| { | |
| "best_metric": 0.5662159323692322, | |
| "best_model_checkpoint": "./vit-mae-demo/checkpoint-7705", | |
| "epoch": 1.0, | |
| "global_step": 7705, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.2143782383419688e-07, | |
| "loss": 1.1962, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.4287564766839375e-07, | |
| "loss": 1.1898, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.6431347150259065e-07, | |
| "loss": 1.1806, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.857512953367875e-07, | |
| "loss": 1.1654, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.071891191709843e-07, | |
| "loss": 1.1466, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.286269430051813e-07, | |
| "loss": 1.1258, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.500647668393782e-07, | |
| "loss": 1.1007, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.71502590673575e-07, | |
| "loss": 1.0794, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.092940414507772e-06, | |
| "loss": 1.0455, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.2143782383419686e-06, | |
| "loss": 1.0274, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.3358160621761657e-06, | |
| "loss": 1.0019, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4572538860103626e-06, | |
| "loss": 0.9826, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.5786917098445593e-06, | |
| "loss": 0.9774, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.7001295336787564e-06, | |
| "loss": 0.9693, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.8215673575129531e-06, | |
| "loss": 0.9425, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.94300518134715e-06, | |
| "loss": 0.9488, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.064443005181347e-06, | |
| "loss": 0.9207, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.185880829015544e-06, | |
| "loss": 0.9075, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.3073186528497407e-06, | |
| "loss": 0.9234, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.428756476683937e-06, | |
| "loss": 0.8749, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5501943005181345e-06, | |
| "loss": 0.8928, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.6716321243523314e-06, | |
| "loss": 0.8831, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.7930699481865283e-06, | |
| "loss": 0.8802, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9145077720207252e-06, | |
| "loss": 0.8831, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 3.0359455958549217e-06, | |
| "loss": 0.8745, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 3.1573834196891186e-06, | |
| "loss": 0.8503, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.278821243523316e-06, | |
| "loss": 0.8672, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.400259067357513e-06, | |
| "loss": 0.8717, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.5216968911917097e-06, | |
| "loss": 0.8471, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.6431347150259062e-06, | |
| "loss": 0.8562, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.764572538860103e-06, | |
| "loss": 0.8275, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 3.8860103626943e-06, | |
| "loss": 0.8412, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.007448186528497e-06, | |
| "loss": 0.8518, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 4.128886010362694e-06, | |
| "loss": 0.8227, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.250323834196891e-06, | |
| "loss": 0.8263, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.371761658031088e-06, | |
| "loss": 0.8236, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.4931994818652845e-06, | |
| "loss": 0.8543, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.6146373056994814e-06, | |
| "loss": 0.79, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.6874965454028015e-06, | |
| "loss": 0.8373, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.687457681301278e-06, | |
| "loss": 0.8063, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.687375635570173e-06, | |
| "loss": 0.8134, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 4.687250409721135e-06, | |
| "loss": 0.7963, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.687082006061389e-06, | |
| "loss": 0.802, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.686870427693687e-06, | |
| "loss": 0.8135, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.686615678516253e-06, | |
| "loss": 0.8251, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.686317763222712e-06, | |
| "loss": 0.7704, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.685976687302003e-06, | |
| "loss": 0.8015, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.685592457038277e-06, | |
| "loss": 0.7791, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.6851650795107834e-06, | |
| "loss": 0.8139, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.684694562593737e-06, | |
| "loss": 0.7643, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.684180914956173e-06, | |
| "loss": 0.7895, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.68362414606179e-06, | |
| "loss": 0.7852, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.683024266168774e-06, | |
| "loss": 0.789, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.682381286329609e-06, | |
| "loss": 0.7944, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.6816952183908725e-06, | |
| "loss": 0.7984, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.680966074993021e-06, | |
| "loss": 0.7705, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 4.6801938695701525e-06, | |
| "loss": 0.7555, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.679378616349761e-06, | |
| "loss": 0.7786, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.678520330352476e-06, | |
| "loss": 0.7742, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.677619027391783e-06, | |
| "loss": 0.7689, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.676674724073732e-06, | |
| "loss": 0.7746, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.675687437796638e-06, | |
| "loss": 0.7735, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.674657186750749e-06, | |
| "loss": 0.7922, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.673583989917919e-06, | |
| "loss": 0.7579, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 4.67246786707126e-06, | |
| "loss": 0.7654, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.671308838774771e-06, | |
| "loss": 0.7437, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.67010692638296e-06, | |
| "loss": 0.7607, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.668862152040458e-06, | |
| "loss": 0.7857, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.667574538681602e-06, | |
| "loss": 0.784, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.666244110030018e-06, | |
| "loss": 0.758, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.664870890598181e-06, | |
| "loss": 0.7385, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.663454905686965e-06, | |
| "loss": 0.7733, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 4.661996181385176e-06, | |
| "loss": 0.7574, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.660494744569075e-06, | |
| "loss": 0.7816, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.658950622901873e-06, | |
| "loss": 0.7538, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.6573638448332325e-06, | |
| "loss": 0.7728, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.655734439598738e-06, | |
| "loss": 0.7852, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.654062437219358e-06, | |
| "loss": 0.7509, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.6523478685008895e-06, | |
| "loss": 0.7577, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 4.650590765033395e-06, | |
| "loss": 0.8028, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.64879115919062e-06, | |
| "loss": 0.7585, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.64694908412939e-06, | |
| "loss": 0.749, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.645064573789009e-06, | |
| "loss": 0.7771, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.643137662890627e-06, | |
| "loss": 0.7645, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.641168386936603e-06, | |
| "loss": 0.7852, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.639156782209853e-06, | |
| "loss": 0.7746, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.6371028857731744e-06, | |
| "loss": 0.7709, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 4.6350067354685745e-06, | |
| "loss": 0.7612, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.632868369916561e-06, | |
| "loss": 0.7601, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.630687828515439e-06, | |
| "loss": 0.7647, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.628465151440582e-06, | |
| "loss": 0.8073, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.626200379643692e-06, | |
| "loss": 0.7374, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.623893554852044e-06, | |
| "loss": 0.7766, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.62154471956772e-06, | |
| "loss": 0.7578, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.619153917066823e-06, | |
| "loss": 0.7671, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 4.61672119139868e-06, | |
| "loss": 0.7761, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.614246587385033e-06, | |
| "loss": 0.7803, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.61173015061921e-06, | |
| "loss": 0.7657, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.6091719274652855e-06, | |
| "loss": 0.7659, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.6065719650572305e-06, | |
| "loss": 0.7458, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.603930311298037e-06, | |
| "loss": 0.7511, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.60124701485884e-06, | |
| "loss": 0.7341, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.598522125178022e-06, | |
| "loss": 0.7825, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 4.595755692460297e-06, | |
| "loss": 0.7773, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.59294776767579e-06, | |
| "loss": 0.783, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.590098402559097e-06, | |
| "loss": 0.7629, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.587207649608328e-06, | |
| "loss": 0.7717, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.584275562084148e-06, | |
| "loss": 0.7135, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.581302194008786e-06, | |
| "loss": 0.7641, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.578287600165046e-06, | |
| "loss": 0.7786, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 4.575231836095298e-06, | |
| "loss": 0.7443, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.5721349581004486e-06, | |
| "loss": 0.7577, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.56899702323891e-06, | |
| "loss": 0.7825, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.565818089325548e-06, | |
| "loss": 0.7511, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.562598214930612e-06, | |
| "loss": 0.7615, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.55933745937866e-06, | |
| "loss": 0.7938, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.556035882747467e-06, | |
| "loss": 0.765, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.552693545866911e-06, | |
| "loss": 0.7689, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 4.549310510317863e-06, | |
| "loss": 0.7487, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.545886838431042e-06, | |
| "loss": 0.7747, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.542422593285872e-06, | |
| "loss": 0.7907, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.538917838709319e-06, | |
| "loss": 0.7508, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.535372639274715e-06, | |
| "loss": 0.7462, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.531787060300568e-06, | |
| "loss": 0.7563, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.528161167849359e-06, | |
| "loss": 0.7192, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.524495028726325e-06, | |
| "loss": 0.7385, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 4.520788710478224e-06, | |
| "loss": 0.7305, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.517042281392097e-06, | |
| "loss": 0.7589, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.513255810494008e-06, | |
| "loss": 0.7603, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.509429367547769e-06, | |
| "loss": 0.7263, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.505563023053659e-06, | |
| "loss": 0.7168, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.501656848247118e-06, | |
| "loss": 0.7261, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.497710915097446e-06, | |
| "loss": 0.7431, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 4.493725296306466e-06, | |
| "loss": 0.7467, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.489700065307189e-06, | |
| "loss": 0.7266, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.485635296262463e-06, | |
| "loss": 0.7231, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.481531064063604e-06, | |
| "loss": 0.7376, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.477387444329015e-06, | |
| "loss": 0.6859, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.473204513402796e-06, | |
| "loss": 0.6906, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.4689823483533345e-06, | |
| "loss": 0.7265, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.464721026971889e-06, | |
| "loss": 0.6932, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 4.460420627771152e-06, | |
| "loss": 0.7083, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.456081229983805e-06, | |
| "loss": 0.6923, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.451702913561059e-06, | |
| "loss": 0.6764, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.447285759171183e-06, | |
| "loss": 0.7025, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.442829848198013e-06, | |
| "loss": 0.667, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.438335262739458e-06, | |
| "loss": 0.6773, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.433802085605981e-06, | |
| "loss": 0.6699, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.4292304003190824e-06, | |
| "loss": 0.6531, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 4.4246202911097526e-06, | |
| "loss": 0.6819, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.4199718429169235e-06, | |
| "loss": 0.678, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.415285141385904e-06, | |
| "loss": 0.6561, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.410560272866801e-06, | |
| "loss": 0.6931, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.40579732441293e-06, | |
| "loss": 0.6609, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.400996383779208e-06, | |
| "loss": 0.6708, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.39615753942054e-06, | |
| "loss": 0.6895, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 4.39128088049019e-06, | |
| "loss": 0.6347, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.386366496838131e-06, | |
| "loss": 0.6387, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.381414479009401e-06, | |
| "loss": 0.6538, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.376424918242424e-06, | |
| "loss": 0.6517, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.371397906467338e-06, | |
| "loss": 0.6585, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.366333536304291e-06, | |
| "loss": 0.6795, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.361231901061746e-06, | |
| "loss": 0.6575, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.356093094734755e-06, | |
| "loss": 0.6594, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 4.350917212003224e-06, | |
| "loss": 0.6646, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.34570434823018e-06, | |
| "loss": 0.6527, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.340454599460004e-06, | |
| "loss": 0.6146, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.335168062416661e-06, | |
| "loss": 0.6632, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.329844834501925e-06, | |
| "loss": 0.6327, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.324485013793582e-06, | |
| "loss": 0.642, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.3190886990436184e-06, | |
| "loss": 0.6518, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.313655989676407e-06, | |
| "loss": 0.6648, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 4.308186985786872e-06, | |
| "loss": 0.6221, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.3026817881386465e-06, | |
| "loss": 0.6491, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.297140498162214e-06, | |
| "loss": 0.6684, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.291563217953044e-06, | |
| "loss": 0.6355, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.285950050269703e-06, | |
| "loss": 0.6151, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.2803010985319704e-06, | |
| "loss": 0.6486, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.2746164668189275e-06, | |
| "loss": 0.6579, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.26889625986704e-06, | |
| "loss": 0.6103, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 4.2631405830682305e-06, | |
| "loss": 0.6243, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.257349542467934e-06, | |
| "loss": 0.6446, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.2515232447631495e-06, | |
| "loss": 0.6307, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.2456617973004655e-06, | |
| "loss": 0.6022, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.23976530807409e-06, | |
| "loss": 0.6184, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.233833885723858e-06, | |
| "loss": 0.6164, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.227867639533227e-06, | |
| "loss": 0.5928, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 4.22186667942727e-06, | |
| "loss": 0.6362, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.215831115970642e-06, | |
| "loss": 0.6311, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.20976106036555e-06, | |
| "loss": 0.6579, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.203656624449703e-06, | |
| "loss": 0.6386, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.197517920694245e-06, | |
| "loss": 0.608, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.1913450622016915e-06, | |
| "loss": 0.6524, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.185138162703841e-06, | |
| "loss": 0.5948, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.178897336559679e-06, | |
| "loss": 0.5944, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 4.1726226987532745e-06, | |
| "loss": 0.6271, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.166314364891657e-06, | |
| "loss": 0.6096, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.15997245120269e-06, | |
| "loss": 0.6078, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.153597074532928e-06, | |
| "loss": 0.5921, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.147188352345466e-06, | |
| "loss": 0.6065, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.1407464027177685e-06, | |
| "loss": 0.5999, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.134271344339505e-06, | |
| "loss": 0.5988, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.127763296510351e-06, | |
| "loss": 0.6057, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 4.121222379137801e-06, | |
| "loss": 0.5926, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.11464871273495e-06, | |
| "loss": 0.6038, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.108042418418282e-06, | |
| "loss": 0.5989, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.1014036179054275e-06, | |
| "loss": 0.5942, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.094732433512933e-06, | |
| "loss": 0.6147, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.088028988154e-06, | |
| "loss": 0.5953, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.081293405336221e-06, | |
| "loss": 0.6255, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 4.0745258091593045e-06, | |
| "loss": 0.5941, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.067726324312792e-06, | |
| "loss": 0.6083, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.060895076073753e-06, | |
| "loss": 0.613, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.054032190304486e-06, | |
| "loss": 0.5829, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.047137793450191e-06, | |
| "loss": 0.5984, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.040212012536648e-06, | |
| "loss": 0.6085, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.033254975167869e-06, | |
| "loss": 0.6008, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.026266809523752e-06, | |
| "loss": 0.6173, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 4.0192476443577164e-06, | |
| "loss": 0.5689, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.012197608994335e-06, | |
| "loss": 0.6124, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 4.005116833326944e-06, | |
| "loss": 0.6094, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.9980054478152595e-06, | |
| "loss": 0.5848, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.990863583482963e-06, | |
| "loss": 0.5825, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.983691371915297e-06, | |
| "loss": 0.5828, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.976488945256635e-06, | |
| "loss": 0.5945, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.969256436208048e-06, | |
| "loss": 0.6047, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 3.961993978024859e-06, | |
| "loss": 0.5872, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.954701704514192e-06, | |
| "loss": 0.6147, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.9473797500325e-06, | |
| "loss": 0.5925, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.940028249483093e-06, | |
| "loss": 0.6357, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.932647338313655e-06, | |
| "loss": 0.6098, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.9252371525137435e-06, | |
| "loss": 0.6184, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.917797828612287e-06, | |
| "loss": 0.6001, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.910329503675069e-06, | |
| "loss": 0.5985, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 3.902832315302201e-06, | |
| "loss": 0.5854, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.895306401625592e-06, | |
| "loss": 0.5967, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.887751901306396e-06, | |
| "loss": 0.5824, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.8801689535324645e-06, | |
| "loss": 0.5695, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.872557698015779e-06, | |
| "loss": 0.6066, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.864918274989874e-06, | |
| "loss": 0.6131, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.857250825207259e-06, | |
| "loss": 0.6004, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 3.849555489936821e-06, | |
| "loss": 0.5908, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.841832410961225e-06, | |
| "loss": 0.5914, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.834081730574294e-06, | |
| "loss": 0.611, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.826303591578402e-06, | |
| "loss": 0.5972, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.818498137281826e-06, | |
| "loss": 0.5953, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.8106655114961175e-06, | |
| "loss": 0.5932, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.802805858533449e-06, | |
| "loss": 0.6316, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.7949193232039534e-06, | |
| "loss": 0.588, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 3.7870060508130585e-06, | |
| "loss": 0.5723, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7790661871588104e-06, | |
| "loss": 0.6306, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7710998785291817e-06, | |
| "loss": 0.6023, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7631072716993856e-06, | |
| "loss": 0.5491, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7550885139291617e-06, | |
| "loss": 0.5676, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7470437529600705e-06, | |
| "loss": 0.5952, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.738973137012767e-06, | |
| "loss": 0.5796, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7308768147842717e-06, | |
| "loss": 0.5717, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 3.7227549354452296e-06, | |
| "loss": 0.5991, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.7146076486371625e-06, | |
| "loss": 0.5906, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.706435104469714e-06, | |
| "loss": 0.5738, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.6982374535178783e-06, | |
| "loss": 0.6033, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.6900148468192326e-06, | |
| "loss": 0.5925, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.6817674358711494e-06, | |
| "loss": 0.6114, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.6734953726280074e-06, | |
| "loss": 0.5925, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 3.665198809498391e-06, | |
| "loss": 0.5856, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.6568778993422835e-06, | |
| "loss": 0.6054, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.64853279546825e-06, | |
| "loss": 0.586, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.6401636516306112e-06, | |
| "loss": 0.5659, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.6317706220266126e-06, | |
| "loss": 0.6194, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.6233538612935835e-06, | |
| "loss": 0.5847, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.6149135245060863e-06, | |
| "loss": 0.5928, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.606449767173061e-06, | |
| "loss": 0.5843, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 3.5979627452349597e-06, | |
| "loss": 0.6179, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.589452615060872e-06, | |
| "loss": 0.547, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.580919533445645e-06, | |
| "loss": 0.5938, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5723636576069957e-06, | |
| "loss": 0.6023, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5637851451826134e-06, | |
| "loss": 0.5723, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5551841542272528e-06, | |
| "loss": 0.5807, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5465608432098257e-06, | |
| "loss": 0.5969, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5379153710104803e-06, | |
| "loss": 0.5992, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 3.5292478969176722e-06, | |
| "loss": 0.5764, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.5205585806252313e-06, | |
| "loss": 0.5862, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.5118475822294193e-06, | |
| "loss": 0.5777, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.5031150622259793e-06, | |
| "loss": 0.5716, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.4943611815071784e-06, | |
| "loss": 0.5729, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.4855861013588455e-06, | |
| "loss": 0.5908, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.4767899834573983e-06, | |
| "loss": 0.6212, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 3.4679729898668627e-06, | |
| "loss": 0.5972, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.4591352830358895e-06, | |
| "loss": 0.6005, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.450277025794762e-06, | |
| "loss": 0.5733, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.441398381352392e-06, | |
| "loss": 0.5631, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.4324995132933178e-06, | |
| "loss": 0.5629, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.423580585574684e-06, | |
| "loss": 0.5931, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.4146417625232287e-06, | |
| "loss": 0.5582, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.405683208832248e-06, | |
| "loss": 0.5946, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 3.3967050895585655e-06, | |
| "loss": 0.5386, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.387707570119493e-06, | |
| "loss": 0.5721, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3786908162897785e-06, | |
| "loss": 0.5882, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.369654994198555e-06, | |
| "loss": 0.6098, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3606002703262762e-06, | |
| "loss": 0.5881, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3515268115016557e-06, | |
| "loss": 0.573, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3424347848985854e-06, | |
| "loss": 0.5673, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3333243580330614e-06, | |
| "loss": 0.6082, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 3.3241956987600945e-06, | |
| "loss": 0.5839, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.315048975270618e-06, | |
| "loss": 0.5966, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.30588435608839e-06, | |
| "loss": 0.6007, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.2967020100668872e-06, | |
| "loss": 0.5822, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.287502106386195e-06, | |
| "loss": 0.5707, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.2782848145498885e-06, | |
| "loss": 0.6007, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.2690503043819114e-06, | |
| "loss": 0.5771, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.259798746023448e-06, | |
| "loss": 0.5913, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 3.2505303099297857e-06, | |
| "loss": 0.5792, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.241245166867174e-06, | |
| "loss": 0.5891, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.231943487909682e-06, | |
| "loss": 0.5751, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.2226254444360448e-06, | |
| "loss": 0.5702, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.213291208126502e-06, | |
| "loss": 0.5906, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.2039409509596418e-06, | |
| "loss": 0.556, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.1945748452092277e-06, | |
| "loss": 0.5541, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 3.1851930634410233e-06, | |
| "loss": 0.5909, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.175795778509618e-06, | |
| "loss": 0.5861, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.166383163555236e-06, | |
| "loss": 0.5759, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.1569553920005527e-06, | |
| "loss": 0.5695, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.147512637547492e-06, | |
| "loss": 0.5654, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.1380550741740343e-06, | |
| "loss": 0.5945, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.1285828761310052e-06, | |
| "loss": 0.5642, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.1190962179388666e-06, | |
| "loss": 0.5874, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 3.109595274384501e-06, | |
| "loss": 0.5721, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.100080220517991e-06, | |
| "loss": 0.5978, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0905512316493976e-06, | |
| "loss": 0.5633, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0810084833455236e-06, | |
| "loss": 0.5646, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0714521514266834e-06, | |
| "loss": 0.5824, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.061882411963465e-06, | |
| "loss": 0.5965, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.05229944127348e-06, | |
| "loss": 0.5749, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0427034159181205e-06, | |
| "loss": 0.5765, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 3.0330945126993043e-06, | |
| "loss": 0.5619, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.0234729086562184e-06, | |
| "loss": 0.5582, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.013838781062053e-06, | |
| "loss": 0.5486, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 3.004192307420741e-06, | |
| "loss": 0.59, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.994533665463684e-06, | |
| "loss": 0.5491, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.9848630331464796e-06, | |
| "loss": 0.6037, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.9751805886456396e-06, | |
| "loss": 0.5722, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 2.9654865103553115e-06, | |
| "loss": 0.587, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.95578097688399e-06, | |
| "loss": 0.5757, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.9460641670512215e-06, | |
| "loss": 0.6111, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.936336259884318e-06, | |
| "loss": 0.5916, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.9265974346150538e-06, | |
| "loss": 0.5614, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.9168478706763613e-06, | |
| "loss": 0.5874, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.90708774769903e-06, | |
| "loss": 0.6018, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.8973172455083925e-06, | |
| "loss": 0.561, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 2.8875365441210138e-06, | |
| "loss": 0.5717, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.8777458237413744e-06, | |
| "loss": 0.5711, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.8679452647585497e-06, | |
| "loss": 0.5632, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.858135047742886e-06, | |
| "loss": 0.5911, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.848315353442673e-06, | |
| "loss": 0.602, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.838486362780817e-06, | |
| "loss": 0.5599, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.828648256851502e-06, | |
| "loss": 0.5685, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.818801216916859e-06, | |
| "loss": 0.598, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 2.808945424403622e-06, | |
| "loss": 0.5906, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.799081060899788e-06, | |
| "loss": 0.5477, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.789208308151268e-06, | |
| "loss": 0.57, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.779327348058543e-06, | |
| "loss": 0.5786, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.7694383626733088e-06, | |
| "loss": 0.5791, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.759541534195124e-06, | |
| "loss": 0.5783, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.7496370449680507e-06, | |
| "loss": 0.6006, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 2.7397250774772976e-06, | |
| "loss": 0.5791, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.7298058143458573e-06, | |
| "loss": 0.5588, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.7198794383311405e-06, | |
| "loss": 0.5801, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.7099461323216063e-06, | |
| "loss": 0.5396, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.7000060793334003e-06, | |
| "loss": 0.5913, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.6900594625069746e-06, | |
| "loss": 0.5806, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.6801064651037175e-06, | |
| "loss": 0.5557, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.670147270502576e-06, | |
| "loss": 0.5556, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 2.6601820621966784e-06, | |
| "loss": 0.574, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.6502110237899505e-06, | |
| "loss": 0.5932, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.6402343389937352e-06, | |
| "loss": 0.5636, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.63025219162341e-06, | |
| "loss": 0.5743, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.6202647655949946e-06, | |
| "loss": 0.5883, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.6102722449217657e-06, | |
| "loss": 0.5784, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.6002748137108672e-06, | |
| "loss": 0.5349, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.5902726561599163e-06, | |
| "loss": 0.5642, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 2.580265956553612e-06, | |
| "loss": 0.5918, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5702548992603357e-06, | |
| "loss": 0.5612, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5602396687287613e-06, | |
| "loss": 0.5738, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5502204494844483e-06, | |
| "loss": 0.5782, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5401974261264476e-06, | |
| "loss": 0.5748, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.530170783323899e-06, | |
| "loss": 0.5803, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5201407058126307e-06, | |
| "loss": 0.6119, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5101073783917504e-06, | |
| "loss": 0.5579, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 2.5000709859202455e-06, | |
| "loss": 0.5793, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.4900317133135754e-06, | |
| "loss": 0.5514, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.479989745540265e-06, | |
| "loss": 0.5795, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.469945267618494e-06, | |
| "loss": 0.5712, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.4598984646126927e-06, | |
| "loss": 0.5922, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.449849521630129e-06, | |
| "loss": 0.5623, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.4397986238175e-06, | |
| "loss": 0.5828, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 2.4297459563575166e-06, | |
| "loss": 0.5921, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.4196917044654974e-06, | |
| "loss": 0.5714, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.4096360533859527e-06, | |
| "loss": 0.5718, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.399579188389171e-06, | |
| "loss": 0.5527, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.3895212947678075e-06, | |
| "loss": 0.5995, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.3794625578334705e-06, | |
| "loss": 0.5862, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.369403162913303e-06, | |
| "loss": 0.5761, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.359343295346573e-06, | |
| "loss": 0.5829, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 2.3492831404812565e-06, | |
| "loss": 0.5687, | |
| "step": 4040 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.3392228836706236e-06, | |
| "loss": 0.5658, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.3291627102698204e-06, | |
| "loss": 0.5741, | |
| "step": 4060 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.3191028056324595e-06, | |
| "loss": 0.5737, | |
| "step": 4070 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.309043355107198e-06, | |
| "loss": 0.5744, | |
| "step": 4080 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.298984544034329e-06, | |
| "loss": 0.5942, | |
| "step": 4090 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.2889265577423624e-06, | |
| "loss": 0.5575, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.2788695815446142e-06, | |
| "loss": 0.5823, | |
| "step": 4110 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 2.2688138007357873e-06, | |
| "loss": 0.5637, | |
| "step": 4120 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.2587594005885604e-06, | |
| "loss": 0.5741, | |
| "step": 4130 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.2487065663501758e-06, | |
| "loss": 0.5866, | |
| "step": 4140 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.238655483239024e-06, | |
| "loss": 0.6083, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.2286063364412324e-06, | |
| "loss": 0.5604, | |
| "step": 4160 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.2185593111072506e-06, | |
| "loss": 0.5496, | |
| "step": 4170 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.208514592348443e-06, | |
