| { | |
| "best_metric": 0.8532649267943023, | |
| "best_model_checkpoint": "result/esimcse-roberta-large-0.15-bpe-0.28-128-bs64", | |
| "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": 0.7558217259678449, | |
| "eval_sickr_spearman": 0.7201285747419712, | |
| "eval_stsb_spearman": 0.7915148771937186, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7742364775992936, | |
| "eval_sickr_spearman": 0.7452505218826546, | |
| "eval_stsb_spearman": 0.8032224333159326, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7750260297296185, | |
| "eval_sickr_spearman": 0.7433483942035024, | |
| "eval_stsb_spearman": 0.8067036652557346, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.680020478689364e-06, | |
| "loss": 0.3135, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7818566453500417, | |
| "eval_sickr_spearman": 0.7477289446624279, | |
| "eval_stsb_spearman": 0.8159843460376556, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.7811770146364453, | |
| "eval_sickr_spearman": 0.7481865670513635, | |
| "eval_stsb_spearman": 0.814167462221527, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7770666807380804, | |
| "eval_sickr_spearman": 0.7352241255196514, | |
| "eval_stsb_spearman": 0.8189092359565092, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7764823954854172, | |
| "eval_sickr_spearman": 0.7331835722061546, | |
| "eval_stsb_spearman": 0.8197812187646797, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.36004095737873e-06, | |
| "loss": 0.0031, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7839001001478747, | |
| "eval_sickr_spearman": 0.7432803341327646, | |
| "eval_stsb_spearman": 0.8245198661629847, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.7943887972940393, | |
| "eval_sickr_spearman": 0.7568865845481532, | |
| "eval_stsb_spearman": 0.8318910100399254, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.7936405273161878, | |
| "eval_sickr_spearman": 0.7561879241466344, | |
| "eval_stsb_spearman": 0.8310931304857413, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.7943520912745958, | |
| "eval_sickr_spearman": 0.7518201139033611, | |
| "eval_stsb_spearman": 0.8368840686458305, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.040061436068092e-06, | |
| "loss": 0.0015, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7887681712959922, | |
| "eval_sickr_spearman": 0.7468924650853165, | |
| "eval_stsb_spearman": 0.8306438775066679, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7897590452490246, | |
| "eval_sickr_spearman": 0.7516917807721141, | |
| "eval_stsb_spearman": 0.8278263097259353, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.7905759145553224, | |
| "eval_sickr_spearman": 0.7542222513202796, | |
| "eval_stsb_spearman": 0.8269295777903652, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.7904633436602717, | |
| "eval_sickr_spearman": 0.7515385135274251, | |
| "eval_stsb_spearman": 0.8293881737931184, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 8.720081914757458e-06, | |
| "loss": 0.0014, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.802801711004685, | |
| "eval_sickr_spearman": 0.7642203093334342, | |
| "eval_stsb_spearman": 0.8413831126759359, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.8055867687511298, | |
| "eval_sickr_spearman": 0.7670204745161748, | |
| "eval_stsb_spearman": 0.8441530629860846, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7998281543125508, | |
| "eval_sickr_spearman": 0.7605069768501523, | |
| "eval_stsb_spearman": 0.8391493317749493, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.793160108534175, | |
| "eval_sickr_spearman": 0.7518937995847628, | |
| "eval_stsb_spearman": 0.8344264174835871, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.400102393446819e-06, | |
| "loss": 0.001, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.7992184752929077, | |
| "eval_sickr_spearman": 0.7565757272596456, | |
| "eval_stsb_spearman": 0.8418612233261697, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.7927251468672808, | |
| "eval_sickr_spearman": 0.7446415355475207, | |
| "eval_stsb_spearman": 0.8408087581870409, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.8028596486819746, | |
| "eval_sickr_spearman": 0.7524543705696469, | |
| "eval_stsb_spearman": 0.8532649267943023, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7893919999211613, | |
| "eval_sickr_spearman": 0.7413113471604106, | |
| "eval_stsb_spearman": 0.837472652681912, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.080122872136184e-06, | |
| "loss": 0.001, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.7934231636232565, | |
| "eval_sickr_spearman": 0.7474009703562418, | |
| "eval_stsb_spearman": 0.8394453568902713, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.7940617961811679, | |
| "eval_sickr_spearman": 0.7509521018099672, | |
| "eval_stsb_spearman": 0.8371714905523686, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.7937132166778321, | |
| "eval_sickr_spearman": 0.7539420378744897, | |
| "eval_stsb_spearman": 0.8334843954811746, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7923308718046048, | |
| "eval_sickr_spearman": 0.7480851733962276, | |
| "eval_stsb_spearman": 0.836576570212982, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.760143350825547e-06, | |
| "loss": 0.0009, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.791635348297863, | |
| "eval_sickr_spearman": 0.748936284513711, | |
| "eval_stsb_spearman": 0.834334412082015, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7912451615688103, | |
| "eval_sickr_spearman": 0.7463851304741407, | |
| "eval_stsb_spearman": 0.83610519266348, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.7903505118672073, | |
| "eval_sickr_spearman": 0.7453605611360493, | |
| "eval_stsb_spearman": 0.8353404625983653, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.7877047830672446, | |
| "eval_sickr_spearman": 0.747770473619605, | |
| "eval_stsb_spearman": 0.8276390925148842, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 7.440163829514912e-06, | |
| "loss": 0.001, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.787528888012516, | |
| "eval_sickr_spearman": 0.7437081951843754, | |
| "eval_stsb_spearman": 0.8313495808406565, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7906651938867704, | |
| "eval_sickr_spearman": 0.7461365516108834, | |
| "eval_stsb_spearman": 0.8351938361626572, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.7920362260308209, | |
| "eval_sickr_spearman": 0.7466354986926226, | |
| "eval_stsb_spearman": 0.8374369533690192, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.7911902163460698, | |
| "eval_sickr_spearman": 0.7485247540365892, | |
| "eval_stsb_spearman": 0.8338556786555504, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.120184308204276e-06, | |
| "loss": 0.0009, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.7911417580464639, | |
| "eval_sickr_spearman": 0.7545039537302142, | |
| "eval_stsb_spearman": 0.8277795623627137, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7895925041055443, | |
| "eval_sickr_spearman": 0.750248782391609, | |
| "eval_stsb_spearman": 0.8289362258194797, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7821260859065025, | |
| "eval_sickr_spearman": 0.7355510732271406, | |
| "eval_stsb_spearman": 0.8287010985858643, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.7922597968364788, | |
| "eval_sickr_spearman": 0.7458612853685416, | |
| "eval_stsb_spearman": 0.838658308304416, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 6.800204786893639e-06, | |
| "loss": 0.0006, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.7854669556183871, | |
| "eval_sickr_spearman": 0.7395174335527985, | |
| "eval_stsb_spearman": 0.8314164776839756, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.7826092624953684, | |
| "eval_sickr_spearman": 0.7410220078053474, | |
| "eval_stsb_spearman": 0.8241965171853896, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7927314656873996, | |
| "eval_sickr_spearman": 0.7533276720559947, | |
| "eval_stsb_spearman": 0.8321352593188046, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7895263185412165, | |
| "eval_sickr_spearman": 0.7509032541798045, | |
| "eval_stsb_spearman": 0.8281493829026283, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 6.480225265583003e-06, | |
| "loss": 0.0008, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.7935029645348786, | |
| "eval_sickr_spearman": 0.7503155936537093, | |
| "eval_stsb_spearman": 0.8366903354160478, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.793606484311646, | |
| "eval_sickr_spearman": 0.7550370989561772, | |
| "eval_stsb_spearman": 0.8321758696671149, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.7888706031258408, | |
| "eval_sickr_spearman": 0.7515478015672222, | |
| "eval_stsb_spearman": 0.8261934046844593, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7897914038552645, | |
| "eval_sickr_spearman": 0.7541371882396327, | |
| "eval_stsb_spearman": 0.8254456194708961, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.1602457442723675e-06, | |
| "loss": 0.0007, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7837650110744263, | |
| "eval_sickr_spearman": 0.7490165925153156, | |
| "eval_stsb_spearman": 0.818513429633537, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.7830368725221115, | |
| "eval_sickr_spearman": 0.752153984061253, | |
| "eval_stsb_spearman": 0.81391976098297, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.7928730911226451, | |
| "eval_sickr_spearman": 0.7560166932700075, | |
| "eval_stsb_spearman": 0.8297294889752829, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.7946611044937584, | |
| "eval_sickr_spearman": 0.7549690388854393, | |
| "eval_stsb_spearman": 0.8343531701020775, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 5.840266222961732e-06, | |
| "loss": 0.0009, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7918018347204214, | |
| "eval_sickr_spearman": 0.7562617479495448, | |
| "eval_stsb_spearman": 0.827341921491298, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7942526450763034, | |
| "eval_sickr_spearman": 0.7569746255570893, | |
| "eval_stsb_spearman": 0.8315306645955175, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.7940324422099587, | |
| "eval_sickr_spearman": 0.757333417884832, | |
| "eval_stsb_spearman": 0.8307314665350856, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.7955177075824724, | |
| "eval_sickr_spearman": 0.7577290980984791, | |
| "eval_stsb_spearman": 0.8333063170664659, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.520286701651095e-06, | |
| "loss": 0.0006, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.7955995637779287, | |
| "eval_sickr_spearman": 0.7530336736840921, | |
| "eval_stsb_spearman": 0.8381654538717653, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7955238196018252, | |
| "eval_sickr_spearman": 0.7530732993427786, | |
| "eval_stsb_spearman": 0.8379743398608718, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7921706195432732, | |
| "eval_sickr_spearman": 0.7532472199610859, | |
| "eval_stsb_spearman": 0.8310940191254605, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.7929086096204032, | |
| "eval_sickr_spearman": 0.7574800088064211, | |
| "eval_stsb_spearman": 0.8283372104343854, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.200307180340458e-06, | |
| "loss": 0.0008, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.785403249216813, | |
| "eval_sickr_spearman": 0.7543678816198399, | |
| "eval_stsb_spearman": 0.8164386168137862, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.7880462597418649, | |
| "eval_sickr_spearman": 0.7476711933329325, | |
| "eval_stsb_spearman": 0.8284213261507974, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7942161541157922, | |
| "eval_sickr_spearman": 0.754079454855704, | |
| "eval_stsb_spearman": 0.8343528533758804, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7968011299207969, | |
| "eval_sickr_spearman": 0.7561006996664228, | |
| "eval_stsb_spearman": 0.837501560175171, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8803276590298225e-06, | |
| "loss": 0.0005, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.7962915878196872, | |
| "eval_sickr_spearman": 0.7541978995518505, | |
| "eval_stsb_spearman": 0.838385276087524, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7903274885336161, | |
| "eval_sickr_spearman": 0.7504886497121908, | |
| "eval_stsb_spearman": 0.8301663273550415, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.7900160038644817, | |
| "eval_sickr_spearman": 0.7502795222965293, | |
| "eval_stsb_spearman": 0.829752485432434, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7954484420566725, | |
| "eval_sickr_spearman": 0.7508721300260727, | |
| "eval_stsb_spearman": 0.8400247540872725, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.560348137719187e-06, | |
| "loss": 0.0005, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7934980610345465, | |
| "eval_sickr_spearman": 0.7509290949123782, | |
| "eval_stsb_spearman": 0.8360670271567147, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.7886755296146708, | |
| "eval_sickr_spearman": 0.747253658968131, | |
| "eval_stsb_spearman": 0.8300974002612107, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.7889761398352569, | |
| "eval_sickr_spearman": 0.7419386813762938, | |
| "eval_stsb_spearman": 0.8360135982942201, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.7898054394244394, | |
| "eval_sickr_spearman": 0.7431272109813798, | |
| "eval_stsb_spearman": 0.836483667867499, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.24036861640855e-06, | |
| "loss": 0.0005, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7904517544876187, | |
| "eval_sickr_spearman": 0.7436593475542127, | |
| "eval_stsb_spearman": 0.8372441614210248, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7851659097948578, | |
| "eval_sickr_spearman": 0.7379476810955982, | |
| "eval_stsb_spearman": 0.8323841384941174, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.7848891215669203, | |
| "eval_sickr_spearman": 0.7407074521220292, | |
| "eval_stsb_spearman": 0.8290707910118115, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7871117814010626, | |
| "eval_sickr_spearman": 0.752143561312241, | |
| "eval_stsb_spearman": 0.8220800014898841, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.920389095097914e-06, | |
| "loss": 0.0008, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.7930397869535002, | |
