| { |
| "best_global_step": 26785, |
| "best_metric": 0.5293099134739596, |
| "best_model_checkpoint": "outputs/smartphone/vinai_phobert-base/checkpoint-26785", |
| "epoch": 55.0, |
| "eval_steps": 500, |
| "global_step": 26785, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "grad_norm": 1.214017629623413, |
| "learning_rate": 2.9992628768482664e-05, |
| "loss": 0.3598, |
| "step": 487 |
| }, |
| { |
| "epoch": 1.0, |
| "eval_f1_aspect": 0.6824618736383442, |
| "eval_f1_macro_aspect": 0.23965991762577465, |
| "eval_f1_sentiment": 0.23965991762577465, |
| "eval_loss": 0.2352765053510666, |
| "eval_runtime": 5.2822, |
| "eval_samples_per_second": 421.033, |
| "eval_steps_per_second": 13.252, |
| "step": 487 |
| }, |
| { |
| "epoch": 2.0, |
| "grad_norm": 1.3990741968154907, |
| "learning_rate": 2.9970461653663515e-05, |
| "loss": 0.2082, |
| "step": 974 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_f1_aspect": 0.771007371007371, |
| "eval_f1_macro_aspect": 0.36928730826344125, |
| "eval_f1_sentiment": 0.36928730826344125, |
| "eval_loss": 0.18111465871334076, |
| "eval_runtime": 5.2643, |
| "eval_samples_per_second": 422.471, |
| "eval_steps_per_second": 13.297, |
| "step": 974 |
| }, |
| { |
| "epoch": 3.0, |
| "grad_norm": 1.6626859903335571, |
| "learning_rate": 2.9933520500597934e-05, |
| "loss": 0.1604, |
| "step": 1461 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_f1_aspect": 0.8001937515136837, |
| "eval_f1_macro_aspect": 0.39273139407455954, |
| "eval_f1_sentiment": 0.39273139407455954, |
| "eval_loss": 0.16435211896896362, |
| "eval_runtime": 5.2672, |
| "eval_samples_per_second": 422.237, |
| "eval_steps_per_second": 13.29, |
| "step": 1461 |
| }, |
| { |
| "epoch": 4.0, |
| "grad_norm": 2.1720831394195557, |
| "learning_rate": 2.9881841765744044e-05, |
| "loss": 0.1325, |
| "step": 1948 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_f1_aspect": 0.791747572815534, |
| "eval_f1_macro_aspect": 0.39735406210910534, |
| "eval_f1_sentiment": 0.39735406210910534, |
| "eval_loss": 0.16188032925128937, |
| "eval_runtime": 5.3277, |
| "eval_samples_per_second": 417.439, |
| "eval_steps_per_second": 13.139, |
| "step": 1948 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 3.8878087997436523, |
| "learning_rate": 2.9815476449773896e-05, |
| "loss": 0.1102, |
| "step": 2435 |
| }, |
| { |
| "epoch": 5.0, |
| "eval_f1_aspect": 0.8246024321796072, |
| "eval_f1_macro_aspect": 0.46887437550765554, |
| "eval_f1_sentiment": 0.46887437550765554, |
| "eval_loss": 0.15876558423042297, |
| "eval_runtime": 5.3286, |
| "eval_samples_per_second": 417.369, |
| "eval_steps_per_second": 13.137, |
| "step": 2435 |
| }, |
| { |
| "epoch": 6.0, |
| "grad_norm": 1.500092625617981, |
| "learning_rate": 2.973449004724197e-05, |
| "loss": 0.0919, |
| "step": 2922 |
| }, |
| { |
| "epoch": 6.0, |
| "eval_f1_aspect": 0.8168293816829382, |
| "eval_f1_macro_aspect": 0.47478350769030153, |
| "eval_f1_sentiment": 0.47478350769030153, |
| "eval_loss": 0.16465845704078674, |
| "eval_runtime": 5.3861, |
| "eval_samples_per_second": 412.914, |
| "eval_steps_per_second": 12.996, |
| "step": 2922 |
| }, |
| { |
| "epoch": 7.0, |
| "grad_norm": 1.246404767036438, |
| "learning_rate": 2.9638962481949954e-05, |
| "loss": 0.078, |
| "step": 3409 |
| }, |
| { |
| "epoch": 7.0, |
| "eval_f1_aspect": 0.8095238095238095, |