| "loss": 0.5869, | |
| "step": 4180 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 2.1984723652336784e-06, | |
| "loss": 0.553, | |
| "step": 4190 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1884328147859143e-06, | |
| "loss": 0.575, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1783961259787957e-06, | |
| "loss": 0.6092, | |
| "step": 4210 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1683624837332402e-06, | |
| "loss": 0.6142, | |
| "step": 4220 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1583320729140354e-06, | |
| "loss": 0.5889, | |
| "step": 4230 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.148305078326432e-06, | |
| "loss": 0.5904, | |
| "step": 4240 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1382816847127392e-06, | |
| "loss": 0.5803, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.1282620767489166e-06, | |
| "loss": 0.5629, | |
| "step": 4260 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 2.118246439041178e-06, | |
| "loss": 0.5552, | |
| "step": 4270 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.1082349561225866e-06, | |
| "loss": 0.5773, | |
| "step": 4280 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0982278124496565e-06, | |
| "loss": 0.5803, | |
| "step": 4290 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.088225192398952e-06, | |
| "loss": 0.5898, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0782272802636917e-06, | |
| "loss": 0.5627, | |
| "step": 4310 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.068234260250355e-06, | |
| "loss": 0.5964, | |
| "step": 4320 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.0582463164752854e-06, | |
| "loss": 0.5625, | |
| "step": 4330 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.048263632961298e-06, | |
| "loss": 0.5695, | |
| "step": 4340 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 2.038286393634294e-06, | |
| "loss": 0.5741, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.028314782319864e-06, | |
| "loss": 0.5664, | |
| "step": 4360 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.018348982739907e-06, | |
| "loss": 0.5539, | |
| "step": 4370 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 2.008389178509244e-06, | |
| "loss": 0.5762, | |
| "step": 4380 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.998435553132236e-06, | |
| "loss": 0.551, | |
| "step": 4390 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.988488289999399e-06, | |
| "loss": 0.5796, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.978547572384028e-06, | |
| "loss": 0.5268, | |
| "step": 4410 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.968613583438823e-06, | |
| "loss": 0.5741, | |
| "step": 4420 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9586865061925098e-06, | |
| "loss": 0.5521, | |
| "step": 4430 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9487665235464706e-06, | |
| "loss": 0.5662, | |
| "step": 4440 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9388538182713696e-06, | |
| "loss": 0.5716, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9289485730037945e-06, | |
| "loss": 0.5942, | |
| "step": 4460 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.919050970242882e-06, | |
| "loss": 0.5511, | |
| "step": 4470 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9091611923469618e-06, | |
| "loss": 0.5591, | |
| "step": 4480 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.899279421530194e-06, | |
| "loss": 0.5656, | |
| "step": 4490 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.889405839859211e-06, | |
| "loss": 0.5657, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8795406292497649e-06, | |
| "loss": 0.5951, | |
| "step": 4510 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8696839714633764e-06, | |
| "loss": 0.5599, | |
| "step": 4520 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8598360481039848e-06, | |
| "loss": 0.5968, | |
| "step": 4530 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8499970406145992e-06, | |
| "loss": 0.5677, | |
| "step": 4540 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8401671302739611e-06, | |
| "loss": 0.5941, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8303464981932012e-06, | |
| "loss": 0.5848, | |
| "step": 4560 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8205353253125017e-06, | |
| "loss": 0.5847, | |
| "step": 4570 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.8107337923977659e-06, | |
| "loss": 0.5591, | |
| "step": 4580 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.8009420800372823e-06, | |
| "loss": 0.5683, | |
| "step": 4590 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7911603686384046e-06, | |
| "loss": 0.5494, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7813888384242203e-06, | |
| "loss": 0.5804, | |
| "step": 4610 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7716276694302366e-06, | |
| "loss": 0.5901, | |
| "step": 4620 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7618770415010592e-06, | |
| "loss": 0.5661, | |
| "step": 4630 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7521371342870795e-06, | |
| "loss": 0.5529, | |
| "step": 4640 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.7424081272411653e-06, | |
| "loss": 0.5664, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.732690199615355e-06, | |
| "loss": 0.5516, | |
| "step": 4660 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.7229835304575547e-06, | |
| "loss": 0.546, | |
| "step": 4670 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.7132882986082373e-06, | |
| "loss": 0.584, | |
| "step": 4680 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.7036046826971493e-06, | |
| "loss": 0.5432, | |
| "step": 4690 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.6939328611400214e-06, | |
| "loss": 0.5955, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.684273012135277e-06, | |
| "loss": 0.5605, | |
| "step": 4710 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.674625313660753e-06, | |
| "loss": 0.5979, | |
| "step": 4720 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.664989943470417e-06, | |
| "loss": 0.6037, | |
| "step": 4730 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.6553670790910962e-06, | |
| "loss": 0.5521, | |
| "step": 4740 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.645756897819203e-06, | |
| "loss": 0.5812, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.6361595767174719e-06, | |
| "loss": 0.5815, | |
| "step": 4760 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.6265752926116924e-06, | |
| "loss": 0.576, | |
| "step": 4770 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.6170042220874564e-06, | |
| "loss": 0.561, | |
| "step": 4780 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.6074465414869008e-06, | |
| "loss": 0.5405, | |
| "step": 4790 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.597902426905461e-06, | |
| "loss": 0.5599, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.5883720541886253e-06, | |
| "loss": 0.5405, | |
| "step": 4810 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5788555989286928e-06, | |
| "loss": 0.5558, | |
| "step": 4820 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5693532364615437e-06, | |
| "loss": 0.5668, | |
| "step": 4830 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5598651418634044e-06, | |
| "loss": 0.5857, | |
| "step": 4840 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5503914899476223e-06, | |
| "loss": 0.5731, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.540932455261447e-06, | |
| "loss": 0.5599, | |
| "step": 4860 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5314882120828105e-06, | |
| "loss": 0.5644, | |
| "step": 4870 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5220589344171221e-06, | |
| "loss": 0.5939, | |
| "step": 4880 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.5126447959940568e-06, | |
| "loss": 0.5872, | |
| "step": 4890 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.5032459702643585e-06, | |
| "loss": 0.5833, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4938626303966397e-06, | |
| "loss": 0.5505, | |
| "step": 4910 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4844949492741964e-06, | |
| "loss": 0.5886, | |
| "step": 4920 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4751430994918185e-06, | |
| "loss": 0.5804, | |
| "step": 4930 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4658072533526132e-06, | |
| "loss": 0.5597, | |
| "step": 4940 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4564875828648266e-06, | |
| "loss": 0.5704, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.4471842597386774e-06, | |
| "loss": 0.5827, | |
| "step": 4960 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.4378974553831926e-06, | |
| "loss": 0.5458, | |
| "step": 4970 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.42862734090305e-06, | |
| "loss": 0.5631, | |
| "step": 4980 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.4193740870954227e-06, | |
| "loss": 0.5672, | |
| "step": 4990 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.410137864446838e-06, | |
| "loss": 0.5692, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.4009188431300279e-06, | |
| "loss": 0.5682, | |
| "step": 5010 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.3917171930008023e-06, | |
| "loss": 0.5557, | |
| "step": 5020 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.3825330835949142e-06, | |
| "loss": 0.5679, | |
| "step": 5030 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.3733666841249394e-06, | |
| "loss": 0.5671, | |
| "step": 5040 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3642181634771537e-06, | |
| "loss": 0.5714, | |
| "step": 5050 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3550876902084279e-06, | |
| "loss": 0.5556, | |
| "step": 5060 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3459754325431172e-06, | |
| "loss": 0.565, | |
| "step": 5070 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3368815583699631e-06, | |
| "loss": 0.553, | |
| "step": 5080 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3278062352390044e-06, | |
| "loss": 0.5736, | |
| "step": 5090 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3187496303584809e-06, | |
| "loss": 0.5827, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3097119105917598e-06, | |
| "loss": 0.5623, | |
| "step": 5110 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.3006932424542622e-06, | |
| "loss": 0.5768, | |
| "step": 5120 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2916937921103908e-06, | |
| "loss": 0.5592, | |
| "step": 5130 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2827137253704695e-06, | |
| "loss": 0.5888, | |
| "step": 5140 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2737532076876895e-06, | |
| "loss": 0.5592, | |
| "step": 5150 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.26481240415506e-06, | |
| "loss": 0.5783, | |
| "step": 5160 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2558914795023695e-06, | |
| "loss": 0.5788, | |
| "step": 5170 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2469905980931466e-06, | |
| "loss": 0.5964, | |
| "step": 5180 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2381099239216298e-06, | |
| "loss": 0.5605, | |
| "step": 5190 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.2292496206097555e-06, | |
| "loss": 0.5575, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.2204098514041339e-06, | |
| "loss": 0.5855, | |
| "step": 5210 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.211590779173044e-06, | |
| "loss": 0.589, | |
| "step": 5220 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.2027925664034376e-06, | |
| "loss": 0.5703, | |
| "step": 5230 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.1940153751979358e-06, | |
| "loss": 0.598, | |
| "step": 5240 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.1852593672718508e-06, | |
| "loss": 0.5744, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.1765247039502054e-06, | |
| "loss": 0.5595, | |
| "step": 5260 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.1678115461647562e-06, | |
| "loss": 0.5845, | |
| "step": 5270 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.159120054451032e-06, | |
| "loss": 0.5748, | |
| "step": 5280 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.1504503889453754e-06, | |
| "loss": 0.5799, | |
| "step": 5290 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.1418027093819907e-06, | |
| "loss": 0.5909, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.1331771750900054e-06, | |
| "loss": 0.5644, | |
| "step": 5310 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.12457394499053e-06, | |
| "loss": 0.5593, | |
| "step": 5320 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.1159931775937283e-06, | |
| "loss": 0.5559, | |
| "step": 5330 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.1074350309959058e-06, | |
| "loss": 0.5893, | |
| "step": 5340 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.0988996628765874e-06, | |
| "loss": 0.583, | |
| "step": 5350 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0903872304956168e-06, | |
| "loss": 0.5815, | |
| "step": 5360 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0818978906902608e-06, | |
| "loss": 0.5769, | |
| "step": 5370 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.073431799872313e-06, | |
| "loss": 0.5678, | |
| "step": 5380 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0649891140252177e-06, | |
| "loss": 0.602, | |
| "step": 5390 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0565699887011967e-06, | |
| "loss": 0.6014, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0481745790183785e-06, | |
| "loss": 0.5416, | |
| "step": 5410 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.039803039657944e-06, | |
| "loss": 0.5702, | |
| "step": 5420 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.0314555248612746e-06, | |
| "loss": 0.5689, | |
| "step": 5430 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.0231321884271116e-06, | |
| "loss": 0.5393, | |
| "step": 5440 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.0148331837087237e-06, | |
| "loss": 0.5694, | |
| "step": 5450 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.0065586636110787e-06, | |
| "loss": 0.534, | |
| "step": 5460 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 9.983087805880254e-07, | |
| "loss": 0.5588, | |
| "step": 5470 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 9.900836866394918e-07, | |
| "loss": 0.5538, | |
| "step": 5480 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 9.818835333086761e-07, | |
| "loss": 0.5974, | |
| "step": 5490 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 9.737084716792582e-07, | |
| "loss": 0.5802, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.655586523726204e-07, | |
| "loss": 0.5757, | |
| "step": 5510 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.574342255450618e-07, | |
| "loss": 0.5545, | |
| "step": 5520 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.493353408850409e-07, | |
| "loss": 0.5949, | |
| "step": 5530 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.412621476104161e-07, | |
| "loss": 0.5759, | |
| "step": 5540 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.332147944656929e-07, | |
| "loss": 0.565, | |
| "step": 5550 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.251934297192861e-07, | |
| "loss": 0.5509, | |
| "step": 5560 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.171982011607873e-07, | |
| "loss": 0.5278, | |
| "step": 5570 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 9.092292560982419e-07, | |
| "loss": 0.5672, | |
| "step": 5580 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 9.01286741355435e-07, | |
| "loss": 0.5837, | |
| "step": 5590 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.933708032691889e-07, | |
| "loss": 0.5698, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.854815876866593e-07, | |
| "loss": 0.5638, | |
| "step": 5610 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.776192399626588e-07, | |
| "loss": 0.555, | |
| "step": 5620 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.697839049569696e-07, | |
| "loss": 0.5713, | |
| "step": 5630 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.61975727031679e-07, | |
| "loss": 0.5568, | |
| "step": 5640 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.541948500485217e-07, | |
| "loss": 0.592, | |
| "step": 5650 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 8.464414173662211e-07, | |
| "loss": 0.5924, | |
| "step": 5660 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.387155718378564e-07, | |
| "loss": 0.5891, | |
| "step": 5670 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.310174558082282e-07, | |
| "loss": 0.5575, | |
| "step": 5680 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.233472111112335e-07, | |
| "loss": 0.5733, | |
| "step": 5690 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.157049790672542e-07, | |
| "loss": 0.5687, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.08090900480554e-07, | |
| "loss": 0.5738, | |
| "step": 5710 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 8.005051156366824e-07, | |
| "loss": 0.5767, | |
| "step": 5720 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.929477642998909e-07, | |
| "loss": 0.5511, | |
| "step": 5730 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.854189857105611e-07, | |
| "loss": 0.5306, | |
| "step": 5740 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.779189185826306e-07, | |
| "loss": 0.5812, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.704477011010477e-07, | |
| "loss": 0.5593, | |
| "step": 5760 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.630054709192186e-07, | |
| "loss": 0.563, | |
| "step": 5770 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.555923651564721e-07, | |
| "loss": 0.5411, | |
| "step": 5780 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.482085203955382e-07, | |
| "loss": 0.6212, | |
| "step": 5790 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.408540726800224e-07, | |
| "loss": 0.6006, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 7.335291575119075e-07, | |
| "loss": 0.5715, | |
| "step": 5810 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 7.262339098490546e-07, | |
| "loss": 0.5517, | |
| "step": 5820 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 7.189684641027155e-07, | |
| "loss": 0.5787, | |
| "step": 5830 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 7.11732954135054e-07, | |
| "loss": 0.58, | |
| "step": 5840 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 7.045275132566867e-07, | |
| "loss": 0.5606, | |
| "step": 5850 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.973522742242199e-07, | |
| "loss": 0.5606, | |
| "step": 5860 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.902073692378057e-07, | |
| "loss": 0.5469, | |
| "step": 5870 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.830929299387097e-07, | |
| "loss": 0.5561, | |
| "step": 5880 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.760090874068779e-07, | |
| "loss": 0.5826, | |
| "step": 5890 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.68955972158531e-07, | |
| "loss": 0.5532, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.619337141437523e-07, | |
| "loss": 0.5563, | |
| "step": 5910 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.549424427440968e-07, | |
| "loss": 0.5595, | |
| "step": 5920 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.479822867702067e-07, | |
| "loss": 0.5698, | |
| "step": 5930 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.41053374459439e-07, | |
| "loss": 0.5702, | |
| "step": 5940 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.341558334735e-07, | |
| "loss": 0.5635, | |
| "step": 5950 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.272897908960978e-07, | |
| "loss": 0.5857, | |
| "step": 5960 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.204553732305978e-07, | |
| "loss": 0.5711, | |
| "step": 5970 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 6.136527063976885e-07, | |
| "loss": 0.5559, | |
| "step": 5980 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 6.068819157330717e-07, | |
| "loss": 0.6099, | |
| "step": 5990 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 6.001431259851415e-07, | |
| "loss": 0.5783, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.934364613126937e-07, | |
| "loss": 0.5771, | |
| "step": 6010 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.867620452826375e-07, | |
| "loss": 0.5711, | |
| "step": 6020 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.801200008677121e-07, | |
| "loss": 0.5851, | |
| "step": 6030 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.735104504442313e-07, | |
| "loss": 0.5726, | |
| "step": 6040 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.669335157898206e-07, | |
| "loss": 0.5488, | |
| "step": 6050 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.603893180811778e-07, | |
| "loss": 0.5644, | |
| "step": 6060 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.538779778918377e-07, | |
| "loss": 0.5686, | |
| "step": 6070 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.473996151899528e-07, | |
| "loss": 0.5601, | |
| "step": 6080 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.409543493360816e-07, | |
| "loss": 0.5825, | |
| "step": 6090 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.345422990809909e-07, | |
| "loss": 0.5875, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.281635825634669e-07, | |
| "loss": 0.5729, | |
| "step": 6110 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 5.218183173081353e-07, | |
| "loss": 0.5807, | |
| "step": 6120 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 5.155066202233038e-07, | |
| "loss": 0.5762, | |
| "step": 6130 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 5.092286075988009e-07, | |
| "loss": 0.5522, | |
| "step": 6140 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 5.029843951038359e-07, | |
| "loss": 0.5893, | |
| "step": 6150 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.9677409778487e-07, | |
| "loss": 0.5541, | |
| "step": 6160 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.905978300634908e-07, | |
| "loss": 0.5514, | |
| "step": 6170 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.844557057343104e-07, | |
| "loss": 0.592, | |
| "step": 6180 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.783478379628659e-07, | |
| "loss": 0.5586, | |
| "step": 6190 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.722743392835343e-07, | |
| "loss": 0.5806, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.66235321597459e-07, | |
| "loss": 0.5745, | |
| "step": 6210 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.6023089617048937e-07, | |
| "loss": 0.595, | |
| "step": 6220 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.542611736311283e-07, | |
| "loss": 0.58, | |
| "step": 6230 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.483262639684977e-07, | |
| "loss": 0.5602, | |
| "step": 6240 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.4242627653030966e-07, | |
| "loss": 0.5682, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.365613200208484e-07, | |
| "loss": 0.5957, | |
| "step": 6260 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 4.307315024989758e-07, | |
| "loss": 0.574, | |
| "step": 6270 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.2493693137613284e-07, | |
| "loss": 0.5369, | |
| "step": 6280 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.191777134143632e-07, | |
| "loss": 0.5248, | |
| "step": 6290 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.1345395472434947e-07, | |
| "loss": 0.5676, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.077657607634508e-07, | |
| "loss": 0.5623, | |
| "step": 6310 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 4.0211323633376616e-07, | |
| "loss": 0.5743, | |
| "step": 6320 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.96496485580203e-07, | |
| "loss": 0.5527, | |
| "step": 6330 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.9091561198855486e-07, | |
| "loss": 0.5762, | |
| "step": 6340 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.85370718383597e-07, | |
| "loss": 0.5795, | |
| "step": 6350 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.79861906927192e-07, | |
| "loss": 0.5694, | |
| "step": 6360 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.743892791164063e-07, | |
| "loss": 0.5891, | |
| "step": 6370 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.6895293578164284e-07, | |
| "loss": 0.571, | |
| "step": 6380 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.6355297708477984e-07, | |
| "loss": 0.5618, | |
| "step": 6390 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.5818950251732617e-07, | |
| "loss": 0.5807, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.5286261089859146e-07, | |
| "loss": 0.5695, | |
| "step": 6410 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.4757240037386145e-07, | |
| "loss": 0.5725, | |
| "step": 6420 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.4231896841259104e-07, | |
| "loss": 0.5618, | |
| "step": 6430 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.3710241180660966e-07, | |
| "loss": 0.5617, | |
| "step": 6440 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.319228266683365e-07, | |
| "loss": 0.5889, | |
| "step": 6450 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.267803084290091e-07, | |
| "loss": 0.5875, | |
| "step": 6460 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.216749518369286e-07, | |
| "loss": 0.5585, | |
| "step": 6470 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.1660685095570927e-07, | |
| "loss": 0.5613, | |
| "step": 6480 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.1157609916254893e-07, | |
| "loss": 0.5827, | |
| "step": 6490 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.0658278914650694e-07, | |
| "loss": 0.536, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 3.016270129067964e-07, | |
| "loss": 0.5667, | |
| "step": 6510 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.9670886175109e-07, | |
| "loss": 0.5759, | |
| "step": 6520 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.9182842629383864e-07, | |
| "loss": 0.5817, | |
| "step": 6530 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.869857964545971e-07, | |
| "loss": 0.5621, | |
| "step": 6540 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.8218106145637425e-07, | |
| "loss": 0.5713, | |
| "step": 6550 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.77414309823984e-07, | |
| "loss": 0.5694, | |
| "step": 6560 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.7268562938241607e-07, | |
| "loss": 0.5928, | |
| "step": 6570 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.6799510725521807e-07, | |
| "loss": 0.5558, | |
| "step": 6580 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.633428298628892e-07, | |
| "loss": 0.5621, | |
| "step": 6590 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.587288829212894e-07, | |
| "loss": 0.5671, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.541533514400597e-07, | |
| "loss": 0.5849, | |
| "step": 6610 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.4961631972105535e-07, | |
| "loss": 0.5439, | |
| "step": 6620 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.451178713567926e-07, | |
| "loss": 0.5871, | |
| "step": 6630 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.4065808922890927e-07, | |
| "loss": 0.6058, | |
| "step": 6640 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.3623705550663684e-07, | |
| "loss": 0.5589, | |
| "step": 6650 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.3185485164528743e-07, | |
| "loss": 0.5592, | |
| "step": 6660 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.275115583847539e-07, | |
| "loss": 0.5976, | |
| "step": 6670 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.2320725574801746e-07, | |
| "loss": 0.5844, | |
| "step": 6680 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.189420230396803e-07, | |
| "loss": 0.5654, | |
| "step": 6690 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.1471593884449914e-07, | |
| "loss": 0.5768, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.1052908102593985e-07, | |
| "loss": 0.5814, | |
| "step": 6710 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.0638152672474132e-07, | |
| "loss": 0.6007, | |
| "step": 6720 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 2.0227335235749564e-07, | |
| "loss": 0.5712, | |
| "step": 6730 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9820463361523924e-07, | |
| "loss": 0.5537, | |
| "step": 6740 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.9417544546205923e-07, | |
| "loss": 0.546, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.9018586213371086e-07, | |
| "loss": 0.5498, | |
| "step": 6760 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8623595713625057e-07, | |
| "loss": 0.5738, | |
| "step": 6770 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.823258032446819e-07, | |
| "loss": 0.5741, | |
| "step": 6780 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.7845547250161414e-07, | |
| "loss": 0.5767, | |
| "step": 6790 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.7462503621593505e-07, | |
| "loss": 0.5909, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.708345649614984e-07, | |
| "loss": 0.5956, | |
| "step": 6810 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.6708412857581967e-07, | |
| "loss": 0.5817, | |
| "step": 6820 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.6337379615879572e-07, | |
| "loss": 0.5775, | |
| "step": 6830 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5970363607142522e-07, | |
| "loss": 0.5778, | |
| "step": 6840 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5607371593455302e-07, | |
| "loss": 0.5534, | |
| "step": 6850 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.5248410262762378e-07, | |
| "loss": 0.5519, | |
| "step": 6860 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.489348622874479e-07, | |
| "loss": 0.5612, | |
| "step": 6870 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4542606030698488e-07, | |
| "loss": 0.5542, | |
| "step": 6880 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.419577613341387e-07, | |
| "loss": 0.5832, | |
| "step": 6890 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.385300292705652e-07, | |
| "loss": 0.578, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.351429272704944e-07, | |
| "loss": 0.5643, | |
| "step": 6910 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.3179651773956942e-07, | |
| "loss": 0.5789, | |
| "step": 6920 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.2849086233369491e-07, | |
| "loss": 0.5766, | |
| "step": 6930 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.2522602195790083e-07, | |
| "loss": 0.5607, | |
| "step": 6940 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.220020567652213e-07, | |
| "loss": 0.5761, | |
| "step": 6950 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1881902615558554e-07, | |
| "loss": 0.5611, | |
| "step": 6960 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1567698877472404e-07, | |
| "loss": 0.5863, | |
| "step": 6970 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.1257600251308756e-07, | |
| "loss": 0.59, | |
| "step": 6980 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.0951612450478108e-07, | |
| "loss": 0.5798, | |
| "step": 6990 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.0649741112651027e-07, | |
| "loss": 0.5696, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.035199179965437e-07, | |
| "loss": 0.5625, | |
| "step": 7010 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.0058369997368691e-07, | |
| "loss": 0.5528, | |
| "step": 7020 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.768881115627403e-08, | |
| "loss": 0.5614, | |
| "step": 7030 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.483530488116837e-08, | |
| "loss": 0.5517, | |
| "step": 7040 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.202323372278011e-08, | |
| "loss": 0.5992, | |
| "step": 7050 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 8.92526494921002e-08, | |
| "loss": 0.5633, | |
| "step": 7060 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 8.652360323574267e-08, | |
| "loss": 0.5581, | |
| "step": 7070 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 8.383614523500516e-08, | |
| "loss": 0.5632, | |
| "step": 7080 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 8.119032500494352e-08, | |
| "loss": 0.567, | |
| "step": 7090 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.8586191293458e-08, | |
| "loss": 0.5361, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.602379208039681e-08, | |
| "loss": 0.608, | |
| "step": 7110 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.350317457667014e-08, | |
| "loss": 0.5594, | |
| "step": 7120 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 7.102438522338184e-08, | |
| "loss": 0.5746, | |
| "step": 7130 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.858746969097272e-08, | |
| "loss": 0.5698, | |
| "step": 7140 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.619247287838001e-08, | |
| "loss": 0.5633, | |
| "step": 7150 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.383943891220943e-08, | |
| "loss": 0.5627, | |
| "step": 7160 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.152841114592296e-08, | |
| "loss": 0.5629, | |
| "step": 7170 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.9259432159039674e-08, | |
| "loss": 0.5933, | |
| "step": 7180 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.703254375634967e-08, | |
| "loss": 0.5361, | |
| "step": 7190 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.484778696714723e-08, | |
| "loss": 0.5623, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 5.270520204447154e-08, | |
| "loss": 0.5728, | |
| "step": 7210 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 5.060482846436743e-08, | |
| "loss": 0.5516, | |
| "step": 7220 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.854670492515627e-08, | |
| "loss": 0.5855, | |
| "step": 7230 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.65308693467243e-08, | |
| "loss": 0.5654, | |
| "step": 7240 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.4557358869823455e-08, | |
| "loss": 0.6076, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.262620985538675e-08, | |
| "loss": 0.5791, | |
| "step": 7260 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.0737457883858756e-08, | |
| "loss": 0.5522, | |
| "step": 7270 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.889113775454016e-08, | |
| "loss": 0.5791, | |
| "step": 7280 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.7087283484946334e-08, | |
| "loss": 0.5661, | |
| "step": 7290 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.532592831018023e-08, | |
| "loss": 0.5753, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.36071046823214e-08, | |
| "loss": 0.5879, | |
| "step": 7310 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.193084426982625e-08, | |
| "loss": 0.5609, | |
| "step": 7320 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 3.029717795694515e-08, | |
| "loss": 0.5779, | |
| "step": 7330 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.870613584315517e-08, | |
| "loss": 0.5748, | |
| "step": 7340 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 2.7157747242602997e-08, | |
| "loss": 0.5437, | |
| "step": 7350 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.565204068356577e-08, | |
| "loss": 0.5945, | |
| "step": 7360 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.4189043907925973e-08, | |
| "loss": 0.5528, | |
| "step": 7370 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.276878387066014e-08, | |
| "loss": 0.5716, | |
| "step": 7380 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.139128673934133e-08, | |
| "loss": 0.6122, | |
| "step": 7390 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 2.0056577893658253e-08, | |
| "loss": 0.5787, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.876468192494691e-08, | |
| "loss": 0.5693, | |
| "step": 7410 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7515622635737304e-08, | |
| "loss": 0.5558, | |
| "step": 7420 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.6309423039316017e-08, | |
| "loss": 0.5533, | |
| "step": 7430 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.5146105359300794e-08, | |
| "loss": 0.5447, | |
| "step": 7440 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.4025691029232247e-08, | |
| "loss": 0.5703, | |
| "step": 7450 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.2948200692177845e-08, | |
| "loss": 0.5734, | |
| "step": 7460 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.191365420035225e-08, | |
| "loss": 0.5465, | |
| "step": 7470 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.0922070614752514e-08, | |
| "loss": 0.5691, | |
| "step": 7480 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.973468204804713e-09, | |
| "loss": 0.5532, | |
| "step": 7490 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 9.0678644480288e-09, | |
| "loss": 0.5912, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.205276029716462e-09, | |
| "loss": 0.5719, | |
| "step": 7510 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 7.385718842623043e-09, | |
| "loss": 0.5561, | |
| "step": 7520 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 6.60920798667584e-09, | |
| "loss": 0.5488, | |
| "step": 7530 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.875757768694904e-09, | |
| "loss": 0.5506, | |
| "step": 7540 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 5.185381702129971e-09, | |
| "loss": 0.5614, | |
| "step": 7550 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.538092506810916e-09, | |
| "loss": 0.5688, | |
| "step": 7560 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.933902108714092e-09, | |
| "loss": 0.5801, | |
| "step": 7570 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.372821639742191e-09, | |
| "loss": 0.5541, | |
| "step": 7580 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.8548614375194582e-09, | |
| "loss": 0.5836, | |
| "step": 7590 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.380031045200962e-09, | |
| "loss": 0.5542, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.948339211296432e-09, | |
| "loss": 0.5689, | |
| "step": 7610 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.5597938895097092e-09, | |
| "loss": 0.5931, | |
| "step": 7620 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2144022385922498e-09, | |
| "loss": 0.5175, | |
| "step": 7630 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.121706222106791e-10, | |
| "loss": 0.5868, | |
| "step": 7640 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.531046088294363e-10, | |
| "loss": 0.5521, | |
| "step": 7650 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.372089716092944e-10, | |
| "loss": 0.5685, | |
| "step": 7660 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.644876883181081e-10, | |
| "loss": 0.5718, | |
| "step": 7670 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.3494394125821567e-10, | |
| "loss": 0.5407, | |
| "step": 7680 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.858011720763194e-11, | |
| "loss": 0.5773, | |
| "step": 7690 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.397807376072982e-12, | |
| "loss": 0.592, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 0.5662159323692322, | |
| "eval_runtime": 58.2198, | |
| "eval_samples_per_second": 186.827, | |
| "eval_steps_per_second": 23.36, | |
| "step": 7705 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 7705, | |
| "total_flos": 6.229344692427817e+18, | |
| "train_loss": 0.6247228051840988, | |
| "train_runtime": 911.3454, | |
| "train_samples_per_second": 67.629, | |
| "train_steps_per_second": 8.455 | |
| } | |
| ], | |
| "max_steps": 7705, | |
| "num_train_epochs": 1, | |
| "total_flos": 6.229344692427817e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |