{ "best_metric": 0.8572538958705256, "best_model_checkpoint": "result/my-unsup-simcse-roberta-large/SupCon-48-0.02-1e-5-0.07-0.0", "epoch": 1.0, "global_step": 3907, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.03, "eval_avg_sts": 0.7782054985417723, "eval_sickr_spearman": 0.7378584393091261, "eval_stsb_spearman": 0.8185525577744187, "step": 125 }, { "epoch": 0.06, "eval_avg_sts": 0.7923362848478215, "eval_sickr_spearman": 0.7560154924924716, "eval_stsb_spearman": 0.8286570772031715, "step": 250 }, { "epoch": 0.1, "eval_avg_sts": 0.7932742301026626, "eval_sickr_spearman": 0.7534467891875612, "eval_stsb_spearman": 0.833101671017764, "step": 375 }, { "epoch": 0.13, "learning_rate": 8.72024571282314e-06, "loss": 0.3301, "step": 500 }, { "epoch": 0.13, "eval_avg_sts": 0.794880955659569, "eval_sickr_spearman": 0.7557729834613105, "eval_stsb_spearman": 0.8339889278578274, "step": 500 }, { "epoch": 0.16, "eval_avg_sts": 0.7976909480948997, "eval_sickr_spearman": 0.7567474864783884, "eval_stsb_spearman": 0.8386344097114109, "step": 625 }, { "epoch": 0.19, "eval_avg_sts": 0.8008783148847614, "eval_sickr_spearman": 0.7603111060184877, "eval_stsb_spearman": 0.8414455237510349, "step": 750 }, { "epoch": 0.22, "eval_avg_sts": 0.7988165156325646, "eval_sickr_spearman": 0.757687311040228, "eval_stsb_spearman": 0.8399457202249012, "step": 875 }, { "epoch": 0.26, "learning_rate": 7.440491425646277e-06, "loss": 0.0007, "step": 1000 }, { "epoch": 0.26, "eval_avg_sts": 0.7996021062049683, "eval_sickr_spearman": 0.755320914734575, "eval_stsb_spearman": 0.8438832976753616, "step": 1000 }, { "epoch": 0.29, "eval_avg_sts": 0.8008678219858503, "eval_sickr_spearman": 0.7550597696360559, "eval_stsb_spearman": 0.8466758743356445, "step": 1125 }, { "epoch": 0.32, "eval_avg_sts": 0.7974316688328598, "eval_sickr_spearman": 0.749437296932812, "eval_stsb_spearman": 0.8454260407329075, "step": 1250 }, { "epoch": 0.35, "eval_avg_sts": 0.8022743240819741, "eval_sickr_spearman": 0.7575770316313261, "eval_stsb_spearman": 0.846971616532622, "step": 1375 }, { "epoch": 0.38, "learning_rate": 6.160737138469415e-06, "loss": 0.0008, "step": 1500 }, { "epoch": 0.38, "eval_avg_sts": 0.801759968136895, "eval_sickr_spearman": 0.7535204688971674, "eval_stsb_spearman": 0.8499994673766226, "step": 1500 }, { "epoch": 0.42, "eval_avg_sts": 0.801312443833835, "eval_sickr_spearman": 0.7532925613208435, "eval_stsb_spearman": 0.8493323263468264, "step": 1625 }, { "epoch": 0.45, "eval_avg_sts": 0.8019017992008193, "eval_sickr_spearman": 0.7542989569892763, "eval_stsb_spearman": 0.8495046414123621, "step": 1750 }, { "epoch": 0.48, "eval_avg_sts": 0.8053453778611492, "eval_sickr_spearman": 0.759368639746069, "eval_stsb_spearman": 0.8513221159762293, "step": 1875 }, { "epoch": 0.51, "learning_rate": 4.8809828512925526e-06, "loss": 0.0007, "step": 2000 }, { "epoch": 0.51, "eval_avg_sts": 0.8074214655871451, "eval_sickr_spearman": 0.7644952393580661, "eval_stsb_spearman": 0.8503476918162242, "step": 2000 }, { "epoch": 0.54, "eval_avg_sts": 0.8043679171172835, "eval_sickr_spearman": 0.760397754125482, "eval_stsb_spearman": 0.848338080109085, "step": 2125 }, { "epoch": 0.58, "eval_avg_sts": 0.8037538137225763, "eval_sickr_spearman": 0.7591252181239807, "eval_stsb_spearman": 0.8483824093211718, "step": 2250 }, { "epoch": 0.61, "eval_avg_sts": 0.8087713317932268, "eval_sickr_spearman": 0.7665475602914152, "eval_stsb_spearman": 0.8509951032950382, "step": 2375 }, { "epoch": 0.64, "learning_rate": 3.60122856411569e-06, "loss": 0.0006, "step": 2500 }, { "epoch": 0.64, "eval_avg_sts": 0.8110685237844888, "eval_sickr_spearman": 0.765394333545886, "eval_stsb_spearman": 0.8567427140230917, "step": 2500 }, { "epoch": 0.67, "eval_avg_sts": 0.8091393361687496, "eval_sickr_spearman": 0.7646847220532393, "eval_stsb_spearman": 0.8535939502842598, "step": 2625 }, { "epoch": 0.7, "eval_avg_sts": 0.8083355725116668, "eval_sickr_spearman": 0.7630815879805395, "eval_stsb_spearman": 0.8535895570427939, "step": 2750 }, { "epoch": 0.74, "eval_avg_sts": 0.8064249843239497, "eval_sickr_spearman": 0.7595829064895845, "eval_stsb_spearman": 0.8532670621583149, "step": 2875 }, { "epoch": 0.77, "learning_rate": 2.321474276938828e-06, "loss": 0.0007, "step": 3000 }, { "epoch": 0.77, "eval_avg_sts": 0.8075703149702256, "eval_sickr_spearman": 0.7607818588436827, "eval_stsb_spearman": 0.8543587710967685, "step": 3000 }, { "epoch": 0.8, "eval_avg_sts": 0.8088746392140569, "eval_sickr_spearman": 0.7618876788920923, "eval_stsb_spearman": 0.8558615995360215, "step": 3125 }, { "epoch": 0.83, "eval_avg_sts": 0.8085515592471004, "eval_sickr_spearman": 0.7616940174910939, "eval_stsb_spearman": 0.855409101003107, "step": 3250 }, { "epoch": 0.86, "eval_avg_sts": 0.8087552005841584, "eval_sickr_spearman": 0.7611086144267661, "eval_stsb_spearman": 0.8564017867415507, "step": 3375 }, { "epoch": 0.9, "learning_rate": 1.0417199897619657e-06, "loss": 0.0006, "step": 3500 }, { "epoch": 0.9, "eval_avg_sts": 0.8090700871643673, "eval_sickr_spearman": 0.7608862784582091, "eval_stsb_spearman": 0.8572538958705256, "step": 3500 }, { "epoch": 0.93, "eval_avg_sts": 0.8087913053132321, "eval_sickr_spearman": 0.7607510228765595, "eval_stsb_spearman": 0.8568315877499046, "step": 3625 }, { "epoch": 0.96, "eval_avg_sts": 0.8089499832999565, "eval_sickr_spearman": 0.7610159144009905, "eval_stsb_spearman": 0.8568840521989225, "step": 3750 }, { "epoch": 0.99, "eval_avg_sts": 0.8088062391857902, "eval_sickr_spearman": 0.7611648108154486, "eval_stsb_spearman": 0.8564476675561318, "step": 3875 }, { "epoch": 1.0, "step": 3907, "total_flos": 137241565209427968, "train_runtime": 6727.1296, "train_samples_per_second": 0.581 } ], "max_steps": 3907, "num_train_epochs": 1, "total_flos": 137241565209427968, "trial_name": null, "trial_params": null }