{ "best_metric": 0.8433710108449586, "best_model_checkpoint": "result/my-unsup-simcse-roberta-base/SupCon-48-0.2-3e-6-0.07-0.00", "epoch": 1.0, "global_step": 7813, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.02, "eval_avg_sts": 0.7648697833708831, "eval_sickr_spearman": 0.7122734723505795, "eval_stsb_spearman": 0.8174660943911868, "step": 125 }, { "epoch": 0.03, "eval_avg_sts": 0.7713268000072269, "eval_sickr_spearman": 0.7159774387690809, "eval_stsb_spearman": 0.8266761612453729, "step": 250 }, { "epoch": 0.05, "eval_avg_sts": 0.7717212045194138, "eval_sickr_spearman": 0.716408085624575, "eval_stsb_spearman": 0.8270343234142526, "step": 375 }, { "epoch": 0.06, "learning_rate": 1.872008191475746e-05, "loss": 0.1356, "step": 500 }, { "epoch": 0.06, "eval_avg_sts": 0.7714897999409676, "eval_sickr_spearman": 0.714245581344527, "eval_stsb_spearman": 0.8287340185374082, "step": 500 }, { "epoch": 0.08, "eval_avg_sts": 0.7683945074315737, "eval_sickr_spearman": 0.707755714980412, "eval_stsb_spearman": 0.8290332998827354, "step": 625 }, { "epoch": 0.1, "eval_avg_sts": 0.7712141135572375, "eval_sickr_spearman": 0.7136527334594763, "eval_stsb_spearman": 0.8287754936549987, "step": 750 }, { "epoch": 0.11, "eval_avg_sts": 0.7754955425035636, "eval_sickr_spearman": 0.7167608740646381, "eval_stsb_spearman": 0.8342302109424891, "step": 875 }, { "epoch": 0.13, "learning_rate": 1.7440163829514915e-05, "loss": 0.0005, "step": 1000 }, { "epoch": 0.13, "eval_avg_sts": 0.7749643365975677, "eval_sickr_spearman": 0.7165658197616978, "eval_stsb_spearman": 0.8333628534334376, "step": 1000 }, { "epoch": 0.14, "eval_avg_sts": 0.7747453579869918, "eval_sickr_spearman": 0.707071944220339, "eval_stsb_spearman": 0.8424187717536445, "step": 1125 }, { "epoch": 0.16, "eval_avg_sts": 0.779538087156316, "eval_sickr_spearman": 0.7169955059951632, "eval_stsb_spearman": 0.8420806683174688, "step": 1250 }, { "epoch": 0.18, "eval_avg_sts": 0.780701521846623, "eval_sickr_spearman": 0.7190910068887052, "eval_stsb_spearman": 0.8423120368045408, "step": 1375 }, { "epoch": 0.19, "learning_rate": 1.616024574427237e-05, "loss": 0.0003, "step": 1500 }, { "epoch": 0.19, "eval_avg_sts": 0.7822508597322496, "eval_sickr_spearman": 0.725160745208541, "eval_stsb_spearman": 0.8393409742559582, "step": 1500 }, { "epoch": 0.21, "eval_avg_sts": 0.7806747469546087, "eval_sickr_spearman": 0.7225573634483924, "eval_stsb_spearman": 0.8387921304608251, "step": 1625 }, { "epoch": 0.22, "eval_avg_sts": 0.783739242864979, "eval_sickr_spearman": 0.7298345731433808, "eval_stsb_spearman": 0.8376439125865772, "step": 1750 }, { "epoch": 0.24, "eval_avg_sts": 0.7819519916053913, "eval_sickr_spearman": 0.7310624707355461, "eval_stsb_spearman": 0.8328415124752365, "step": 1875 }, { "epoch": 0.26, "learning_rate": 1.4880327659029823e-05, "loss": 0.0003, "step": 2000 }, { "epoch": 0.26, "eval_avg_sts": 0.7824129920157619, "eval_sickr_spearman": 0.731245325138721, "eval_stsb_spearman": 0.8335806588928026, "step": 2000 }, { "epoch": 0.27, "eval_avg_sts": 0.786146631865762, "eval_sickr_spearman": 0.7289222528865652, "eval_stsb_spearman": 0.8433710108449586, "step": 2125 }, { "epoch": 0.29, "eval_avg_sts": 0.7797286001800978, "eval_sickr_spearman": 0.7224563060109667, "eval_stsb_spearman": 0.8370008943492291, "step": 2250 }, { "epoch": 0.3, "eval_avg_sts": 0.780005893079553, "eval_sickr_spearman": 0.7195845264559818, "eval_stsb_spearman": 0.8404272597031243, "step": 2375 }, { "epoch": 0.32, "learning_rate": 1.3600409573787278e-05, "loss": 0.0003, "step": 2500 }, { "epoch": 0.32, "eval_avg_sts": 0.7815935533979442, "eval_sickr_spearman": 0.7247260157094245, "eval_stsb_spearman": 0.8384610910864639, "step": 2500 }, { "epoch": 0.34, "eval_avg_sts": 0.7790885863000977, "eval_sickr_spearman": 0.719636736263245, "eval_stsb_spearman": 0.8385404363369504, "step": 2625 }, { "epoch": 0.35, "eval_avg_sts": 0.7786611915735007, "eval_sickr_spearman": 0.720594044146008, "eval_stsb_spearman": 0.8367283390009934, "step": 2750 }, { "epoch": 0.37, "eval_avg_sts": 0.7740761960283342, "eval_sickr_spearman": 0.7133921647341745, "eval_stsb_spearman": 0.8347602273224938, "step": 2875 }, { "epoch": 0.38, "learning_rate": 1.2320491488544735e-05, "loss": 0.0003, "step": 3000 }, { "epoch": 0.38, "eval_avg_sts": 0.7730594558706145, "eval_sickr_spearman": 0.7107198103123609, "eval_stsb_spearman": 0.8353991014288681, "step": 3000 }, { "epoch": 0.4, "eval_avg_sts": 0.7735160682870633, "eval_sickr_spearman": 0.7115497397141101, "eval_stsb_spearman": 0.8354823968600166, "step": 3125 }, { "epoch": 0.42, "eval_avg_sts": 0.7696881215537197, "eval_sickr_spearman": 0.7034571235261068, "eval_stsb_spearman": 0.8359191195813325, "step": 3250 }, { "epoch": 0.43, "eval_avg_sts": 0.7731307195393922, "eval_sickr_spearman": 0.7140373184886913, "eval_stsb_spearman": 0.8322241205900931, "step": 3375 }, { "epoch": 0.45, "learning_rate": 1.104057340330219e-05, "loss": 0.0003, "step": 3500 }, { "epoch": 0.45, "eval_avg_sts": 0.7744383399418961, "eval_sickr_spearman": 0.7103547259103297, "eval_stsb_spearman": 0.8385219539734627, "step": 3500 }, { "epoch": 0.46, "eval_avg_sts": 0.7749032723185794, "eval_sickr_spearman": 0.7106885420653246, "eval_stsb_spearman": 0.839118002571834, "step": 3625 }, { "epoch": 0.48, "eval_avg_sts": 0.7729790005154229, "eval_sickr_spearman": 0.7092490979863258, "eval_stsb_spearman": 0.8367089030445198, "step": 3750 }, { "epoch": 0.5, "eval_avg_sts": 0.7675103682931668, "eval_sickr_spearman": 0.7009157018868134, "eval_stsb_spearman": 0.83410503469952, "step": 3875 }, { "epoch": 0.51, "learning_rate": 9.760655318059645e-06, "loss": 0.0003, "step": 4000 }, { "epoch": 0.51, "eval_avg_sts": 0.7600618545185335, "eval_sickr_spearman": 0.6939802029635556, "eval_stsb_spearman": 0.8261435060735114, "step": 4000 }, { "epoch": 0.53, "eval_avg_sts": 0.7611442447284995, "eval_sickr_spearman": 0.6946870766834202, "eval_stsb_spearman": 0.8276014127735787, "step": 4125 }, { "epoch": 0.54, "eval_avg_sts": 0.765201470252675, "eval_sickr_spearman": 0.6980777842583424, "eval_stsb_spearman": 0.8323251562470075, "step": 4250 }, { "epoch": 0.56, "eval_avg_sts": 0.7601537112935712, "eval_sickr_spearman": 0.6892433277086275, "eval_stsb_spearman": 0.831064094878515, "step": 4375 }, { "epoch": 0.58, "learning_rate": 8.4807372328171e-06, "loss": 0.0002, "step": 4500 }, { "epoch": 0.58, "eval_avg_sts": 0.7572602743940218, "eval_sickr_spearman": 0.6881403895263041, "eval_stsb_spearman": 0.8263801592617394, "step": 4500 }, { "epoch": 0.59, "eval_avg_sts": 0.7620805832344215, "eval_sickr_spearman": 0.6991238944111545, "eval_stsb_spearman": 0.8250372720576886, "step": 4625 }, { "epoch": 0.61, "eval_avg_sts": 0.7638147878715696, "eval_sickr_spearman": 0.6998619475523625, "eval_stsb_spearman": 0.8277676281907768, "step": 4750 }, { "epoch": 0.62, "eval_avg_sts": 0.762036570031333, "eval_sickr_spearman": 0.6968174001255049, "eval_stsb_spearman": 0.8272557399371612, "step": 4875 }, { "epoch": 0.64, "learning_rate": 7.200819147574556e-06, "loss": 0.0002, "step": 5000 }, { "epoch": 0.64, "eval_avg_sts": 0.7616123775261012, "eval_sickr_spearman": 0.6975154841538064, "eval_stsb_spearman": 0.8257092708983962, "step": 5000 }, { "epoch": 0.66, "eval_avg_sts": 0.7572440478626198, "eval_sickr_spearman": 0.6909725434226024, "eval_stsb_spearman": 0.8235155523026372, "step": 5125 }, { "epoch": 0.67, "eval_avg_sts": 0.7615420687641667, "eval_sickr_spearman": 0.7007691109652242, "eval_stsb_spearman": 0.8223150265631094, "step": 5250 }, { "epoch": 0.69, "eval_avg_sts": 0.7622736227072948, "eval_sickr_spearman": 0.70394151718411, "eval_stsb_spearman": 0.8206057282304795, "step": 5375 }, { "epoch": 0.7, "learning_rate": 5.920901062332012e-06, "loss": 0.0002, "step": 5500 }, { "epoch": 0.7, "eval_avg_sts": 0.7598959005042131, "eval_sickr_spearman": 0.7014216615093624, "eval_stsb_spearman": 0.8183701394990638, "step": 5500 }, { "epoch": 0.72, "eval_avg_sts": 0.7617340962416614, "eval_sickr_spearman": 0.705636198536151, "eval_stsb_spearman": 0.8178319939471719, "step": 5625 }, { "epoch": 0.74, "eval_avg_sts": 0.7602681846424262, "eval_sickr_spearman": 0.6992537297148518, "eval_stsb_spearman": 0.8212826395700007, "step": 5750 }, { "epoch": 0.75, "eval_avg_sts": 0.7600256686195976, "eval_sickr_spearman": 0.6983496883235836, "eval_stsb_spearman": 0.8217016489156117, "step": 5875 }, { "epoch": 0.77, "learning_rate": 4.640982977089467e-06, "loss": 0.0003, "step": 6000 }, { "epoch": 0.77, "eval_avg_sts": 0.7581145273334985, "eval_sickr_spearman": 0.6957373727785674, "eval_stsb_spearman": 0.8204916818884296, "step": 6000 }, { "epoch": 0.78, "eval_avg_sts": 0.759467788986251, "eval_sickr_spearman": 0.6969048167301223, "eval_stsb_spearman": 0.8220307612423797, "step": 6125 }, { "epoch": 0.8, "eval_avg_sts": 0.758567409617105, "eval_sickr_spearman": 0.6932303894390052, "eval_stsb_spearman": 0.8239044297952048, "step": 6250 }, { "epoch": 0.82, "eval_avg_sts": 0.7593640477073755, "eval_sickr_spearman": 0.6939898572149447, "eval_stsb_spearman": 0.8247382381998064, "step": 6375 }, { "epoch": 0.83, "learning_rate": 3.3610648918469217e-06, "loss": 0.0002, "step": 6500 }, { "epoch": 0.83, "eval_avg_sts": 0.7596970769949833, "eval_sickr_spearman": 0.6966290701767661, "eval_stsb_spearman": 0.8227650838132007, "step": 6500 }, { "epoch": 0.85, "eval_avg_sts": 0.7608506682812288, "eval_sickr_spearman": 0.6957939053849601, "eval_stsb_spearman": 0.8259074311774975, "step": 6625 }, { "epoch": 0.86, "eval_avg_sts": 0.7601011463122427, "eval_sickr_spearman": 0.6948619579237564, "eval_stsb_spearman": 0.8253403347007289, "step": 6750 }, { "epoch": 0.88, "eval_avg_sts": 0.7613624100018648, "eval_sickr_spearman": 0.6960120626476921, "eval_stsb_spearman": 0.8267127573560374, "step": 6875 }, { "epoch": 0.9, "learning_rate": 2.0811468066043775e-06, "loss": 0.0002, "step": 7000 }, { "epoch": 0.9, "eval_avg_sts": 0.7632853767372183, "eval_sickr_spearman": 0.6985597763612739, "eval_stsb_spearman": 0.8280109771131627, "step": 7000 }, { "epoch": 0.91, "eval_avg_sts": 0.761317305690945, "eval_sickr_spearman": 0.6965424220697717, "eval_stsb_spearman": 0.8260921893121185, "step": 7125 }, { "epoch": 0.93, "eval_avg_sts": 0.7612243104925751, "eval_sickr_spearman": 0.6965610101060282, "eval_stsb_spearman": 0.8258876108791219, "step": 7250 }, { "epoch": 0.94, "eval_avg_sts": 0.7611869303629479, "eval_sickr_spearman": 0.6964085113589622, "eval_stsb_spearman": 0.8259653493669338, "step": 7375 }, { "epoch": 0.96, "learning_rate": 8.012287213618329e-07, "loss": 0.0002, "step": 7500 }, { "epoch": 0.96, "eval_avg_sts": 0.761389925660918, "eval_sickr_spearman": 0.6968965073495735, "eval_stsb_spearman": 0.8258833439722626, "step": 7500 }, { "epoch": 0.98, "eval_avg_sts": 0.7614202638478765, "eval_sickr_spearman": 0.6971455005794286, "eval_stsb_spearman": 0.8256950271163245, "step": 7625 }, { "epoch": 0.99, "eval_avg_sts": 0.7612860433632234, "eval_sickr_spearman": 0.6969965561338691, "eval_stsb_spearman": 0.8255755305925775, "step": 7750 }, { "epoch": 1.0, "step": 7813, "total_flos": 48546397602349056, "train_runtime": 6045.9604, "train_samples_per_second": 1.292 } ], "max_steps": 7813, "num_train_epochs": 1, "total_flos": 48546397602349056, "trial_name": null, "trial_params": null }