| "eval_sickr_spearman": 0.7565603092760841, | |
| "eval_stsb_spearman": 0.8295192646309164, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7979788678504024, | |
| "eval_sickr_spearman": 0.7627290877270851, | |
| "eval_stsb_spearman": 0.8332286479737196, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7967114262262445, | |
| "eval_sickr_spearman": 0.7644874102885316, | |
| "eval_stsb_spearman": 0.8289354421639574, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.801864222640388, | |
| "eval_sickr_spearman": 0.7659917924166748, | |
| "eval_stsb_spearman": 0.8377366528641014, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.600409573787278e-06, | |
| "loss": 0.0005, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7984090692179152, | |
| "eval_sickr_spearman": 0.7636251079244132, | |
| "eval_stsb_spearman": 0.8331930305114172, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7950614235855115, | |
| "eval_sickr_spearman": 0.761550308435591, | |
| "eval_stsb_spearman": 0.8285725387354319, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7995666707496905, | |
| "eval_sickr_spearman": 0.7639466281174399, | |
| "eval_stsb_spearman": 0.8351867133819411, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.794625632859824, | |
| "eval_sickr_spearman": 0.761795891457244, | |
| "eval_stsb_spearman": 0.8274553742624039, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.280430052476642e-06, | |
| "loss": 0.0005, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7935143889832998, | |
| "eval_sickr_spearman": 0.7613707384202849, | |
| "eval_stsb_spearman": 0.8256580395463149, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.7943479719067924, | |
| "eval_sickr_spearman": 0.7625838416763363, | |
| "eval_stsb_spearman": 0.8261121021372485, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.7887052397337098, | |
| "eval_sickr_spearman": 0.7615231708632785, | |
| "eval_stsb_spearman": 0.815887308604141, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.794244759928596, | |
| "eval_sickr_spearman": 0.7613072230312128, | |
| "eval_stsb_spearman": 0.8271822968259792, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 2.960450531166006e-06, | |
| "loss": 0.0006, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7874100790996466, | |
| "eval_sickr_spearman": 0.7576639198938275, | |
| "eval_stsb_spearman": 0.8171562383054657, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.789179157033471, | |
| "eval_sickr_spearman": 0.7584775187210878, | |
| "eval_stsb_spearman": 0.819880795345854, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.7873417867649358, | |
| "eval_sickr_spearman": 0.7538412686236725, | |
| "eval_stsb_spearman": 0.8208423049061991, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.7870785735870268, | |
| "eval_sickr_spearman": 0.7543877664958354, | |
| "eval_stsb_spearman": 0.8197693806782183, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 2.640471009855369e-06, | |
| "loss": 0.0006, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7886389449841755, | |
| "eval_sickr_spearman": 0.7516758344464366, | |
| "eval_stsb_spearman": 0.8256020555219146, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7816407061330272, | |
| "eval_sickr_spearman": 0.7511766952402917, | |
| "eval_stsb_spearman": 0.8121047170257626, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.7880937817548945, | |
| "eval_sickr_spearman": 0.7527955835142635, | |
| "eval_stsb_spearman": 0.8233919799955256, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.7923766055265167, | |
| "eval_sickr_spearman": 0.7552486279269106, | |
| "eval_stsb_spearman": 0.8295045831261226, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.3204914885447333e-06, | |
| "loss": 0.0006, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.7926331664582384, | |
| "eval_sickr_spearman": 0.7545561155063759, | |
| "eval_stsb_spearman": 0.8307102174101011, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7931451615056908, | |
| "eval_sickr_spearman": 0.7558778833868514, | |
| "eval_stsb_spearman": 0.8304124396245304, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7883567651086225, | |
| "eval_sickr_spearman": 0.7542461708087957, | |
| "eval_stsb_spearman": 0.8224673594084494, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.7862547538448916, | |
| "eval_sickr_spearman": 0.7544667296265997, | |
| "eval_stsb_spearman": 0.8180427780631837, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.000511967234097e-06, | |
| "loss": 0.0008, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.7841379910987387, | |
| "eval_sickr_spearman": 0.7531644143422066, | |
| "eval_stsb_spearman": 0.815111567855271, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.7835845835747262, | |
| "eval_sickr_spearman": 0.7517147876697029, | |
| "eval_stsb_spearman": 0.8154543794797493, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.785837092232381, | |
| "eval_sickr_spearman": 0.7525228148891964, | |
| "eval_stsb_spearman": 0.8191513695755656, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7832659241868551, | |
| "eval_sickr_spearman": 0.750127647953782, | |
| "eval_stsb_spearman": 0.8164042004199282, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6805324459234608e-06, | |
| "loss": 0.0006, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.7814624983749696, | |
| "eval_sickr_spearman": 0.7456563366587051, | |
| "eval_stsb_spearman": 0.8172686600912342, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7813485495429527, | |
| "eval_sickr_spearman": 0.7430078536943062, | |
| "eval_stsb_spearman": 0.8196892453915993, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.7848142530915927, | |
| "eval_sickr_spearman": 0.7457980764390489, | |
| "eval_stsb_spearman": 0.8238304297441366, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7872867184061136, | |
| "eval_sickr_spearman": 0.7477951135746429, | |
| "eval_stsb_spearman": 0.8267783232375843, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3605529246128248e-06, | |
| "loss": 0.0008, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7841077000969573, | |
| "eval_sickr_spearman": 0.7476563997536896, | |
| "eval_stsb_spearman": 0.820559000440225, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.7858331122771427, | |
| "eval_sickr_spearman": 0.7483037808982353, | |
| "eval_stsb_spearman": 0.8233624436560499, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.7841019114435142, | |
| "eval_sickr_spearman": 0.7481370469857809, | |
| "eval_stsb_spearman": 0.8200667759012474, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.7799615698577176, | |
| "eval_sickr_spearman": 0.745557104403134, | |
| "eval_stsb_spearman": 0.8143660353123011, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0405734033021888e-06, | |
| "loss": 0.0006, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7825679820043924, | |
| "eval_sickr_spearman": 0.7460993934448276, | |
| "eval_stsb_spearman": 0.8190365705639572, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7831020374187931, | |
| "eval_sickr_spearman": 0.7452405314135555, | |
| "eval_stsb_spearman": 0.8209635434240309, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.7813864099961566, | |
| "eval_sickr_spearman": 0.7446007571423997, | |
| "eval_stsb_spearman": 0.8181720628499135, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.7834609263550933, | |
| "eval_sickr_spearman": 0.7452780917348799, | |
| "eval_stsb_spearman": 0.8216437609753068, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 7.205938819915525e-07, | |
| "loss": 0.0004, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.7842031580372257, | |
| "eval_sickr_spearman": 0.7454702641717339, | |
| "eval_stsb_spearman": 0.8229360519027176, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7831083936431449, | |
| "eval_sickr_spearman": 0.744312234316061, | |
| "eval_stsb_spearman": 0.8219045529702287, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7840839730054485, | |
| "eval_sickr_spearman": 0.7447129097642577, | |
| "eval_stsb_spearman": 0.8234550362466394, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.7836831213121471, | |
| "eval_sickr_spearman": 0.7446981642161162, | |
| "eval_stsb_spearman": 0.8226680784081779, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.0061436068091647e-07, | |
| "loss": 0.0008, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.7842518604266702, | |
| "eval_sickr_spearman": 0.7449969176670613, | |
| "eval_stsb_spearman": 0.823506803186279, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.7840099446787474, | |
| "eval_sickr_spearman": 0.744850855087588, | |
| "eval_stsb_spearman": 0.823169034269907, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7847295434857142, | |
| "eval_sickr_spearman": 0.745309888324032, | |
| "eval_stsb_spearman": 0.8241491986473966, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7847406171241809, | |
| "eval_sickr_spearman": 0.7451867846110461, | |
| "eval_stsb_spearman": 0.8242944496373156, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.06348393702803e-08, | |
| "loss": 0.0005, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.7846749915698714, | |
| "eval_sickr_spearman": 0.7452754500243008, | |
| "eval_stsb_spearman": 0.824074533115442, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.7844513467514725, | |
| "eval_sickr_spearman": 0.7450856249779936, | |
| "eval_stsb_spearman": 0.8238170685249515, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 11057.6017, | |
| "train_samples_per_second": 1.413 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 332584911022755840, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |