| { | |
| "best_metric": 0.8458953427959743, | |
| "best_model_checkpoint": "result/202112181001/unsup-50", | |
| "epoch": 1.0, | |
| "global_step": 15626, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "eval_avg_sts": 0.761248005315763, | |
| "eval_sickr_spearman": 0.7213024068300173, | |
| "eval_stsb_spearman": 0.8011936038015086, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7746352581861129, | |
| "eval_sickr_spearman": 0.7301758166029986, | |
| "eval_stsb_spearman": 0.8190946997692271, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7744791612857793, | |
| "eval_sickr_spearman": 0.7277358846810439, | |
| "eval_stsb_spearman": 0.8212224378905147, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.680020478689364e-06, | |
| "loss": 1.7462, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7754321639768694, | |
| "eval_sickr_spearman": 0.7178552146798041, | |
| "eval_stsb_spearman": 0.8330091132739348, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.7756061119368747, | |
| "eval_sickr_spearman": 0.7211281980251013, | |
| "eval_stsb_spearman": 0.830084025848648, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7761725091988075, | |
| "eval_sickr_spearman": 0.7143484639638074, | |
| "eval_stsb_spearman": 0.8379965544338077, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.77636456856168, | |
| "eval_sickr_spearman": 0.7136611869333294, | |
| "eval_stsb_spearman": 0.8390679501900307, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.36004095737873e-06, | |
| "loss": 1.3861, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7815101681729256, | |
| "eval_sickr_spearman": 0.720763641965186, | |
| "eval_stsb_spearman": 0.8422566943806652, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.782742997532409, | |
| "eval_sickr_spearman": 0.7218028601992392, | |
| "eval_stsb_spearman": 0.8436831348655786, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.7794339691410397, | |
| "eval_sickr_spearman": 0.7191296238942614, | |
| "eval_stsb_spearman": 0.839738314387818, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.779702256639139, | |
| "eval_sickr_spearman": 0.7198173812357537, | |
| "eval_stsb_spearman": 0.8395871320425241, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.040061436068092e-06, | |
| "loss": 1.3857, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7785124378341957, | |
| "eval_sickr_spearman": 0.7208927495658516, | |
| "eval_stsb_spearman": 0.8361321261025397, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7815668593754403, | |
| "eval_sickr_spearman": 0.7249143456581633, | |
| "eval_stsb_spearman": 0.8382193730927173, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.7780682651684501, | |
| "eval_sickr_spearman": 0.7186580065092406, | |
| "eval_stsb_spearman": 0.8374785238276595, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.7764755234700329, | |
| "eval_sickr_spearman": 0.7121685243939373, | |
| "eval_stsb_spearman": 0.8407825225461284, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 8.720081914757458e-06, | |
| "loss": 1.3856, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.7787685310346691, | |
| "eval_sickr_spearman": 0.711641719273364, | |
| "eval_stsb_spearman": 0.8458953427959743, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.777459506394991, | |
| "eval_sickr_spearman": 0.7114851378826758, | |
| "eval_stsb_spearman": 0.8434338749073061, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7748485468900443, | |
| "eval_sickr_spearman": 0.7098947320518569, | |
| "eval_stsb_spearman": 0.8398023617282316, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.7709473209139031, | |
| "eval_sickr_spearman": 0.7093069754635588, | |
| "eval_stsb_spearman": 0.8325876663642473, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.400102393446819e-06, | |
| "loss": 1.3846, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.7701267446908748, | |
| "eval_sickr_spearman": 0.7100784510148577, | |
| "eval_stsb_spearman": 0.830175038366892, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.7680303310013604, | |
| "eval_sickr_spearman": 0.7085754617886564, | |
| "eval_stsb_spearman": 0.8274852002140645, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7692721367240678, | |
| "eval_sickr_spearman": 0.7100761935530902, | |
| "eval_stsb_spearman": 0.8284680798950453, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7687344261791832, | |
| "eval_sickr_spearman": 0.7127513818098886, | |
| "eval_stsb_spearman": 0.8247174705484778, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.080122872136184e-06, | |
| "loss": 1.3839, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.7679441883899669, | |
| "eval_sickr_spearman": 0.7105865410600155, | |
| "eval_stsb_spearman": 0.8253018357199182, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.7656167470448696, | |
| "eval_sickr_spearman": 0.7087553862946435, | |
| "eval_stsb_spearman": 0.8224781077950958, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.7663780332257234, | |
| "eval_sickr_spearman": 0.7098858943291922, | |
| "eval_stsb_spearman": 0.8228701721222546, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7674357066052843, | |
| "eval_sickr_spearman": 0.7114235139795306, | |
| "eval_stsb_spearman": 0.8234478992310379, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.760143350825547e-06, | |
| "loss": 1.3838, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7701785710496845, | |
| "eval_sickr_spearman": 0.7123642031011961, | |
| "eval_stsb_spearman": 0.8279929389981728, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7656151765041461, | |
| "eval_sickr_spearman": 0.7059885546963011, | |
| "eval_stsb_spearman": 0.8252417983119913, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.7621620825341495, | |
| "eval_sickr_spearman": 0.7033147113103427, | |
| "eval_stsb_spearman": 0.8210094537579563, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.7633276550955901, | |
| "eval_sickr_spearman": 0.7018189267493571, | |
| "eval_stsb_spearman": 0.824836383441823, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 7.440163829514912e-06, | |
| "loss": 1.3838, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7638754922978754, | |
| "eval_sickr_spearman": 0.704708381749671, | |
| "eval_stsb_spearman": 0.8230426028460798, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7629738705536501, | |
| "eval_sickr_spearman": 0.7085977482197237, | |
| "eval_stsb_spearman": 0.8173499928875765, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.7627301530711117, | |
| "eval_sickr_spearman": 0.7088170101977885, | |
| "eval_stsb_spearman": 0.8166432959444349, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.7628966433663935, | |
| "eval_sickr_spearman": 0.7097353648572853, | |
| "eval_stsb_spearman": 0.8160579218755019, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.120184308204276e-06, | |
| "loss": 1.3836, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.7678625208909737, | |
| "eval_sickr_spearman": 0.711450411396336, | |
| "eval_stsb_spearman": 0.8242746303856114, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7655480126563503, | |
| "eval_sickr_spearman": 0.7153905467406086, | |
| "eval_stsb_spearman": 0.8157054785720921, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7596495835137121, | |
| "eval_sickr_spearman": 0.7168872919236232, | |
| "eval_stsb_spearman": 0.8024118751038011, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.759915830574287, | |
| "eval_sickr_spearman": 0.716031713913706, | |
| "eval_stsb_spearman": 0.8037999472348679, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 6.800204786893639e-06, | |
| "loss": 1.3839, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.7631274164456352, | |
| "eval_sickr_spearman": 0.7173388323082428, | |
| "eval_stsb_spearman": 0.8089160005830275, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.7644212632530025, | |
| "eval_sickr_spearman": 0.7177001222532603, | |
| "eval_stsb_spearman": 0.8111424042527446, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7650758154422789, | |
| "eval_sickr_spearman": 0.7161933385700453, | |
| "eval_stsb_spearman": 0.8139582923145124, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7623450154586153, | |
| "eval_sickr_spearman": 0.7126757328251238, | |
| "eval_stsb_spearman": 0.812014298092107, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 6.480225265583003e-06, | |
| "loss": 1.3837, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.760563393080759, | |
| "eval_sickr_spearman": 0.7095261413794209, | |
| "eval_stsb_spearman": 0.8116006447820971, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.7619760416313872, | |
| "eval_sickr_spearman": 0.7102908925765185, | |
| "eval_stsb_spearman": 0.8136611906862561, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.7724230141569579, | |
| "eval_sickr_spearman": 0.7169229310008902, | |
| "eval_stsb_spearman": 0.8279230973130257, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7702102651422748, | |
| "eval_sickr_spearman": 0.7178937356231574, | |
| "eval_stsb_spearman": 0.8225267946613921, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.1602457442723675e-06, | |
| "loss": 1.3838, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.770665987431916, | |
| "eval_sickr_spearman": 0.7171555936562564, | |
| "eval_stsb_spearman": 0.8241763812075757, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.7691855971919378, | |
| "eval_sickr_spearman": 0.715682864023961, | |
| "eval_stsb_spearman": 0.8226883303599146, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.767784144291997, | |
| "eval_sickr_spearman": 0.7138758859567579, | |
| "eval_stsb_spearman": 0.8216924026272361, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.7649847098109797, | |
| "eval_sickr_spearman": 0.7114461846594095, | |
| "eval_stsb_spearman": 0.8185232349625498, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 5.840266222961732e-06, | |
| "loss": 1.3836, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7637744131826092, | |
| "eval_sickr_spearman": 0.710355302283547, | |
| "eval_stsb_spearman": 0.8171935240816712, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7608727838172745, | |
| "eval_sickr_spearman": 0.7098643763957481, | |
| "eval_stsb_spearman": 0.811881191238801, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.7646512924332959, | |
| "eval_sickr_spearman": 0.7139204107877909, | |
| "eval_stsb_spearman": 0.8153821740788009, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.7601972138062232, | |
| "eval_sickr_spearman": 0.7111435887203493, | |
| "eval_stsb_spearman": 0.8092508388920971, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.520286701651095e-06, | |
| "loss": 1.3835, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.7611323377854309, | |
| "eval_sickr_spearman": 0.7118220280281626, | |
| "eval_stsb_spearman": 0.810442647542699, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7604355676265102, | |
| "eval_sickr_spearman": 0.7119947959000356, | |
| "eval_stsb_spearman": 0.808876339352985, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7612569771069564, | |
| "eval_sickr_spearman": 0.7121121358808489, | |
| "eval_stsb_spearman": 0.810401818333064, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.7580965487519077, | |
| "eval_sickr_spearman": 0.7099162499853012, | |
| "eval_stsb_spearman": 0.8062768475185141, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.200307180340458e-06, | |
| "loss": 1.3834, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.7603187166135681, | |
| "eval_sickr_spearman": 0.7112826867901141, | |
| "eval_stsb_spearman": 0.809354746437022, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.7666064632874305, | |
| "eval_sickr_spearman": 0.7150916972274607, | |
| "eval_stsb_spearman": 0.8181212293474004, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7667637614191116, | |
| "eval_sickr_spearman": 0.7163257603167101, | |
| "eval_stsb_spearman": 0.817201762521513, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.76575579617678, | |
| "eval_sickr_spearman": 0.7169438245300158, | |
| "eval_stsb_spearman": 0.8145677678235441, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8803276590298225e-06, | |
| "loss": 1.3835, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.7645504111230637, | |
| "eval_sickr_spearman": 0.7151874712437283, | |
| "eval_stsb_spearman": 0.8139133510023991, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7636472351766994, | |
| "eval_sickr_spearman": 0.7151762319659918, | |
| "eval_stsb_spearman": 0.812118238387407, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.763386535627875, | |
| "eval_sickr_spearman": 0.7155925655532575, | |
| "eval_stsb_spearman": 0.8111805057024924, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7640074826605159, | |
| "eval_sickr_spearman": 0.7152915546405447, | |
| "eval_stsb_spearman": 0.8127234106804869, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.560348137719187e-06, | |
| "loss": 1.3834, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7649386730175963, | |
| "eval_sickr_spearman": 0.717294883850427, | |
| "eval_stsb_spearman": 0.8125824621847657, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.7640366893274448, | |
| "eval_sickr_spearman": 0.7173166899704798, | |
| "eval_stsb_spearman": 0.81075668868441, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.7659452440774945, | |
| "eval_sickr_spearman": 0.7160896874531419, | |
| "eval_stsb_spearman": 0.8158008007018472, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.7648165929469989, | |
| "eval_sickr_spearman": 0.7170738447216088, | |
| "eval_stsb_spearman": 0.8125593411723889, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.24036861640855e-06, | |
| "loss": 1.3833, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7648317013147197, | |
| "eval_sickr_spearman": 0.7197047483028814, | |
| "eval_stsb_spearman": 0.809958654326558, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7646856930039588, | |
| "eval_sickr_spearman": 0.7153951096952452, | |
| "eval_stsb_spearman": 0.8139762763126724, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.7605850473405789, | |
| "eval_sickr_spearman": 0.7148776706394525, | |
| "eval_stsb_spearman": 0.8062924240417051, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7585442400027406, | |
| "eval_sickr_spearman": 0.712004930462439, | |
| "eval_stsb_spearman": 0.8050835495430422, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.920389095097914e-06, | |
| "loss": 1.3836, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.7579964843404219, | |
| "eval_sickr_spearman": 0.7117327382105892, | |
| "eval_stsb_spearman": 0.8042602304702546, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7572053974073514, | |
| "eval_sickr_spearman": 0.7118436420238098, | |
| "eval_stsb_spearman": 0.8025671527908931, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7554038631509703, | |
| "eval_sickr_spearman": 0.7089095180991583, | |
| "eval_stsb_spearman": 0.8018982082027825, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.7562469961681211, | |
| "eval_sickr_spearman": 0.7074827542309392, | |
| "eval_stsb_spearman": 0.8050112381053031, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.600409573787278e-06, | |
| "loss": 1.3838, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7552535332565927, | |
| "eval_sickr_spearman": 0.7048757870552695, | |
| "eval_stsb_spearman": 0.8056312794579159, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7556796351772364, | |
| "eval_sickr_spearman": 0.7052522859423561, | |
| "eval_stsb_spearman": 0.8061069844121167, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7548247282438733, | |
| "eval_sickr_spearman": 0.7052348986836356, | |
| "eval_stsb_spearman": 0.8044145578041109, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7550666157689844, | |
| "eval_sickr_spearman": 0.7055118460145278, | |
| "eval_stsb_spearman": 0.804621385523441, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.280430052476642e-06, | |
| "loss": 1.3834, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7537972221274167, | |
| "eval_sickr_spearman": 0.7036938207939937, | |
| "eval_stsb_spearman": 0.8039006234608399, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.7562406389240146, | |
| "eval_sickr_spearman": 0.705437541900603, | |
| "eval_stsb_spearman": 0.8070437359474261, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.754961024854511, | |
| "eval_sickr_spearman": 0.7057035861714687, | |
| "eval_stsb_spearman": 0.8042184635375533, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7551902781685098, | |
| "eval_sickr_spearman": 0.7065001339577184, | |
| "eval_stsb_spearman": 0.8038804223793012, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 2.960450531166006e-06, | |
| "loss": 1.3835, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7554401053339275, | |
| "eval_sickr_spearman": 0.7060114175005856, | |
| "eval_stsb_spearman": 0.8048687931672693, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.7534755447971544, | |
| "eval_sickr_spearman": 0.7050457522061723, | |
| "eval_stsb_spearman": 0.8019053373881364, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.7539382725635217, | |
| "eval_sickr_spearman": 0.7049087675048707, | |
| "eval_stsb_spearman": 0.8029677776221726, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.753868215480303, | |
| "eval_sickr_spearman": 0.704858142723955, | |
| "eval_stsb_spearman": 0.8028782882366511, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 2.640471009855369e-06, | |
| "loss": 1.3834, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7557061287206795, | |
| "eval_sickr_spearman": 0.7051221216574589, | |
| "eval_stsb_spearman": 0.8062901357839002, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7559397122519471, | |
| "eval_sickr_spearman": 0.7052638134067013, | |
| "eval_stsb_spearman": 0.8066156110971928, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.7552144818328008, | |
| "eval_sickr_spearman": 0.7066991748420779, | |
| "eval_stsb_spearman": 0.8037297888235237, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.7580575974049937, | |
| "eval_sickr_spearman": 0.7095217225180886, | |
| "eval_stsb_spearman": 0.8065934722918988, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.3204914885447333e-06, | |
| "loss": 1.3837, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.7593639629333797, | |
| "eval_sickr_spearman": 0.7092687907379155, | |
| "eval_stsb_spearman": 0.809459135128844, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.758702221139986, | |
| "eval_sickr_spearman": 0.7086140787942126, | |
| "eval_stsb_spearman": 0.8087903634857594, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7586744411177226, | |
| "eval_sickr_spearman": 0.7085018781412531, | |
| "eval_stsb_spearman": 0.8088470040941923, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.7563138538872969, | |
| "eval_sickr_spearman": 0.7065571468751254, | |
| "eval_stsb_spearman": 0.8060705608994685, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.000511967234097e-06, | |
| "loss": 1.3834, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.7560181589528421, | |
| "eval_sickr_spearman": 0.7051710653498243, | |
| "eval_stsb_spearman": 0.8068652525558597, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.7557672310008925, | |
| "eval_sickr_spearman": 0.7053724597581543, | |
| "eval_stsb_spearman": 0.8061620022436305, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7570608961078531, | |
| "eval_sickr_spearman": 0.7063670878067347, | |
| "eval_stsb_spearman": 0.8077547044089713, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.754422551565998, | |
| "eval_sickr_spearman": 0.7042020378783104, | |
| "eval_stsb_spearman": 0.8046430652536857, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6805324459234608e-06, | |
| "loss": 1.3833, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.7542485855006298, | |
| "eval_sickr_spearman": 0.7042514138505888, | |
| "eval_stsb_spearman": 0.8042457571506708, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7565743715828739, | |
| "eval_sickr_spearman": 0.7055018075143271, | |
| "eval_stsb_spearman": 0.8076469356514206, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.7554365827378587, | |
| "eval_sickr_spearman": 0.7045894567425102, | |
| "eval_stsb_spearman": 0.8062837087332073, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7558174021768111, | |
| "eval_sickr_spearman": 0.7048116005866615, | |
| "eval_stsb_spearman": 0.8068232037669607, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3605529246128248e-06, | |
| "loss": 1.3836, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7569484443145752, | |
| "eval_sickr_spearman": 0.705718763999523, | |
| "eval_stsb_spearman": 0.8081781246296273, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.7568877768032388, | |
| "eval_sickr_spearman": 0.7058967672614527, | |
| "eval_stsb_spearman": 0.807878786345025, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.7568799416714773, | |
| "eval_sickr_spearman": 0.7063571453687371, | |
| "eval_stsb_spearman": 0.8074027379742176, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.7568042749881129, | |
| "eval_sickr_spearman": 0.7065208833935396, | |
| "eval_stsb_spearman": 0.8070876665826862, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0405734033021888e-06, | |
| "loss": 1.3832, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7566981479496052, | |
| "eval_sickr_spearman": 0.7065628145450951, | |
| "eval_stsb_spearman": 0.8068334813541154, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.756387866564892, | |
| "eval_sickr_spearman": 0.7066071472517225, | |
| "eval_stsb_spearman": 0.8061685858780614, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.7562446066417168, | |
| "eval_sickr_spearman": 0.7063944655345545, | |
| "eval_stsb_spearman": 0.8060947477488789, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.7563928060572327, | |
| "eval_sickr_spearman": 0.7064530154472075, | |
| "eval_stsb_spearman": 0.8063325966672579, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 7.205938819915525e-07, | |
| "loss": 1.3832, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.756455127789967, | |
| "eval_sickr_spearman": 0.7067425469266765, | |
| "eval_stsb_spearman": 0.8061677086532576, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7557241070509702, | |
| "eval_sickr_spearman": 0.7057858634482321, | |
| "eval_stsb_spearman": 0.8056623506537083, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7557558426468225, | |
| "eval_sickr_spearman": 0.7064631500096109, | |
| "eval_stsb_spearman": 0.805048535284034, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.7558703144820087, | |
| "eval_sickr_spearman": 0.7066919221457608, | |
| "eval_stsb_spearman": 0.8050487068182565, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.0061436068091647e-07, | |
| "loss": 1.3835, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.7560711558386244, | |
| "eval_sickr_spearman": 0.7066483579367564, | |
| "eval_stsb_spearman": 0.8054939537404925, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.7559005428331718, | |
| "eval_sickr_spearman": 0.706265598089396, | |
| "eval_stsb_spearman": 0.8055354875769476, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7558756101540065, | |
| "eval_sickr_spearman": 0.706157720235566, | |
| "eval_stsb_spearman": 0.8055935000724471, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7559496122876235, | |
| "eval_sickr_spearman": 0.706316799243529, | |
| "eval_stsb_spearman": 0.8055824253317181, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.06348393702803e-08, | |
| "loss": 1.3833, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.7561548756681951, | |
| "eval_sickr_spearman": 0.7064748695983618, | |
| "eval_stsb_spearman": 0.8058348817380282, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.7561501455323656, | |
| "eval_sickr_spearman": 0.7064547925979608, | |
| "eval_stsb_spearman": 0.8058454984667704, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 9210.9807, | |
| "train_samples_per_second": 1.696 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 478639376454758400, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |