|
{ |
|
"best_metric": 41.1359, |
|
"best_model_checkpoint": "/kaggle/working/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted-amr-generation-v2-fted/checkpoint-3600", |
|
"epoch": 39.735099337748345, |
|
"eval_steps": 3600, |
|
"global_step": 3600, |
|
"is_hyper_param_search": false, |
|
"is_local_process_zero": true, |
|
"is_world_process_zero": true, |
|
"log_history": [ |
|
{ |
|
"epoch": 0.011037527593818985, |
|
"learning_rate": 5e-09, |
|
"loss": 2.1638, |
|
"step": 1 |
|
}, |
|
{ |
|
"epoch": 0.22075055187637968, |
|
"learning_rate": 1e-07, |
|
"loss": 1.9727, |
|
"step": 20 |
|
}, |
|
{ |
|
"epoch": 0.44150110375275936, |
|
"learning_rate": 2e-07, |
|
"loss": 2.0028, |
|
"step": 40 |
|
}, |
|
{ |
|
"epoch": 0.6622516556291391, |
|
"learning_rate": 3e-07, |
|
"loss": 2.0448, |
|
"step": 60 |
|
}, |
|
{ |
|
"epoch": 0.8830022075055187, |
|
"learning_rate": 4e-07, |
|
"loss": 1.9696, |
|
"step": 80 |
|
}, |
|
{ |
|
"epoch": 1.1037527593818985, |
|
"learning_rate": 5e-07, |
|
"loss": 1.9375, |
|
"step": 100 |
|
}, |
|
{ |
|
"epoch": 1.3245033112582782, |
|
"learning_rate": 6e-07, |
|
"loss": 1.9075, |
|
"step": 120 |
|
}, |
|
{ |
|
"epoch": 1.5452538631346577, |
|
"learning_rate": 7e-07, |
|
"loss": 1.9866, |
|
"step": 140 |
|
}, |
|
{ |
|
"epoch": 1.7660044150110377, |
|
"learning_rate": 8e-07, |
|
"loss": 1.882, |
|
"step": 160 |
|
}, |
|
{ |
|
"epoch": 1.9867549668874172, |
|
"learning_rate": 9e-07, |
|
"loss": 1.9916, |
|
"step": 180 |
|
}, |
|
{ |
|
"epoch": 2.207505518763797, |
|
"learning_rate": 1e-06, |
|
"loss": 1.9132, |
|
"step": 200 |
|
}, |
|
{ |
|
"epoch": 2.4282560706401766, |
|
"learning_rate": 9.996864111498258e-07, |
|
"loss": 1.9858, |
|
"step": 220 |
|
}, |
|
{ |
|
"epoch": 2.6490066225165565, |
|
"learning_rate": 9.993728222996515e-07, |
|
"loss": 1.9373, |
|
"step": 240 |
|
}, |
|
{ |
|
"epoch": 2.869757174392936, |
|
"learning_rate": 9.990592334494773e-07, |
|
"loss": 1.9146, |
|
"step": 260 |
|
}, |
|
{ |
|
"epoch": 3.0905077262693155, |
|
"learning_rate": 9.987456445993032e-07, |
|
"loss": 1.8665, |
|
"step": 280 |
|
}, |
|
{ |
|
"epoch": 3.3112582781456954, |
|
"learning_rate": 9.98432055749129e-07, |
|
"loss": 1.9419, |
|
"step": 300 |
|
}, |
|
{ |
|
"epoch": 3.5320088300220753, |
|
"learning_rate": 9.981184668989545e-07, |
|
"loss": 1.8332, |
|
"step": 320 |
|
}, |
|
{ |
|
"epoch": 3.752759381898455, |
|
"learning_rate": 9.978048780487803e-07, |
|
"loss": 1.8534, |
|
"step": 340 |
|
}, |
|
{ |
|
"epoch": 3.9735099337748343, |
|
"learning_rate": 9.974912891986062e-07, |
|
"loss": 1.9421, |
|
"step": 360 |
|
}, |
|
{ |
|
"epoch": 4.194260485651214, |
|
"learning_rate": 9.97177700348432e-07, |
|
"loss": 1.9374, |
|
"step": 380 |
|
}, |
|
{ |
|
"epoch": 4.415011037527594, |
|
"learning_rate": 9.96864111498258e-07, |
|
"loss": 1.8876, |
|
"step": 400 |
|
}, |
|
{ |
|
"epoch": 4.635761589403973, |
|
"learning_rate": 9.965505226480835e-07, |
|
"loss": 1.8593, |
|
"step": 420 |
|
}, |
|
{ |
|
"epoch": 4.856512141280353, |
|
"learning_rate": 9.962369337979094e-07, |
|
"loss": 1.8243, |
|
"step": 440 |
|
}, |
|
{ |
|
"epoch": 5.077262693156733, |
|
"learning_rate": 9.95923344947735e-07, |
|
"loss": 1.8791, |
|
"step": 460 |
|
}, |
|
{ |
|
"epoch": 5.298013245033113, |
|
"learning_rate": 9.95609756097561e-07, |
|
"loss": 1.8878, |
|
"step": 480 |
|
}, |
|
{ |
|
"epoch": 5.518763796909492, |
|
"learning_rate": 9.952961672473868e-07, |
|
"loss": 1.8985, |
|
"step": 500 |
|
}, |
|
{ |
|
"epoch": 5.739514348785872, |
|
"learning_rate": 9.949825783972126e-07, |
|
"loss": 1.84, |
|
"step": 520 |
|
}, |
|
{ |
|
"epoch": 5.960264900662252, |
|
"learning_rate": 9.946689895470383e-07, |
|
"loss": 1.8468, |
|
"step": 540 |
|
}, |
|
{ |
|
"epoch": 6.181015452538631, |
|
"learning_rate": 9.94355400696864e-07, |
|
"loss": 1.8469, |
|
"step": 560 |
|
}, |
|
{ |
|
"epoch": 6.401766004415011, |
|
"learning_rate": 9.940418118466898e-07, |
|
"loss": 1.8528, |
|
"step": 580 |
|
}, |
|
{ |
|
"epoch": 6.622516556291391, |
|
"learning_rate": 9.937282229965156e-07, |
|
"loss": 1.8269, |
|
"step": 600 |
|
}, |
|
{ |
|
"epoch": 6.843267108167771, |
|
"learning_rate": 9.934146341463415e-07, |
|
"loss": 1.8324, |
|
"step": 620 |
|
}, |
|
{ |
|
"epoch": 7.06401766004415, |
|
"learning_rate": 9.931010452961673e-07, |
|
"loss": 1.808, |
|
"step": 640 |
|
}, |
|
{ |
|
"epoch": 7.28476821192053, |
|
"learning_rate": 9.92787456445993e-07, |
|
"loss": 1.9003, |
|
"step": 660 |
|
}, |
|
{ |
|
"epoch": 7.50551876379691, |
|
"learning_rate": 9.924738675958186e-07, |
|
"loss": 1.908, |
|
"step": 680 |
|
}, |
|
{ |
|
"epoch": 7.72626931567329, |
|
"learning_rate": 9.921602787456445e-07, |
|
"loss": 1.8165, |
|
"step": 700 |
|
}, |
|
{ |
|
"epoch": 7.947019867549669, |
|
"learning_rate": 9.918466898954704e-07, |
|
"loss": 1.7924, |
|
"step": 720 |
|
}, |
|
{ |
|
"epoch": 8.167770419426049, |
|
"learning_rate": 9.915331010452962e-07, |
|
"loss": 1.8275, |
|
"step": 740 |
|
}, |
|
{ |
|
"epoch": 8.388520971302428, |
|
"learning_rate": 9.912195121951219e-07, |
|
"loss": 1.8033, |
|
"step": 760 |
|
}, |
|
{ |
|
"epoch": 8.609271523178808, |
|
"learning_rate": 9.909059233449477e-07, |
|
"loss": 1.8283, |
|
"step": 780 |
|
}, |
|
{ |
|
"epoch": 8.830022075055188, |
|
"learning_rate": 9.905923344947734e-07, |
|
"loss": 1.8361, |
|
"step": 800 |
|
}, |
|
{ |
|
"epoch": 9.050772626931566, |
|
"learning_rate": 9.902787456445992e-07, |
|
"loss": 1.8264, |
|
"step": 820 |
|
}, |
|
{ |
|
"epoch": 9.271523178807946, |
|
"learning_rate": 9.89965156794425e-07, |
|
"loss": 1.8157, |
|
"step": 840 |
|
}, |
|
{ |
|
"epoch": 9.492273730684326, |
|
"learning_rate": 9.89651567944251e-07, |
|
"loss": 1.8038, |
|
"step": 860 |
|
}, |
|
{ |
|
"epoch": 9.713024282560706, |
|
"learning_rate": 9.893379790940768e-07, |
|
"loss": 1.8495, |
|
"step": 880 |
|
}, |
|
{ |
|
"epoch": 9.933774834437086, |
|
"learning_rate": 9.890243902439024e-07, |
|
"loss": 1.8379, |
|
"step": 900 |
|
}, |
|
{ |
|
"epoch": 10.154525386313466, |
|
"learning_rate": 9.88710801393728e-07, |
|
"loss": 1.807, |
|
"step": 920 |
|
}, |
|
{ |
|
"epoch": 10.375275938189846, |
|
"learning_rate": 9.88397212543554e-07, |
|
"loss": 1.7354, |
|
"step": 940 |
|
}, |
|
{ |
|
"epoch": 10.596026490066226, |
|
"learning_rate": 9.880836236933798e-07, |
|
"loss": 1.7909, |
|
"step": 960 |
|
}, |
|
{ |
|
"epoch": 10.816777041942604, |
|
"learning_rate": 9.877700348432054e-07, |
|
"loss": 1.8504, |
|
"step": 980 |
|
}, |
|
{ |
|
"epoch": 11.037527593818984, |
|
"learning_rate": 9.874564459930313e-07, |
|
"loss": 1.7565, |
|
"step": 1000 |
|
}, |
|
{ |
|
"epoch": 11.258278145695364, |
|
"learning_rate": 9.871428571428572e-07, |
|
"loss": 1.8464, |
|
"step": 1020 |
|
}, |
|
{ |
|
"epoch": 11.479028697571744, |
|
"learning_rate": 9.868292682926828e-07, |
|
"loss": 1.7463, |
|
"step": 1040 |
|
}, |
|
{ |
|
"epoch": 11.699779249448124, |
|
"learning_rate": 9.865156794425087e-07, |
|
"loss": 1.8286, |
|
"step": 1060 |
|
}, |
|
{ |
|
"epoch": 11.920529801324504, |
|
"learning_rate": 9.862020905923345e-07, |
|
"loss": 1.7821, |
|
"step": 1080 |
|
}, |
|
{ |
|
"epoch": 12.141280353200884, |
|
"learning_rate": 9.858885017421604e-07, |
|
"loss": 1.8069, |
|
"step": 1100 |
|
}, |
|
{ |
|
"epoch": 12.362030905077262, |
|
"learning_rate": 9.85574912891986e-07, |
|
"loss": 1.7943, |
|
"step": 1120 |
|
}, |
|
{ |
|
"epoch": 12.582781456953642, |
|
"learning_rate": 9.852613240418117e-07, |
|
"loss": 1.7535, |
|
"step": 1140 |
|
}, |
|
{ |
|
"epoch": 12.803532008830022, |
|
"learning_rate": 9.849477351916375e-07, |
|
"loss": 1.7881, |
|
"step": 1160 |
|
}, |
|
{ |
|
"epoch": 13.024282560706402, |
|
"learning_rate": 9.846341463414634e-07, |
|
"loss": 1.815, |
|
"step": 1180 |
|
}, |
|
{ |
|
"epoch": 13.245033112582782, |
|
"learning_rate": 9.843205574912892e-07, |
|
"loss": 1.8258, |
|
"step": 1200 |
|
}, |
|
{ |
|
"epoch": 13.465783664459162, |
|
"learning_rate": 9.840069686411149e-07, |
|
"loss": 1.7228, |
|
"step": 1220 |
|
}, |
|
{ |
|
"epoch": 13.686534216335541, |
|
"learning_rate": 9.836933797909407e-07, |
|
"loss": 1.7866, |
|
"step": 1240 |
|
}, |
|
{ |
|
"epoch": 13.90728476821192, |
|
"learning_rate": 9.833797909407664e-07, |
|
"loss": 1.7871, |
|
"step": 1260 |
|
}, |
|
{ |
|
"epoch": 14.1280353200883, |
|
"learning_rate": 9.830662020905923e-07, |
|
"loss": 1.7238, |
|
"step": 1280 |
|
}, |
|
{ |
|
"epoch": 14.34878587196468, |
|
"learning_rate": 9.827526132404181e-07, |
|
"loss": 1.805, |
|
"step": 1300 |
|
}, |
|
{ |
|
"epoch": 14.56953642384106, |
|
"learning_rate": 9.82439024390244e-07, |
|
"loss": 1.7909, |
|
"step": 1320 |
|
}, |
|
{ |
|
"epoch": 14.79028697571744, |
|
"learning_rate": 9.821254355400698e-07, |
|
"loss": 1.7861, |
|
"step": 1340 |
|
}, |
|
{ |
|
"epoch": 15.01103752759382, |
|
"learning_rate": 9.818118466898953e-07, |
|
"loss": 1.7651, |
|
"step": 1360 |
|
}, |
|
{ |
|
"epoch": 15.2317880794702, |
|
"learning_rate": 9.814982578397211e-07, |
|
"loss": 1.7666, |
|
"step": 1380 |
|
}, |
|
{ |
|
"epoch": 15.45253863134658, |
|
"learning_rate": 9.81184668989547e-07, |
|
"loss": 1.7159, |
|
"step": 1400 |
|
}, |
|
{ |
|
"epoch": 15.673289183222957, |
|
"learning_rate": 9.808710801393728e-07, |
|
"loss": 1.7701, |
|
"step": 1420 |
|
}, |
|
{ |
|
"epoch": 15.894039735099337, |
|
"learning_rate": 9.805574912891987e-07, |
|
"loss": 1.7977, |
|
"step": 1440 |
|
}, |
|
{ |
|
"epoch": 16.11479028697572, |
|
"learning_rate": 9.802439024390243e-07, |
|
"loss": 1.7805, |
|
"step": 1460 |
|
}, |
|
{ |
|
"epoch": 16.335540838852097, |
|
"learning_rate": 9.799303135888502e-07, |
|
"loss": 1.7711, |
|
"step": 1480 |
|
}, |
|
{ |
|
"epoch": 16.556291390728475, |
|
"learning_rate": 9.796167247386758e-07, |
|
"loss": 1.6767, |
|
"step": 1500 |
|
}, |
|
{ |
|
"epoch": 16.777041942604857, |
|
"learning_rate": 9.793031358885017e-07, |
|
"loss": 1.722, |
|
"step": 1520 |
|
}, |
|
{ |
|
"epoch": 16.997792494481235, |
|
"learning_rate": 9.789895470383276e-07, |
|
"loss": 1.783, |
|
"step": 1540 |
|
}, |
|
{ |
|
"epoch": 17.218543046357617, |
|
"learning_rate": 9.786759581881534e-07, |
|
"loss": 1.7233, |
|
"step": 1560 |
|
}, |
|
{ |
|
"epoch": 17.439293598233995, |
|
"learning_rate": 9.78362369337979e-07, |
|
"loss": 1.7209, |
|
"step": 1580 |
|
}, |
|
{ |
|
"epoch": 17.660044150110377, |
|
"learning_rate": 9.780487804878047e-07, |
|
"loss": 1.6968, |
|
"step": 1600 |
|
}, |
|
{ |
|
"epoch": 17.880794701986755, |
|
"learning_rate": 9.777351916376306e-07, |
|
"loss": 1.6818, |
|
"step": 1620 |
|
}, |
|
{ |
|
"epoch": 18.101545253863133, |
|
"learning_rate": 9.774216027874564e-07, |
|
"loss": 1.7819, |
|
"step": 1640 |
|
}, |
|
{ |
|
"epoch": 18.322295805739515, |
|
"learning_rate": 9.771080139372823e-07, |
|
"loss": 1.7445, |
|
"step": 1660 |
|
}, |
|
{ |
|
"epoch": 18.543046357615893, |
|
"learning_rate": 9.76794425087108e-07, |
|
"loss": 1.6888, |
|
"step": 1680 |
|
}, |
|
{ |
|
"epoch": 18.763796909492275, |
|
"learning_rate": 9.764808362369338e-07, |
|
"loss": 1.7562, |
|
"step": 1700 |
|
}, |
|
{ |
|
"epoch": 18.984547461368653, |
|
"learning_rate": 9.761672473867594e-07, |
|
"loss": 1.7368, |
|
"step": 1720 |
|
}, |
|
{ |
|
"epoch": 19.205298013245034, |
|
"learning_rate": 9.758536585365853e-07, |
|
"loss": 1.6816, |
|
"step": 1740 |
|
}, |
|
{ |
|
"epoch": 19.426048565121413, |
|
"learning_rate": 9.755400696864111e-07, |
|
"loss": 1.7307, |
|
"step": 1760 |
|
}, |
|
{ |
|
"epoch": 19.64679911699779, |
|
"learning_rate": 9.75226480836237e-07, |
|
"loss": 1.7313, |
|
"step": 1780 |
|
}, |
|
{ |
|
"epoch": 19.867549668874172, |
|
"learning_rate": 9.749128919860627e-07, |
|
"loss": 1.7407, |
|
"step": 1800 |
|
}, |
|
{ |
|
"epoch": 20.08830022075055, |
|
"learning_rate": 9.745993031358883e-07, |
|
"loss": 1.6885, |
|
"step": 1820 |
|
}, |
|
{ |
|
"epoch": 20.309050772626932, |
|
"learning_rate": 9.742857142857142e-07, |
|
"loss": 1.6741, |
|
"step": 1840 |
|
}, |
|
{ |
|
"epoch": 20.52980132450331, |
|
"learning_rate": 9.7397212543554e-07, |
|
"loss": 1.6958, |
|
"step": 1860 |
|
}, |
|
{ |
|
"epoch": 20.750551876379692, |
|
"learning_rate": 9.736585365853659e-07, |
|
"loss": 1.7036, |
|
"step": 1880 |
|
}, |
|
{ |
|
"epoch": 20.97130242825607, |
|
"learning_rate": 9.733449477351917e-07, |
|
"loss": 1.7489, |
|
"step": 1900 |
|
}, |
|
{ |
|
"epoch": 21.192052980132452, |
|
"learning_rate": 9.730313588850174e-07, |
|
"loss": 1.7719, |
|
"step": 1920 |
|
}, |
|
{ |
|
"epoch": 21.41280353200883, |
|
"learning_rate": 9.72717770034843e-07, |
|
"loss": 1.7401, |
|
"step": 1940 |
|
}, |
|
{ |
|
"epoch": 21.63355408388521, |
|
"learning_rate": 9.724041811846689e-07, |
|
"loss": 1.7247, |
|
"step": 1960 |
|
}, |
|
{ |
|
"epoch": 21.85430463576159, |
|
"learning_rate": 9.720905923344947e-07, |
|
"loss": 1.7175, |
|
"step": 1980 |
|
}, |
|
{ |
|
"epoch": 22.075055187637968, |
|
"learning_rate": 9.717770034843206e-07, |
|
"loss": 1.7288, |
|
"step": 2000 |
|
}, |
|
{ |
|
"epoch": 22.29580573951435, |
|
"learning_rate": 9.714634146341462e-07, |
|
"loss": 1.7068, |
|
"step": 2020 |
|
}, |
|
{ |
|
"epoch": 22.516556291390728, |
|
"learning_rate": 9.71149825783972e-07, |
|
"loss": 1.6812, |
|
"step": 2040 |
|
}, |
|
{ |
|
"epoch": 22.73730684326711, |
|
"learning_rate": 9.708362369337977e-07, |
|
"loss": 1.6772, |
|
"step": 2060 |
|
}, |
|
{ |
|
"epoch": 22.958057395143488, |
|
"learning_rate": 9.705226480836236e-07, |
|
"loss": 1.7103, |
|
"step": 2080 |
|
}, |
|
{ |
|
"epoch": 23.178807947019866, |
|
"learning_rate": 9.702090592334495e-07, |
|
"loss": 1.6984, |
|
"step": 2100 |
|
}, |
|
{ |
|
"epoch": 23.399558498896248, |
|
"learning_rate": 9.698954703832753e-07, |
|
"loss": 1.7281, |
|
"step": 2120 |
|
}, |
|
{ |
|
"epoch": 23.620309050772626, |
|
"learning_rate": 9.695818815331012e-07, |
|
"loss": 1.7123, |
|
"step": 2140 |
|
}, |
|
{ |
|
"epoch": 23.841059602649008, |
|
"learning_rate": 9.692682926829266e-07, |
|
"loss": 1.7174, |
|
"step": 2160 |
|
}, |
|
{ |
|
"epoch": 24.061810154525386, |
|
"learning_rate": 9.689547038327525e-07, |
|
"loss": 1.7281, |
|
"step": 2180 |
|
}, |
|
{ |
|
"epoch": 24.282560706401767, |
|
"learning_rate": 9.686411149825783e-07, |
|
"loss": 1.7189, |
|
"step": 2200 |
|
}, |
|
{ |
|
"epoch": 24.503311258278146, |
|
"learning_rate": 9.683275261324042e-07, |
|
"loss": 1.6272, |
|
"step": 2220 |
|
}, |
|
{ |
|
"epoch": 24.724061810154524, |
|
"learning_rate": 9.6801393728223e-07, |
|
"loss": 1.6588, |
|
"step": 2240 |
|
}, |
|
{ |
|
"epoch": 24.944812362030905, |
|
"learning_rate": 9.677003484320557e-07, |
|
"loss": 1.6627, |
|
"step": 2260 |
|
}, |
|
{ |
|
"epoch": 25.165562913907284, |
|
"learning_rate": 9.673867595818815e-07, |
|
"loss": 1.6504, |
|
"step": 2280 |
|
}, |
|
{ |
|
"epoch": 25.386313465783665, |
|
"learning_rate": 9.670731707317072e-07, |
|
"loss": 1.6847, |
|
"step": 2300 |
|
}, |
|
{ |
|
"epoch": 25.607064017660043, |
|
"learning_rate": 9.66759581881533e-07, |
|
"loss": 1.6793, |
|
"step": 2320 |
|
}, |
|
{ |
|
"epoch": 25.827814569536425, |
|
"learning_rate": 9.66445993031359e-07, |
|
"loss": 1.7009, |
|
"step": 2340 |
|
}, |
|
{ |
|
"epoch": 26.048565121412803, |
|
"learning_rate": 9.661324041811848e-07, |
|
"loss": 1.6581, |
|
"step": 2360 |
|
}, |
|
{ |
|
"epoch": 26.26931567328918, |
|
"learning_rate": 9.658188153310104e-07, |
|
"loss": 1.6648, |
|
"step": 2380 |
|
}, |
|
{ |
|
"epoch": 26.490066225165563, |
|
"learning_rate": 9.65505226480836e-07, |
|
"loss": 1.6303, |
|
"step": 2400 |
|
}, |
|
{ |
|
"epoch": 26.71081677704194, |
|
"learning_rate": 9.65191637630662e-07, |
|
"loss": 1.7222, |
|
"step": 2420 |
|
}, |
|
{ |
|
"epoch": 26.931567328918323, |
|
"learning_rate": 9.648780487804878e-07, |
|
"loss": 1.7164, |
|
"step": 2440 |
|
}, |
|
{ |
|
"epoch": 27.1523178807947, |
|
"learning_rate": 9.645644599303136e-07, |
|
"loss": 1.6894, |
|
"step": 2460 |
|
}, |
|
{ |
|
"epoch": 27.373068432671083, |
|
"learning_rate": 9.642508710801393e-07, |
|
"loss": 1.6904, |
|
"step": 2480 |
|
}, |
|
{ |
|
"epoch": 27.59381898454746, |
|
"learning_rate": 9.639372822299651e-07, |
|
"loss": 1.6133, |
|
"step": 2500 |
|
}, |
|
{ |
|
"epoch": 27.814569536423843, |
|
"learning_rate": 9.636236933797908e-07, |
|
"loss": 1.6445, |
|
"step": 2520 |
|
}, |
|
{ |
|
"epoch": 28.03532008830022, |
|
"learning_rate": 9.633101045296166e-07, |
|
"loss": 1.6421, |
|
"step": 2540 |
|
}, |
|
{ |
|
"epoch": 28.2560706401766, |
|
"learning_rate": 9.629965156794425e-07, |
|
"loss": 1.672, |
|
"step": 2560 |
|
}, |
|
{ |
|
"epoch": 28.47682119205298, |
|
"learning_rate": 9.626829268292684e-07, |
|
"loss": 1.6592, |
|
"step": 2580 |
|
}, |
|
{ |
|
"epoch": 28.69757174392936, |
|
"learning_rate": 9.62369337979094e-07, |
|
"loss": 1.649, |
|
"step": 2600 |
|
}, |
|
{ |
|
"epoch": 28.91832229580574, |
|
"learning_rate": 9.620557491289199e-07, |
|
"loss": 1.6667, |
|
"step": 2620 |
|
}, |
|
{ |
|
"epoch": 29.13907284768212, |
|
"learning_rate": 9.617421602787455e-07, |
|
"loss": 1.6172, |
|
"step": 2640 |
|
}, |
|
{ |
|
"epoch": 29.3598233995585, |
|
"learning_rate": 9.614285714285714e-07, |
|
"loss": 1.6696, |
|
"step": 2660 |
|
}, |
|
{ |
|
"epoch": 29.58057395143488, |
|
"learning_rate": 9.611149825783972e-07, |
|
"loss": 1.7091, |
|
"step": 2680 |
|
}, |
|
{ |
|
"epoch": 29.801324503311257, |
|
"learning_rate": 9.60801393728223e-07, |
|
"loss": 1.6111, |
|
"step": 2700 |
|
}, |
|
{ |
|
"epoch": 30.02207505518764, |
|
"learning_rate": 9.604878048780487e-07, |
|
"loss": 1.6274, |
|
"step": 2720 |
|
}, |
|
{ |
|
"epoch": 30.242825607064017, |
|
"learning_rate": 9.601742160278746e-07, |
|
"loss": 1.6186, |
|
"step": 2740 |
|
}, |
|
{ |
|
"epoch": 30.4635761589404, |
|
"learning_rate": 9.598606271777002e-07, |
|
"loss": 1.6336, |
|
"step": 2760 |
|
}, |
|
{ |
|
"epoch": 30.684326710816777, |
|
"learning_rate": 9.59547038327526e-07, |
|
"loss": 1.6414, |
|
"step": 2780 |
|
}, |
|
{ |
|
"epoch": 30.90507726269316, |
|
"learning_rate": 9.59233449477352e-07, |
|
"loss": 1.6628, |
|
"step": 2800 |
|
}, |
|
{ |
|
"epoch": 31.125827814569536, |
|
"learning_rate": 9.589198606271776e-07, |
|
"loss": 1.6092, |
|
"step": 2820 |
|
}, |
|
{ |
|
"epoch": 31.346578366445915, |
|
"learning_rate": 9.586062717770034e-07, |
|
"loss": 1.669, |
|
"step": 2840 |
|
}, |
|
{ |
|
"epoch": 31.567328918322296, |
|
"learning_rate": 9.58292682926829e-07, |
|
"loss": 1.6597, |
|
"step": 2860 |
|
}, |
|
{ |
|
"epoch": 31.788079470198674, |
|
"learning_rate": 9.57979094076655e-07, |
|
"loss": 1.5658, |
|
"step": 2880 |
|
}, |
|
{ |
|
"epoch": 32.00883002207505, |
|
"learning_rate": 9.576655052264808e-07, |
|
"loss": 1.6662, |
|
"step": 2900 |
|
}, |
|
{ |
|
"epoch": 32.22958057395144, |
|
"learning_rate": 9.573519163763067e-07, |
|
"loss": 1.6241, |
|
"step": 2920 |
|
}, |
|
{ |
|
"epoch": 32.450331125827816, |
|
"learning_rate": 9.570383275261325e-07, |
|
"loss": 1.6181, |
|
"step": 2940 |
|
}, |
|
{ |
|
"epoch": 32.671081677704194, |
|
"learning_rate": 9.567247386759582e-07, |
|
"loss": 1.6561, |
|
"step": 2960 |
|
}, |
|
{ |
|
"epoch": 32.89183222958057, |
|
"learning_rate": 9.564111498257838e-07, |
|
"loss": 1.6176, |
|
"step": 2980 |
|
}, |
|
{ |
|
"epoch": 33.11258278145695, |
|
"learning_rate": 9.560975609756097e-07, |
|
"loss": 1.686, |
|
"step": 3000 |
|
}, |
|
{ |
|
"epoch": 33.333333333333336, |
|
"learning_rate": 9.557839721254355e-07, |
|
"loss": 1.5938, |
|
"step": 3020 |
|
}, |
|
{ |
|
"epoch": 33.554083885209714, |
|
"learning_rate": 9.554703832752614e-07, |
|
"loss": 1.5952, |
|
"step": 3040 |
|
}, |
|
{ |
|
"epoch": 33.77483443708609, |
|
"learning_rate": 9.55156794425087e-07, |
|
"loss": 1.5772, |
|
"step": 3060 |
|
}, |
|
{ |
|
"epoch": 33.99558498896247, |
|
"learning_rate": 9.548432055749129e-07, |
|
"loss": 1.5889, |
|
"step": 3080 |
|
}, |
|
{ |
|
"epoch": 34.216335540838855, |
|
"learning_rate": 9.545296167247385e-07, |
|
"loss": 1.5953, |
|
"step": 3100 |
|
}, |
|
{ |
|
"epoch": 34.437086092715234, |
|
"learning_rate": 9.542160278745644e-07, |
|
"loss": 1.6379, |
|
"step": 3120 |
|
}, |
|
{ |
|
"epoch": 34.65783664459161, |
|
"learning_rate": 9.539024390243903e-07, |
|
"loss": 1.6252, |
|
"step": 3140 |
|
}, |
|
{ |
|
"epoch": 34.87858719646799, |
|
"learning_rate": 9.53588850174216e-07, |
|
"loss": 1.628, |
|
"step": 3160 |
|
}, |
|
{ |
|
"epoch": 35.09933774834437, |
|
"learning_rate": 9.532752613240419e-07, |
|
"loss": 1.6547, |
|
"step": 3180 |
|
}, |
|
{ |
|
"epoch": 35.32008830022075, |
|
"learning_rate": 9.529616724738675e-07, |
|
"loss": 1.6308, |
|
"step": 3200 |
|
}, |
|
{ |
|
"epoch": 35.54083885209713, |
|
"learning_rate": 9.526480836236935e-07, |
|
"loss": 1.599, |
|
"step": 3220 |
|
}, |
|
{ |
|
"epoch": 35.76158940397351, |
|
"learning_rate": 9.523344947735191e-07, |
|
"loss": 1.6198, |
|
"step": 3240 |
|
}, |
|
{ |
|
"epoch": 35.98233995584989, |
|
"learning_rate": 9.520209059233449e-07, |
|
"loss": 1.6248, |
|
"step": 3260 |
|
}, |
|
{ |
|
"epoch": 36.203090507726266, |
|
"learning_rate": 9.517073170731706e-07, |
|
"loss": 1.576, |
|
"step": 3280 |
|
}, |
|
{ |
|
"epoch": 36.42384105960265, |
|
"learning_rate": 9.513937282229965e-07, |
|
"loss": 1.6648, |
|
"step": 3300 |
|
}, |
|
{ |
|
"epoch": 36.64459161147903, |
|
"learning_rate": 9.510801393728223e-07, |
|
"loss": 1.6035, |
|
"step": 3320 |
|
}, |
|
{ |
|
"epoch": 36.86534216335541, |
|
"learning_rate": 9.50766550522648e-07, |
|
"loss": 1.6051, |
|
"step": 3340 |
|
}, |
|
{ |
|
"epoch": 37.086092715231786, |
|
"learning_rate": 9.504529616724738e-07, |
|
"loss": 1.6052, |
|
"step": 3360 |
|
}, |
|
{ |
|
"epoch": 37.30684326710817, |
|
"learning_rate": 9.501393728222996e-07, |
|
"loss": 1.6081, |
|
"step": 3380 |
|
}, |
|
{ |
|
"epoch": 37.52759381898455, |
|
"learning_rate": 9.498257839721255e-07, |
|
"loss": 1.6066, |
|
"step": 3400 |
|
}, |
|
{ |
|
"epoch": 37.74834437086093, |
|
"learning_rate": 9.495121951219511e-07, |
|
"loss": 1.5763, |
|
"step": 3420 |
|
}, |
|
{ |
|
"epoch": 37.969094922737305, |
|
"learning_rate": 9.49198606271777e-07, |
|
"loss": 1.6039, |
|
"step": 3440 |
|
}, |
|
{ |
|
"epoch": 38.18984547461368, |
|
"learning_rate": 9.488850174216028e-07, |
|
"loss": 1.6203, |
|
"step": 3460 |
|
}, |
|
{ |
|
"epoch": 38.41059602649007, |
|
"learning_rate": 9.485714285714285e-07, |
|
"loss": 1.619, |
|
"step": 3480 |
|
}, |
|
{ |
|
"epoch": 38.63134657836645, |
|
"learning_rate": 9.482578397212543e-07, |
|
"loss": 1.6055, |
|
"step": 3500 |
|
}, |
|
{ |
|
"epoch": 38.852097130242825, |
|
"learning_rate": 9.479442508710801e-07, |
|
"loss": 1.6037, |
|
"step": 3520 |
|
}, |
|
{ |
|
"epoch": 39.0728476821192, |
|
"learning_rate": 9.476306620209059e-07, |
|
"loss": 1.6259, |
|
"step": 3540 |
|
}, |
|
{ |
|
"epoch": 39.29359823399559, |
|
"learning_rate": 9.473170731707316e-07, |
|
"loss": 1.5917, |
|
"step": 3560 |
|
}, |
|
{ |
|
"epoch": 39.51434878587197, |
|
"learning_rate": 9.470034843205574e-07, |
|
"loss": 1.6039, |
|
"step": 3580 |
|
}, |
|
{ |
|
"epoch": 39.735099337748345, |
|
"learning_rate": 9.466898954703833e-07, |
|
"loss": 1.5364, |
|
"step": 3600 |
|
}, |
|
{ |
|
"epoch": 39.735099337748345, |
|
"eval_bleu": 41.1359, |
|
"eval_gen_len": 9.5667, |
|
"eval_loss": 2.0473527908325195, |
|
"eval_runtime": 3.9497, |
|
"eval_samples_per_second": 7.596, |
|
"eval_steps_per_second": 1.519, |
|
"step": 3600 |
|
} |
|
], |
|
"logging_steps": 20, |
|
"max_steps": 57600, |
|
"num_input_tokens_seen": 0, |
|
"num_train_epochs": 640, |
|
"save_steps": 3600, |
|
"stateful_callbacks": { |
|
"TrainerControl": { |
|
"args": { |
|
"should_epoch_stop": false, |
|
"should_evaluate": false, |
|
"should_log": false, |
|
"should_save": true, |
|
"should_training_stop": false |
|
}, |
|
"attributes": {} |
|
} |
|
}, |
|
"total_flos": 1.0641987012722688e+16, |
|
"train_batch_size": 5, |
|
"trial_name": null, |
|
"trial_params": null |
|
} |
|
|