| { | |
| "best_metric": 0.8220295594432419, | |
| "best_model_checkpoint": "result/my-unsup-simcse-bert-base-uncased-d0.2-rs192-std0.5-t0.05", | |
| "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": 0.7039632579138562, | |
| "eval_sickr_spearman": 0.6894170562025291, | |
| "eval_stsb_spearman": 0.7185094596251831, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7320830312686424, | |
| "eval_sickr_spearman": 0.7157956410501377, | |
| "eval_stsb_spearman": 0.7483704214871472, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7306267531474842, | |
| "eval_sickr_spearman": 0.7151454920610715, | |
| "eval_stsb_spearman": 0.7461080142338968, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9040061436068092e-05, | |
| "loss": 0.0316, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7490673075570297, | |
| "eval_sickr_spearman": 0.7306142925280271, | |
| "eval_stsb_spearman": 0.7675203225860322, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.751267358789496, | |
| "eval_sickr_spearman": 0.7340745011067302, | |
| "eval_stsb_spearman": 0.7684602164722617, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7786649440391495, | |
| "eval_sickr_spearman": 0.7455865474683157, | |
| "eval_stsb_spearman": 0.8117433406099835, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7696928806594809, | |
| "eval_sickr_spearman": 0.7452508100692632, | |
| "eval_stsb_spearman": 0.7941349512496988, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8080122872136184e-05, | |
| "loss": 0.0008, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7651441102319789, | |
| "eval_sickr_spearman": 0.7429110710249084, | |
| "eval_stsb_spearman": 0.7873771494390496, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.7657607879714892, | |
| "eval_sickr_spearman": 0.7408295471818849, | |
| "eval_stsb_spearman": 0.7906920287610936, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.7673800101461663, | |
| "eval_sickr_spearman": 0.7415319540093158, | |
| "eval_stsb_spearman": 0.7932280662830168, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.7657849020870391, | |
| "eval_sickr_spearman": 0.7396460128113483, | |
| "eval_stsb_spearman": 0.7919237913627298, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.7120184308204276e-05, | |
| "loss": 0.0006, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7564555230487979, | |
| "eval_sickr_spearman": 0.7337248346882608, | |
| "eval_stsb_spearman": 0.7791862114093351, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7663342569223988, | |
| "eval_sickr_spearman": 0.740796117535284, | |
| "eval_stsb_spearman": 0.7918723963095137, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.7650001912920963, | |
| "eval_sickr_spearman": 0.7428139521378004, | |
| "eval_stsb_spearman": 0.787186430446392, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.7697843702026972, | |
| "eval_sickr_spearman": 0.7472399220731197, | |
| "eval_stsb_spearman": 0.7923288183322746, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.6160245744272367e-05, | |
| "loss": 0.0004, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.772637186091681, | |
| "eval_sickr_spearman": 0.7453534525330364, | |
| "eval_stsb_spearman": 0.7999209196503256, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7844981403056908, | |
| "eval_sickr_spearman": 0.7469667211681398, | |
| "eval_stsb_spearman": 0.8220295594432419, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7825301157194102, | |
| "eval_sickr_spearman": 0.745924590360237, | |
| "eval_stsb_spearman": 0.8191356410785836, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.7786888226299735, | |
| "eval_sickr_spearman": 0.7462584545463333, | |
| "eval_stsb_spearman": 0.8111191907136137, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.5200307180340456e-05, | |
| "loss": 0.0004, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.7798782002704093, | |
| "eval_sickr_spearman": 0.7466363632524484, | |
| "eval_stsb_spearman": 0.8131200372883702, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.7751115325157576, | |
| "eval_sickr_spearman": 0.7426031436335886, | |
| "eval_stsb_spearman": 0.8076199213979265, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7666810908791618, | |
| "eval_sickr_spearman": 0.7362270149176792, | |
| "eval_stsb_spearman": 0.7971351668406444, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7732616540318422, | |
| "eval_sickr_spearman": 0.7442702551334042, | |
| "eval_stsb_spearman": 0.8022530529302804, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.424036861640855e-05, | |
| "loss": 0.0004, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.7733068432231915, | |
| "eval_sickr_spearman": 0.7420109201528567, | |
| "eval_stsb_spearman": 0.8046027662935263, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.7672720656943643, | |
| "eval_sickr_spearman": 0.7361518462439285, | |
| "eval_stsb_spearman": 0.7983922851448001, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.7652170530985714, | |
| "eval_sickr_spearman": 0.7340480840009392, | |
| "eval_stsb_spearman": 0.7963860221962036, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7769025861727988, | |
| "eval_sickr_spearman": 0.7457697079566438, | |
| "eval_stsb_spearman": 0.8080354643889539, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3280430052476642e-05, | |
| "loss": 0.0003, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7745062864644923, | |
| "eval_sickr_spearman": 0.743565975093017, | |
| "eval_stsb_spearman": 0.8054465978359675, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7713562889079433, | |
| "eval_sickr_spearman": 0.7430872010738819, | |
| "eval_stsb_spearman": 0.7996253767420045, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.7689988388424054, | |
| "eval_sickr_spearman": 0.7403553361173865, | |
| "eval_stsb_spearman": 0.7976423415674243, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.7691667949901446, | |
| "eval_sickr_spearman": 0.7390877953504347, | |
| "eval_stsb_spearman": 0.7992457946298546, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2320491488544734e-05, | |
| "loss": 0.0004, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7640678967425198, | |
| "eval_sickr_spearman": 0.7361696177514606, | |
| "eval_stsb_spearman": 0.7919661757335791, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7720104233010221, | |
| "eval_sickr_spearman": 0.7409091347169678, | |
| "eval_stsb_spearman": 0.8031117118850762, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.7716821019144677, | |
| "eval_sickr_spearman": 0.7378609849575023, | |
| "eval_stsb_spearman": 0.8055032188714333, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.7749099391602485, | |
| "eval_sickr_spearman": 0.7425721155420596, | |
| "eval_stsb_spearman": 0.8072477627784373, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1360552924612826e-05, | |
| "loss": 0.0004, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.7668234253805365, | |
| "eval_sickr_spearman": 0.7378467197203752, | |
| "eval_stsb_spearman": 0.7958001310406978, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7705218895648999, | |
| "eval_sickr_spearman": 0.7391226659300787, | |
| "eval_stsb_spearman": 0.801921113199721, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7678707529208644, | |
| "eval_sickr_spearman": 0.7376356230295548, | |
| "eval_stsb_spearman": 0.798105882812174, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.768346434351147, | |
| "eval_sickr_spearman": 0.7373303853799155, | |
| "eval_stsb_spearman": 0.7993624833223786, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0400614360680917e-05, | |
| "loss": 0.0002, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.7714937923042149, | |
| "eval_sickr_spearman": 0.7393215146900324, | |
| "eval_stsb_spearman": 0.8036660699183974, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.7705501211555204, | |
| "eval_sickr_spearman": 0.7388855363822787, | |
| "eval_stsb_spearman": 0.802214705928762, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.769454128932314, | |
| "eval_sickr_spearman": 0.7382745327408845, | |
| "eval_stsb_spearman": 0.8006337251237436, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7705108541596231, | |
| "eval_sickr_spearman": 0.7397600866772639, | |
| "eval_stsb_spearman": 0.8012616216419822, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.944067579674901e-05, | |
| "loss": 0.0004, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.767759877044913, | |
| "eval_sickr_spearman": 0.7378074783105003, | |
| "eval_stsb_spearman": 0.7977122757793257, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.7698443583732223, | |
| "eval_sickr_spearman": 0.7391045582048366, | |
| "eval_stsb_spearman": 0.800584158541608, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.7712564343816768, | |
| "eval_sickr_spearman": 0.741173257743776, | |
| "eval_stsb_spearman": 0.8013396110195776, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7664448809595665, | |
| "eval_sickr_spearman": 0.739177853665631, | |
| "eval_stsb_spearman": 0.7937119082535019, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.84807372328171e-05, | |
| "loss": 0.0002, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7638769604550799, | |
| "eval_sickr_spearman": 0.7379414370524111, | |
| "eval_stsb_spearman": 0.7898124838577486, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.766509076179771, | |
| "eval_sickr_spearman": 0.7385067150852364, | |
| "eval_stsb_spearman": 0.7945114372743056, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.7666375164161757, | |
| "eval_sickr_spearman": 0.7395296334525638, | |
| "eval_stsb_spearman": 0.7937453993797876, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.7675116415394574, | |
| "eval_sickr_spearman": 0.7385751113736841, | |
| "eval_stsb_spearman": 0.7964481717052306, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.7520798668885192e-05, | |
| "loss": 0.0003, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7676886943660834, | |
| "eval_sickr_spearman": 0.737567130678904, | |
| "eval_stsb_spearman": 0.797810258053263, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7661323845941221, | |
| "eval_sickr_spearman": 0.7377277947132145, | |
| "eval_stsb_spearman": 0.7945369744750297, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.7657108387605938, | |
| "eval_sickr_spearman": 0.7375463332119813, | |
| "eval_stsb_spearman": 0.7938753443092063, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.7634754316352123, | |
| "eval_sickr_spearman": 0.7353273443566419, | |
| "eval_stsb_spearman": 0.7916235189137826, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.6560860104953284e-05, | |
| "loss": 0.0002, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.7659995201775873, | |
| "eval_sickr_spearman": 0.7410373297267062, | |
| "eval_stsb_spearman": 0.7909617106284685, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7603361178127348, | |
| "eval_sickr_spearman": 0.7367544444425711, | |
| "eval_stsb_spearman": 0.7839177911828984, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7697008342068297, | |
| "eval_sickr_spearman": 0.7388405312402312, | |
| "eval_stsb_spearman": 0.800561137173428, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.7683036722669765, | |
| "eval_sickr_spearman": 0.7386270810254403, | |
| "eval_stsb_spearman": 0.7979802635085127, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.5600921541021372e-05, | |
| "loss": 0.0002, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.7732594696328385, | |
| "eval_sickr_spearman": 0.7422230254968075, | |
| "eval_stsb_spearman": 0.8042959137688697, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.7737749390548014, | |
| "eval_sickr_spearman": 0.742280086445316, | |
| "eval_stsb_spearman": 0.8052697916642868, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.772584123544011, | |
| "eval_sickr_spearman": 0.741925424792297, | |
| "eval_stsb_spearman": 0.8032428222957251, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.772843322519777, | |
| "eval_sickr_spearman": 0.7422232656523147, | |
| "eval_stsb_spearman": 0.8034633793872396, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.4640982977089467e-05, | |
| "loss": 0.0002, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.7715935954543726, | |
| "eval_sickr_spearman": 0.7422030925897106, | |
| "eval_stsb_spearman": 0.8009840983190345, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7690169428580813, | |
| "eval_sickr_spearman": 0.7407306511440238, | |
| "eval_stsb_spearman": 0.7973032345721387, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.7675907004689784, | |
| "eval_sickr_spearman": 0.7394478364868148, | |
| "eval_stsb_spearman": 0.795733564451142, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7677635816632358, | |
| "eval_sickr_spearman": 0.7398567252533574, | |
| "eval_stsb_spearman": 0.7956704380731142, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.368104441315756e-05, | |
| "loss": 0.0001, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7665064992996365, | |
| "eval_sickr_spearman": 0.7390165171959004, | |
| "eval_stsb_spearman": 0.7939964814033726, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.768501738751435, | |
| "eval_sickr_spearman": 0.7403488519186923, | |
| "eval_stsb_spearman": 0.7966546255841779, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.7705756275356871, | |
| "eval_sickr_spearman": 0.7442900919782981, | |
| "eval_stsb_spearman": 0.7968611630930761, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.7708996463902738, | |
| "eval_sickr_spearman": 0.7443713125708299, | |
| "eval_stsb_spearman": 0.7974279802097175, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.2721105849225649e-05, | |
| "loss": 0.0002, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7690762421498543, | |
| "eval_sickr_spearman": 0.7437613656136672, | |
| "eval_stsb_spearman": 0.7943911186860414, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7632497822206319, | |
| "eval_sickr_spearman": 0.7418991037487088, | |
| "eval_stsb_spearman": 0.7846004606925551, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.7661106731884346, | |
| "eval_sickr_spearman": 0.7425510779196297, | |
| "eval_stsb_spearman": 0.7896702684572396, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7582809505084913, | |
| "eval_sickr_spearman": 0.734091696241045, | |
| "eval_stsb_spearman": 0.7824702047759376, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.176116728529374e-05, | |
| "loss": 0.0003, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.760148082899698, | |
| "eval_sickr_spearman": 0.735014469761874, | |
| "eval_stsb_spearman": 0.785281696037522, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7629335466481396, | |
| "eval_sickr_spearman": 0.7371143414856469, | |
| "eval_stsb_spearman": 0.7887527518106323, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7642452574201212, | |
| "eval_sickr_spearman": 0.738628137709672, | |
| "eval_stsb_spearman": 0.7898623771305704, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.7661101842825881, | |
| "eval_sickr_spearman": 0.7410933820220844, | |
| "eval_stsb_spearman": 0.791126986543092, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.0801228721361832e-05, | |
| "loss": 0.0002, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7672844353835271, | |
| "eval_sickr_spearman": 0.741854819073183, | |
| "eval_stsb_spearman": 0.7927140516938713, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7695572795608598, | |
| "eval_sickr_spearman": 0.7413840662479879, | |
| "eval_stsb_spearman": 0.7977304928737319, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7687590638385537, | |
| "eval_sickr_spearman": 0.739025354918565, | |
| "eval_stsb_spearman": 0.7984927727585424, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7669714587155453, | |
| "eval_sickr_spearman": 0.7381937444282656, | |
| "eval_stsb_spearman": 0.795749173002825, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.841290157429926e-06, | |
| "loss": 0.0002, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7675942026136746, | |
| "eval_sickr_spearman": 0.7389821269272708, | |
| "eval_stsb_spearman": 0.7962062783000783, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.7678054650263499, | |
| "eval_sickr_spearman": 0.738494515185471, | |
| "eval_stsb_spearman": 0.7971164148672288, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.7682414535062771, | |
| "eval_sickr_spearman": 0.7377959508461551, | |
| "eval_stsb_spearman": 0.7986869561663991, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7680341988448937, | |
| "eval_sickr_spearman": 0.7380161254151475, | |
| "eval_stsb_spearman": 0.7980522722746398, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.881351593498018e-06, | |
| "loss": 0.0001, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7603629068459727, | |
| "eval_sickr_spearman": 0.7332741588634669, | |
| "eval_stsb_spearman": 0.7874516548284786, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.7621832287379494, | |
| "eval_sickr_spearman": 0.7349048083795262, | |
| "eval_stsb_spearman": 0.7894616490963726, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.7621016943200358, | |
| "eval_sickr_spearman": 0.7353910816282503, | |
| "eval_stsb_spearman": 0.7888123070118214, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.7610119893777948, | |
| "eval_sickr_spearman": 0.7316196795433297, | |
| "eval_stsb_spearman": 0.7904042992122599, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.921413029566108e-06, | |
| "loss": 0.0002, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7617757929098798, | |
| "eval_sickr_spearman": 0.7318357714686997, | |
| "eval_stsb_spearman": 0.7917158143510601, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7611330064890793, | |
| "eval_sickr_spearman": 0.7338293983960914, | |
| "eval_stsb_spearman": 0.7884366145820674, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.7621633110041289, | |
| "eval_sickr_spearman": 0.7341784404102422, | |
| "eval_stsb_spearman": 0.7901481815980157, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.7595196438845424, | |
| "eval_sickr_spearman": 0.7331927461465293, | |
| "eval_stsb_spearman": 0.7858465416225555, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.961474465634199e-06, | |
| "loss": 0.0003, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.7635656813674752, | |
| "eval_sickr_spearman": 0.7330679133138917, | |
| "eval_stsb_spearman": 0.7940634494210588, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7621975138758124, | |
| "eval_sickr_spearman": 0.7340756058220633, | |
| "eval_stsb_spearman": 0.7903194219295613, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7631065024648397, | |
| "eval_sickr_spearman": 0.7321919220858634, | |
| "eval_stsb_spearman": 0.794021082843816, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.7628859990174557, | |
| "eval_sickr_spearman": 0.7333722383726035, | |
| "eval_stsb_spearman": 0.7923997596623078, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 6.001535901702292e-06, | |
| "loss": 0.0003, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.7629723942710446, | |
| "eval_sickr_spearman": 0.7341199385286906, | |
| "eval_stsb_spearman": 0.7918248500133985, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.7576056121896545, | |
| "eval_sickr_spearman": 0.7317101701384391, | |
| "eval_stsb_spearman": 0.7835010542408699, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7578230668275763, | |
| "eval_sickr_spearman": 0.7317774136804525, | |
| "eval_stsb_spearman": 0.7838687199747002, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7592080133938004, | |
| "eval_sickr_spearman": 0.7330049445399063, | |
| "eval_stsb_spearman": 0.7854110822476944, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 5.0415973377703825e-06, | |
| "loss": 0.0002, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.7649310794949764, | |
| "eval_sickr_spearman": 0.7366114078224884, | |
| "eval_stsb_spearman": 0.7932507511674645, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7663911796944379, | |
| "eval_sickr_spearman": 0.7376808683271094, | |
| "eval_stsb_spearman": 0.7951014910617664, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.7666777328866676, | |
| "eval_sickr_spearman": 0.7386608949208526, | |
| "eval_stsb_spearman": 0.7946945708524826, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.767106492489369, | |
| "eval_sickr_spearman": 0.7391275651024254, | |
| "eval_stsb_spearman": 0.7950854198763126, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.081658773838474e-06, | |
| "loss": 0.0002, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7663641777603831, | |
| "eval_sickr_spearman": 0.7384899041997329, | |
| "eval_stsb_spearman": 0.7942384513210333, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.7662444760718037, | |
| "eval_sickr_spearman": 0.7383694902284276, | |
| "eval_stsb_spearman": 0.7941194619151797, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.766969739589136, | |
| "eval_sickr_spearman": 0.7386586374590851, | |
| "eval_stsb_spearman": 0.7952808417191867, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.766963601043843, | |
| "eval_sickr_spearman": 0.7385023442550055, | |
| "eval_stsb_spearman": 0.7954248578326806, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.121720209906566e-06, | |
| "loss": 0.0001, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7667105573155901, | |
| "eval_sickr_spearman": 0.7384424975026136, | |
| "eval_stsb_spearman": 0.7949786171285665, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7681149645930665, | |
| "eval_sickr_spearman": 0.7393782874519323, | |
| "eval_stsb_spearman": 0.7968516417342008, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.7680143099599532, | |
| "eval_sickr_spearman": 0.7393469231426932, | |
| "eval_stsb_spearman": 0.7966816967772132, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.7676174458337446, | |
| "eval_sickr_spearman": 0.7389574389411316, | |
| "eval_stsb_spearman": 0.7962774527263575, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.1617816459746575e-06, | |
| "loss": 0.0001, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.7678090950201439, | |
| "eval_sickr_spearman": 0.7391150770160515, | |
| "eval_stsb_spearman": 0.7965031130242363, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7674776747079344, | |
| "eval_sickr_spearman": 0.7389988417505713, | |
| "eval_stsb_spearman": 0.7959565076652975, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7655137295819681, | |
| "eval_sickr_spearman": 0.7381747721431975, | |
| "eval_stsb_spearman": 0.7928526870207386, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.7657447555359481, | |
| "eval_sickr_spearman": 0.7382629572454379, | |
| "eval_stsb_spearman": 0.7932265538264583, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2018430820427492e-06, | |
| "loss": 0.0002, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.7661720244295546, | |
| "eval_sickr_spearman": 0.7381314480897004, | |
| "eval_stsb_spearman": 0.7942126007694087, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.7660620762670376, | |
| "eval_sickr_spearman": 0.7380774611316839, | |
| "eval_stsb_spearman": 0.7940466914023911, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.766592222198617, | |
| "eval_sickr_spearman": 0.738653161913521, | |
| "eval_stsb_spearman": 0.794531282483713, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.766138712195428, | |
| "eval_sickr_spearman": 0.7384393754810201, | |
| "eval_stsb_spearman": 0.7938380489098358, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.4190451811084087e-07, | |
| "loss": 0.0002, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.7661330517073972, | |
| "eval_sickr_spearman": 0.7384201630404448, | |
| "eval_stsb_spearman": 0.7938459403743495, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.7660975303030011, | |
| "eval_sickr_spearman": 0.7383983569203919, | |
| "eval_stsb_spearman": 0.7937967036856103, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 3834.5201, | |
| "train_samples_per_second": 4.075 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 42267980696739840, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |