2.2755, + "step": 41145 + }, + { + "epoch": 0.11, + "learning_rate": 9.703895316480802e-05, + "loss": 2.4875, + "step": 41150 + }, + { + "epoch": 0.11, + "learning_rate": 9.703823900157778e-05, + "loss": 2.3603, + "step": 41155 + }, + { + "epoch": 0.11, + "learning_rate": 9.703752475486341e-05, + "loss": 2.2652, + "step": 41160 + }, + { + "epoch": 0.11, + "learning_rate": 9.703681042466619e-05, + "loss": 2.3714, + "step": 41165 + }, + { + "epoch": 0.11, + "learning_rate": 9.703609601098739e-05, + "loss": 2.4042, + "step": 41170 + }, + { + "epoch": 0.11, + "learning_rate": 9.703538151382827e-05, + "loss": 2.2791, + "step": 41175 + }, + { + "epoch": 0.11, + "learning_rate": 9.70346669331901e-05, + "loss": 2.4334, + "step": 41180 + }, + { + "epoch": 0.11, + "learning_rate": 9.703395226907417e-05, + "loss": 2.3356, + "step": 41185 + }, + { + "epoch": 0.11, + "learning_rate": 9.70332375214817e-05, + "loss": 2.3211, + "step": 41190 + }, + { + "epoch": 0.11, + "learning_rate": 9.703252269041401e-05, + "loss": 2.4375, + "step": 41195 + }, + { + "epoch": 0.11, + "learning_rate": 9.703180777587233e-05, + "loss": 2.4143, + "step": 41200 + }, + { + "epoch": 0.11, + "learning_rate": 9.703109277785793e-05, + "loss": 2.4059, + "step": 41205 + }, + { + "epoch": 0.11, + "learning_rate": 9.703037769637211e-05, + "loss": 2.3658, + "step": 41210 + }, + { + "epoch": 0.11, + "learning_rate": 9.702966253141612e-05, + "loss": 2.4238, + "step": 41215 + }, + { + "epoch": 0.11, + "learning_rate": 9.702894728299123e-05, + "loss": 2.3577, + "step": 41220 + }, + { + "epoch": 0.11, + "learning_rate": 9.702823195109872e-05, + "loss": 2.4403, + "step": 41225 + }, + { + "epoch": 0.11, + "learning_rate": 9.702751653573983e-05, + "loss": 2.4412, + "step": 41230 + }, + { + "epoch": 0.11, + "learning_rate": 9.702680103691586e-05, + "loss": 2.5128, + "step": 41235 + }, + { + "epoch": 0.11, + "learning_rate": 9.702608545462808e-05, + "loss": 2.3165, + "step": 41240 + }, + { + "epoch": 0.11, + "learning_rate": 9.702536978887774e-05, + "loss": 2.3996, + "step": 41245 + }, + { + "epoch": 0.11, + "learning_rate": 9.702465403966612e-05, + "loss": 2.2904, + "step": 41250 + }, + { + "epoch": 0.11, + "learning_rate": 9.702393820699448e-05, + "loss": 2.4239, + "step": 41255 + }, + { + "epoch": 0.11, + "learning_rate": 9.702322229086411e-05, + "loss": 2.3433, + "step": 41260 + }, + { + "epoch": 0.11, + "learning_rate": 9.702250629127629e-05, + "loss": 2.4594, + "step": 41265 + }, + { + "epoch": 0.11, + "learning_rate": 9.702179020823224e-05, + "loss": 2.2752, + "step": 41270 + }, + { + "epoch": 0.11, + "learning_rate": 9.702107404173328e-05, + "loss": 2.367, + "step": 41275 + }, + { + "epoch": 0.11, + "learning_rate": 9.702035779178066e-05, + "loss": 2.4795, + "step": 41280 + }, + { + "epoch": 0.11, + "learning_rate": 9.701964145837565e-05, + "loss": 2.3694, + "step": 41285 + }, + { + "epoch": 0.11, + "learning_rate": 9.701892504151954e-05, + "loss": 2.3943, + "step": 41290 + }, + { + "epoch": 0.11, + "learning_rate": 9.701820854121358e-05, + "loss": 2.3904, + "step": 41295 + }, + { + "epoch": 0.11, + "learning_rate": 9.701749195745903e-05, + "loss": 2.3833, + "step": 41300 + }, + { + "epoch": 0.11, + "learning_rate": 9.701677529025721e-05, + "loss": 2.2944, + "step": 41305 + }, + { + "epoch": 0.11, + "learning_rate": 9.701605853960935e-05, + "loss": 2.438, + "step": 41310 + }, + { + "epoch": 0.11, + "learning_rate": 9.701534170551677e-05, + "loss": 2.3315, + "step": 41315 + }, + { + "epoch": 0.11, + "learning_rate": 9.701462478798067e-05, + "loss": 2.4909, + "step": 41320 + }, + { + "epoch": 0.11, + "learning_rate": 9.701390778700236e-05, + "loss": 2.3607, + "step": 41325 + }, + { + "epoch": 0.11, + "learning_rate": 9.701319070258312e-05, + "loss": 2.3033, + "step": 41330 + }, + { + "epoch": 0.11, + "learning_rate": 9.701247353472422e-05, + "loss": 2.2586, + "step": 41335 + }, + { + "epoch": 0.11, + "learning_rate": 9.701175628342693e-05, + "loss": 2.3517, + "step": 41340 + }, + { + "epoch": 0.11, + "learning_rate": 9.701103894869251e-05, + "loss": 2.4349, + "step": 41345 + }, + { + "epoch": 0.11, + "learning_rate": 9.701032153052225e-05, + "loss": 2.4301, + "step": 41350 + }, + { + "epoch": 0.11, + "learning_rate": 9.700960402891741e-05, + "loss": 2.3103, + "step": 41355 + }, + { + "epoch": 0.11, + "learning_rate": 9.700888644387928e-05, + "loss": 2.4244, + "step": 41360 + }, + { + "epoch": 0.11, + "learning_rate": 9.700816877540913e-05, + "loss": 2.3213, + "step": 41365 + }, + { + "epoch": 0.11, + "learning_rate": 9.700745102350823e-05, + "loss": 2.4305, + "step": 41370 + }, + { + "epoch": 0.11, + "learning_rate": 9.700673318817784e-05, + "loss": 2.4212, + "step": 41375 + }, + { + "epoch": 0.11, + "learning_rate": 9.700601526941925e-05, + "loss": 2.4003, + "step": 41380 + }, + { + "epoch": 0.11, + "learning_rate": 9.700529726723373e-05, + "loss": 2.4009, + "step": 41385 + }, + { + "epoch": 0.11, + "learning_rate": 9.700457918162255e-05, + "loss": 2.3766, + "step": 41390 + }, + { + "epoch": 0.11, + "learning_rate": 9.7003861012587e-05, + "loss": 2.3676, + "step": 41395 + }, + { + "epoch": 0.11, + "learning_rate": 9.700314276012834e-05, + "loss": 2.3787, + "step": 41400 + }, + { + "epoch": 0.11, + "learning_rate": 9.700242442424785e-05, + "loss": 2.3581, + "step": 41405 + }, + { + "epoch": 0.11, + "learning_rate": 9.700170600494681e-05, + "loss": 2.2816, + "step": 41410 + }, + { + "epoch": 0.11, + "learning_rate": 9.700098750222647e-05, + "loss": 2.3225, + "step": 41415 + }, + { + "epoch": 0.11, + "learning_rate": 9.700026891608813e-05, + "loss": 2.4106, + "step": 41420 + }, + { + "epoch": 0.11, + "learning_rate": 9.699955024653307e-05, + "loss": 2.4753, + "step": 41425 + }, + { + "epoch": 0.11, + "learning_rate": 9.699883149356254e-05, + "loss": 2.4565, + "step": 41430 + }, + { + "epoch": 0.11, + "learning_rate": 9.699811265717784e-05, + "loss": 2.2785, + "step": 41435 + }, + { + "epoch": 0.11, + "learning_rate": 9.699739373738024e-05, + "loss": 2.5132, + "step": 41440 + }, + { + "epoch": 0.11, + "learning_rate": 9.6996674734171e-05, + "loss": 2.4615, + "step": 41445 + }, + { + "epoch": 0.11, + "learning_rate": 9.699595564755141e-05, + "loss": 2.2863, + "step": 41450 + }, + { + "epoch": 0.11, + "learning_rate": 9.699523647752276e-05, + "loss": 2.4283, + "step": 41455 + }, + { + "epoch": 0.11, + "learning_rate": 9.69945172240863e-05, + "loss": 2.3547, + "step": 41460 + }, + { + "epoch": 0.11, + "learning_rate": 9.699379788724331e-05, + "loss": 2.3288, + "step": 41465 + }, + { + "epoch": 0.11, + "learning_rate": 9.699307846699508e-05, + "loss": 2.4537, + "step": 41470 + }, + { + "epoch": 0.11, + "learning_rate": 9.699235896334289e-05, + "loss": 2.3398, + "step": 41475 + }, + { + "epoch": 0.11, + "learning_rate": 9.699163937628799e-05, + "loss": 2.4168, + "step": 41480 + }, + { + "epoch": 0.11, + "learning_rate": 9.699091970583169e-05, + "loss": 2.3878, + "step": 41485 + }, + { + "epoch": 0.11, + "learning_rate": 9.699019995197524e-05, + "loss": 2.3971, + "step": 41490 + }, + { + "epoch": 0.11, + "learning_rate": 9.698948011471994e-05, + "loss": 2.4308, + "step": 41495 + }, + { + "epoch": 0.11, + "learning_rate": 9.698876019406707e-05, + "loss": 2.3864, + "step": 41500 + }, + { + "epoch": 0.11, + "learning_rate": 9.698804019001787e-05, + "loss": 2.2593, + "step": 41505 + }, + { + "epoch": 0.11, + "learning_rate": 9.698732010257365e-05, + "loss": 2.4469, + "step": 41510 + }, + { + "epoch": 0.11, + "learning_rate": 9.698659993173569e-05, + "loss": 2.3937, + "step": 41515 + }, + { + "epoch": 0.11, + "learning_rate": 9.698587967750526e-05, + "loss": 2.4816, + "step": 41520 + }, + { + "epoch": 0.11, + "learning_rate": 9.698515933988361e-05, + "loss": 2.448, + "step": 41525 + }, + { + "epoch": 0.11, + "learning_rate": 9.698443891887208e-05, + "loss": 2.3689, + "step": 41530 + }, + { + "epoch": 0.11, + "learning_rate": 9.698371841447189e-05, + "loss": 2.4362, + "step": 41535 + }, + { + "epoch": 0.11, + "learning_rate": 9.698299782668436e-05, + "loss": 2.2787, + "step": 41540 + }, + { + "epoch": 0.11, + "learning_rate": 9.698227715551074e-05, + "loss": 2.3057, + "step": 41545 + }, + { + "epoch": 0.11, + "learning_rate": 9.698155640095233e-05, + "loss": 2.3504, + "step": 41550 + }, + { + "epoch": 0.11, + "learning_rate": 9.698083556301038e-05, + "loss": 2.3695, + "step": 41555 + }, + { + "epoch": 0.11, + "learning_rate": 9.698011464168621e-05, + "loss": 2.2906, + "step": 41560 + }, + { + "epoch": 0.11, + "learning_rate": 9.697939363698109e-05, + "loss": 2.422, + "step": 41565 + }, + { + "epoch": 0.11, + "learning_rate": 9.697867254889627e-05, + "loss": 2.2944, + "step": 41570 + }, + { + "epoch": 0.11, + "learning_rate": 9.697795137743307e-05, + "loss": 2.3521, + "step": 41575 + }, + { + "epoch": 0.11, + "learning_rate": 9.697723012259272e-05, + "loss": 2.3614, + "step": 41580 + }, + { + "epoch": 0.11, + "learning_rate": 9.697650878437654e-05, + "loss": 2.4489, + "step": 41585 + }, + { + "epoch": 0.11, + "learning_rate": 9.69757873627858e-05, + "loss": 2.361, + "step": 41590 + }, + { + "epoch": 0.11, + "learning_rate": 9.697506585782177e-05, + "loss": 2.3197, + "step": 41595 + }, + { + "epoch": 0.11, + "learning_rate": 9.697434426948575e-05, + "loss": 2.4409, + "step": 41600 + }, + { + "epoch": 0.11, + "learning_rate": 9.697362259777902e-05, + "loss": 2.2785, + "step": 41605 + }, + { + "epoch": 0.11, + "learning_rate": 9.697290084270283e-05, + "loss": 2.3804, + "step": 41610 + }, + { + "epoch": 0.11, + "learning_rate": 9.69721790042585e-05, + "loss": 2.2834, + "step": 41615 + }, + { + "epoch": 0.11, + "learning_rate": 9.69714570824473e-05, + "loss": 2.3327, + "step": 41620 + }, + { + "epoch": 0.11, + "learning_rate": 9.697073507727048e-05, + "loss": 2.4056, + "step": 41625 + }, + { + "epoch": 0.11, + "learning_rate": 9.697001298872937e-05, + "loss": 2.4639, + "step": 41630 + }, + { + "epoch": 0.11, + "learning_rate": 9.696929081682521e-05, + "loss": 2.3797, + "step": 41635 + }, + { + "epoch": 0.11, + "learning_rate": 9.696856856155931e-05, + "loss": 2.4088, + "step": 41640 + }, + { + "epoch": 0.11, + "learning_rate": 9.696784622293294e-05, + "loss": 2.2387, + "step": 41645 + }, + { + "epoch": 0.11, + "learning_rate": 9.696712380094737e-05, + "loss": 2.2256, + "step": 41650 + }, + { + "epoch": 0.11, + "learning_rate": 9.69664012956039e-05, + "loss": 2.3832, + "step": 41655 + }, + { + "epoch": 0.11, + "learning_rate": 9.696567870690383e-05, + "loss": 2.2585, + "step": 41660 + }, + { + "epoch": 0.11, + "learning_rate": 9.69649560348484e-05, + "loss": 2.3484, + "step": 41665 + }, + { + "epoch": 0.11, + "learning_rate": 9.696423327943891e-05, + "loss": 2.3794, + "step": 41670 + }, + { + "epoch": 0.11, + "learning_rate": 9.696351044067666e-05, + "loss": 2.5256, + "step": 41675 + }, + { + "epoch": 0.11, + "learning_rate": 9.696278751856292e-05, + "loss": 2.403, + "step": 41680 + }, + { + "epoch": 0.11, + "learning_rate": 9.696206451309895e-05, + "loss": 2.4365, + "step": 41685 + }, + { + "epoch": 0.11, + "learning_rate": 9.696134142428608e-05, + "loss": 2.4019, + "step": 41690 + }, + { + "epoch": 0.11, + "learning_rate": 9.696061825212554e-05, + "loss": 2.3548, + "step": 41695 + }, + { + "epoch": 0.11, + "learning_rate": 9.695989499661868e-05, + "loss": 2.3857, + "step": 41700 + }, + { + "epoch": 0.11, + "learning_rate": 9.695917165776671e-05, + "loss": 2.4034, + "step": 41705 + }, + { + "epoch": 0.11, + "learning_rate": 9.695844823557096e-05, + "loss": 2.4022, + "step": 41710 + }, + { + "epoch": 0.11, + "learning_rate": 9.69577247300327e-05, + "loss": 2.3076, + "step": 41715 + }, + { + "epoch": 0.11, + "learning_rate": 9.695700114115324e-05, + "loss": 2.3956, + "step": 41720 + }, + { + "epoch": 0.11, + "learning_rate": 9.695627746893383e-05, + "loss": 2.3772, + "step": 41725 + }, + { + "epoch": 0.11, + "learning_rate": 9.695555371337575e-05, + "loss": 2.3783, + "step": 41730 + }, + { + "epoch": 0.11, + "learning_rate": 9.695482987448031e-05, + "loss": 2.3948, + "step": 41735 + }, + { + "epoch": 0.11, + "learning_rate": 9.695410595224879e-05, + "loss": 2.3902, + "step": 41740 + }, + { + "epoch": 0.11, + "learning_rate": 9.695338194668247e-05, + "loss": 2.3643, + "step": 41745 + }, + { + "epoch": 0.11, + "learning_rate": 9.695265785778263e-05, + "loss": 2.4038, + "step": 41750 + }, + { + "epoch": 0.11, + "learning_rate": 9.695193368555056e-05, + "loss": 2.4694, + "step": 41755 + }, + { + "epoch": 0.11, + "learning_rate": 9.695120942998754e-05, + "loss": 2.3811, + "step": 41760 + }, + { + "epoch": 0.11, + "learning_rate": 9.695048509109488e-05, + "loss": 2.3842, + "step": 41765 + }, + { + "epoch": 0.11, + "learning_rate": 9.694976066887382e-05, + "loss": 2.4114, + "step": 41770 + }, + { + "epoch": 0.11, + "learning_rate": 9.69490361633257e-05, + "loss": 2.3165, + "step": 41775 + }, + { + "epoch": 0.11, + "learning_rate": 9.694831157445174e-05, + "loss": 2.3596, + "step": 41780 + }, + { + "epoch": 0.11, + "learning_rate": 9.69475869022533e-05, + "loss": 2.3333, + "step": 41785 + }, + { + "epoch": 0.11, + "learning_rate": 9.694686214673161e-05, + "loss": 2.3441, + "step": 41790 + }, + { + "epoch": 0.11, + "learning_rate": 9.694613730788799e-05, + "loss": 2.361, + "step": 41795 + }, + { + "epoch": 0.11, + "learning_rate": 9.69454123857237e-05, + "loss": 2.3631, + "step": 41800 + }, + { + "epoch": 0.11, + "learning_rate": 9.694468738024005e-05, + "loss": 2.319, + "step": 41805 + }, + { + "epoch": 0.11, + "learning_rate": 9.69439622914383e-05, + "loss": 2.4054, + "step": 41810 + }, + { + "epoch": 0.11, + "learning_rate": 9.694323711931977e-05, + "loss": 2.426, + "step": 41815 + }, + { + "epoch": 0.11, + "learning_rate": 9.694251186388572e-05, + "loss": 2.2616, + "step": 41820 + }, + { + "epoch": 0.11, + "learning_rate": 9.694178652513745e-05, + "loss": 2.3254, + "step": 41825 + }, + { + "epoch": 0.11, + "learning_rate": 9.694106110307626e-05, + "loss": 2.2477, + "step": 41830 + }, + { + "epoch": 0.11, + "learning_rate": 9.69403355977034e-05, + "loss": 2.4382, + "step": 41835 + }, + { + "epoch": 0.11, + "learning_rate": 9.693961000902019e-05, + "loss": 2.4449, + "step": 41840 + }, + { + "epoch": 0.11, + "learning_rate": 9.693888433702791e-05, + "loss": 2.3704, + "step": 41845 + }, + { + "epoch": 0.11, + "learning_rate": 9.693815858172784e-05, + "loss": 2.2935, + "step": 41850 + }, + { + "epoch": 0.11, + "learning_rate": 9.693743274312125e-05, + "loss": 2.4429, + "step": 41855 + }, + { + "epoch": 0.11, + "learning_rate": 9.69367068212095e-05, + "loss": 2.3503, + "step": 41860 + }, + { + "epoch": 0.11, + "learning_rate": 9.69359808159938e-05, + "loss": 2.242, + "step": 41865 + }, + { + "epoch": 0.11, + "learning_rate": 9.693525472747546e-05, + "loss": 2.5087, + "step": 41870 + }, + { + "epoch": 0.11, + "learning_rate": 9.69345285556558e-05, + "loss": 2.3539, + "step": 41875 + }, + { + "epoch": 0.11, + "learning_rate": 9.693380230053608e-05, + "loss": 2.434, + "step": 41880 + }, + { + "epoch": 0.11, + "learning_rate": 9.693307596211759e-05, + "loss": 2.2954, + "step": 41885 + }, + { + "epoch": 0.11, + "learning_rate": 9.693234954040163e-05, + "loss": 2.4357, + "step": 41890 + }, + { + "epoch": 0.11, + "learning_rate": 9.693162303538947e-05, + "loss": 2.3316, + "step": 41895 + }, + { + "epoch": 0.11, + "learning_rate": 9.693089644708243e-05, + "loss": 2.3529, + "step": 41900 + }, + { + "epoch": 0.11, + "learning_rate": 9.693016977548177e-05, + "loss": 2.3888, + "step": 41905 + }, + { + "epoch": 0.11, + "learning_rate": 9.692944302058881e-05, + "loss": 2.3047, + "step": 41910 + }, + { + "epoch": 0.11, + "learning_rate": 9.69287161824048e-05, + "loss": 2.2911, + "step": 41915 + }, + { + "epoch": 0.11, + "learning_rate": 9.692798926093107e-05, + "loss": 2.2737, + "step": 41920 + }, + { + "epoch": 0.11, + "learning_rate": 9.692726225616888e-05, + "loss": 2.2616, + "step": 41925 + }, + { + "epoch": 0.11, + "learning_rate": 9.692653516811954e-05, + "loss": 2.2884, + "step": 41930 + }, + { + "epoch": 0.11, + "learning_rate": 9.692580799678432e-05, + "loss": 2.3526, + "step": 41935 + }, + { + "epoch": 0.11, + "learning_rate": 9.692508074216453e-05, + "loss": 2.2315, + "step": 41940 + }, + { + "epoch": 0.11, + "learning_rate": 9.692435340426147e-05, + "loss": 2.4604, + "step": 41945 + }, + { + "epoch": 0.11, + "learning_rate": 9.692362598307639e-05, + "loss": 2.3732, + "step": 41950 + }, + { + "epoch": 0.11, + "learning_rate": 9.692289847861062e-05, + "loss": 2.3796, + "step": 41955 + }, + { + "epoch": 0.11, + "learning_rate": 9.692217089086544e-05, + "loss": 2.3406, + "step": 41960 + }, + { + "epoch": 0.11, + "learning_rate": 9.692144321984214e-05, + "loss": 2.2935, + "step": 41965 + }, + { + "epoch": 0.11, + "learning_rate": 9.6920715465542e-05, + "loss": 2.3702, + "step": 41970 + }, + { + "epoch": 0.11, + "learning_rate": 9.691998762796632e-05, + "loss": 2.2199, + "step": 41975 + }, + { + "epoch": 0.11, + "learning_rate": 9.691925970711639e-05, + "loss": 2.3963, + "step": 41980 + }, + { + "epoch": 0.11, + "learning_rate": 9.691853170299352e-05, + "loss": 2.3302, + "step": 41985 + }, + { + "epoch": 0.11, + "learning_rate": 9.691780361559899e-05, + "loss": 2.3708, + "step": 41990 + }, + { + "epoch": 0.11, + "learning_rate": 9.691707544493407e-05, + "loss": 2.3834, + "step": 41995 + }, + { + "epoch": 0.11, + "learning_rate": 9.691634719100009e-05, + "loss": 2.4149, + "step": 42000 + }, + { + "epoch": 0.11, + "learning_rate": 9.691561885379831e-05, + "loss": 2.4038, + "step": 42005 + }, + { + "epoch": 0.11, + "learning_rate": 9.691489043333006e-05, + "loss": 2.3449, + "step": 42010 + }, + { + "epoch": 0.11, + "learning_rate": 9.691416192959659e-05, + "loss": 2.3645, + "step": 42015 + }, + { + "epoch": 0.11, + "learning_rate": 9.691343334259922e-05, + "loss": 2.3762, + "step": 42020 + }, + { + "epoch": 0.11, + "learning_rate": 9.691270467233923e-05, + "loss": 2.3619, + "step": 42025 + }, + { + "epoch": 0.11, + "learning_rate": 9.691197591881793e-05, + "loss": 2.4302, + "step": 42030 + }, + { + "epoch": 0.11, + "learning_rate": 9.69112470820366e-05, + "loss": 2.4659, + "step": 42035 + }, + { + "epoch": 0.11, + "learning_rate": 9.691051816199654e-05, + "loss": 2.2095, + "step": 42040 + }, + { + "epoch": 0.11, + "learning_rate": 9.690978915869903e-05, + "loss": 2.329, + "step": 42045 + }, + { + "epoch": 0.11, + "learning_rate": 9.690906007214537e-05, + "loss": 2.3637, + "step": 42050 + }, + { + "epoch": 0.11, + "learning_rate": 9.690833090233686e-05, + "loss": 2.2535, + "step": 42055 + }, + { + "epoch": 0.11, + "learning_rate": 9.69076016492748e-05, + "loss": 2.3028, + "step": 42060 + }, + { + "epoch": 0.11, + "learning_rate": 9.690687231296047e-05, + "loss": 2.449, + "step": 42065 + }, + { + "epoch": 0.11, + "learning_rate": 9.690614289339519e-05, + "loss": 2.365, + "step": 42070 + }, + { + "epoch": 0.11, + "learning_rate": 9.690541339058021e-05, + "loss": 2.4888, + "step": 42075 + }, + { + "epoch": 0.11, + "learning_rate": 9.690468380451687e-05, + "loss": 2.3433, + "step": 42080 + }, + { + "epoch": 0.11, + "learning_rate": 9.690395413520643e-05, + "loss": 2.4337, + "step": 42085 + }, + { + "epoch": 0.11, + "learning_rate": 9.69032243826502e-05, + "loss": 2.3654, + "step": 42090 + }, + { + "epoch": 0.11, + "learning_rate": 9.690249454684949e-05, + "loss": 2.4058, + "step": 42095 + }, + { + "epoch": 0.11, + "learning_rate": 9.690176462780557e-05, + "loss": 2.3747, + "step": 42100 + }, + { + "epoch": 0.11, + "learning_rate": 9.690103462551974e-05, + "loss": 2.3602, + "step": 42105 + }, + { + "epoch": 0.11, + "learning_rate": 9.690030453999332e-05, + "loss": 2.3502, + "step": 42110 + }, + { + "epoch": 0.11, + "learning_rate": 9.689957437122756e-05, + "loss": 2.4409, + "step": 42115 + }, + { + "epoch": 0.11, + "learning_rate": 9.689884411922381e-05, + "loss": 2.2908, + "step": 42120 + }, + { + "epoch": 0.11, + "learning_rate": 9.689811378398333e-05, + "loss": 2.3005, + "step": 42125 + }, + { + "epoch": 0.11, + "learning_rate": 9.689738336550742e-05, + "loss": 2.3733, + "step": 42130 + }, + { + "epoch": 0.11, + "learning_rate": 9.689665286379738e-05, + "loss": 2.4388, + "step": 42135 + }, + { + "epoch": 0.11, + "learning_rate": 9.689592227885452e-05, + "loss": 2.4715, + "step": 42140 + }, + { + "epoch": 0.11, + "learning_rate": 9.689519161068012e-05, + "loss": 2.4764, + "step": 42145 + }, + { + "epoch": 0.11, + "learning_rate": 9.689446085927548e-05, + "loss": 2.4088, + "step": 42150 + }, + { + "epoch": 0.11, + "learning_rate": 9.68937300246419e-05, + "loss": 2.2917, + "step": 42155 + }, + { + "epoch": 0.11, + "learning_rate": 9.689299910678069e-05, + "loss": 2.3766, + "step": 42160 + }, + { + "epoch": 0.11, + "learning_rate": 9.68922681056931e-05, + "loss": 2.4246, + "step": 42165 + }, + { + "epoch": 0.11, + "learning_rate": 9.689153702138048e-05, + "loss": 2.4399, + "step": 42170 + }, + { + "epoch": 0.11, + "learning_rate": 9.689080585384412e-05, + "loss": 2.2656, + "step": 42175 + }, + { + "epoch": 0.11, + "learning_rate": 9.689007460308528e-05, + "loss": 2.3985, + "step": 42180 + }, + { + "epoch": 0.11, + "learning_rate": 9.68893432691053e-05, + "loss": 2.3418, + "step": 42185 + }, + { + "epoch": 0.11, + "learning_rate": 9.688861185190547e-05, + "loss": 2.2907, + "step": 42190 + }, + { + "epoch": 0.11, + "learning_rate": 9.688788035148706e-05, + "loss": 2.4016, + "step": 42195 + }, + { + "epoch": 0.11, + "learning_rate": 9.68871487678514e-05, + "loss": 2.3354, + "step": 42200 + }, + { + "epoch": 0.11, + "learning_rate": 9.688641710099977e-05, + "loss": 2.2936, + "step": 42205 + }, + { + "epoch": 0.11, + "learning_rate": 9.688568535093348e-05, + "loss": 2.352, + "step": 42210 + }, + { + "epoch": 0.11, + "learning_rate": 9.688495351765382e-05, + "loss": 2.3069, + "step": 42215 + }, + { + "epoch": 0.11, + "learning_rate": 9.688422160116209e-05, + "loss": 2.3108, + "step": 42220 + }, + { + "epoch": 0.11, + "learning_rate": 9.68834896014596e-05, + "loss": 2.3437, + "step": 42225 + }, + { + "epoch": 0.11, + "learning_rate": 9.688275751854763e-05, + "loss": 2.3532, + "step": 42230 + }, + { + "epoch": 0.11, + "learning_rate": 9.68820253524275e-05, + "loss": 2.297, + "step": 42235 + }, + { + "epoch": 0.11, + "learning_rate": 9.68812931031005e-05, + "loss": 2.3231, + "step": 42240 + }, + { + "epoch": 0.11, + "learning_rate": 9.688056077056791e-05, + "loss": 2.2839, + "step": 42245 + }, + { + "epoch": 0.11, + "learning_rate": 9.687982835483108e-05, + "loss": 2.2461, + "step": 42250 + }, + { + "epoch": 0.11, + "learning_rate": 9.687909585589125e-05, + "loss": 2.4217, + "step": 42255 + }, + { + "epoch": 0.11, + "learning_rate": 9.687836327374978e-05, + "loss": 2.2502, + "step": 42260 + }, + { + "epoch": 0.11, + "learning_rate": 9.687763060840791e-05, + "loss": 2.4195, + "step": 42265 + }, + { + "epoch": 0.11, + "learning_rate": 9.687689785986698e-05, + "loss": 2.2284, + "step": 42270 + }, + { + "epoch": 0.11, + "learning_rate": 9.687616502812829e-05, + "loss": 2.2766, + "step": 42275 + }, + { + "epoch": 0.11, + "learning_rate": 9.687543211319311e-05, + "loss": 2.3114, + "step": 42280 + }, + { + "epoch": 0.11, + "learning_rate": 9.687469911506278e-05, + "loss": 2.438, + "step": 42285 + }, + { + "epoch": 0.11, + "learning_rate": 9.687396603373856e-05, + "loss": 2.2155, + "step": 42290 + }, + { + "epoch": 0.11, + "learning_rate": 9.687323286922178e-05, + "loss": 2.3344, + "step": 42295 + }, + { + "epoch": 0.11, + "learning_rate": 9.687249962151375e-05, + "loss": 2.3727, + "step": 42300 + }, + { + "epoch": 0.11, + "learning_rate": 9.687176629061575e-05, + "loss": 2.4118, + "step": 42305 + }, + { + "epoch": 0.11, + "learning_rate": 9.687103287652908e-05, + "loss": 2.2794, + "step": 42310 + }, + { + "epoch": 0.11, + "learning_rate": 9.687029937925506e-05, + "loss": 2.3612, + "step": 42315 + }, + { + "epoch": 0.11, + "learning_rate": 9.686956579879497e-05, + "loss": 2.322, + "step": 42320 + }, + { + "epoch": 0.11, + "learning_rate": 9.686883213515013e-05, + "loss": 2.4173, + "step": 42325 + }, + { + "epoch": 0.11, + "learning_rate": 9.686809838832183e-05, + "loss": 2.4084, + "step": 42330 + }, + { + "epoch": 0.11, + "learning_rate": 9.686736455831137e-05, + "loss": 2.3863, + "step": 42335 + }, + { + "epoch": 0.11, + "learning_rate": 9.686663064512009e-05, + "loss": 2.3424, + "step": 42340 + }, + { + "epoch": 0.11, + "learning_rate": 9.686589664874923e-05, + "loss": 2.4518, + "step": 42345 + }, + { + "epoch": 0.11, + "learning_rate": 9.686516256920014e-05, + "loss": 2.2345, + "step": 42350 + }, + { + "epoch": 0.11, + "learning_rate": 9.68644284064741e-05, + "loss": 2.3176, + "step": 42355 + }, + { + "epoch": 0.11, + "learning_rate": 9.686369416057243e-05, + "loss": 2.3251, + "step": 42360 + }, + { + "epoch": 0.11, + "learning_rate": 9.686295983149642e-05, + "loss": 2.2406, + "step": 42365 + }, + { + "epoch": 0.11, + "learning_rate": 9.686222541924738e-05, + "loss": 2.233, + "step": 42370 + }, + { + "epoch": 0.11, + "learning_rate": 9.686149092382661e-05, + "loss": 2.3245, + "step": 42375 + }, + { + "epoch": 0.11, + "learning_rate": 9.686075634523542e-05, + "loss": 2.512, + "step": 42380 + }, + { + "epoch": 0.11, + "learning_rate": 9.686002168347509e-05, + "loss": 2.3418, + "step": 42385 + }, + { + "epoch": 0.11, + "learning_rate": 9.685928693854695e-05, + "loss": 2.3036, + "step": 42390 + }, + { + "epoch": 0.11, + "learning_rate": 9.68585521104523e-05, + "loss": 2.3848, + "step": 42395 + }, + { + "epoch": 0.11, + "learning_rate": 9.685781719919244e-05, + "loss": 2.335, + "step": 42400 + }, + { + "epoch": 0.11, + "learning_rate": 9.685708220476867e-05, + "loss": 2.4088, + "step": 42405 + }, + { + "epoch": 0.11, + "learning_rate": 9.68563471271823e-05, + "loss": 2.3648, + "step": 42410 + }, + { + "epoch": 0.11, + "learning_rate": 9.685561196643463e-05, + "loss": 2.4163, + "step": 42415 + }, + { + "epoch": 0.11, + "learning_rate": 9.685487672252698e-05, + "loss": 2.353, + "step": 42420 + }, + { + "epoch": 0.11, + "learning_rate": 9.685414139546063e-05, + "loss": 2.3682, + "step": 42425 + }, + { + "epoch": 0.11, + "learning_rate": 9.685340598523691e-05, + "loss": 2.4974, + "step": 42430 + }, + { + "epoch": 0.11, + "learning_rate": 9.685267049185712e-05, + "loss": 2.3043, + "step": 42435 + }, + { + "epoch": 0.11, + "learning_rate": 9.685193491532254e-05, + "loss": 2.3457, + "step": 42440 + }, + { + "epoch": 0.11, + "learning_rate": 9.685119925563449e-05, + "loss": 2.4007, + "step": 42445 + }, + { + "epoch": 0.11, + "learning_rate": 9.68504635127943e-05, + "loss": 2.2975, + "step": 42450 + }, + { + "epoch": 0.11, + "learning_rate": 9.684972768680325e-05, + "loss": 2.4916, + "step": 42455 + }, + { + "epoch": 0.11, + "learning_rate": 9.684899177766265e-05, + "loss": 2.3335, + "step": 42460 + }, + { + "epoch": 0.11, + "learning_rate": 9.68482557853738e-05, + "loss": 2.5299, + "step": 42465 + }, + { + "epoch": 0.11, + "learning_rate": 9.684751970993801e-05, + "loss": 2.3424, + "step": 42470 + }, + { + "epoch": 0.11, + "learning_rate": 9.684678355135661e-05, + "loss": 2.3758, + "step": 42475 + }, + { + "epoch": 0.11, + "learning_rate": 9.684604730963088e-05, + "loss": 2.4151, + "step": 42480 + }, + { + "epoch": 0.11, + "learning_rate": 9.684531098476214e-05, + "loss": 2.4476, + "step": 42485 + }, + { + "epoch": 0.11, + "learning_rate": 9.684457457675169e-05, + "loss": 2.2642, + "step": 42490 + }, + { + "epoch": 0.11, + "learning_rate": 9.684383808560082e-05, + "loss": 2.416, + "step": 42495 + }, + { + "epoch": 0.11, + "learning_rate": 9.684310151131087e-05, + "loss": 2.2487, + "step": 42500 + }, + { + "epoch": 0.11, + "learning_rate": 9.684236485388313e-05, + "loss": 2.2489, + "step": 42505 + }, + { + "epoch": 0.11, + "learning_rate": 9.68416281133189e-05, + "loss": 2.35, + "step": 42510 + }, + { + "epoch": 0.11, + "learning_rate": 9.68408912896195e-05, + "loss": 2.3989, + "step": 42515 + }, + { + "epoch": 0.11, + "learning_rate": 9.684015438278625e-05, + "loss": 2.4074, + "step": 42520 + }, + { + "epoch": 0.11, + "learning_rate": 9.683941739282043e-05, + "loss": 2.3408, + "step": 42525 + }, + { + "epoch": 0.11, + "learning_rate": 9.683868031972336e-05, + "loss": 2.2951, + "step": 42530 + }, + { + "epoch": 0.11, + "learning_rate": 9.683794316349635e-05, + "loss": 2.374, + "step": 42535 + }, + { + "epoch": 0.11, + "learning_rate": 9.683720592414072e-05, + "loss": 2.3435, + "step": 42540 + }, + { + "epoch": 0.11, + "learning_rate": 9.683646860165775e-05, + "loss": 2.3804, + "step": 42545 + }, + { + "epoch": 0.11, + "learning_rate": 9.683573119604877e-05, + "loss": 2.4475, + "step": 42550 + }, + { + "epoch": 0.11, + "learning_rate": 9.683499370731508e-05, + "loss": 2.4404, + "step": 42555 + }, + { + "epoch": 0.11, + "learning_rate": 9.6834256135458e-05, + "loss": 2.2461, + "step": 42560 + }, + { + "epoch": 0.11, + "learning_rate": 9.683351848047882e-05, + "loss": 2.3902, + "step": 42565 + }, + { + "epoch": 0.11, + "learning_rate": 9.683278074237887e-05, + "loss": 2.4019, + "step": 42570 + }, + { + "epoch": 0.11, + "learning_rate": 9.683204292115944e-05, + "loss": 2.4575, + "step": 42575 + }, + { + "epoch": 0.11, + "learning_rate": 9.683130501682187e-05, + "loss": 2.3129, + "step": 42580 + }, + { + "epoch": 0.11, + "learning_rate": 9.683056702936742e-05, + "loss": 2.2908, + "step": 42585 + }, + { + "epoch": 0.11, + "learning_rate": 9.682982895879744e-05, + "loss": 2.285, + "step": 42590 + }, + { + "epoch": 0.11, + "learning_rate": 9.682909080511323e-05, + "loss": 2.4267, + "step": 42595 + }, + { + "epoch": 0.11, + "learning_rate": 9.682835256831611e-05, + "loss": 2.5257, + "step": 42600 + }, + { + "epoch": 0.11, + "learning_rate": 9.682761424840736e-05, + "loss": 2.3975, + "step": 42605 + }, + { + "epoch": 0.11, + "learning_rate": 9.682687584538832e-05, + "loss": 2.1724, + "step": 42610 + }, + { + "epoch": 0.11, + "learning_rate": 9.682613735926028e-05, + "loss": 2.1919, + "step": 42615 + }, + { + "epoch": 0.11, + "learning_rate": 9.682539879002457e-05, + "loss": 2.3509, + "step": 42620 + }, + { + "epoch": 0.11, + "learning_rate": 9.682466013768248e-05, + "loss": 2.4392, + "step": 42625 + }, + { + "epoch": 0.11, + "learning_rate": 9.682392140223534e-05, + "loss": 2.3798, + "step": 42630 + }, + { + "epoch": 0.11, + "learning_rate": 9.682318258368446e-05, + "loss": 2.2473, + "step": 42635 + }, + { + "epoch": 0.11, + "learning_rate": 9.682244368203113e-05, + "loss": 2.294, + "step": 42640 + }, + { + "epoch": 0.11, + "learning_rate": 9.682170469727669e-05, + "loss": 2.4481, + "step": 42645 + }, + { + "epoch": 0.11, + "learning_rate": 9.682096562942242e-05, + "loss": 2.4727, + "step": 42650 + }, + { + "epoch": 0.11, + "learning_rate": 9.682022647846967e-05, + "loss": 2.3026, + "step": 42655 + }, + { + "epoch": 0.11, + "learning_rate": 9.681948724441972e-05, + "loss": 2.4342, + "step": 42660 + }, + { + "epoch": 0.11, + "learning_rate": 9.681874792727387e-05, + "loss": 2.3642, + "step": 42665 + }, + { + "epoch": 0.11, + "learning_rate": 9.681800852703348e-05, + "loss": 2.3658, + "step": 42670 + }, + { + "epoch": 0.11, + "learning_rate": 9.681726904369985e-05, + "loss": 2.358, + "step": 42675 + }, + { + "epoch": 0.11, + "learning_rate": 9.681652947727427e-05, + "loss": 2.3997, + "step": 42680 + }, + { + "epoch": 0.11, + "learning_rate": 9.681578982775805e-05, + "loss": 2.354, + "step": 42685 + }, + { + "epoch": 0.11, + "learning_rate": 9.681505009515253e-05, + "loss": 2.4579, + "step": 42690 + }, + { + "epoch": 0.11, + "learning_rate": 9.6814310279459e-05, + "loss": 2.3124, + "step": 42695 + }, + { + "epoch": 0.11, + "learning_rate": 9.681357038067877e-05, + "loss": 2.3335, + "step": 42700 + }, + { + "epoch": 0.11, + "learning_rate": 9.68128303988132e-05, + "loss": 2.3625, + "step": 42705 + }, + { + "epoch": 0.11, + "learning_rate": 9.681209033386354e-05, + "loss": 2.3259, + "step": 42710 + }, + { + "epoch": 0.11, + "learning_rate": 9.681135018583113e-05, + "loss": 2.3726, + "step": 42715 + }, + { + "epoch": 0.11, + "learning_rate": 9.681060995471728e-05, + "loss": 2.4036, + "step": 42720 + }, + { + "epoch": 0.11, + "learning_rate": 9.680986964052332e-05, + "loss": 2.3845, + "step": 42725 + }, + { + "epoch": 0.11, + "learning_rate": 9.680912924325056e-05, + "loss": 2.2922, + "step": 42730 + }, + { + "epoch": 0.11, + "learning_rate": 9.68083887629003e-05, + "loss": 2.4001, + "step": 42735 + }, + { + "epoch": 0.11, + "learning_rate": 9.680764819947386e-05, + "loss": 2.2821, + "step": 42740 + }, + { + "epoch": 0.11, + "learning_rate": 9.680690755297255e-05, + "loss": 2.4266, + "step": 42745 + }, + { + "epoch": 0.11, + "learning_rate": 9.680616682339769e-05, + "loss": 2.4447, + "step": 42750 + }, + { + "epoch": 0.11, + "learning_rate": 9.68054260107506e-05, + "loss": 2.4932, + "step": 42755 + }, + { + "epoch": 0.11, + "learning_rate": 9.68046851150326e-05, + "loss": 2.44, + "step": 42760 + }, + { + "epoch": 0.11, + "learning_rate": 9.680394413624498e-05, + "loss": 2.3516, + "step": 42765 + }, + { + "epoch": 0.11, + "learning_rate": 9.680320307438907e-05, + "loss": 2.3404, + "step": 42770 + }, + { + "epoch": 0.11, + "learning_rate": 9.680246192946618e-05, + "loss": 2.3606, + "step": 42775 + }, + { + "epoch": 0.11, + "learning_rate": 9.680172070147763e-05, + "loss": 2.389, + "step": 42780 + }, + { + "epoch": 0.11, + "learning_rate": 9.680097939042474e-05, + "loss": 2.3999, + "step": 42785 + }, + { + "epoch": 0.11, + "learning_rate": 9.680023799630881e-05, + "loss": 2.346, + "step": 42790 + }, + { + "epoch": 0.11, + "learning_rate": 9.679949651913118e-05, + "loss": 2.3901, + "step": 42795 + }, + { + "epoch": 0.11, + "learning_rate": 9.679875495889316e-05, + "loss": 2.5008, + "step": 42800 + }, + { + "epoch": 0.11, + "learning_rate": 9.679801331559605e-05, + "loss": 2.3019, + "step": 42805 + }, + { + "epoch": 0.11, + "learning_rate": 9.679727158924117e-05, + "loss": 2.4654, + "step": 42810 + }, + { + "epoch": 0.11, + "learning_rate": 9.679652977982984e-05, + "loss": 2.3716, + "step": 42815 + }, + { + "epoch": 0.11, + "learning_rate": 9.679578788736338e-05, + "loss": 2.463, + "step": 42820 + }, + { + "epoch": 0.11, + "learning_rate": 9.679504591184312e-05, + "loss": 2.3686, + "step": 42825 + }, + { + "epoch": 0.11, + "learning_rate": 9.679430385327035e-05, + "loss": 2.403, + "step": 42830 + }, + { + "epoch": 0.11, + "learning_rate": 9.67935617116464e-05, + "loss": 2.3466, + "step": 42835 + }, + { + "epoch": 0.11, + "learning_rate": 9.679281948697259e-05, + "loss": 2.3247, + "step": 42840 + }, + { + "epoch": 0.11, + "learning_rate": 9.679207717925023e-05, + "loss": 2.366, + "step": 42845 + }, + { + "epoch": 0.11, + "learning_rate": 9.679133478848063e-05, + "loss": 2.4597, + "step": 42850 + }, + { + "epoch": 0.11, + "learning_rate": 9.679059231466514e-05, + "loss": 2.3215, + "step": 42855 + }, + { + "epoch": 0.11, + "learning_rate": 9.678984975780505e-05, + "loss": 2.3979, + "step": 42860 + }, + { + "epoch": 0.11, + "learning_rate": 9.678910711790168e-05, + "loss": 2.3926, + "step": 42865 + }, + { + "epoch": 0.11, + "learning_rate": 9.678836439495634e-05, + "loss": 2.3784, + "step": 42870 + }, + { + "epoch": 0.11, + "learning_rate": 9.678762158897037e-05, + "loss": 2.467, + "step": 42875 + }, + { + "epoch": 0.11, + "learning_rate": 9.678687869994509e-05, + "loss": 2.4183, + "step": 42880 + }, + { + "epoch": 0.11, + "learning_rate": 9.67861357278818e-05, + "loss": 2.3275, + "step": 42885 + }, + { + "epoch": 0.11, + "learning_rate": 9.678539267278181e-05, + "loss": 2.3333, + "step": 42890 + }, + { + "epoch": 0.12, + "learning_rate": 9.678464953464647e-05, + "loss": 2.361, + "step": 42895 + }, + { + "epoch": 0.12, + "learning_rate": 9.678390631347709e-05, + "loss": 2.4262, + "step": 42900 + }, + { + "epoch": 0.12, + "learning_rate": 9.678316300927497e-05, + "loss": 2.5222, + "step": 42905 + }, + { + "epoch": 0.12, + "learning_rate": 9.678241962204144e-05, + "loss": 2.444, + "step": 42910 + }, + { + "epoch": 0.12, + "learning_rate": 9.678167615177782e-05, + "loss": 2.2533, + "step": 42915 + }, + { + "epoch": 0.12, + "learning_rate": 9.678093259848543e-05, + "loss": 2.3063, + "step": 42920 + }, + { + "epoch": 0.12, + "learning_rate": 9.678018896216561e-05, + "loss": 2.3328, + "step": 42925 + }, + { + "epoch": 0.12, + "learning_rate": 9.677944524281965e-05, + "loss": 2.3794, + "step": 42930 + }, + { + "epoch": 0.12, + "learning_rate": 9.677870144044887e-05, + "loss": 2.3627, + "step": 42935 + }, + { + "epoch": 0.12, + "learning_rate": 9.677795755505462e-05, + "loss": 2.272, + "step": 42940 + }, + { + "epoch": 0.12, + "learning_rate": 9.677721358663819e-05, + "loss": 2.373, + "step": 42945 + }, + { + "epoch": 0.12, + "learning_rate": 9.677646953520089e-05, + "loss": 2.4183, + "step": 42950 + }, + { + "epoch": 0.12, + "learning_rate": 9.677572540074409e-05, + "loss": 2.2662, + "step": 42955 + }, + { + "epoch": 0.12, + "learning_rate": 9.677498118326908e-05, + "loss": 2.3032, + "step": 42960 + }, + { + "epoch": 0.12, + "learning_rate": 9.677423688277717e-05, + "loss": 2.3188, + "step": 42965 + }, + { + "epoch": 0.12, + "learning_rate": 9.677349249926971e-05, + "loss": 2.4009, + "step": 42970 + }, + { + "epoch": 0.12, + "learning_rate": 9.677274803274798e-05, + "loss": 2.4193, + "step": 42975 + }, + { + "epoch": 0.12, + "learning_rate": 9.677200348321335e-05, + "loss": 2.2923, + "step": 42980 + }, + { + "epoch": 0.12, + "learning_rate": 9.67712588506671e-05, + "loss": 2.2597, + "step": 42985 + }, + { + "epoch": 0.12, + "learning_rate": 9.677051413511058e-05, + "loss": 2.3956, + "step": 42990 + }, + { + "epoch": 0.12, + "learning_rate": 9.676976933654511e-05, + "loss": 2.24, + "step": 42995 + }, + { + "epoch": 0.12, + "learning_rate": 9.676902445497198e-05, + "loss": 2.4098, + "step": 43000 + }, + { + "epoch": 0.12, + "learning_rate": 9.676827949039256e-05, + "loss": 2.3823, + "step": 43005 + }, + { + "epoch": 0.12, + "learning_rate": 9.676753444280813e-05, + "loss": 2.3916, + "step": 43010 + }, + { + "epoch": 0.12, + "learning_rate": 9.676678931222002e-05, + "loss": 2.3486, + "step": 43015 + }, + { + "epoch": 0.12, + "learning_rate": 9.676604409862959e-05, + "loss": 2.3998, + "step": 43020 + }, + { + "epoch": 0.12, + "learning_rate": 9.676529880203812e-05, + "loss": 2.4561, + "step": 43025 + }, + { + "epoch": 0.12, + "learning_rate": 9.676455342244695e-05, + "loss": 2.4767, + "step": 43030 + }, + { + "epoch": 0.12, + "learning_rate": 9.67638079598574e-05, + "loss": 2.3853, + "step": 43035 + }, + { + "epoch": 0.12, + "learning_rate": 9.67630624142708e-05, + "loss": 2.3634, + "step": 43040 + }, + { + "epoch": 0.12, + "learning_rate": 9.676231678568844e-05, + "loss": 2.3344, + "step": 43045 + }, + { + "epoch": 0.12, + "learning_rate": 9.67615710741117e-05, + "loss": 2.2964, + "step": 43050 + }, + { + "epoch": 0.12, + "learning_rate": 9.676082527954185e-05, + "loss": 2.347, + "step": 43055 + }, + { + "epoch": 0.12, + "learning_rate": 9.676007940198026e-05, + "loss": 2.3301, + "step": 43060 + }, + { + "epoch": 0.12, + "learning_rate": 9.675933344142822e-05, + "loss": 2.3614, + "step": 43065 + }, + { + "epoch": 0.12, + "learning_rate": 9.675858739788706e-05, + "loss": 2.1586, + "step": 43070 + }, + { + "epoch": 0.12, + "learning_rate": 9.675784127135811e-05, + "loss": 2.3594, + "step": 43075 + }, + { + "epoch": 0.12, + "learning_rate": 9.675709506184271e-05, + "loss": 2.4094, + "step": 43080 + }, + { + "epoch": 0.12, + "learning_rate": 9.675634876934215e-05, + "loss": 2.324, + "step": 43085 + }, + { + "epoch": 0.12, + "learning_rate": 9.675560239385777e-05, + "loss": 2.2544, + "step": 43090 + }, + { + "epoch": 0.12, + "learning_rate": 9.675485593539091e-05, + "loss": 2.2634, + "step": 43095 + }, + { + "epoch": 0.12, + "learning_rate": 9.675410939394287e-05, + "loss": 2.3706, + "step": 43100 + }, + { + "epoch": 0.12, + "learning_rate": 9.675336276951499e-05, + "loss": 2.3371, + "step": 43105 + }, + { + "epoch": 0.12, + "learning_rate": 9.675261606210859e-05, + "loss": 2.4054, + "step": 43110 + }, + { + "epoch": 0.12, + "learning_rate": 9.675186927172502e-05, + "loss": 2.3842, + "step": 43115 + }, + { + "epoch": 0.12, + "learning_rate": 9.675112239836554e-05, + "loss": 2.3057, + "step": 43120 + }, + { + "epoch": 0.12, + "learning_rate": 9.675037544203154e-05, + "loss": 2.3438, + "step": 43125 + }, + { + "epoch": 0.12, + "learning_rate": 9.674962840272434e-05, + "loss": 2.2941, + "step": 43130 + }, + { + "epoch": 0.12, + "learning_rate": 9.674888128044524e-05, + "loss": 2.2475, + "step": 43135 + }, + { + "epoch": 0.12, + "learning_rate": 9.674813407519556e-05, + "loss": 2.2669, + "step": 43140 + }, + { + "epoch": 0.12, + "learning_rate": 9.674738678697665e-05, + "loss": 2.3848, + "step": 43145 + }, + { + "epoch": 0.12, + "learning_rate": 9.674663941578983e-05, + "loss": 2.3189, + "step": 43150 + }, + { + "epoch": 0.12, + "learning_rate": 9.674589196163641e-05, + "loss": 2.2983, + "step": 43155 + }, + { + "epoch": 0.12, + "learning_rate": 9.674514442451775e-05, + "loss": 2.4566, + "step": 43160 + }, + { + "epoch": 0.12, + "learning_rate": 9.674439680443516e-05, + "loss": 2.3624, + "step": 43165 + }, + { + "epoch": 0.12, + "learning_rate": 9.674364910138995e-05, + "loss": 2.3372, + "step": 43170 + }, + { + "epoch": 0.12, + "learning_rate": 9.674290131538346e-05, + "loss": 2.4554, + "step": 43175 + }, + { + "epoch": 0.12, + "learning_rate": 9.674215344641703e-05, + "loss": 2.4341, + "step": 43180 + }, + { + "epoch": 0.12, + "learning_rate": 9.674140549449198e-05, + "loss": 2.3467, + "step": 43185 + }, + { + "epoch": 0.12, + "learning_rate": 9.674065745960963e-05, + "loss": 2.4598, + "step": 43190 + }, + { + "epoch": 0.12, + "learning_rate": 9.673990934177131e-05, + "loss": 2.5013, + "step": 43195 + }, + { + "epoch": 0.12, + "learning_rate": 9.673916114097834e-05, + "loss": 2.3192, + "step": 43200 + }, + { + "epoch": 0.12, + "learning_rate": 9.673841285723207e-05, + "loss": 2.4604, + "step": 43205 + }, + { + "epoch": 0.12, + "learning_rate": 9.67376644905338e-05, + "loss": 2.4506, + "step": 43210 + }, + { + "epoch": 0.12, + "learning_rate": 9.673691604088489e-05, + "loss": 2.3188, + "step": 43215 + }, + { + "epoch": 0.12, + "learning_rate": 9.673616750828662e-05, + "loss": 2.194, + "step": 43220 + }, + { + "epoch": 0.12, + "learning_rate": 9.673541889274039e-05, + "loss": 2.4017, + "step": 43225 + }, + { + "epoch": 0.12, + "learning_rate": 9.673467019424747e-05, + "loss": 2.3213, + "step": 43230 + }, + { + "epoch": 0.12, + "learning_rate": 9.673392141280921e-05, + "loss": 2.37, + "step": 43235 + }, + { + "epoch": 0.12, + "learning_rate": 9.673317254842693e-05, + "loss": 2.3703, + "step": 43240 + }, + { + "epoch": 0.12, + "learning_rate": 9.673242360110197e-05, + "loss": 2.3785, + "step": 43245 + }, + { + "epoch": 0.12, + "learning_rate": 9.673167457083565e-05, + "loss": 2.3204, + "step": 43250 + }, + { + "epoch": 0.12, + "learning_rate": 9.673092545762932e-05, + "loss": 2.2914, + "step": 43255 + }, + { + "epoch": 0.12, + "learning_rate": 9.673017626148428e-05, + "loss": 2.379, + "step": 43260 + }, + { + "epoch": 0.12, + "learning_rate": 9.672942698240188e-05, + "loss": 2.2943, + "step": 43265 + }, + { + "epoch": 0.12, + "learning_rate": 9.672867762038343e-05, + "loss": 2.4301, + "step": 43270 + }, + { + "epoch": 0.12, + "learning_rate": 9.672792817543028e-05, + "loss": 2.4068, + "step": 43275 + }, + { + "epoch": 0.12, + "learning_rate": 9.672717864754378e-05, + "loss": 2.4918, + "step": 43280 + }, + { + "epoch": 0.12, + "learning_rate": 9.67264290367252e-05, + "loss": 2.1364, + "step": 43285 + }, + { + "epoch": 0.12, + "learning_rate": 9.672567934297592e-05, + "loss": 2.3425, + "step": 43290 + }, + { + "epoch": 0.12, + "learning_rate": 9.672492956629725e-05, + "loss": 2.3175, + "step": 43295 + }, + { + "epoch": 0.12, + "learning_rate": 9.672417970669052e-05, + "loss": 2.4179, + "step": 43300 + }, + { + "epoch": 0.12, + "learning_rate": 9.672342976415707e-05, + "loss": 2.1816, + "step": 43305 + }, + { + "epoch": 0.12, + "learning_rate": 9.672267973869822e-05, + "loss": 2.479, + "step": 43310 + }, + { + "epoch": 0.12, + "learning_rate": 9.672192963031532e-05, + "loss": 2.3657, + "step": 43315 + }, + { + "epoch": 0.12, + "learning_rate": 9.67211794390097e-05, + "loss": 2.472, + "step": 43320 + }, + { + "epoch": 0.12, + "learning_rate": 9.672042916478266e-05, + "loss": 2.3213, + "step": 43325 + }, + { + "epoch": 0.12, + "learning_rate": 9.671967880763556e-05, + "loss": 2.4233, + "step": 43330 + }, + { + "epoch": 0.12, + "learning_rate": 9.67189283675697e-05, + "loss": 2.3005, + "step": 43335 + }, + { + "epoch": 0.12, + "learning_rate": 9.671817784458647e-05, + "loss": 2.3285, + "step": 43340 + }, + { + "epoch": 0.12, + "learning_rate": 9.671742723868715e-05, + "loss": 2.373, + "step": 43345 + }, + { + "epoch": 0.12, + "learning_rate": 9.67166765498731e-05, + "loss": 2.2039, + "step": 43350 + }, + { + "epoch": 0.12, + "learning_rate": 9.671592577814566e-05, + "loss": 2.3846, + "step": 43355 + }, + { + "epoch": 0.12, + "learning_rate": 9.671517492350613e-05, + "loss": 2.4075, + "step": 43360 + }, + { + "epoch": 0.12, + "learning_rate": 9.671442398595585e-05, + "loss": 2.2907, + "step": 43365 + }, + { + "epoch": 0.12, + "learning_rate": 9.671367296549616e-05, + "loss": 2.4489, + "step": 43370 + }, + { + "epoch": 0.12, + "learning_rate": 9.671292186212839e-05, + "loss": 2.3779, + "step": 43375 + }, + { + "epoch": 0.12, + "learning_rate": 9.67121706758539e-05, + "loss": 2.3621, + "step": 43380 + }, + { + "epoch": 0.12, + "learning_rate": 9.671141940667398e-05, + "loss": 2.3212, + "step": 43385 + }, + { + "epoch": 0.12, + "learning_rate": 9.671066805458999e-05, + "loss": 2.3382, + "step": 43390 + }, + { + "epoch": 0.12, + "learning_rate": 9.670991661960326e-05, + "loss": 2.3327, + "step": 43395 + }, + { + "epoch": 0.12, + "learning_rate": 9.670916510171511e-05, + "loss": 2.3668, + "step": 43400 + }, + { + "epoch": 0.12, + "learning_rate": 9.67084135009269e-05, + "loss": 2.3149, + "step": 43405 + }, + { + "epoch": 0.12, + "learning_rate": 9.670766181723993e-05, + "loss": 2.303, + "step": 43410 + }, + { + "epoch": 0.12, + "learning_rate": 9.670691005065557e-05, + "loss": 2.342, + "step": 43415 + }, + { + "epoch": 0.12, + "learning_rate": 9.670615820117512e-05, + "loss": 2.3384, + "step": 43420 + }, + { + "epoch": 0.12, + "learning_rate": 9.670540626879994e-05, + "loss": 2.4247, + "step": 43425 + }, + { + "epoch": 0.12, + "learning_rate": 9.670465425353136e-05, + "loss": 2.3547, + "step": 43430 + }, + { + "epoch": 0.12, + "learning_rate": 9.670390215537071e-05, + "loss": 2.243, + "step": 43435 + }, + { + "epoch": 0.12, + "learning_rate": 9.670314997431931e-05, + "loss": 2.4099, + "step": 43440 + }, + { + "epoch": 0.12, + "learning_rate": 9.670239771037853e-05, + "loss": 2.4085, + "step": 43445 + }, + { + "epoch": 0.12, + "learning_rate": 9.670164536354966e-05, + "loss": 2.3289, + "step": 43450 + }, + { + "epoch": 0.12, + "learning_rate": 9.670089293383408e-05, + "loss": 2.3854, + "step": 43455 + }, + { + "epoch": 0.12, + "learning_rate": 9.67001404212331e-05, + "loss": 2.4651, + "step": 43460 + }, + { + "epoch": 0.12, + "learning_rate": 9.669938782574807e-05, + "loss": 2.437, + "step": 43465 + }, + { + "epoch": 0.12, + "learning_rate": 9.669863514738032e-05, + "loss": 2.3737, + "step": 43470 + }, + { + "epoch": 0.12, + "learning_rate": 9.669788238613116e-05, + "loss": 2.3864, + "step": 43475 + }, + { + "epoch": 0.12, + "learning_rate": 9.669712954200196e-05, + "loss": 2.3534, + "step": 43480 + }, + { + "epoch": 0.12, + "learning_rate": 9.669637661499404e-05, + "loss": 2.3332, + "step": 43485 + }, + { + "epoch": 0.12, + "learning_rate": 9.669562360510875e-05, + "loss": 2.3689, + "step": 43490 + }, + { + "epoch": 0.12, + "learning_rate": 9.669487051234742e-05, + "loss": 2.3971, + "step": 43495 + }, + { + "epoch": 0.12, + "learning_rate": 9.669411733671138e-05, + "loss": 2.4118, + "step": 43500 + }, + { + "epoch": 0.12, + "learning_rate": 9.669336407820197e-05, + "loss": 2.3889, + "step": 43505 + }, + { + "epoch": 0.12, + "learning_rate": 9.669261073682055e-05, + "loss": 2.4576, + "step": 43510 + }, + { + "epoch": 0.12, + "learning_rate": 9.66918573125684e-05, + "loss": 2.2972, + "step": 43515 + }, + { + "epoch": 0.12, + "learning_rate": 9.66911038054469e-05, + "loss": 2.3184, + "step": 43520 + }, + { + "epoch": 0.12, + "learning_rate": 9.669035021545741e-05, + "loss": 2.384, + "step": 43525 + }, + { + "epoch": 0.12, + "learning_rate": 9.66895965426012e-05, + "loss": 2.427, + "step": 43530 + }, + { + "epoch": 0.12, + "learning_rate": 9.668884278687967e-05, + "loss": 2.4438, + "step": 43535 + }, + { + "epoch": 0.12, + "learning_rate": 9.66880889482941e-05, + "loss": 2.4582, + "step": 43540 + }, + { + "epoch": 0.12, + "learning_rate": 9.668733502684591e-05, + "loss": 2.3819, + "step": 43545 + }, + { + "epoch": 0.12, + "learning_rate": 9.668658102253635e-05, + "loss": 2.2845, + "step": 43550 + }, + { + "epoch": 0.12, + "learning_rate": 9.668582693536681e-05, + "loss": 2.3377, + "step": 43555 + }, + { + "epoch": 0.12, + "learning_rate": 9.668507276533861e-05, + "loss": 2.3649, + "step": 43560 + }, + { + "epoch": 0.12, + "learning_rate": 9.66843185124531e-05, + "loss": 2.352, + "step": 43565 + }, + { + "epoch": 0.12, + "learning_rate": 9.668356417671161e-05, + "loss": 2.3569, + "step": 43570 + }, + { + "epoch": 0.12, + "learning_rate": 9.668280975811548e-05, + "loss": 2.3784, + "step": 43575 + }, + { + "epoch": 0.12, + "learning_rate": 9.668205525666604e-05, + "loss": 2.4461, + "step": 43580 + }, + { + "epoch": 0.12, + "learning_rate": 9.668130067236466e-05, + "loss": 2.2534, + "step": 43585 + }, + { + "epoch": 0.12, + "learning_rate": 9.668054600521264e-05, + "loss": 2.3945, + "step": 43590 + }, + { + "epoch": 0.12, + "learning_rate": 9.667979125521136e-05, + "loss": 2.2874, + "step": 43595 + }, + { + "epoch": 0.12, + "learning_rate": 9.667903642236212e-05, + "loss": 2.3609, + "step": 43600 + }, + { + "epoch": 0.12, + "learning_rate": 9.667828150666628e-05, + "loss": 2.4444, + "step": 43605 + }, + { + "epoch": 0.12, + "learning_rate": 9.667752650812518e-05, + "loss": 2.1067, + "step": 43610 + }, + { + "epoch": 0.12, + "learning_rate": 9.667677142674015e-05, + "loss": 2.3064, + "step": 43615 + }, + { + "epoch": 0.12, + "learning_rate": 9.667601626251254e-05, + "loss": 2.2533, + "step": 43620 + }, + { + "epoch": 0.12, + "learning_rate": 9.667526101544368e-05, + "loss": 2.4061, + "step": 43625 + }, + { + "epoch": 0.12, + "learning_rate": 9.667450568553494e-05, + "loss": 2.263, + "step": 43630 + }, + { + "epoch": 0.12, + "learning_rate": 9.667375027278762e-05, + "loss": 2.315, + "step": 43635 + }, + { + "epoch": 0.12, + "learning_rate": 9.667299477720309e-05, + "loss": 2.4209, + "step": 43640 + }, + { + "epoch": 0.12, + "learning_rate": 9.667223919878267e-05, + "loss": 2.2059, + "step": 43645 + }, + { + "epoch": 0.12, + "learning_rate": 9.667148353752771e-05, + "loss": 2.2815, + "step": 43650 + }, + { + "epoch": 0.12, + "learning_rate": 9.667072779343956e-05, + "loss": 2.366, + "step": 43655 + }, + { + "epoch": 0.12, + "learning_rate": 9.666997196651955e-05, + "loss": 2.24, + "step": 43660 + }, + { + "epoch": 0.12, + "learning_rate": 9.666921605676903e-05, + "loss": 2.4144, + "step": 43665 + }, + { + "epoch": 0.12, + "learning_rate": 9.666846006418933e-05, + "loss": 2.4501, + "step": 43670 + }, + { + "epoch": 0.12, + "learning_rate": 9.66677039887818e-05, + "loss": 2.2941, + "step": 43675 + }, + { + "epoch": 0.12, + "learning_rate": 9.666694783054777e-05, + "loss": 2.3954, + "step": 43680 + }, + { + "epoch": 0.12, + "learning_rate": 9.66661915894886e-05, + "loss": 2.3603, + "step": 43685 + }, + { + "epoch": 0.12, + "learning_rate": 9.666543526560563e-05, + "loss": 2.3951, + "step": 43690 + }, + { + "epoch": 0.12, + "learning_rate": 9.66646788589002e-05, + "loss": 2.3706, + "step": 43695 + }, + { + "epoch": 0.12, + "learning_rate": 9.666392236937363e-05, + "loss": 2.3236, + "step": 43700 + }, + { + "epoch": 0.12, + "learning_rate": 9.66631657970273e-05, + "loss": 2.4786, + "step": 43705 + }, + { + "epoch": 0.12, + "learning_rate": 9.666240914186254e-05, + "loss": 2.3951, + "step": 43710 + }, + { + "epoch": 0.12, + "learning_rate": 9.666165240388066e-05, + "loss": 2.3453, + "step": 43715 + }, + { + "epoch": 0.12, + "learning_rate": 9.666089558308306e-05, + "loss": 2.3339, + "step": 43720 + }, + { + "epoch": 0.12, + "learning_rate": 9.666013867947104e-05, + "loss": 2.2045, + "step": 43725 + }, + { + "epoch": 0.12, + "learning_rate": 9.665938169304594e-05, + "loss": 2.2895, + "step": 43730 + }, + { + "epoch": 0.12, + "learning_rate": 9.665862462380915e-05, + "loss": 2.4109, + "step": 43735 + }, + { + "epoch": 0.12, + "learning_rate": 9.665786747176198e-05, + "loss": 2.4072, + "step": 43740 + }, + { + "epoch": 0.12, + "learning_rate": 9.665711023690575e-05, + "loss": 2.3213, + "step": 43745 + }, + { + "epoch": 0.12, + "learning_rate": 9.665635291924186e-05, + "loss": 2.3751, + "step": 43750 + }, + { + "epoch": 0.12, + "learning_rate": 9.665559551877161e-05, + "loss": 2.5108, + "step": 43755 + }, + { + "epoch": 0.12, + "learning_rate": 9.665483803549637e-05, + "loss": 2.2815, + "step": 43760 + }, + { + "epoch": 0.12, + "learning_rate": 9.665408046941748e-05, + "loss": 2.537, + "step": 43765 + }, + { + "epoch": 0.12, + "learning_rate": 9.665332282053628e-05, + "loss": 2.3371, + "step": 43770 + }, + { + "epoch": 0.12, + "learning_rate": 9.66525650888541e-05, + "loss": 2.2583, + "step": 43775 + }, + { + "epoch": 0.12, + "learning_rate": 9.66518072743723e-05, + "loss": 2.3434, + "step": 43780 + }, + { + "epoch": 0.12, + "learning_rate": 9.665104937709222e-05, + "loss": 2.4983, + "step": 43785 + }, + { + "epoch": 0.12, + "learning_rate": 9.665029139701522e-05, + "loss": 2.3136, + "step": 43790 + }, + { + "epoch": 0.12, + "learning_rate": 9.664953333414264e-05, + "loss": 2.2701, + "step": 43795 + }, + { + "epoch": 0.12, + "learning_rate": 9.66487751884758e-05, + "loss": 2.2671, + "step": 43800 + }, + { + "epoch": 0.12, + "learning_rate": 9.664801696001609e-05, + "loss": 2.3275, + "step": 43805 + }, + { + "epoch": 0.12, + "learning_rate": 9.66472586487648e-05, + "loss": 2.4816, + "step": 43810 + }, + { + "epoch": 0.12, + "learning_rate": 9.664650025472333e-05, + "loss": 2.3826, + "step": 43815 + }, + { + "epoch": 0.12, + "learning_rate": 9.664574177789301e-05, + "loss": 2.3932, + "step": 43820 + }, + { + "epoch": 0.12, + "learning_rate": 9.664498321827515e-05, + "loss": 2.5349, + "step": 43825 + }, + { + "epoch": 0.12, + "learning_rate": 9.664422457587114e-05, + "loss": 2.473, + "step": 43830 + }, + { + "epoch": 0.12, + "learning_rate": 9.664346585068232e-05, + "loss": 2.3178, + "step": 43835 + }, + { + "epoch": 0.12, + "learning_rate": 9.664270704271001e-05, + "loss": 2.2868, + "step": 43840 + }, + { + "epoch": 0.12, + "learning_rate": 9.66419481519556e-05, + "loss": 2.2686, + "step": 43845 + }, + { + "epoch": 0.12, + "learning_rate": 9.664118917842039e-05, + "loss": 2.34, + "step": 43850 + }, + { + "epoch": 0.12, + "learning_rate": 9.664043012210575e-05, + "loss": 2.3803, + "step": 43855 + }, + { + "epoch": 0.12, + "learning_rate": 9.663967098301303e-05, + "loss": 2.3859, + "step": 43860 + }, + { + "epoch": 0.12, + "learning_rate": 9.663891176114357e-05, + "loss": 2.3277, + "step": 43865 + }, + { + "epoch": 0.12, + "learning_rate": 9.663815245649874e-05, + "loss": 2.2703, + "step": 43870 + }, + { + "epoch": 0.12, + "learning_rate": 9.663739306907985e-05, + "loss": 2.2886, + "step": 43875 + }, + { + "epoch": 0.12, + "learning_rate": 9.663663359888828e-05, + "loss": 2.3272, + "step": 43880 + }, + { + "epoch": 0.12, + "learning_rate": 9.663587404592536e-05, + "loss": 2.4017, + "step": 43885 + }, + { + "epoch": 0.12, + "learning_rate": 9.663511441019243e-05, + "loss": 2.2565, + "step": 43890 + }, + { + "epoch": 0.12, + "learning_rate": 9.663435469169085e-05, + "loss": 2.3476, + "step": 43895 + }, + { + "epoch": 0.12, + "learning_rate": 9.663359489042198e-05, + "loss": 2.3135, + "step": 43900 + }, + { + "epoch": 0.12, + "learning_rate": 9.663283500638716e-05, + "loss": 2.2684, + "step": 43905 + }, + { + "epoch": 0.12, + "learning_rate": 9.663207503958774e-05, + "loss": 2.221, + "step": 43910 + }, + { + "epoch": 0.12, + "learning_rate": 9.663131499002505e-05, + "loss": 2.4248, + "step": 43915 + }, + { + "epoch": 0.12, + "learning_rate": 9.663055485770045e-05, + "loss": 2.3644, + "step": 43920 + }, + { + "epoch": 0.12, + "learning_rate": 9.66297946426153e-05, + "loss": 2.2445, + "step": 43925 + }, + { + "epoch": 0.12, + "learning_rate": 9.662903434477095e-05, + "loss": 2.2776, + "step": 43930 + }, + { + "epoch": 0.12, + "learning_rate": 9.662827396416873e-05, + "loss": 2.4384, + "step": 43935 + }, + { + "epoch": 0.12, + "learning_rate": 9.662751350081001e-05, + "loss": 2.4471, + "step": 43940 + }, + { + "epoch": 0.12, + "learning_rate": 9.662675295469612e-05, + "loss": 2.4276, + "step": 43945 + }, + { + "epoch": 0.12, + "learning_rate": 9.662599232582843e-05, + "loss": 2.4376, + "step": 43950 + }, + { + "epoch": 0.12, + "learning_rate": 9.662523161420827e-05, + "loss": 2.3091, + "step": 43955 + }, + { + "epoch": 0.12, + "learning_rate": 9.662447081983702e-05, + "loss": 2.3579, + "step": 43960 + }, + { + "epoch": 0.12, + "learning_rate": 9.6623709942716e-05, + "loss": 2.3075, + "step": 43965 + }, + { + "epoch": 0.12, + "learning_rate": 9.662294898284657e-05, + "loss": 2.2847, + "step": 43970 + }, + { + "epoch": 0.12, + "learning_rate": 9.662218794023008e-05, + "loss": 2.2569, + "step": 43975 + }, + { + "epoch": 0.12, + "learning_rate": 9.662142681486788e-05, + "loss": 2.3916, + "step": 43980 + }, + { + "epoch": 0.12, + "learning_rate": 9.662066560676132e-05, + "loss": 2.3779, + "step": 43985 + }, + { + "epoch": 0.12, + "learning_rate": 9.661990431591178e-05, + "loss": 2.3566, + "step": 43990 + }, + { + "epoch": 0.12, + "learning_rate": 9.661914294232057e-05, + "loss": 2.3086, + "step": 43995 + }, + { + "epoch": 0.12, + "learning_rate": 9.661838148598905e-05, + "loss": 2.4258, + "step": 44000 + }, + { + "epoch": 0.12, + "learning_rate": 9.661761994691858e-05, + "loss": 2.2572, + "step": 44005 + }, + { + "epoch": 0.12, + "learning_rate": 9.661685832511052e-05, + "loss": 2.429, + "step": 44010 + }, + { + "epoch": 0.12, + "learning_rate": 9.661609662056621e-05, + "loss": 2.3967, + "step": 44015 + }, + { + "epoch": 0.12, + "learning_rate": 9.6615334833287e-05, + "loss": 2.3337, + "step": 44020 + }, + { + "epoch": 0.12, + "learning_rate": 9.661457296327424e-05, + "loss": 2.4003, + "step": 44025 + }, + { + "epoch": 0.12, + "learning_rate": 9.661381101052929e-05, + "loss": 2.3111, + "step": 44030 + }, + { + "epoch": 0.12, + "learning_rate": 9.66130489750535e-05, + "loss": 2.2599, + "step": 44035 + }, + { + "epoch": 0.12, + "learning_rate": 9.661228685684823e-05, + "loss": 2.4116, + "step": 44040 + }, + { + "epoch": 0.12, + "learning_rate": 9.661152465591481e-05, + "loss": 2.4401, + "step": 44045 + }, + { + "epoch": 0.12, + "learning_rate": 9.661076237225463e-05, + "loss": 2.4232, + "step": 44050 + }, + { + "epoch": 0.12, + "learning_rate": 9.6610000005869e-05, + "loss": 2.4195, + "step": 44055 + }, + { + "epoch": 0.12, + "learning_rate": 9.660923755675931e-05, + "loss": 2.5208, + "step": 44060 + }, + { + "epoch": 0.12, + "learning_rate": 9.66084750249269e-05, + "loss": 2.2888, + "step": 44065 + }, + { + "epoch": 0.12, + "learning_rate": 9.66077124103731e-05, + "loss": 2.3689, + "step": 44070 + }, + { + "epoch": 0.12, + "learning_rate": 9.660694971309929e-05, + "loss": 2.2206, + "step": 44075 + }, + { + "epoch": 0.12, + "learning_rate": 9.660618693310681e-05, + "loss": 2.379, + "step": 44080 + }, + { + "epoch": 0.12, + "learning_rate": 9.660542407039704e-05, + "loss": 2.3108, + "step": 44085 + }, + { + "epoch": 0.12, + "learning_rate": 9.660466112497131e-05, + "loss": 2.4226, + "step": 44090 + }, + { + "epoch": 0.12, + "learning_rate": 9.660389809683098e-05, + "loss": 2.4445, + "step": 44095 + }, + { + "epoch": 0.12, + "learning_rate": 9.660313498597739e-05, + "loss": 2.3886, + "step": 44100 + }, + { + "epoch": 0.12, + "learning_rate": 9.660237179241193e-05, + "loss": 2.4343, + "step": 44105 + }, + { + "epoch": 0.12, + "learning_rate": 9.660160851613592e-05, + "loss": 2.2873, + "step": 44110 + }, + { + "epoch": 0.12, + "learning_rate": 9.660084515715073e-05, + "loss": 2.3933, + "step": 44115 + }, + { + "epoch": 0.12, + "learning_rate": 9.660008171545771e-05, + "loss": 2.3968, + "step": 44120 + }, + { + "epoch": 0.12, + "learning_rate": 9.659931819105822e-05, + "loss": 2.3781, + "step": 44125 + }, + { + "epoch": 0.12, + "learning_rate": 9.659855458395362e-05, + "loss": 2.2431, + "step": 44130 + }, + { + "epoch": 0.12, + "learning_rate": 9.659779089414523e-05, + "loss": 2.4057, + "step": 44135 + }, + { + "epoch": 0.12, + "learning_rate": 9.659702712163446e-05, + "loss": 2.3415, + "step": 44140 + }, + { + "epoch": 0.12, + "learning_rate": 9.659626326642263e-05, + "loss": 2.3171, + "step": 44145 + }, + { + "epoch": 0.12, + "learning_rate": 9.65954993285111e-05, + "loss": 2.3662, + "step": 44150 + }, + { + "epoch": 0.12, + "learning_rate": 9.659473530790123e-05, + "loss": 2.3416, + "step": 44155 + }, + { + "epoch": 0.12, + "learning_rate": 9.659397120459437e-05, + "loss": 2.2792, + "step": 44160 + }, + { + "epoch": 0.12, + "learning_rate": 9.659320701859189e-05, + "loss": 2.2248, + "step": 44165 + }, + { + "epoch": 0.12, + "learning_rate": 9.659244274989514e-05, + "loss": 2.4612, + "step": 44170 + }, + { + "epoch": 0.12, + "learning_rate": 9.659167839850549e-05, + "loss": 2.3324, + "step": 44175 + }, + { + "epoch": 0.12, + "learning_rate": 9.659091396442425e-05, + "loss": 2.3666, + "step": 44180 + }, + { + "epoch": 0.12, + "learning_rate": 9.659014944765283e-05, + "loss": 2.2361, + "step": 44185 + }, + { + "epoch": 0.12, + "learning_rate": 9.658938484819256e-05, + "loss": 2.3896, + "step": 44190 + }, + { + "epoch": 0.12, + "learning_rate": 9.65886201660448e-05, + "loss": 2.2814, + "step": 44195 + }, + { + "epoch": 0.12, + "learning_rate": 9.65878554012109e-05, + "loss": 2.2187, + "step": 44200 + }, + { + "epoch": 0.12, + "learning_rate": 9.658709055369223e-05, + "loss": 2.1172, + "step": 44205 + }, + { + "epoch": 0.12, + "learning_rate": 9.658632562349015e-05, + "loss": 2.3673, + "step": 44210 + }, + { + "epoch": 0.12, + "learning_rate": 9.658556061060602e-05, + "loss": 2.3373, + "step": 44215 + }, + { + "epoch": 0.12, + "learning_rate": 9.658479551504118e-05, + "loss": 2.3706, + "step": 44220 + }, + { + "epoch": 0.12, + "learning_rate": 9.6584030336797e-05, + "loss": 2.4424, + "step": 44225 + }, + { + "epoch": 0.12, + "learning_rate": 9.658326507587483e-05, + "loss": 2.258, + "step": 44230 + }, + { + "epoch": 0.12, + "learning_rate": 9.658249973227603e-05, + "loss": 2.3596, + "step": 44235 + }, + { + "epoch": 0.12, + "learning_rate": 9.658173430600197e-05, + "loss": 2.4208, + "step": 44240 + }, + { + "epoch": 0.12, + "learning_rate": 9.6580968797054e-05, + "loss": 2.3891, + "step": 44245 + }, + { + "epoch": 0.12, + "learning_rate": 9.658020320543348e-05, + "loss": 2.359, + "step": 44250 + }, + { + "epoch": 0.12, + "learning_rate": 9.657943753114175e-05, + "loss": 2.3391, + "step": 44255 + }, + { + "epoch": 0.12, + "learning_rate": 9.657867177418021e-05, + "loss": 2.2743, + "step": 44260 + }, + { + "epoch": 0.12, + "learning_rate": 9.657790593455019e-05, + "loss": 2.2254, + "step": 44265 + }, + { + "epoch": 0.12, + "learning_rate": 9.657714001225305e-05, + "loss": 2.2084, + "step": 44270 + }, + { + "epoch": 0.12, + "learning_rate": 9.657637400729015e-05, + "loss": 2.2911, + "step": 44275 + }, + { + "epoch": 0.12, + "learning_rate": 9.657560791966285e-05, + "loss": 2.3809, + "step": 44280 + }, + { + "epoch": 0.12, + "learning_rate": 9.657484174937253e-05, + "loss": 2.3914, + "step": 44285 + }, + { + "epoch": 0.12, + "learning_rate": 9.657407549642053e-05, + "loss": 2.3505, + "step": 44290 + }, + { + "epoch": 0.12, + "learning_rate": 9.65733091608082e-05, + "loss": 2.275, + "step": 44295 + }, + { + "epoch": 0.12, + "learning_rate": 9.657254274253693e-05, + "loss": 2.3718, + "step": 44300 + }, + { + "epoch": 0.12, + "learning_rate": 9.657177624160804e-05, + "loss": 2.3796, + "step": 44305 + }, + { + "epoch": 0.12, + "learning_rate": 9.657100965802295e-05, + "loss": 2.2576, + "step": 44310 + }, + { + "epoch": 0.12, + "learning_rate": 9.657024299178296e-05, + "loss": 2.4218, + "step": 44315 + }, + { + "epoch": 0.12, + "learning_rate": 9.656947624288946e-05, + "loss": 2.3936, + "step": 44320 + }, + { + "epoch": 0.12, + "learning_rate": 9.656870941134381e-05, + "loss": 2.2699, + "step": 44325 + }, + { + "epoch": 0.12, + "learning_rate": 9.656794249714736e-05, + "loss": 2.4117, + "step": 44330 + }, + { + "epoch": 0.12, + "learning_rate": 9.656717550030148e-05, + "loss": 2.3678, + "step": 44335 + }, + { + "epoch": 0.12, + "learning_rate": 9.656640842080753e-05, + "loss": 2.3112, + "step": 44340 + }, + { + "epoch": 0.12, + "learning_rate": 9.656564125866688e-05, + "loss": 2.2646, + "step": 44345 + }, + { + "epoch": 0.12, + "learning_rate": 9.656487401388086e-05, + "loss": 2.3235, + "step": 44350 + }, + { + "epoch": 0.12, + "learning_rate": 9.656410668645086e-05, + "loss": 2.4196, + "step": 44355 + }, + { + "epoch": 0.12, + "learning_rate": 9.656333927637825e-05, + "loss": 2.3661, + "step": 44360 + }, + { + "epoch": 0.12, + "learning_rate": 9.656257178366436e-05, + "loss": 2.361, + "step": 44365 + }, + { + "epoch": 0.12, + "learning_rate": 9.656180420831061e-05, + "loss": 2.3512, + "step": 44370 + }, + { + "epoch": 0.12, + "learning_rate": 9.656103655031827e-05, + "loss": 2.3264, + "step": 44375 + }, + { + "epoch": 0.12, + "learning_rate": 9.656026880968878e-05, + "loss": 2.382, + "step": 44380 + }, + { + "epoch": 0.12, + "learning_rate": 9.655950098642348e-05, + "loss": 2.4408, + "step": 44385 + }, + { + "epoch": 0.12, + "learning_rate": 9.655873308052373e-05, + "loss": 2.3863, + "step": 44390 + }, + { + "epoch": 0.12, + "learning_rate": 9.655796509199089e-05, + "loss": 2.5051, + "step": 44395 + }, + { + "epoch": 0.12, + "learning_rate": 9.655719702082633e-05, + "loss": 2.442, + "step": 44400 + }, + { + "epoch": 0.12, + "learning_rate": 9.65564288670314e-05, + "loss": 2.2947, + "step": 44405 + }, + { + "epoch": 0.12, + "learning_rate": 9.655566063060748e-05, + "loss": 2.2378, + "step": 44410 + }, + { + "epoch": 0.12, + "learning_rate": 9.655489231155593e-05, + "loss": 2.3475, + "step": 44415 + }, + { + "epoch": 0.12, + "learning_rate": 9.65541239098781e-05, + "loss": 2.3488, + "step": 44420 + }, + { + "epoch": 0.12, + "learning_rate": 9.655335542557536e-05, + "loss": 2.3475, + "step": 44425 + }, + { + "epoch": 0.12, + "learning_rate": 9.65525868586491e-05, + "loss": 2.3383, + "step": 44430 + }, + { + "epoch": 0.12, + "learning_rate": 9.655181820910064e-05, + "loss": 2.4289, + "step": 44435 + }, + { + "epoch": 0.12, + "learning_rate": 9.655104947693136e-05, + "loss": 2.4331, + "step": 44440 + }, + { + "epoch": 0.12, + "learning_rate": 9.655028066214264e-05, + "loss": 2.3518, + "step": 44445 + }, + { + "epoch": 0.12, + "learning_rate": 9.654951176473584e-05, + "loss": 2.4034, + "step": 44450 + }, + { + "epoch": 0.12, + "learning_rate": 9.654874278471233e-05, + "loss": 2.2439, + "step": 44455 + }, + { + "epoch": 0.12, + "learning_rate": 9.654797372207345e-05, + "loss": 2.4435, + "step": 44460 + }, + { + "epoch": 0.12, + "learning_rate": 9.654720457682057e-05, + "loss": 2.3771, + "step": 44465 + }, + { + "epoch": 0.12, + "learning_rate": 9.654643534895508e-05, + "loss": 2.2329, + "step": 44470 + }, + { + "epoch": 0.12, + "learning_rate": 9.654566603847832e-05, + "loss": 2.2382, + "step": 44475 + }, + { + "epoch": 0.12, + "learning_rate": 9.654489664539167e-05, + "loss": 2.2527, + "step": 44480 + }, + { + "epoch": 0.12, + "learning_rate": 9.654412716969649e-05, + "loss": 2.4272, + "step": 44485 + }, + { + "epoch": 0.12, + "learning_rate": 9.654335761139414e-05, + "loss": 2.4845, + "step": 44490 + }, + { + "epoch": 0.12, + "learning_rate": 9.654258797048599e-05, + "loss": 2.349, + "step": 44495 + }, + { + "epoch": 0.12, + "learning_rate": 9.654181824697342e-05, + "loss": 2.415, + "step": 44500 + }, + { + "epoch": 0.12, + "learning_rate": 9.654104844085777e-05, + "loss": 2.4, + "step": 44505 + }, + { + "epoch": 0.12, + "learning_rate": 9.654027855214043e-05, + "loss": 2.2994, + "step": 44510 + }, + { + "epoch": 0.12, + "learning_rate": 9.653950858082275e-05, + "loss": 2.2965, + "step": 44515 + }, + { + "epoch": 0.12, + "learning_rate": 9.653873852690611e-05, + "loss": 2.4116, + "step": 44520 + }, + { + "epoch": 0.12, + "learning_rate": 9.653796839039187e-05, + "loss": 2.4828, + "step": 44525 + }, + { + "epoch": 0.12, + "learning_rate": 9.653719817128139e-05, + "loss": 2.3723, + "step": 44530 + }, + { + "epoch": 0.12, + "learning_rate": 9.653642786957604e-05, + "loss": 2.3293, + "step": 44535 + }, + { + "epoch": 0.12, + "learning_rate": 9.65356574852772e-05, + "loss": 2.2854, + "step": 44540 + }, + { + "epoch": 0.12, + "learning_rate": 9.653488701838623e-05, + "loss": 2.3099, + "step": 44545 + }, + { + "epoch": 0.12, + "learning_rate": 9.653411646890449e-05, + "loss": 2.3529, + "step": 44550 + }, + { + "epoch": 0.12, + "learning_rate": 9.653334583683334e-05, + "loss": 2.4359, + "step": 44555 + }, + { + "epoch": 0.12, + "learning_rate": 9.653257512217417e-05, + "loss": 2.5481, + "step": 44560 + }, + { + "epoch": 0.12, + "learning_rate": 9.653180432492836e-05, + "loss": 2.3522, + "step": 44565 + }, + { + "epoch": 0.12, + "learning_rate": 9.653103344509722e-05, + "loss": 2.4951, + "step": 44570 + }, + { + "epoch": 0.12, + "learning_rate": 9.653026248268218e-05, + "loss": 2.4038, + "step": 44575 + }, + { + "epoch": 0.12, + "learning_rate": 9.652949143768458e-05, + "loss": 2.2938, + "step": 44580 + }, + { + "epoch": 0.12, + "learning_rate": 9.652872031010579e-05, + "loss": 2.3929, + "step": 44585 + }, + { + "epoch": 0.12, + "learning_rate": 9.652794909994717e-05, + "loss": 2.4029, + "step": 44590 + }, + { + "epoch": 0.12, + "learning_rate": 9.652717780721012e-05, + "loss": 2.3807, + "step": 44595 + }, + { + "epoch": 0.12, + "learning_rate": 9.652640643189598e-05, + "loss": 2.3487, + "step": 44600 + }, + { + "epoch": 0.12, + "learning_rate": 9.652563497400612e-05, + "loss": 2.2793, + "step": 44605 + }, + { + "epoch": 0.12, + "learning_rate": 9.652486343354192e-05, + "loss": 2.3069, + "step": 44610 + }, + { + "epoch": 0.12, + "learning_rate": 9.652409181050474e-05, + "loss": 2.2613, + "step": 44615 + }, + { + "epoch": 0.12, + "learning_rate": 9.652332010489597e-05, + "loss": 2.4113, + "step": 44620 + }, + { + "epoch": 0.12, + "learning_rate": 9.652254831671695e-05, + "loss": 2.306, + "step": 44625 + }, + { + "epoch": 0.12, + "learning_rate": 9.652177644596908e-05, + "loss": 2.2678, + "step": 44630 + }, + { + "epoch": 0.12, + "learning_rate": 9.652100449265369e-05, + "loss": 2.3803, + "step": 44635 + }, + { + "epoch": 0.12, + "learning_rate": 9.652023245677219e-05, + "loss": 2.2212, + "step": 44640 + }, + { + "epoch": 0.12, + "learning_rate": 9.651946033832593e-05, + "loss": 2.4066, + "step": 44645 + }, + { + "epoch": 0.12, + "learning_rate": 9.651868813731628e-05, + "loss": 2.3952, + "step": 44650 + }, + { + "epoch": 0.12, + "learning_rate": 9.651791585374462e-05, + "loss": 2.2486, + "step": 44655 + }, + { + "epoch": 0.12, + "learning_rate": 9.651714348761232e-05, + "loss": 2.2735, + "step": 44660 + }, + { + "epoch": 0.12, + "learning_rate": 9.651637103892075e-05, + "loss": 2.3204, + "step": 44665 + }, + { + "epoch": 0.12, + "learning_rate": 9.651559850767127e-05, + "loss": 2.3677, + "step": 44670 + }, + { + "epoch": 0.12, + "learning_rate": 9.651482589386526e-05, + "loss": 2.3563, + "step": 44675 + }, + { + "epoch": 0.12, + "learning_rate": 9.651405319750409e-05, + "loss": 2.2404, + "step": 44680 + }, + { + "epoch": 0.12, + "learning_rate": 9.651328041858914e-05, + "loss": 2.2943, + "step": 44685 + }, + { + "epoch": 0.12, + "learning_rate": 9.651250755712175e-05, + "loss": 2.3605, + "step": 44690 + }, + { + "epoch": 0.12, + "learning_rate": 9.651173461310334e-05, + "loss": 2.4711, + "step": 44695 + }, + { + "epoch": 0.12, + "learning_rate": 9.651096158653525e-05, + "loss": 2.3309, + "step": 44700 + }, + { + "epoch": 0.12, + "learning_rate": 9.651018847741886e-05, + "loss": 2.3646, + "step": 44705 + }, + { + "epoch": 0.12, + "learning_rate": 9.650941528575554e-05, + "loss": 2.4088, + "step": 44710 + }, + { + "epoch": 0.12, + "learning_rate": 9.650864201154666e-05, + "loss": 2.4034, + "step": 44715 + }, + { + "epoch": 0.12, + "learning_rate": 9.650786865479359e-05, + "loss": 2.3878, + "step": 44720 + }, + { + "epoch": 0.12, + "learning_rate": 9.650709521549772e-05, + "loss": 2.3304, + "step": 44725 + }, + { + "epoch": 0.12, + "learning_rate": 9.650632169366041e-05, + "loss": 2.3204, + "step": 44730 + }, + { + "epoch": 0.12, + "learning_rate": 9.650554808928302e-05, + "loss": 2.4118, + "step": 44735 + }, + { + "epoch": 0.12, + "learning_rate": 9.650477440236695e-05, + "loss": 2.3401, + "step": 44740 + }, + { + "epoch": 0.12, + "learning_rate": 9.650400063291354e-05, + "loss": 2.4329, + "step": 44745 + }, + { + "epoch": 0.12, + "learning_rate": 9.65032267809242e-05, + "loss": 2.4068, + "step": 44750 + }, + { + "epoch": 0.12, + "learning_rate": 9.650245284640029e-05, + "loss": 2.3038, + "step": 44755 + }, + { + "epoch": 0.12, + "learning_rate": 9.650167882934318e-05, + "loss": 2.4328, + "step": 44760 + }, + { + "epoch": 0.12, + "learning_rate": 9.650090472975424e-05, + "loss": 2.3073, + "step": 44765 + }, + { + "epoch": 0.12, + "learning_rate": 9.650013054763485e-05, + "loss": 2.2355, + "step": 44770 + }, + { + "epoch": 0.12, + "learning_rate": 9.649935628298636e-05, + "loss": 2.3568, + "step": 44775 + }, + { + "epoch": 0.12, + "learning_rate": 9.649858193581019e-05, + "loss": 2.231, + "step": 44780 + }, + { + "epoch": 0.12, + "learning_rate": 9.649780750610768e-05, + "loss": 2.3397, + "step": 44785 + }, + { + "epoch": 0.12, + "learning_rate": 9.649703299388022e-05, + "loss": 2.3234, + "step": 44790 + }, + { + "epoch": 0.12, + "learning_rate": 9.649625839912918e-05, + "loss": 2.3656, + "step": 44795 + }, + { + "epoch": 0.12, + "learning_rate": 9.649548372185592e-05, + "loss": 2.3101, + "step": 44800 + }, + { + "epoch": 0.12, + "learning_rate": 9.649470896206186e-05, + "loss": 2.3461, + "step": 44805 + }, + { + "epoch": 0.12, + "learning_rate": 9.649393411974832e-05, + "loss": 2.2635, + "step": 44810 + }, + { + "epoch": 0.12, + "learning_rate": 9.64931591949167e-05, + "loss": 2.3926, + "step": 44815 + }, + { + "epoch": 0.12, + "learning_rate": 9.649238418756838e-05, + "loss": 2.2665, + "step": 44820 + }, + { + "epoch": 0.12, + "learning_rate": 9.649160909770474e-05, + "loss": 2.3476, + "step": 44825 + }, + { + "epoch": 0.12, + "learning_rate": 9.649083392532711e-05, + "loss": 2.3951, + "step": 44830 + }, + { + "epoch": 0.12, + "learning_rate": 9.649005867043695e-05, + "loss": 2.2819, + "step": 44835 + }, + { + "epoch": 0.12, + "learning_rate": 9.648928333303554e-05, + "loss": 2.3227, + "step": 44840 + }, + { + "epoch": 0.12, + "learning_rate": 9.648850791312433e-05, + "loss": 2.4404, + "step": 44845 + }, + { + "epoch": 0.12, + "learning_rate": 9.648773241070468e-05, + "loss": 2.293, + "step": 44850 + }, + { + "epoch": 0.12, + "learning_rate": 9.648695682577794e-05, + "loss": 2.332, + "step": 44855 + }, + { + "epoch": 0.12, + "learning_rate": 9.648618115834551e-05, + "loss": 2.3157, + "step": 44860 + }, + { + "epoch": 0.12, + "learning_rate": 9.648540540840876e-05, + "loss": 2.2614, + "step": 44865 + }, + { + "epoch": 0.12, + "learning_rate": 9.648462957596904e-05, + "loss": 2.4069, + "step": 44870 + }, + { + "epoch": 0.12, + "learning_rate": 9.648385366102779e-05, + "loss": 2.3548, + "step": 44875 + }, + { + "epoch": 0.12, + "learning_rate": 9.648307766358633e-05, + "loss": 2.437, + "step": 44880 + }, + { + "epoch": 0.12, + "learning_rate": 9.648230158364607e-05, + "loss": 2.3883, + "step": 44885 + }, + { + "epoch": 0.12, + "learning_rate": 9.648152542120836e-05, + "loss": 2.3502, + "step": 44890 + }, + { + "epoch": 0.12, + "learning_rate": 9.648074917627459e-05, + "loss": 2.3976, + "step": 44895 + }, + { + "epoch": 0.12, + "learning_rate": 9.647997284884615e-05, + "loss": 2.4517, + "step": 44900 + }, + { + "epoch": 0.12, + "learning_rate": 9.647919643892441e-05, + "loss": 2.3693, + "step": 44905 + }, + { + "epoch": 0.12, + "learning_rate": 9.647841994651075e-05, + "loss": 2.4205, + "step": 44910 + }, + { + "epoch": 0.12, + "learning_rate": 9.647764337160652e-05, + "loss": 2.4366, + "step": 44915 + }, + { + "epoch": 0.12, + "learning_rate": 9.647686671421314e-05, + "loss": 2.3437, + "step": 44920 + }, + { + "epoch": 0.12, + "learning_rate": 9.647608997433197e-05, + "loss": 2.4229, + "step": 44925 + }, + { + "epoch": 0.12, + "learning_rate": 9.647531315196439e-05, + "loss": 2.3507, + "step": 44930 + }, + { + "epoch": 0.12, + "learning_rate": 9.647453624711178e-05, + "loss": 2.2656, + "step": 44935 + }, + { + "epoch": 0.12, + "learning_rate": 9.64737592597755e-05, + "loss": 2.4038, + "step": 44940 + }, + { + "epoch": 0.12, + "learning_rate": 9.647298218995698e-05, + "loss": 2.4372, + "step": 44945 + }, + { + "epoch": 0.12, + "learning_rate": 9.647220503765754e-05, + "loss": 2.311, + "step": 44950 + }, + { + "epoch": 0.12, + "learning_rate": 9.647142780287859e-05, + "loss": 2.3091, + "step": 44955 + }, + { + "epoch": 0.12, + "learning_rate": 9.647065048562149e-05, + "loss": 2.3564, + "step": 44960 + }, + { + "epoch": 0.12, + "learning_rate": 9.646987308588765e-05, + "loss": 2.3771, + "step": 44965 + }, + { + "epoch": 0.12, + "learning_rate": 9.646909560367843e-05, + "loss": 2.3095, + "step": 44970 + }, + { + "epoch": 0.12, + "learning_rate": 9.64683180389952e-05, + "loss": 2.2844, + "step": 44975 + }, + { + "epoch": 0.12, + "learning_rate": 9.646754039183939e-05, + "loss": 2.2961, + "step": 44980 + }, + { + "epoch": 0.12, + "learning_rate": 9.646676266221231e-05, + "loss": 2.4019, + "step": 44985 + }, + { + "epoch": 0.12, + "learning_rate": 9.646598485011538e-05, + "loss": 2.3301, + "step": 44990 + }, + { + "epoch": 0.12, + "learning_rate": 9.646520695554997e-05, + "loss": 2.2654, + "step": 44995 + }, + { + "epoch": 0.12, + "learning_rate": 9.646442897851748e-05, + "loss": 2.3446, + "step": 45000 + }, + { + "epoch": 0.12, + "learning_rate": 9.646365091901926e-05, + "loss": 2.507, + "step": 45005 + }, + { + "epoch": 0.12, + "learning_rate": 9.646287277705671e-05, + "loss": 2.371, + "step": 45010 + }, + { + "epoch": 0.12, + "learning_rate": 9.64620945526312e-05, + "loss": 2.3125, + "step": 45015 + }, + { + "epoch": 0.12, + "learning_rate": 9.646131624574412e-05, + "loss": 2.4115, + "step": 45020 + }, + { + "epoch": 0.12, + "learning_rate": 9.646053785639687e-05, + "loss": 2.3112, + "step": 45025 + }, + { + "epoch": 0.12, + "learning_rate": 9.645975938459079e-05, + "loss": 2.3009, + "step": 45030 + }, + { + "epoch": 0.12, + "learning_rate": 9.64589808303273e-05, + "loss": 2.257, + "step": 45035 + }, + { + "epoch": 0.12, + "learning_rate": 9.645820219360775e-05, + "loss": 2.4497, + "step": 45040 + }, + { + "epoch": 0.12, + "learning_rate": 9.645742347443355e-05, + "loss": 2.3545, + "step": 45045 + }, + { + "epoch": 0.12, + "learning_rate": 9.645664467280606e-05, + "loss": 2.3157, + "step": 45050 + }, + { + "epoch": 0.12, + "learning_rate": 9.645586578872668e-05, + "loss": 2.3306, + "step": 45055 + }, + { + "epoch": 0.12, + "learning_rate": 9.645508682219677e-05, + "loss": 2.4572, + "step": 45060 + }, + { + "epoch": 0.12, + "learning_rate": 9.645430777321773e-05, + "loss": 2.2377, + "step": 45065 + }, + { + "epoch": 0.12, + "learning_rate": 9.645352864179095e-05, + "loss": 2.4364, + "step": 45070 + }, + { + "epoch": 0.12, + "learning_rate": 9.645274942791779e-05, + "loss": 2.4186, + "step": 45075 + }, + { + "epoch": 0.12, + "learning_rate": 9.645197013159965e-05, + "loss": 2.4116, + "step": 45080 + }, + { + "epoch": 0.12, + "learning_rate": 9.645119075283791e-05, + "loss": 2.3959, + "step": 45085 + }, + { + "epoch": 0.12, + "learning_rate": 9.645041129163393e-05, + "loss": 2.4242, + "step": 45090 + }, + { + "epoch": 0.12, + "learning_rate": 9.644963174798914e-05, + "loss": 2.2781, + "step": 45095 + }, + { + "epoch": 0.12, + "learning_rate": 9.644885212190489e-05, + "loss": 2.3298, + "step": 45100 + }, + { + "epoch": 0.12, + "learning_rate": 9.644807241338256e-05, + "loss": 2.3, + "step": 45105 + }, + { + "epoch": 0.12, + "learning_rate": 9.644729262242356e-05, + "loss": 2.3129, + "step": 45110 + }, + { + "epoch": 0.12, + "learning_rate": 9.644651274902925e-05, + "loss": 2.2896, + "step": 45115 + }, + { + "epoch": 0.12, + "learning_rate": 9.644573279320101e-05, + "loss": 2.407, + "step": 45120 + }, + { + "epoch": 0.12, + "learning_rate": 9.644495275494026e-05, + "loss": 2.3913, + "step": 45125 + }, + { + "epoch": 0.12, + "learning_rate": 9.644417263424837e-05, + "loss": 2.3098, + "step": 45130 + }, + { + "epoch": 0.12, + "learning_rate": 9.64433924311267e-05, + "loss": 2.3844, + "step": 45135 + }, + { + "epoch": 0.12, + "learning_rate": 9.644261214557664e-05, + "loss": 2.2384, + "step": 45140 + }, + { + "epoch": 0.12, + "learning_rate": 9.64418317775996e-05, + "loss": 2.4104, + "step": 45145 + }, + { + "epoch": 0.12, + "learning_rate": 9.644105132719696e-05, + "loss": 2.2568, + "step": 45150 + }, + { + "epoch": 0.12, + "learning_rate": 9.644027079437009e-05, + "loss": 2.3216, + "step": 45155 + }, + { + "epoch": 0.12, + "learning_rate": 9.643949017912037e-05, + "loss": 2.2977, + "step": 45160 + }, + { + "epoch": 0.12, + "learning_rate": 9.64387094814492e-05, + "loss": 2.3564, + "step": 45165 + }, + { + "epoch": 0.12, + "learning_rate": 9.643792870135797e-05, + "loss": 2.3612, + "step": 45170 + }, + { + "epoch": 0.12, + "learning_rate": 9.643714783884807e-05, + "loss": 2.3105, + "step": 45175 + }, + { + "epoch": 0.12, + "learning_rate": 9.643636689392085e-05, + "loss": 2.3207, + "step": 45180 + }, + { + "epoch": 0.12, + "learning_rate": 9.643558586657774e-05, + "loss": 2.2485, + "step": 45185 + }, + { + "epoch": 0.12, + "learning_rate": 9.64348047568201e-05, + "loss": 2.2607, + "step": 45190 + }, + { + "epoch": 0.12, + "learning_rate": 9.643402356464933e-05, + "loss": 2.3872, + "step": 45195 + }, + { + "epoch": 0.12, + "learning_rate": 9.643324229006678e-05, + "loss": 2.3502, + "step": 45200 + }, + { + "epoch": 0.12, + "learning_rate": 9.64324609330739e-05, + "loss": 2.4472, + "step": 45205 + }, + { + "epoch": 0.12, + "learning_rate": 9.643167949367202e-05, + "loss": 2.3743, + "step": 45210 + }, + { + "epoch": 0.12, + "learning_rate": 9.643089797186255e-05, + "loss": 2.4308, + "step": 45215 + }, + { + "epoch": 0.12, + "learning_rate": 9.643011636764689e-05, + "loss": 2.3952, + "step": 45220 + }, + { + "epoch": 0.12, + "learning_rate": 9.642933468102642e-05, + "loss": 2.2913, + "step": 45225 + }, + { + "epoch": 0.12, + "learning_rate": 9.64285529120025e-05, + "loss": 2.3107, + "step": 45230 + }, + { + "epoch": 0.12, + "learning_rate": 9.642777106057656e-05, + "loss": 2.3476, + "step": 45235 + }, + { + "epoch": 0.12, + "learning_rate": 9.642698912674995e-05, + "loss": 2.3845, + "step": 45240 + }, + { + "epoch": 0.12, + "learning_rate": 9.642620711052409e-05, + "loss": 2.309, + "step": 45245 + }, + { + "epoch": 0.12, + "learning_rate": 9.642542501190033e-05, + "loss": 2.3025, + "step": 45250 + }, + { + "epoch": 0.12, + "learning_rate": 9.642464283088009e-05, + "loss": 2.4153, + "step": 45255 + }, + { + "epoch": 0.12, + "learning_rate": 9.642386056746475e-05, + "loss": 2.3795, + "step": 45260 + }, + { + "epoch": 0.12, + "learning_rate": 9.642307822165571e-05, + "loss": 2.4601, + "step": 45265 + }, + { + "epoch": 0.12, + "learning_rate": 9.642229579345433e-05, + "loss": 2.317, + "step": 45270 + }, + { + "epoch": 0.12, + "learning_rate": 9.642151328286202e-05, + "loss": 2.4719, + "step": 45275 + }, + { + "epoch": 0.12, + "learning_rate": 9.642073068988016e-05, + "loss": 2.5272, + "step": 45280 + }, + { + "epoch": 0.12, + "learning_rate": 9.641994801451013e-05, + "loss": 2.1936, + "step": 45285 + }, + { + "epoch": 0.12, + "learning_rate": 9.641916525675334e-05, + "loss": 2.4446, + "step": 45290 + }, + { + "epoch": 0.12, + "learning_rate": 9.641838241661118e-05, + "loss": 2.4288, + "step": 45295 + }, + { + "epoch": 0.12, + "learning_rate": 9.641759949408503e-05, + "loss": 2.3797, + "step": 45300 + }, + { + "epoch": 0.12, + "learning_rate": 9.641681648917626e-05, + "loss": 2.3875, + "step": 45305 + }, + { + "epoch": 0.12, + "learning_rate": 9.641603340188627e-05, + "loss": 2.3826, + "step": 45310 + }, + { + "epoch": 0.12, + "learning_rate": 9.641525023221649e-05, + "loss": 2.456, + "step": 45315 + }, + { + "epoch": 0.12, + "learning_rate": 9.641446698016827e-05, + "loss": 2.3191, + "step": 45320 + }, + { + "epoch": 0.12, + "learning_rate": 9.641368364574299e-05, + "loss": 2.3362, + "step": 45325 + }, + { + "epoch": 0.12, + "learning_rate": 9.641290022894209e-05, + "loss": 2.413, + "step": 45330 + }, + { + "epoch": 0.12, + "learning_rate": 9.641211672976691e-05, + "loss": 2.2565, + "step": 45335 + }, + { + "epoch": 0.12, + "learning_rate": 9.641133314821886e-05, + "loss": 2.2928, + "step": 45340 + }, + { + "epoch": 0.12, + "learning_rate": 9.641054948429932e-05, + "loss": 2.3289, + "step": 45345 + }, + { + "epoch": 0.12, + "learning_rate": 9.640976573800971e-05, + "loss": 2.3093, + "step": 45350 + }, + { + "epoch": 0.12, + "learning_rate": 9.64089819093514e-05, + "loss": 2.3365, + "step": 45355 + }, + { + "epoch": 0.12, + "learning_rate": 9.640819799832577e-05, + "loss": 2.2543, + "step": 45360 + }, + { + "epoch": 0.12, + "learning_rate": 9.640741400493424e-05, + "loss": 2.2752, + "step": 45365 + }, + { + "epoch": 0.12, + "learning_rate": 9.64066299291782e-05, + "loss": 2.2855, + "step": 45370 + }, + { + "epoch": 0.12, + "learning_rate": 9.6405845771059e-05, + "loss": 2.4043, + "step": 45375 + }, + { + "epoch": 0.12, + "learning_rate": 9.640506153057807e-05, + "loss": 2.3604, + "step": 45380 + }, + { + "epoch": 0.12, + "learning_rate": 9.640427720773678e-05, + "loss": 2.4204, + "step": 45385 + }, + { + "epoch": 0.12, + "learning_rate": 9.640349280253655e-05, + "loss": 2.3452, + "step": 45390 + }, + { + "epoch": 0.12, + "learning_rate": 9.640270831497875e-05, + "loss": 2.3325, + "step": 45395 + }, + { + "epoch": 0.12, + "learning_rate": 9.640192374506478e-05, + "loss": 2.4446, + "step": 45400 + }, + { + "epoch": 0.12, + "learning_rate": 9.640113909279602e-05, + "loss": 2.2893, + "step": 45405 + }, + { + "epoch": 0.12, + "learning_rate": 9.640035435817388e-05, + "loss": 2.4057, + "step": 45410 + }, + { + "epoch": 0.12, + "learning_rate": 9.639956954119975e-05, + "loss": 2.2546, + "step": 45415 + }, + { + "epoch": 0.12, + "learning_rate": 9.639878464187501e-05, + "loss": 2.4091, + "step": 45420 + }, + { + "epoch": 0.12, + "learning_rate": 9.639799966020107e-05, + "loss": 2.4299, + "step": 45425 + }, + { + "epoch": 0.12, + "learning_rate": 9.639721459617932e-05, + "loss": 2.2499, + "step": 45430 + }, + { + "epoch": 0.12, + "learning_rate": 9.639642944981114e-05, + "loss": 2.3115, + "step": 45435 + }, + { + "epoch": 0.12, + "learning_rate": 9.639564422109794e-05, + "loss": 2.3881, + "step": 45440 + }, + { + "epoch": 0.12, + "learning_rate": 9.63948589100411e-05, + "loss": 2.3338, + "step": 45445 + }, + { + "epoch": 0.12, + "learning_rate": 9.639407351664202e-05, + "loss": 2.4027, + "step": 45450 + }, + { + "epoch": 0.12, + "learning_rate": 9.639328804090208e-05, + "loss": 2.3187, + "step": 45455 + }, + { + "epoch": 0.12, + "learning_rate": 9.63925024828227e-05, + "loss": 2.2378, + "step": 45460 + }, + { + "epoch": 0.12, + "learning_rate": 9.639171684240527e-05, + "loss": 2.3066, + "step": 45465 + }, + { + "epoch": 0.12, + "learning_rate": 9.639093111965117e-05, + "loss": 2.4003, + "step": 45470 + }, + { + "epoch": 0.12, + "learning_rate": 9.639014531456179e-05, + "loss": 2.4893, + "step": 45475 + }, + { + "epoch": 0.12, + "learning_rate": 9.638935942713854e-05, + "loss": 2.2988, + "step": 45480 + }, + { + "epoch": 0.12, + "learning_rate": 9.638857345738282e-05, + "loss": 2.257, + "step": 45485 + }, + { + "epoch": 0.12, + "learning_rate": 9.638778740529599e-05, + "loss": 2.3926, + "step": 45490 + }, + { + "epoch": 0.12, + "learning_rate": 9.638700127087949e-05, + "loss": 2.342, + "step": 45495 + }, + { + "epoch": 0.12, + "learning_rate": 9.638621505413469e-05, + "loss": 2.5016, + "step": 45500 + }, + { + "epoch": 0.12, + "learning_rate": 9.6385428755063e-05, + "loss": 2.3365, + "step": 45505 + }, + { + "epoch": 0.12, + "learning_rate": 9.638464237366579e-05, + "loss": 2.329, + "step": 45510 + }, + { + "epoch": 0.12, + "learning_rate": 9.638385590994447e-05, + "loss": 2.33, + "step": 45515 + }, + { + "epoch": 0.12, + "learning_rate": 9.638306936390045e-05, + "loss": 2.3784, + "step": 45520 + }, + { + "epoch": 0.12, + "learning_rate": 9.638228273553511e-05, + "loss": 2.2974, + "step": 45525 + }, + { + "epoch": 0.12, + "learning_rate": 9.638149602484985e-05, + "loss": 2.3789, + "step": 45530 + }, + { + "epoch": 0.12, + "learning_rate": 9.638070923184607e-05, + "loss": 2.2802, + "step": 45535 + }, + { + "epoch": 0.12, + "learning_rate": 9.637992235652515e-05, + "loss": 2.3811, + "step": 45540 + }, + { + "epoch": 0.12, + "learning_rate": 9.637913539888852e-05, + "loss": 2.2526, + "step": 45545 + }, + { + "epoch": 0.12, + "learning_rate": 9.637834835893753e-05, + "loss": 2.4455, + "step": 45550 + }, + { + "epoch": 0.12, + "learning_rate": 9.637756123667363e-05, + "loss": 2.237, + "step": 45555 + }, + { + "epoch": 0.12, + "learning_rate": 9.637677403209817e-05, + "loss": 2.4062, + "step": 45560 + }, + { + "epoch": 0.12, + "learning_rate": 9.637598674521258e-05, + "loss": 2.2907, + "step": 45565 + }, + { + "epoch": 0.12, + "learning_rate": 9.637519937601822e-05, + "loss": 2.3615, + "step": 45570 + }, + { + "epoch": 0.12, + "learning_rate": 9.637441192451653e-05, + "loss": 2.4264, + "step": 45575 + }, + { + "epoch": 0.12, + "learning_rate": 9.637362439070889e-05, + "loss": 2.3642, + "step": 45580 + }, + { + "epoch": 0.12, + "learning_rate": 9.637283677459668e-05, + "loss": 2.3729, + "step": 45585 + }, + { + "epoch": 0.12, + "learning_rate": 9.637204907618134e-05, + "loss": 2.3341, + "step": 45590 + }, + { + "epoch": 0.12, + "learning_rate": 9.637126129546423e-05, + "loss": 2.3296, + "step": 45595 + }, + { + "epoch": 0.12, + "learning_rate": 9.637047343244676e-05, + "loss": 2.4035, + "step": 45600 + }, + { + "epoch": 0.12, + "learning_rate": 9.636968548713033e-05, + "loss": 2.204, + "step": 45605 + }, + { + "epoch": 0.12, + "learning_rate": 9.636889745951633e-05, + "loss": 2.3426, + "step": 45610 + }, + { + "epoch": 0.12, + "learning_rate": 9.636810934960618e-05, + "loss": 2.2807, + "step": 45615 + }, + { + "epoch": 0.12, + "learning_rate": 9.636732115740124e-05, + "loss": 2.3303, + "step": 45620 + }, + { + "epoch": 0.12, + "learning_rate": 9.636653288290296e-05, + "loss": 2.4258, + "step": 45625 + }, + { + "epoch": 0.12, + "learning_rate": 9.63657445261127e-05, + "loss": 2.3809, + "step": 45630 + }, + { + "epoch": 0.12, + "learning_rate": 9.636495608703188e-05, + "loss": 2.2567, + "step": 45635 + }, + { + "epoch": 0.12, + "learning_rate": 9.63641675656619e-05, + "loss": 2.2627, + "step": 45640 + }, + { + "epoch": 0.12, + "learning_rate": 9.636337896200411e-05, + "loss": 2.3435, + "step": 45645 + }, + { + "epoch": 0.12, + "learning_rate": 9.636259027605999e-05, + "loss": 2.465, + "step": 45650 + }, + { + "epoch": 0.12, + "learning_rate": 9.636180150783088e-05, + "loss": 2.3882, + "step": 45655 + }, + { + "epoch": 0.12, + "learning_rate": 9.636101265731821e-05, + "loss": 2.2413, + "step": 45660 + }, + { + "epoch": 0.12, + "learning_rate": 9.636022372452337e-05, + "loss": 2.3534, + "step": 45665 + }, + { + "epoch": 0.12, + "learning_rate": 9.635943470944775e-05, + "loss": 2.3433, + "step": 45670 + }, + { + "epoch": 0.12, + "learning_rate": 9.635864561209277e-05, + "loss": 2.3021, + "step": 45675 + }, + { + "epoch": 0.12, + "learning_rate": 9.635785643245982e-05, + "loss": 2.4815, + "step": 45680 + }, + { + "epoch": 0.12, + "learning_rate": 9.63570671705503e-05, + "loss": 2.4016, + "step": 45685 + }, + { + "epoch": 0.12, + "learning_rate": 9.635627782636563e-05, + "loss": 2.3602, + "step": 45690 + }, + { + "epoch": 0.12, + "learning_rate": 9.635548839990717e-05, + "loss": 2.3216, + "step": 45695 + }, + { + "epoch": 0.12, + "learning_rate": 9.635469889117634e-05, + "loss": 2.3978, + "step": 45700 + }, + { + "epoch": 0.12, + "learning_rate": 9.635390930017456e-05, + "loss": 2.378, + "step": 45705 + }, + { + "epoch": 0.12, + "learning_rate": 9.635311962690322e-05, + "loss": 2.4407, + "step": 45710 + }, + { + "epoch": 0.12, + "learning_rate": 9.63523298713637e-05, + "loss": 2.4059, + "step": 45715 + }, + { + "epoch": 0.12, + "learning_rate": 9.635154003355745e-05, + "loss": 2.5078, + "step": 45720 + }, + { + "epoch": 0.12, + "learning_rate": 9.635075011348582e-05, + "loss": 2.272, + "step": 45725 + }, + { + "epoch": 0.12, + "learning_rate": 9.634996011115025e-05, + "loss": 2.3801, + "step": 45730 + }, + { + "epoch": 0.12, + "learning_rate": 9.634917002655212e-05, + "loss": 2.2978, + "step": 45735 + }, + { + "epoch": 0.12, + "learning_rate": 9.634837985969283e-05, + "loss": 2.2176, + "step": 45740 + }, + { + "epoch": 0.12, + "learning_rate": 9.63475896105738e-05, + "loss": 2.3807, + "step": 45745 + }, + { + "epoch": 0.12, + "learning_rate": 9.634679927919642e-05, + "loss": 2.3366, + "step": 45750 + }, + { + "epoch": 0.12, + "learning_rate": 9.634600886556208e-05, + "loss": 2.3989, + "step": 45755 + }, + { + "epoch": 0.12, + "learning_rate": 9.634521836967223e-05, + "loss": 2.3084, + "step": 45760 + }, + { + "epoch": 0.12, + "learning_rate": 9.634442779152822e-05, + "loss": 2.3672, + "step": 45765 + }, + { + "epoch": 0.12, + "learning_rate": 9.634363713113149e-05, + "loss": 2.3924, + "step": 45770 + }, + { + "epoch": 0.12, + "learning_rate": 9.634284638848342e-05, + "loss": 2.2511, + "step": 45775 + }, + { + "epoch": 0.12, + "learning_rate": 9.634205556358542e-05, + "loss": 2.4471, + "step": 45780 + }, + { + "epoch": 0.12, + "learning_rate": 9.63412646564389e-05, + "loss": 2.357, + "step": 45785 + }, + { + "epoch": 0.12, + "learning_rate": 9.634047366704526e-05, + "loss": 2.4154, + "step": 45790 + }, + { + "epoch": 0.12, + "learning_rate": 9.63396825954059e-05, + "loss": 2.3636, + "step": 45795 + }, + { + "epoch": 0.12, + "learning_rate": 9.633889144152223e-05, + "loss": 2.2685, + "step": 45800 + }, + { + "epoch": 0.12, + "learning_rate": 9.633810020539564e-05, + "loss": 2.4151, + "step": 45805 + }, + { + "epoch": 0.12, + "learning_rate": 9.633730888702755e-05, + "loss": 2.4156, + "step": 45810 + }, + { + "epoch": 0.12, + "learning_rate": 9.633651748641936e-05, + "loss": 2.3199, + "step": 45815 + }, + { + "epoch": 0.12, + "learning_rate": 9.633572600357249e-05, + "loss": 2.364, + "step": 45820 + }, + { + "epoch": 0.12, + "learning_rate": 9.633493443848831e-05, + "loss": 2.4262, + "step": 45825 + }, + { + "epoch": 0.12, + "learning_rate": 9.633414279116826e-05, + "loss": 2.3707, + "step": 45830 + }, + { + "epoch": 0.12, + "learning_rate": 9.633335106161371e-05, + "loss": 2.3208, + "step": 45835 + }, + { + "epoch": 0.12, + "learning_rate": 9.633255924982611e-05, + "loss": 2.3857, + "step": 45840 + }, + { + "epoch": 0.12, + "learning_rate": 9.633176735580682e-05, + "loss": 2.2787, + "step": 45845 + }, + { + "epoch": 0.12, + "learning_rate": 9.633097537955727e-05, + "loss": 2.434, + "step": 45850 + }, + { + "epoch": 0.12, + "learning_rate": 9.633018332107887e-05, + "loss": 2.4041, + "step": 45855 + }, + { + "epoch": 0.12, + "learning_rate": 9.632939118037298e-05, + "loss": 2.3955, + "step": 45860 + }, + { + "epoch": 0.12, + "learning_rate": 9.632859895744109e-05, + "loss": 2.2931, + "step": 45865 + }, + { + "epoch": 0.12, + "learning_rate": 9.632780665228453e-05, + "loss": 2.5058, + "step": 45870 + }, + { + "epoch": 0.12, + "learning_rate": 9.632701426490474e-05, + "loss": 2.4346, + "step": 45875 + }, + { + "epoch": 0.12, + "learning_rate": 9.632622179530312e-05, + "loss": 2.3635, + "step": 45880 + }, + { + "epoch": 0.12, + "learning_rate": 9.632542924348107e-05, + "loss": 2.4274, + "step": 45885 + }, + { + "epoch": 0.12, + "learning_rate": 9.632463660944002e-05, + "loss": 2.3578, + "step": 45890 + }, + { + "epoch": 0.12, + "learning_rate": 9.632384389318136e-05, + "loss": 2.2914, + "step": 45895 + }, + { + "epoch": 0.12, + "learning_rate": 9.63230510947065e-05, + "loss": 2.2482, + "step": 45900 + }, + { + "epoch": 0.12, + "learning_rate": 9.632225821401684e-05, + "loss": 2.3332, + "step": 45905 + }, + { + "epoch": 0.12, + "learning_rate": 9.632146525111379e-05, + "loss": 2.2775, + "step": 45910 + }, + { + "epoch": 0.12, + "learning_rate": 9.632067220599875e-05, + "loss": 2.4171, + "step": 45915 + }, + { + "epoch": 0.12, + "learning_rate": 9.631987907867315e-05, + "loss": 2.3344, + "step": 45920 + }, + { + "epoch": 0.12, + "learning_rate": 9.631908586913838e-05, + "loss": 2.3756, + "step": 45925 + }, + { + "epoch": 0.12, + "learning_rate": 9.631829257739585e-05, + "loss": 2.4086, + "step": 45930 + }, + { + "epoch": 0.12, + "learning_rate": 9.631749920344696e-05, + "loss": 2.3944, + "step": 45935 + }, + { + "epoch": 0.12, + "learning_rate": 9.631670574729314e-05, + "loss": 2.2662, + "step": 45940 + }, + { + "epoch": 0.12, + "learning_rate": 9.63159122089358e-05, + "loss": 2.4066, + "step": 45945 + }, + { + "epoch": 0.12, + "learning_rate": 9.631511858837632e-05, + "loss": 2.4026, + "step": 45950 + }, + { + "epoch": 0.12, + "learning_rate": 9.631432488561613e-05, + "loss": 2.4191, + "step": 45955 + }, + { + "epoch": 0.12, + "learning_rate": 9.631353110065661e-05, + "loss": 2.4222, + "step": 45960 + }, + { + "epoch": 0.12, + "learning_rate": 9.631273723349923e-05, + "loss": 2.2984, + "step": 45965 + }, + { + "epoch": 0.12, + "learning_rate": 9.631194328414534e-05, + "loss": 2.3167, + "step": 45970 + }, + { + "epoch": 0.12, + "learning_rate": 9.631114925259637e-05, + "loss": 2.2379, + "step": 45975 + }, + { + "epoch": 0.12, + "learning_rate": 9.631035513885371e-05, + "loss": 2.3424, + "step": 45980 + }, + { + "epoch": 0.12, + "learning_rate": 9.630956094291881e-05, + "loss": 2.3416, + "step": 45985 + }, + { + "epoch": 0.12, + "learning_rate": 9.630876666479305e-05, + "loss": 2.4365, + "step": 45990 + }, + { + "epoch": 0.12, + "learning_rate": 9.630797230447785e-05, + "loss": 2.3327, + "step": 45995 + }, + { + "epoch": 0.12, + "learning_rate": 9.630717786197462e-05, + "loss": 2.3288, + "step": 46000 + }, + { + "epoch": 0.12, + "learning_rate": 9.630638333728476e-05, + "loss": 2.319, + "step": 46005 + }, + { + "epoch": 0.12, + "learning_rate": 9.630558873040969e-05, + "loss": 2.3093, + "step": 46010 + }, + { + "epoch": 0.12, + "learning_rate": 9.630479404135082e-05, + "loss": 2.4213, + "step": 46015 + }, + { + "epoch": 0.12, + "learning_rate": 9.630399927010956e-05, + "loss": 2.2131, + "step": 46020 + }, + { + "epoch": 0.12, + "learning_rate": 9.63032044166873e-05, + "loss": 2.4041, + "step": 46025 + }, + { + "epoch": 0.12, + "learning_rate": 9.630240948108549e-05, + "loss": 2.4344, + "step": 46030 + }, + { + "epoch": 0.12, + "learning_rate": 9.63016144633055e-05, + "loss": 2.3629, + "step": 46035 + }, + { + "epoch": 0.12, + "learning_rate": 9.630081936334879e-05, + "loss": 2.4463, + "step": 46040 + }, + { + "epoch": 0.12, + "learning_rate": 9.63000241812167e-05, + "loss": 2.3809, + "step": 46045 + }, + { + "epoch": 0.12, + "learning_rate": 9.629922891691072e-05, + "loss": 2.3605, + "step": 46050 + }, + { + "epoch": 0.12, + "learning_rate": 9.629843357043222e-05, + "loss": 2.3474, + "step": 46055 + }, + { + "epoch": 0.12, + "learning_rate": 9.629763814178261e-05, + "loss": 2.4027, + "step": 46060 + }, + { + "epoch": 0.12, + "learning_rate": 9.62968426309633e-05, + "loss": 2.2945, + "step": 46065 + }, + { + "epoch": 0.12, + "learning_rate": 9.629604703797573e-05, + "loss": 2.3149, + "step": 46070 + }, + { + "epoch": 0.12, + "learning_rate": 9.629525136282128e-05, + "loss": 2.357, + "step": 46075 + }, + { + "epoch": 0.12, + "learning_rate": 9.629445560550136e-05, + "loss": 2.3762, + "step": 46080 + }, + { + "epoch": 0.12, + "learning_rate": 9.629365976601741e-05, + "loss": 2.2773, + "step": 46085 + }, + { + "epoch": 0.12, + "learning_rate": 9.629286384437082e-05, + "loss": 2.3979, + "step": 46090 + }, + { + "epoch": 0.12, + "learning_rate": 9.629206784056302e-05, + "loss": 2.3962, + "step": 46095 + }, + { + "epoch": 0.12, + "learning_rate": 9.629127175459542e-05, + "loss": 2.345, + "step": 46100 + }, + { + "epoch": 0.12, + "learning_rate": 9.629047558646942e-05, + "loss": 2.3514, + "step": 46105 + }, + { + "epoch": 0.12, + "learning_rate": 9.628967933618644e-05, + "loss": 2.4102, + "step": 46110 + }, + { + "epoch": 0.12, + "learning_rate": 9.62888830037479e-05, + "loss": 2.354, + "step": 46115 + }, + { + "epoch": 0.12, + "learning_rate": 9.628808658915521e-05, + "loss": 2.3603, + "step": 46120 + }, + { + "epoch": 0.12, + "learning_rate": 9.628729009240975e-05, + "loss": 2.2407, + "step": 46125 + }, + { + "epoch": 0.12, + "learning_rate": 9.6286493513513e-05, + "loss": 2.3521, + "step": 46130 + }, + { + "epoch": 0.12, + "learning_rate": 9.62856968524663e-05, + "loss": 2.4799, + "step": 46135 + }, + { + "epoch": 0.12, + "learning_rate": 9.628490010927114e-05, + "loss": 2.3392, + "step": 46140 + }, + { + "epoch": 0.12, + "learning_rate": 9.628410328392888e-05, + "loss": 2.3609, + "step": 46145 + }, + { + "epoch": 0.12, + "learning_rate": 9.628330637644097e-05, + "loss": 2.2909, + "step": 46150 + }, + { + "epoch": 0.12, + "learning_rate": 9.628250938680877e-05, + "loss": 2.4896, + "step": 46155 + }, + { + "epoch": 0.12, + "learning_rate": 9.628171231503375e-05, + "loss": 2.2718, + "step": 46160 + }, + { + "epoch": 0.12, + "learning_rate": 9.628091516111729e-05, + "loss": 2.2078, + "step": 46165 + }, + { + "epoch": 0.12, + "learning_rate": 9.628011792506082e-05, + "loss": 2.3481, + "step": 46170 + }, + { + "epoch": 0.12, + "learning_rate": 9.627932060686577e-05, + "loss": 2.4607, + "step": 46175 + }, + { + "epoch": 0.12, + "learning_rate": 9.627852320653352e-05, + "loss": 2.3917, + "step": 46180 + }, + { + "epoch": 0.12, + "learning_rate": 9.627772572406552e-05, + "loss": 2.3586, + "step": 46185 + }, + { + "epoch": 0.12, + "learning_rate": 9.627692815946316e-05, + "loss": 2.339, + "step": 46190 + }, + { + "epoch": 0.12, + "learning_rate": 9.627613051272786e-05, + "loss": 2.3375, + "step": 46195 + }, + { + "epoch": 0.12, + "learning_rate": 9.627533278386104e-05, + "loss": 2.1799, + "step": 46200 + }, + { + "epoch": 0.12, + "learning_rate": 9.627453497286411e-05, + "loss": 2.5121, + "step": 46205 + }, + { + "epoch": 0.12, + "learning_rate": 9.627373707973852e-05, + "loss": 2.3604, + "step": 46210 + }, + { + "epoch": 0.12, + "learning_rate": 9.627293910448562e-05, + "loss": 2.1591, + "step": 46215 + }, + { + "epoch": 0.12, + "learning_rate": 9.627214104710689e-05, + "loss": 2.3334, + "step": 46220 + }, + { + "epoch": 0.12, + "learning_rate": 9.62713429076037e-05, + "loss": 2.419, + "step": 46225 + }, + { + "epoch": 0.12, + "learning_rate": 9.627054468597751e-05, + "loss": 2.3972, + "step": 46230 + }, + { + "epoch": 0.12, + "learning_rate": 9.626974638222973e-05, + "loss": 2.4229, + "step": 46235 + }, + { + "epoch": 0.12, + "learning_rate": 9.626894799636173e-05, + "loss": 2.4478, + "step": 46240 + }, + { + "epoch": 0.12, + "learning_rate": 9.626814952837495e-05, + "loss": 2.2295, + "step": 46245 + }, + { + "epoch": 0.12, + "learning_rate": 9.626735097827084e-05, + "loss": 2.253, + "step": 46250 + }, + { + "epoch": 0.12, + "learning_rate": 9.626655234605078e-05, + "loss": 2.3774, + "step": 46255 + }, + { + "epoch": 0.12, + "learning_rate": 9.62657536317162e-05, + "loss": 2.2915, + "step": 46260 + }, + { + "epoch": 0.12, + "learning_rate": 9.626495483526852e-05, + "loss": 2.4106, + "step": 46265 + }, + { + "epoch": 0.12, + "learning_rate": 9.626415595670915e-05, + "loss": 2.326, + "step": 46270 + }, + { + "epoch": 0.12, + "learning_rate": 9.626335699603951e-05, + "loss": 2.3556, + "step": 46275 + }, + { + "epoch": 0.12, + "learning_rate": 9.626255795326102e-05, + "loss": 2.4617, + "step": 46280 + }, + { + "epoch": 0.12, + "learning_rate": 9.626175882837511e-05, + "loss": 2.3441, + "step": 46285 + }, + { + "epoch": 0.12, + "learning_rate": 9.626095962138319e-05, + "loss": 2.2578, + "step": 46290 + }, + { + "epoch": 0.12, + "learning_rate": 9.626016033228666e-05, + "loss": 2.3043, + "step": 46295 + }, + { + "epoch": 0.12, + "learning_rate": 9.625936096108696e-05, + "loss": 2.274, + "step": 46300 + }, + { + "epoch": 0.12, + "learning_rate": 9.625856150778551e-05, + "loss": 2.4493, + "step": 46305 + }, + { + "epoch": 0.12, + "learning_rate": 9.625776197238372e-05, + "loss": 2.2998, + "step": 46310 + }, + { + "epoch": 0.12, + "learning_rate": 9.6256962354883e-05, + "loss": 2.3276, + "step": 46315 + }, + { + "epoch": 0.12, + "learning_rate": 9.625616265528478e-05, + "loss": 2.3438, + "step": 46320 + }, + { + "epoch": 0.12, + "learning_rate": 9.625536287359049e-05, + "loss": 2.347, + "step": 46325 + }, + { + "epoch": 0.12, + "learning_rate": 9.625456300980154e-05, + "loss": 2.4624, + "step": 46330 + }, + { + "epoch": 0.12, + "learning_rate": 9.625376306391936e-05, + "loss": 2.3586, + "step": 46335 + }, + { + "epoch": 0.12, + "learning_rate": 9.625296303594533e-05, + "loss": 2.3218, + "step": 46340 + }, + { + "epoch": 0.12, + "learning_rate": 9.62521629258809e-05, + "loss": 2.3903, + "step": 46345 + }, + { + "epoch": 0.12, + "learning_rate": 9.62513627337275e-05, + "loss": 2.4251, + "step": 46350 + }, + { + "epoch": 0.12, + "learning_rate": 9.625056245948653e-05, + "loss": 2.376, + "step": 46355 + }, + { + "epoch": 0.12, + "learning_rate": 9.624976210315944e-05, + "loss": 2.3971, + "step": 46360 + }, + { + "epoch": 0.12, + "learning_rate": 9.624896166474762e-05, + "loss": 2.2967, + "step": 46365 + }, + { + "epoch": 0.12, + "learning_rate": 9.624816114425249e-05, + "loss": 2.2814, + "step": 46370 + }, + { + "epoch": 0.12, + "learning_rate": 9.62473605416755e-05, + "loss": 2.214, + "step": 46375 + }, + { + "epoch": 0.12, + "learning_rate": 9.624655985701803e-05, + "loss": 2.2409, + "step": 46380 + }, + { + "epoch": 0.12, + "learning_rate": 9.624575909028152e-05, + "loss": 2.394, + "step": 46385 + }, + { + "epoch": 0.12, + "learning_rate": 9.62449582414674e-05, + "loss": 2.4079, + "step": 46390 + }, + { + "epoch": 0.12, + "learning_rate": 9.624415731057709e-05, + "loss": 2.2285, + "step": 46395 + }, + { + "epoch": 0.12, + "learning_rate": 9.624335629761201e-05, + "loss": 2.4258, + "step": 46400 + }, + { + "epoch": 0.12, + "learning_rate": 9.624255520257358e-05, + "loss": 2.4167, + "step": 46405 + }, + { + "epoch": 0.12, + "learning_rate": 9.62417540254632e-05, + "loss": 2.3191, + "step": 46410 + }, + { + "epoch": 0.12, + "learning_rate": 9.624095276628232e-05, + "loss": 2.3842, + "step": 46415 + }, + { + "epoch": 0.12, + "learning_rate": 9.624015142503236e-05, + "loss": 2.3884, + "step": 46420 + }, + { + "epoch": 0.12, + "learning_rate": 9.623935000171473e-05, + "loss": 2.3997, + "step": 46425 + }, + { + "epoch": 0.12, + "learning_rate": 9.623854849633087e-05, + "loss": 2.3342, + "step": 46430 + }, + { + "epoch": 0.12, + "learning_rate": 9.623774690888219e-05, + "loss": 2.2799, + "step": 46435 + }, + { + "epoch": 0.12, + "learning_rate": 9.62369452393701e-05, + "loss": 2.3467, + "step": 46440 + }, + { + "epoch": 0.12, + "learning_rate": 9.623614348779603e-05, + "loss": 2.3009, + "step": 46445 + }, + { + "epoch": 0.12, + "learning_rate": 9.623534165416144e-05, + "loss": 2.2503, + "step": 46450 + }, + { + "epoch": 0.12, + "learning_rate": 9.62345397384677e-05, + "loss": 2.2992, + "step": 46455 + }, + { + "epoch": 0.12, + "learning_rate": 9.623373774071627e-05, + "loss": 2.3808, + "step": 46460 + }, + { + "epoch": 0.12, + "learning_rate": 9.623293566090856e-05, + "loss": 2.391, + "step": 46465 + }, + { + "epoch": 0.12, + "learning_rate": 9.623213349904599e-05, + "loss": 2.3373, + "step": 46470 + }, + { + "epoch": 0.12, + "learning_rate": 9.623133125512997e-05, + "loss": 2.3919, + "step": 46475 + }, + { + "epoch": 0.12, + "learning_rate": 9.623052892916196e-05, + "loss": 2.4501, + "step": 46480 + }, + { + "epoch": 0.12, + "learning_rate": 9.622972652114335e-05, + "loss": 2.3262, + "step": 46485 + }, + { + "epoch": 0.12, + "learning_rate": 9.622892403107559e-05, + "loss": 2.2507, + "step": 46490 + }, + { + "epoch": 0.12, + "learning_rate": 9.622812145896008e-05, + "loss": 2.2937, + "step": 46495 + }, + { + "epoch": 0.12, + "learning_rate": 9.622731880479827e-05, + "loss": 2.2891, + "step": 46500 + }, + { + "epoch": 0.12, + "learning_rate": 9.622651606859157e-05, + "loss": 2.2659, + "step": 46505 + }, + { + "epoch": 0.12, + "learning_rate": 9.622571325034139e-05, + "loss": 2.2703, + "step": 46510 + }, + { + "epoch": 0.12, + "learning_rate": 9.622491035004919e-05, + "loss": 2.3615, + "step": 46515 + }, + { + "epoch": 0.12, + "learning_rate": 9.622410736771638e-05, + "loss": 2.3237, + "step": 46520 + }, + { + "epoch": 0.12, + "learning_rate": 9.622330430334437e-05, + "loss": 2.315, + "step": 46525 + }, + { + "epoch": 0.12, + "learning_rate": 9.62225011569346e-05, + "loss": 2.289, + "step": 46530 + }, + { + "epoch": 0.12, + "learning_rate": 9.622169792848849e-05, + "loss": 2.3733, + "step": 46535 + }, + { + "epoch": 0.12, + "learning_rate": 9.622089461800748e-05, + "loss": 2.4106, + "step": 46540 + }, + { + "epoch": 0.12, + "learning_rate": 9.622009122549296e-05, + "loss": 2.2966, + "step": 46545 + }, + { + "epoch": 0.12, + "learning_rate": 9.62192877509464e-05, + "loss": 2.3195, + "step": 46550 + }, + { + "epoch": 0.12, + "learning_rate": 9.62184841943692e-05, + "loss": 2.3864, + "step": 46555 + }, + { + "epoch": 0.12, + "learning_rate": 9.621768055576279e-05, + "loss": 2.3809, + "step": 46560 + }, + { + "epoch": 0.12, + "learning_rate": 9.62168768351286e-05, + "loss": 2.3313, + "step": 46565 + }, + { + "epoch": 0.12, + "learning_rate": 9.621607303246804e-05, + "loss": 2.4381, + "step": 46570 + }, + { + "epoch": 0.12, + "learning_rate": 9.621526914778257e-05, + "loss": 2.3842, + "step": 46575 + }, + { + "epoch": 0.12, + "learning_rate": 9.621446518107359e-05, + "loss": 2.3944, + "step": 46580 + }, + { + "epoch": 0.12, + "learning_rate": 9.621366113234254e-05, + "loss": 2.3607, + "step": 46585 + }, + { + "epoch": 0.12, + "learning_rate": 9.621285700159083e-05, + "loss": 2.3158, + "step": 46590 + }, + { + "epoch": 0.12, + "learning_rate": 9.621205278881991e-05, + "loss": 2.3122, + "step": 46595 + }, + { + "epoch": 0.12, + "learning_rate": 9.621124849403118e-05, + "loss": 2.299, + "step": 46600 + }, + { + "epoch": 0.12, + "learning_rate": 9.621044411722611e-05, + "loss": 2.3026, + "step": 46605 + }, + { + "epoch": 0.12, + "learning_rate": 9.620963965840608e-05, + "loss": 2.426, + "step": 46610 + }, + { + "epoch": 0.12, + "learning_rate": 9.620883511757255e-05, + "loss": 2.4365, + "step": 46615 + }, + { + "epoch": 0.13, + "learning_rate": 9.620803049472693e-05, + "loss": 2.2231, + "step": 46620 + }, + { + "epoch": 0.13, + "learning_rate": 9.620722578987064e-05, + "loss": 2.2711, + "step": 46625 + }, + { + "epoch": 0.13, + "learning_rate": 9.620642100300516e-05, + "loss": 2.3044, + "step": 46630 + }, + { + "epoch": 0.13, + "learning_rate": 9.620561613413186e-05, + "loss": 2.1949, + "step": 46635 + }, + { + "epoch": 0.13, + "learning_rate": 9.620481118325219e-05, + "loss": 2.2951, + "step": 46640 + }, + { + "epoch": 0.13, + "learning_rate": 9.620400615036757e-05, + "loss": 2.4083, + "step": 46645 + }, + { + "epoch": 0.13, + "learning_rate": 9.620320103547946e-05, + "loss": 2.2992, + "step": 46650 + }, + { + "epoch": 0.13, + "learning_rate": 9.620239583858925e-05, + "loss": 2.3549, + "step": 46655 + }, + { + "epoch": 0.13, + "learning_rate": 9.62015905596984e-05, + "loss": 2.4112, + "step": 46660 + }, + { + "epoch": 0.13, + "learning_rate": 9.620078519880831e-05, + "loss": 2.3881, + "step": 46665 + }, + { + "epoch": 0.13, + "learning_rate": 9.619997975592043e-05, + "loss": 2.3186, + "step": 46670 + }, + { + "epoch": 0.13, + "learning_rate": 9.619917423103617e-05, + "loss": 2.3096, + "step": 46675 + }, + { + "epoch": 0.13, + "learning_rate": 9.619836862415699e-05, + "loss": 2.2889, + "step": 46680 + }, + { + "epoch": 0.13, + "learning_rate": 9.61975629352843e-05, + "loss": 2.402, + "step": 46685 + }, + { + "epoch": 0.13, + "learning_rate": 9.619675716441954e-05, + "loss": 2.3825, + "step": 46690 + }, + { + "epoch": 0.13, + "learning_rate": 9.619595131156412e-05, + "loss": 2.3753, + "step": 46695 + }, + { + "epoch": 0.13, + "learning_rate": 9.619514537671949e-05, + "loss": 2.293, + "step": 46700 + }, + { + "epoch": 0.13, + "learning_rate": 9.619433935988707e-05, + "loss": 2.3726, + "step": 46705 + }, + { + "epoch": 0.13, + "learning_rate": 9.61935332610683e-05, + "loss": 2.4428, + "step": 46710 + }, + { + "epoch": 0.13, + "learning_rate": 9.619272708026458e-05, + "loss": 2.2229, + "step": 46715 + }, + { + "epoch": 0.13, + "learning_rate": 9.619192081747739e-05, + "loss": 2.4557, + "step": 46720 + }, + { + "epoch": 0.13, + "learning_rate": 9.619111447270814e-05, + "loss": 2.373, + "step": 46725 + }, + { + "epoch": 0.13, + "learning_rate": 9.619030804595826e-05, + "loss": 2.3413, + "step": 46730 + }, + { + "epoch": 0.13, + "learning_rate": 9.618950153722917e-05, + "loss": 2.3588, + "step": 46735 + }, + { + "epoch": 0.13, + "learning_rate": 9.618869494652229e-05, + "loss": 2.2827, + "step": 46740 + }, + { + "epoch": 0.13, + "learning_rate": 9.61878882738391e-05, + "loss": 2.3611, + "step": 46745 + }, + { + "epoch": 0.13, + "learning_rate": 9.618708151918099e-05, + "loss": 2.3977, + "step": 46750 + }, + { + "epoch": 0.13, + "learning_rate": 9.618627468254941e-05, + "loss": 2.3747, + "step": 46755 + }, + { + "epoch": 0.13, + "learning_rate": 9.61854677639458e-05, + "loss": 2.2928, + "step": 46760 + }, + { + "epoch": 0.13, + "learning_rate": 9.618466076337156e-05, + "loss": 2.3165, + "step": 46765 + }, + { + "epoch": 0.13, + "learning_rate": 9.618385368082815e-05, + "loss": 2.3263, + "step": 46770 + }, + { + "epoch": 0.13, + "learning_rate": 9.6183046516317e-05, + "loss": 2.3465, + "step": 46775 + }, + { + "epoch": 0.13, + "learning_rate": 9.618223926983952e-05, + "loss": 2.2808, + "step": 46780 + }, + { + "epoch": 0.13, + "learning_rate": 9.618143194139717e-05, + "loss": 2.2537, + "step": 46785 + }, + { + "epoch": 0.13, + "learning_rate": 9.618062453099137e-05, + "loss": 2.3377, + "step": 46790 + }, + { + "epoch": 0.13, + "learning_rate": 9.617981703862357e-05, + "loss": 2.3425, + "step": 46795 + }, + { + "epoch": 0.13, + "learning_rate": 9.617900946429517e-05, + "loss": 2.3397, + "step": 46800 + }, + { + "epoch": 0.13, + "learning_rate": 9.617820180800763e-05, + "loss": 2.4327, + "step": 46805 + }, + { + "epoch": 0.13, + "learning_rate": 9.617739406976239e-05, + "loss": 2.3981, + "step": 46810 + }, + { + "epoch": 0.13, + "learning_rate": 9.617658624956084e-05, + "loss": 2.3847, + "step": 46815 + }, + { + "epoch": 0.13, + "learning_rate": 9.617577834740445e-05, + "loss": 2.3896, + "step": 46820 + }, + { + "epoch": 0.13, + "learning_rate": 9.617497036329466e-05, + "loss": 2.373, + "step": 46825 + }, + { + "epoch": 0.13, + "learning_rate": 9.617416229723288e-05, + "loss": 2.4137, + "step": 46830 + }, + { + "epoch": 0.13, + "learning_rate": 9.617335414922054e-05, + "loss": 2.2046, + "step": 46835 + }, + { + "epoch": 0.13, + "learning_rate": 9.617254591925913e-05, + "loss": 2.3157, + "step": 46840 + }, + { + "epoch": 0.13, + "learning_rate": 9.617173760735e-05, + "loss": 2.4246, + "step": 46845 + }, + { + "epoch": 0.13, + "learning_rate": 9.617092921349466e-05, + "loss": 2.4711, + "step": 46850 + }, + { + "epoch": 0.13, + "learning_rate": 9.61701207376945e-05, + "loss": 2.4615, + "step": 46855 + }, + { + "epoch": 0.13, + "learning_rate": 9.616931217995098e-05, + "loss": 2.303, + "step": 46860 + }, + { + "epoch": 0.13, + "learning_rate": 9.616850354026549e-05, + "loss": 2.3377, + "step": 46865 + }, + { + "epoch": 0.13, + "learning_rate": 9.616769481863953e-05, + "loss": 2.3051, + "step": 46870 + }, + { + "epoch": 0.13, + "learning_rate": 9.616688601507449e-05, + "loss": 2.3907, + "step": 46875 + }, + { + "epoch": 0.13, + "learning_rate": 9.616607712957181e-05, + "loss": 2.4815, + "step": 46880 + }, + { + "epoch": 0.13, + "learning_rate": 9.616526816213296e-05, + "loss": 2.3699, + "step": 46885 + }, + { + "epoch": 0.13, + "learning_rate": 9.616445911275934e-05, + "loss": 2.456, + "step": 46890 + }, + { + "epoch": 0.13, + "learning_rate": 9.616364998145239e-05, + "loss": 2.3177, + "step": 46895 + }, + { + "epoch": 0.13, + "learning_rate": 9.616284076821356e-05, + "loss": 2.4051, + "step": 46900 + }, + { + "epoch": 0.13, + "learning_rate": 9.616203147304426e-05, + "loss": 2.44, + "step": 46905 + }, + { + "epoch": 0.13, + "learning_rate": 9.616122209594595e-05, + "loss": 2.2011, + "step": 46910 + }, + { + "epoch": 0.13, + "learning_rate": 9.616041263692008e-05, + "loss": 2.3017, + "step": 46915 + }, + { + "epoch": 0.13, + "learning_rate": 9.615960309596805e-05, + "loss": 2.2741, + "step": 46920 + }, + { + "epoch": 0.13, + "learning_rate": 9.615879347309132e-05, + "loss": 2.3764, + "step": 46925 + }, + { + "epoch": 0.13, + "learning_rate": 9.615798376829132e-05, + "loss": 2.3808, + "step": 46930 + }, + { + "epoch": 0.13, + "learning_rate": 9.615717398156949e-05, + "loss": 2.3542, + "step": 46935 + }, + { + "epoch": 0.13, + "learning_rate": 9.615636411292727e-05, + "loss": 2.3396, + "step": 46940 + }, + { + "epoch": 0.13, + "learning_rate": 9.615555416236609e-05, + "loss": 2.3701, + "step": 46945 + }, + { + "epoch": 0.13, + "learning_rate": 9.615474412988739e-05, + "loss": 2.4176, + "step": 46950 + }, + { + "epoch": 0.13, + "learning_rate": 9.61539340154926e-05, + "loss": 2.2088, + "step": 46955 + }, + { + "epoch": 0.13, + "learning_rate": 9.615312381918318e-05, + "loss": 2.3218, + "step": 46960 + }, + { + "epoch": 0.13, + "learning_rate": 9.615231354096055e-05, + "loss": 2.2841, + "step": 46965 + }, + { + "epoch": 0.13, + "learning_rate": 9.615150318082616e-05, + "loss": 2.3107, + "step": 46970 + }, + { + "epoch": 0.13, + "learning_rate": 9.615069273878144e-05, + "loss": 2.4256, + "step": 46975 + }, + { + "epoch": 0.13, + "learning_rate": 9.614988221482782e-05, + "loss": 2.27, + "step": 46980 + }, + { + "epoch": 0.13, + "learning_rate": 9.614907160896674e-05, + "loss": 2.277, + "step": 46985 + }, + { + "epoch": 0.13, + "learning_rate": 9.614826092119965e-05, + "loss": 2.4039, + "step": 46990 + }, + { + "epoch": 0.13, + "learning_rate": 9.614745015152799e-05, + "loss": 2.2609, + "step": 46995 + }, + { + "epoch": 0.13, + "learning_rate": 9.614663929995318e-05, + "loss": 2.3187, + "step": 47000 + } + ], + "logging_steps": 5, + "max_steps": 372958, + "num_train_epochs": 1, + "save_steps": 1000, + "total_flos": 6.326848812220416e+18, + "trial_name": null, + "trial_params": null +}