| "eval_f1_macro_aspect": 0.4796789474797879, |
| "eval_f1_sentiment": 0.4796789474797879, |
| "eval_loss": 0.17780332267284393, |
| "eval_runtime": 5.2676, |
| "eval_samples_per_second": 422.206, |
| "eval_steps_per_second": 13.289, |
| "step": 3409 |
| }, |
| { |
| "epoch": 8.0, |
| "grad_norm": 0.4573521018028259, |
| "learning_rate": 2.9528988028071616e-05, |
| "loss": 0.0677, |
| "step": 3896 |
| }, |
| { |
| "epoch": 8.0, |
| "eval_f1_aspect": 0.8132741294695022, |
| "eval_f1_macro_aspect": 0.48194296253877855, |
| "eval_f1_sentiment": 0.48194296253877855, |
| "eval_loss": 0.17726708948612213, |
| "eval_runtime": 5.2769, |
| "eval_samples_per_second": 421.462, |
| "eval_steps_per_second": 13.265, |
| "step": 3896 |
| }, |
| { |
| "epoch": 9.0, |
| "grad_norm": 1.0054028034210205, |
| "learning_rate": 2.9404675217115552e-05, |
| "loss": 0.0582, |
| "step": 4383 |
| }, |
| { |
| "epoch": 9.0, |
| "eval_f1_aspect": 0.8148489739451233, |
| "eval_f1_macro_aspect": 0.49901186603018977, |
| "eval_f1_sentiment": 0.49901186603018977, |
| "eval_loss": 0.18432779610157013, |
| "eval_runtime": 5.2788, |
| "eval_samples_per_second": 421.307, |
| "eval_steps_per_second": 13.261, |
| "step": 4383 |
| }, |
| { |
| "epoch": 10.0, |
| "grad_norm": 0.597675621509552, |
| "learning_rate": 2.926614673081771e-05, |
| "loss": 0.05, |
| "step": 4870 |
| }, |
| { |
| "epoch": 10.0, |
| "eval_f1_aspect": 0.8054654932839277, |
| "eval_f1_macro_aspect": 0.48681724911620844, |
| "eval_f1_sentiment": 0.48681724911620844, |
| "eval_loss": 0.19903235137462616, |
| "eval_runtime": 5.3605, |
| "eval_samples_per_second": 414.887, |
| "eval_steps_per_second": 13.059, |
| "step": 4870 |
| }, |
| { |
| "epoch": 11.0, |
| "grad_norm": 1.099621057510376, |
| "learning_rate": 2.9113539280069318e-05, |
| "loss": 0.0443, |
| "step": 5357 |
| }, |
| { |
| "epoch": 11.0, |
| "eval_f1_aspect": 0.803812180381218, |
| "eval_f1_macro_aspect": 0.49064332461219184, |
| "eval_f1_sentiment": 0.49064332461219184, |
| "eval_loss": 0.20838035643100739, |
| "eval_runtime": 5.3056, |
| "eval_samples_per_second": 419.18, |
| "eval_steps_per_second": 13.194, |
| "step": 5357 |
| }, |
| { |
| "epoch": 12.0, |
| "grad_norm": 8.490036010742188, |
| "learning_rate": 2.8947003469999752e-05, |
| "loss": 0.0385, |
| "step": 5844 |
| }, |
| { |
| "epoch": 12.0, |
| "eval_f1_aspect": 0.8149005090236002, |
| "eval_f1_macro_aspect": 0.5031596962140534, |
| "eval_f1_sentiment": 0.5031596962140534, |
| "eval_loss": 0.2110637128353119, |
| "eval_runtime": 5.2792, |
| "eval_samples_per_second": 421.277, |
| "eval_steps_per_second": 13.26, |
| "step": 5844 |
| }, |
| { |
| "epoch": 13.0, |
| "grad_norm": 1.8095080852508545, |
| "learning_rate": 2.8766703651347423e-05, |
| "loss": 0.0327, |
| "step": 6331 |
| }, |
| { |
| "epoch": 13.0, |
| "eval_f1_aspect": 0.8119677790563866, |
| "eval_f1_macro_aspect": 0.49863733694553214, |
| "eval_f1_sentiment": 0.49863733694553214, |
| "eval_loss": 0.2265346497297287, |
| "eval_runtime": 5.381, |
| "eval_samples_per_second": 413.307, |
| "eval_steps_per_second": 13.009, |
| "step": 6331 |
| }, |
| { |
| "epoch": 14.0, |
| "grad_norm": 1.6875431537628174, |
| "learning_rate": 2.857281775826549e-05, |
| "loss": 0.0283, |
| "step": 6818 |
| }, |
| { |
| "epoch": 14.0, |
| "eval_f1_aspect": 0.8086857142857143, |
| "eval_f1_macro_aspect": 0.5133505127639592, |
| "eval_f1_sentiment": 0.5133505127639592, |
| "eval_loss": 0.23116181790828705, |
| "eval_runtime": 5.2712, |
| "eval_samples_per_second": 421.915, |
| "eval_steps_per_second": 13.28, |
| "step": 6818 |
| }, |
| { |
| "epoch": 15.0, |
| "grad_norm": 1.34958016872406, |
| "learning_rate": 2.8365537132722293e-05, |
| "loss": 0.0251, |
| "step": 7305 |
| }, |
| { |
| "epoch": 15.0, |
| "eval_f1_aspect": 0.8102658111824015, |
| "eval_f1_macro_aspect": 0.5025552457560816, |
| "eval_f1_sentiment": 0.5025552457560816, |
| "eval_loss": 0.23273742198944092, |
| "eval_runtime": 5.2784, |
| "eval_samples_per_second": 421.337, |
| "eval_steps_per_second": 13.262, |
| "step": 7305 |
| }, |
| { |
| "epoch": 16.0, |
| "grad_norm": 4.683711051940918, |
| "learning_rate": 2.8145066335669954e-05, |
| "loss": 0.0207, |
| "step": 7792 |
| }, |
| { |
| "epoch": 16.0, |
| "eval_f1_aspect": 0.8045182111572153, |
| "eval_f1_macro_aspect": 0.4980081283232479, |
| "eval_f1_sentiment": 0.4980081283232479, |
| "eval_loss": 0.2506747841835022, |
| "eval_runtime": 5.2898, |
| "eval_samples_per_second": 420.428, |
| "eval_steps_per_second": 13.233, |
| "step": 7792 |
| }, |
| { |
| "epoch": 17.0, |
| "grad_norm": 0.895351231098175, |
| "learning_rate": 2.7911622945167402e-05, |
| "loss": 0.0187, |
| "step": 8279 |
| }, |
| { |
| "epoch": 17.0, |
| "eval_f1_aspect": 0.8070496681162738, |
| "eval_f1_macro_aspect": 0.5174731474301049, |
| "eval_f1_sentiment": 0.5174731474301049, |
| "eval_loss": 0.24516579508781433, |
| "eval_runtime": 5.2809, |
| "eval_samples_per_second": 421.137, |
| "eval_steps_per_second": 13.255, |
| "step": 8279 |
| }, |
| { |
| "epoch": 18.0, |
| "grad_norm": 0.40506213903427124, |
| "learning_rate": 2.76654373416571e-05, |
| "loss": 0.0164, |
| "step": 8766 |
| }, |
| { |
| "epoch": 18.0, |
| "eval_f1_aspect": 0.80653324131585, |
| "eval_f1_macro_aspect": 0.4992792703618323, |
| "eval_f1_sentiment": 0.4992792703618323, |
| "eval_loss": 0.2601372003555298, |
| "eval_runtime": 5.2829, |
| "eval_samples_per_second": 420.982, |
| "eval_steps_per_second": 13.25, |
| "step": 8766 |
| }, |
| { |
| "epoch": 19.0, |
| "grad_norm": 0.15703117847442627, |
| "learning_rate": 2.740675248060736e-05, |
| "loss": 0.0141, |
| "step": 9253 |
| }, |
| { |
| "epoch": 19.0, |
| "eval_f1_aspect": 0.8097674418604651, |
| "eval_f1_macro_aspect": 0.49406251042693405, |
| "eval_f1_sentiment": 0.49406251042693405, |
| "eval_loss": 0.2688779830932617, |
| "eval_runtime": 5.2799, |
| "eval_samples_per_second": 421.222, |
| "eval_steps_per_second": 13.258, |
| "step": 9253 |
| }, |
| { |
| "epoch": 20.0, |
| "grad_norm": 0.614011287689209, |
| "learning_rate": 2.7135823652744645e-05, |
| "loss": 0.0135, |
| "step": 9740 |
| }, |
| { |
| "epoch": 20.0, |
| "eval_f1_aspect": 0.8078146297137665, |
| "eval_f1_macro_aspect": 0.5061512331950289, |
| "eval_f1_sentiment": 0.5061512331950289, |
| "eval_loss": 0.2663342356681824, |
| "eval_runtime": 5.2404, |
| "eval_samples_per_second": 424.392, |
| "eval_steps_per_second": 13.358, |
| "step": 9740 |
| }, |
| { |
| "epoch": 21.0, |
| "grad_norm": 0.04745159670710564, |
| "learning_rate": 2.6852918232112412e-05, |
| "loss": 0.0118, |
| "step": 10227 |
| }, |
| { |
| "epoch": 21.0, |
| "eval_f1_aspect": 0.8046875, |
| "eval_f1_macro_aspect": 0.4998370067335585, |
| "eval_f1_sentiment": 0.4998370067335585, |
| "eval_loss": 0.2750583589076996, |
| "eval_runtime": 5.2661, |
| "eval_samples_per_second": 422.324, |
| "eval_steps_per_second": 13.293, |
| "step": 10227 |
| }, |
| { |
| "epoch": 22.0, |
| "grad_norm": 0.6339995265007019, |
| "learning_rate": 2.6558315412205247e-05, |
| "loss": 0.0104, |
| "step": 10714 |
| }, |
| { |
| "epoch": 22.0, |
| "eval_f1_aspect": 0.8146297137664698, |
| "eval_f1_macro_aspect": 0.5046009870856859, |
| "eval_f1_sentiment": 0.5046009870856859, |
| "eval_loss": 0.2895522117614746, |
| "eval_runtime": 5.2588, |
| "eval_samples_per_second": 422.906, |
| "eval_steps_per_second": 13.311, |
| "step": 10714 |
| }, |
| { |
| "epoch": 23.0, |
| "grad_norm": 0.04146909713745117, |
| "learning_rate": 2.6252305930438563e-05, |
| "loss": 0.0092, |
| "step": 11201 |
| }, |
| { |
| "epoch": 23.0, |
| "eval_f1_aspect": 0.8131566925536775, |
| "eval_f1_macro_aspect": 0.5065319656270583, |
| "eval_f1_sentiment": 0.5065319656270583, |
| "eval_loss": 0.294137567281723, |
| "eval_runtime": 5.3335, |
| "eval_samples_per_second": 416.987, |
| "eval_steps_per_second": 13.125, |
| "step": 11201 |
| }, |
| { |
| "epoch": 24.0, |
| "grad_norm": 11.924123764038086, |
| "learning_rate": 2.5935191781225898e-05, |
| "loss": 0.0081, |
| "step": 11688 |
| }, |
| { |
| "epoch": 24.0, |
| "eval_f1_aspect": 0.8160550458715596, |
| "eval_f1_macro_aspect": 0.5096299311354855, |
| "eval_f1_sentiment": 0.5096299311354855, |
| "eval_loss": 0.2907048463821411, |
| "eval_runtime": 5.3212, |
| "eval_samples_per_second": 417.948, |
| "eval_steps_per_second": 13.155, |
| "step": 11688 |
| }, |
| { |
| "epoch": 25.0, |
| "grad_norm": 0.7775519490242004, |
| "learning_rate": 2.5607285917946864e-05, |
| "loss": 0.0083, |
| "step": 12175 |
| }, |
| { |
| "epoch": 25.0, |
| "eval_f1_aspect": 0.818926296633303, |
| "eval_f1_macro_aspect": 0.5084465129276063, |
| "eval_f1_sentiment": 0.5084465129276063, |
| "eval_loss": 0.30101829767227173, |
| "eval_runtime": 5.3775, |
| "eval_samples_per_second": 413.572, |
| "eval_steps_per_second": 13.017, |
| "step": 12175 |
| }, |
| { |
| "epoch": 26.0, |
| "grad_norm": 2.315323829650879, |
| "learning_rate": 2.5268911944099967e-05, |
| "loss": 0.008, |
| "step": 12662 |
| }, |
| { |
| "epoch": 26.0, |
| "eval_f1_aspect": 0.8159158472444547, |
| "eval_f1_macro_aspect": 0.5100191821282204, |
| "eval_f1_sentiment": 0.5100191821282204, |
| "eval_loss": 0.3063946068286896, |
| "eval_runtime": 5.3501, |
| "eval_samples_per_second": 415.691, |
| "eval_steps_per_second": 13.084, |
| "step": 12662 |
| }, |
| { |
| "epoch": 27.0, |
| "grad_norm": 1.4911683797836304, |
| "learning_rate": 2.492040379394501e-05, |
| "loss": 0.0066, |
| "step": 13149 |
| }, |
| { |
| "epoch": 27.0, |
| "eval_f1_aspect": 0.8137142857142857, |
| "eval_f1_macro_aspect": 0.5107237241401983, |
| "eval_f1_sentiment": 0.5107237241401983, |
| "eval_loss": 0.31042352318763733, |
| "eval_runtime": 5.2518, |
| "eval_samples_per_second": 423.477, |
| "eval_steps_per_second": 13.329, |
| "step": 13149 |
| }, |
| { |
| "epoch": 28.0, |
| "grad_norm": 0.9785826802253723, |
| "learning_rate": 2.4562105402950287e-05, |
| "loss": 0.0054, |
| "step": 13636 |
| }, |
| { |
| "epoch": 28.0, |
| "eval_f1_aspect": 0.8164295548416705, |
| "eval_f1_macro_aspect": 0.5113764177223036, |
| "eval_f1_sentiment": 0.5113764177223036, |
| "eval_loss": 0.30704399943351746, |
| "eval_runtime": 5.3102, |
| "eval_samples_per_second": 418.818, |
| "eval_steps_per_second": 13.182, |
| "step": 13636 |
| }, |
| { |
| "epoch": 29.0, |
| "grad_norm": 0.01176587212830782, |
| "learning_rate": 2.419437036836983e-05, |
| "loss": 0.0059, |
| "step": 14123 |
| }, |
| { |
| "epoch": 29.0, |
| "eval_f1_aspect": 0.810004589261129, |
| "eval_f1_macro_aspect": 0.51595438584814, |
| "eval_f1_sentiment": 0.51595438584814, |
| "eval_loss": 0.3131285309791565, |
| "eval_runtime": 5.3519, |
| "eval_samples_per_second": 415.555, |
| "eval_steps_per_second": 13.08, |
| "step": 14123 |
| }, |
| { |
| "epoch": 30.0, |
| "grad_norm": 0.22203879058361053, |
| "learning_rate": 2.3817561600285582e-05, |
| "loss": 0.0051, |
| "step": 14610 |
| }, |
| { |
| "epoch": 30.0, |
| "eval_f1_aspect": 0.8104515241806096, |
| "eval_f1_macro_aspect": 0.5174387744262563, |
| "eval_f1_sentiment": 0.5174387744262563, |
| "eval_loss": 0.32132473587989807, |
| "eval_runtime": 5.3173, |
| "eval_samples_per_second": 418.256, |
| "eval_steps_per_second": 13.165, |
| "step": 14610 |
| }, |
| { |
| "epoch": 31.0, |
| "grad_norm": 0.06852120161056519, |
| "learning_rate": 2.343205096345896e-05, |
| "loss": 0.0054, |
| "step": 15097 |
| }, |
| { |
| "epoch": 31.0, |
| "eval_f1_aspect": 0.8126286890871655, |
| "eval_f1_macro_aspect": 0.505160075382636, |
| "eval_f1_sentiment": 0.505160075382636, |
| "eval_loss": 0.31839850544929504, |
| "eval_runtime": 5.2676, |
| "eval_samples_per_second": 422.201, |
| "eval_steps_per_second": 13.289, |
| "step": 15097 |
| }, |
| { |
| "epoch": 32.0, |
| "grad_norm": 2.4210550785064697, |
| "learning_rate": 2.303821891034526e-05, |
| "loss": 0.0055, |
| "step": 15584 |
| }, |
| { |
| "epoch": 32.0, |
| "eval_f1_aspect": 0.8065116279069767, |
| "eval_f1_macro_aspect": 0.5054766945577382, |
| "eval_f1_sentiment": 0.5054766945577382, |
| "eval_loss": 0.32863593101501465, |
| "eval_runtime": 5.2714, |
| "eval_samples_per_second": 421.901, |
| "eval_steps_per_second": 13.279, |
| "step": 15584 |
| }, |
| { |
| "epoch": 33.0, |
| "grad_norm": 0.003676912747323513, |
| "learning_rate": 2.2636454105632974e-05, |
| "loss": 0.0047, |
| "step": 16071 |
| }, |
| { |
| "epoch": 33.0, |
| "eval_f1_aspect": 0.8185381600184459, |
| "eval_f1_macro_aspect": 0.5082259114157306, |
| "eval_f1_sentiment": 0.5082259114157306, |
| "eval_loss": 0.3176405429840088, |
| "eval_runtime": 5.2759, |
| "eval_samples_per_second": 421.542, |
| "eval_steps_per_second": 13.268, |
| "step": 16071 |
| }, |
| { |
| "epoch": 34.0, |
| "grad_norm": 0.3813173472881317, |
| "learning_rate": 2.2227153042678708e-05, |
| "loss": 0.0043, |
| "step": 16558 |
| }, |
| { |
| "epoch": 34.0, |
| "eval_f1_aspect": 0.822883295194508, |
| "eval_f1_macro_aspect": 0.5175972532342072, |
| "eval_f1_sentiment": 0.5175972532342072, |
| "eval_loss": 0.31699639558792114, |
| "eval_runtime": 5.2786, |
| "eval_samples_per_second": 421.322, |
| "eval_steps_per_second": 13.261, |
| "step": 16558 |
| }, |
| { |
| "epoch": 35.0, |
| "grad_norm": 0.017288848757743835, |
| "learning_rate": 2.1810719652216077e-05, |
| "loss": 0.0039, |
| "step": 17045 |
| }, |
| { |
| "epoch": 35.0, |
| "eval_f1_aspect": 0.821256038647343, |
| "eval_f1_macro_aspect": 0.5079532801084986, |
| "eval_f1_sentiment": 0.5079532801084986, |
| "eval_loss": 0.339198499917984, |
| "eval_runtime": 5.3292, |
| "eval_samples_per_second": 417.322, |
| "eval_steps_per_second": 13.135, |
| "step": 17045 |
| }, |
| { |
| "epoch": 36.0, |
| "grad_norm": 0.43333691358566284, |
| "learning_rate": 2.138756490372486e-05, |
| "loss": 0.0036, |
| "step": 17532 |
| }, |
| { |
| "epoch": 36.0, |
| "eval_f1_aspect": 0.8142365097588978, |
| "eval_f1_macro_aspect": 0.5104098398330861, |
| "eval_f1_sentiment": 0.5104098398330861, |
| "eval_loss": 0.33852654695510864, |
| "eval_runtime": 5.2774, |
| "eval_samples_per_second": 421.424, |
| "eval_steps_per_second": 13.264, |
| "step": 17532 |
| }, |
| { |
| "epoch": 37.0, |
| "grad_norm": 0.13463492691516876, |
| "learning_rate": 2.0958106399853726e-05, |
| "loss": 0.0035, |
| "step": 18019 |
| }, |
| { |
| "epoch": 37.0, |
| "eval_f1_aspect": 0.8179949760219228, |
| "eval_f1_macro_aspect": 0.5061536037175701, |
| "eval_f1_sentiment": 0.5061536037175701, |
| "eval_loss": 0.3427109718322754, |
| "eval_runtime": 5.2483, |
| "eval_samples_per_second": 423.758, |
| "eval_steps_per_second": 13.338, |
| "step": 18019 |
| }, |
| { |
| "epoch": 38.0, |
| "grad_norm": 0.14240214228630066, |
| "learning_rate": 2.0522767964296835e-05, |
| "loss": 0.0044, |
| "step": 18506 |
| }, |
| { |
| "epoch": 38.0, |
| "eval_f1_aspect": 0.813799405985835, |
| "eval_f1_macro_aspect": 0.4926077943615257, |
| "eval_f1_sentiment": 0.4926077943615257, |
| "eval_loss": 0.34948092699050903, |
| "eval_runtime": 5.2669, |
| "eval_samples_per_second": 422.261, |
| "eval_steps_per_second": 13.291, |
| "step": 18506 |
| }, |
| { |
| "epoch": 39.0, |
| "grad_norm": 0.3045540153980255, |
| "learning_rate": 2.0081979223531034e-05, |
| "loss": 0.005, |
| "step": 18993 |
| }, |
| { |
| "epoch": 39.0, |
| "eval_f1_aspect": 0.8197553657973691, |
| "eval_f1_macro_aspect": 0.5165866468068806, |
| "eval_f1_sentiment": 0.5165866468068806, |
| "eval_loss": 0.33417609333992004, |
| "eval_runtime": 5.334, |
| "eval_samples_per_second": 416.95, |
| "eval_steps_per_second": 13.123, |
| "step": 18993 |
| }, |
| { |
| "epoch": 40.0, |
| "grad_norm": 0.10395223647356033, |
| "learning_rate": 1.9636175182826402e-05, |
| "loss": 0.0026, |
| "step": 19480 |
| }, |
| { |
| "epoch": 40.0, |
| "eval_f1_aspect": 0.8171650308148825, |
| "eval_f1_macro_aspect": 0.5192047014309698, |
| "eval_f1_sentiment": 0.5192047014309698, |
| "eval_loss": 0.36051589250564575, |
| "eval_runtime": 5.3449, |
| "eval_samples_per_second": 416.099, |
| "eval_steps_per_second": 13.097, |
| "step": 19480 |
| }, |
| { |
| "epoch": 41.0, |
| "grad_norm": 0.47768187522888184, |
| "learning_rate": 1.918579579694853e-05, |
| "loss": 0.0031, |
| "step": 19967 |
| }, |
| { |
| "epoch": 41.0, |
| "eval_f1_aspect": 0.8147638697845025, |
| "eval_f1_macro_aspect": 0.5071432459726105, |
| "eval_f1_sentiment": 0.5071432459726105, |
| "eval_loss": 0.3499327600002289, |
| "eval_runtime": 5.2702, |
| "eval_samples_per_second": 421.995, |
| "eval_steps_per_second": 13.282, |
| "step": 19967 |
| }, |
| { |
| "epoch": 42.0, |
| "grad_norm": 3.665292978286743, |
| "learning_rate": 1.8731285535976327e-05, |
| "loss": 0.0024, |
| "step": 20454 |
| }, |
| { |
| "epoch": 42.0, |
| "eval_f1_aspect": 0.8167467398764585, |
| "eval_f1_macro_aspect": 0.5079256124525199, |
| "eval_f1_sentiment": 0.5079256124525199, |
| "eval_loss": 0.3565986454486847, |
| "eval_runtime": 5.2524, |
| "eval_samples_per_second": 423.429, |
| "eval_steps_per_second": 13.327, |
| "step": 20454 |
| }, |
| { |
| "epoch": 43.0, |
| "grad_norm": 0.041507333517074585, |
| "learning_rate": 1.8273092946663674e-05, |
| "loss": 0.0024, |
| "step": 20941 |
| }, |
| { |
| "epoch": 43.0, |
| "eval_f1_aspect": 0.8205010342450012, |
| "eval_f1_macro_aspect": 0.5057632659931312, |
| "eval_f1_sentiment": 0.5057632659931312, |
| "eval_loss": 0.353305459022522, |
| "eval_runtime": 5.268, |
| "eval_samples_per_second": 422.171, |
| "eval_steps_per_second": 13.288, |
| "step": 20941 |
| }, |
| { |
| "epoch": 44.0, |
| "grad_norm": 11.411001205444336, |
| "learning_rate": 1.7811670209777964e-05, |
| "loss": 0.0019, |
| "step": 21428 |
| }, |
| { |
| "epoch": 44.0, |
| "eval_f1_aspect": 0.8167010782289515, |
| "eval_f1_macro_aspect": 0.5160780166340071, |
| "eval_f1_sentiment": 0.5160780166340071, |
| "eval_loss": 0.3536563515663147, |
| "eval_runtime": 5.2634, |
| "eval_samples_per_second": 422.542, |
| "eval_steps_per_second": 13.299, |
| "step": 21428 |
| }, |
| { |
| "epoch": 45.0, |
| "grad_norm": 0.008572953753173351, |
| "learning_rate": 1.734747269385226e-05, |
| "loss": 0.0026, |
| "step": 21915 |
| }, |
| { |
| "epoch": 45.0, |
| "eval_f1_aspect": 0.8217799130633723, |
| "eval_f1_macro_aspect": 0.5217043640896729, |
| "eval_f1_sentiment": 0.5217043640896729, |
| "eval_loss": 0.35641050338745117, |
| "eval_runtime": 5.2869, |
| "eval_samples_per_second": 420.661, |
| "eval_steps_per_second": 13.24, |
| "step": 21915 |
| }, |
| { |
| "epoch": 46.0, |
| "grad_norm": 4.907774448394775, |
| "learning_rate": 1.6880958505791537e-05, |
| "loss": 0.0023, |
| "step": 22402 |
| }, |
| { |
| "epoch": 46.0, |
| "eval_f1_aspect": 0.8204426192105864, |
| "eval_f1_macro_aspect": 0.5082924274745652, |
| "eval_f1_sentiment": 0.5082924274745652, |
| "eval_loss": 0.36888423562049866, |
| "eval_runtime": 5.2716, |
| "eval_samples_per_second": 421.884, |
| "eval_steps_per_second": 13.279, |
| "step": 22402 |
| }, |
| { |
| "epoch": 47.0, |
| "grad_norm": 0.04665122181177139, |
| "learning_rate": 1.641258803877647e-05, |
| "loss": 0.0021, |
| "step": 22889 |
| }, |
| { |
| "epoch": 47.0, |
| "eval_f1_aspect": 0.8187053674268602, |
| "eval_f1_macro_aspect": 0.509624571236536, |
| "eval_f1_sentiment": 0.509624571236536, |
| "eval_loss": 0.36498305201530457, |
| "eval_runtime": 5.301, |
| "eval_samples_per_second": 419.544, |
| "eval_steps_per_second": 13.205, |
| "step": 22889 |
| }, |
| { |
| "epoch": 48.0, |
| "grad_norm": 4.29740571975708, |
| "learning_rate": 1.5942823517910956e-05, |
| "loss": 0.0026, |
| "step": 23376 |
| }, |
| { |
| "epoch": 48.0, |
| "eval_f1_aspect": 0.8177655677655677, |
| "eval_f1_macro_aspect": 0.5109795293385091, |
| "eval_f1_sentiment": 0.5109795293385091, |
| "eval_loss": 0.368932843208313, |
| "eval_runtime": 5.2831, |
| "eval_samples_per_second": 420.965, |
| "eval_steps_per_second": 13.25, |
| "step": 23376 |
| }, |
| { |
| "epoch": 49.0, |
| "grad_norm": 0.8426123857498169, |
| "learning_rate": 1.5472128544061724e-05, |
| "loss": 0.0019, |
| "step": 23863 |
| }, |
| { |
| "epoch": 49.0, |
| "eval_f1_aspect": 0.8196042337781868, |
| "eval_f1_macro_aspect": 0.5148117758549687, |
| "eval_f1_sentiment": 0.5148117758549687, |
| "eval_loss": 0.3668632209300995, |
| "eval_runtime": 5.331, |
| "eval_samples_per_second": 417.179, |
| "eval_steps_per_second": 13.131, |
| "step": 23863 |
| }, |
| { |
| "epoch": 50.0, |
| "grad_norm": 0.0009361854754388332, |
| "learning_rate": 1.5000967636340272e-05, |
| "loss": 0.0015, |
| "step": 24350 |
| }, |
| { |
| "epoch": 50.0, |
| "eval_f1_aspect": 0.8164571428571429, |
| "eval_f1_macro_aspect": 0.516046635549457, |
| "eval_f1_sentiment": 0.516046635549457, |
| "eval_loss": 0.3719753623008728, |
| "eval_runtime": 5.3067, |
| "eval_samples_per_second": 419.092, |
| "eval_steps_per_second": 13.191, |
| "step": 24350 |
| }, |
| { |
| "epoch": 51.0, |
| "grad_norm": 0.024893753230571747, |
| "learning_rate": 1.4529805773678577e-05, |
| "loss": 0.0013, |
| "step": 24837 |
| }, |
| { |
| "epoch": 51.0, |
| "eval_f1_aspect": 0.8113293741434445, |
| "eval_f1_macro_aspect": 0.5109673118759412, |
| "eval_f1_sentiment": 0.5109673118759412, |
| "eval_loss": 0.38190165162086487, |
| "eval_runtime": 5.2708, |
| "eval_samples_per_second": 421.949, |
| "eval_steps_per_second": 13.281, |
| "step": 24837 |
| }, |
| { |
| "epoch": 52.0, |
| "grad_norm": 0.0006926740752533078, |
| "learning_rate": 1.4059107935951025e-05, |
| "loss": 0.0016, |
| "step": 25324 |
| }, |
| { |
| "epoch": 52.0, |
| "eval_f1_aspect": 0.8156450137236962, |
| "eval_f1_macro_aspect": 0.517296046782514, |
| "eval_f1_sentiment": 0.517296046782514, |
| "eval_loss": 0.3835242986679077, |
| "eval_runtime": 5.3451, |
| "eval_samples_per_second": 416.079, |
| "eval_steps_per_second": 13.096, |
| "step": 25324 |
| }, |
| { |
| "epoch": 53.0, |
| "grad_norm": 0.08918596804141998, |
| "learning_rate": 1.3589338645095411e-05, |
| "loss": 0.0022, |
| "step": 25811 |
| }, |
| { |
| "epoch": 53.0, |
| "eval_f1_aspect": 0.8180987202925045, |
| "eval_f1_macro_aspect": 0.5149898328099092, |
| "eval_f1_sentiment": 0.5149898328099092, |
| "eval_loss": 0.3808538615703583, |
| "eval_runtime": 5.2576, |
| "eval_samples_per_second": 423.008, |
| "eval_steps_per_second": 13.314, |
| "step": 25811 |
| }, |
| { |
| "epoch": 54.0, |
| "grad_norm": 0.01354946754872799, |
| "learning_rate": 1.312096150668588e-05, |
| "loss": 0.0014, |
| "step": 26298 |
| }, |
| { |
| "epoch": 54.0, |
| "eval_f1_aspect": 0.8204540243063517, |
| "eval_f1_macro_aspect": 0.5117108970702628, |
| "eval_f1_sentiment": 0.5117108970702628, |
| "eval_loss": 0.3776213824748993, |
| "eval_runtime": 5.2933, |
| "eval_samples_per_second": 420.157, |
| "eval_steps_per_second": 13.224, |
| "step": 26298 |
| }, |
| { |
| "epoch": 55.0, |
| "grad_norm": 0.4617435038089752, |
| "learning_rate": 1.2654438752410183e-05, |
| "loss": 0.001, |
| "step": 26785 |
| }, |
| { |
| "epoch": 55.0, |
| "eval_f1_aspect": 0.8272706474490963, |
| "eval_f1_macro_aspect": 0.5293099134739596, |
| "eval_f1_sentiment": 0.5293099134739596, |
| "eval_loss": 0.37219125032424927, |
| "eval_runtime": 5.2642, |
| "eval_samples_per_second": 422.478, |
| "eval_steps_per_second": 13.297, |
| "step": 26785 |
| } |
| ], |
| "logging_steps": 487, |
| "max_steps": 48700, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 100, |
| "save_steps": 500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": false |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 5.508401968847866e+16, |
| "train_batch_size": 32, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
