{ "best_metric": 0.7694667033321299, "best_model_checkpoint": "/praid/simcse", "epoch": 0.40088254442863686, "global_step": 66500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 9.999969858455306e-09, "loss": 5.4599, "step": 1 }, { "epoch": 0.0, "eval_avg_sts": 0.5995849851747349, "eval_sickr_spearman": 0.5697828282953606, "eval_stsb_spearman": 0.6293871420541092, "step": 250 }, { "epoch": 0.0, "learning_rate": 9.984929227653058e-09, "loss": 5.4858, "step": 500 }, { "epoch": 0.0, "eval_avg_sts": 0.603632450991215, "eval_sickr_spearman": 0.5695678305286125, "eval_stsb_spearman": 0.6376970714538175, "step": 500 }, { "epoch": 0.0, "eval_avg_sts": 0.6086904770175188, "eval_sickr_spearman": 0.5718612396457673, "eval_stsb_spearman": 0.6455197143892701, "step": 750 }, { "epoch": 0.01, "learning_rate": 9.969858455306118e-09, "loss": 5.2729, "step": 1000 }, { "epoch": 0.01, "eval_avg_sts": 0.612799381419272, "eval_sickr_spearman": 0.5742047501651341, "eval_stsb_spearman": 0.6513940126734101, "step": 1000 }, { "epoch": 0.01, "eval_avg_sts": 0.6162570263582352, "eval_sickr_spearman": 0.5758439027009196, "eval_stsb_spearman": 0.656670150015551, "step": 1250 }, { "epoch": 0.01, "learning_rate": 9.954787682959177e-09, "loss": 4.9236, "step": 1500 }, { "epoch": 0.01, "eval_avg_sts": 0.6186805435983472, "eval_sickr_spearman": 0.5769587834066011, "eval_stsb_spearman": 0.6604023037900933, "step": 1500 }, { "epoch": 0.01, "eval_avg_sts": 0.6207924657118333, "eval_sickr_spearman": 0.5785645184067163, "eval_stsb_spearman": 0.6630204130169505, "step": 1750 }, { "epoch": 0.01, "learning_rate": 9.939716910612235e-09, "loss": 4.3524, "step": 2000 }, { "epoch": 0.01, "eval_avg_sts": 0.6228190394905305, "eval_sickr_spearman": 0.5800011736250131, "eval_stsb_spearman": 0.6656369053560479, "step": 2000 }, { "epoch": 0.01, "eval_avg_sts": 0.6242788079574524, "eval_sickr_spearman": 0.5802247872392591, "eval_stsb_spearman": 0.6683328286756457, "step": 2250 }, { "epoch": 0.02, "learning_rate": 9.924646138265295e-09, "loss": 3.8783, "step": 2500 }, { "epoch": 0.02, "eval_avg_sts": 0.6263156249834569, "eval_sickr_spearman": 0.5810882644873206, "eval_stsb_spearman": 0.6715429854795933, "step": 2500 }, { "epoch": 0.02, "eval_avg_sts": 0.6282653788172143, "eval_sickr_spearman": 0.5811133151884266, "eval_stsb_spearman": 0.6754174424460021, "step": 2750 }, { "epoch": 0.02, "learning_rate": 9.909575365918353e-09, "loss": 3.3578, "step": 3000 }, { "epoch": 0.02, "eval_avg_sts": 0.6304970270247872, "eval_sickr_spearman": 0.5811867242012102, "eval_stsb_spearman": 0.6798073298483642, "step": 3000 }, { "epoch": 0.02, "eval_avg_sts": 0.6327063870946847, "eval_sickr_spearman": 0.5807497065029901, "eval_stsb_spearman": 0.6846630676863793, "step": 3250 }, { "epoch": 0.02, "learning_rate": 9.894504593571412e-09, "loss": 2.9718, "step": 3500 }, { "epoch": 0.02, "eval_avg_sts": 0.635125450375027, "eval_sickr_spearman": 0.5803658779673568, "eval_stsb_spearman": 0.6898850227826971, "step": 3500 }, { "epoch": 0.02, "eval_avg_sts": 0.6375971333458483, "eval_sickr_spearman": 0.5801840487233453, "eval_stsb_spearman": 0.6950102179683513, "step": 3750 }, { "epoch": 0.02, "learning_rate": 9.87943382122447e-09, "loss": 2.4584, "step": 4000 }, { "epoch": 0.02, "eval_avg_sts": 0.6397488256977145, "eval_sickr_spearman": 0.5797635156813201, "eval_stsb_spearman": 0.6997341357141089, "step": 4000 }, { "epoch": 0.03, "eval_avg_sts": 0.6416441729368727, "eval_sickr_spearman": 0.5791708648996884, "eval_stsb_spearman": 0.7041174809740571, "step": 4250 }, { "epoch": 0.03, "learning_rate": 9.86436304887753e-09, "loss": 1.6636, "step": 4500 }, { "epoch": 0.03, "eval_avg_sts": 0.6439610982515633, "eval_sickr_spearman": 0.5797300483430633, "eval_stsb_spearman": 0.7081921481600633, "step": 4500 }, { "epoch": 0.03, "eval_avg_sts": 0.6455335224686861, "eval_sickr_spearman": 0.5791015895016001, "eval_stsb_spearman": 0.7119654554357722, "step": 4750 }, { "epoch": 0.03, "learning_rate": 9.849292276530588e-09, "loss": 1.2653, "step": 5000 }, { "epoch": 0.03, "eval_avg_sts": 0.6473309813927987, "eval_sickr_spearman": 0.5790988005567453, "eval_stsb_spearman": 0.715563162228852, "step": 5000 }, { "epoch": 0.03, "eval_avg_sts": 0.6491464635996412, "eval_sickr_spearman": 0.5793477636876175, "eval_stsb_spearman": 0.7189451635116649, "step": 5250 }, { "epoch": 0.03, "learning_rate": 9.834221504183646e-09, "loss": 1.0416, "step": 5500 }, { "epoch": 0.03, "eval_avg_sts": 0.6504410144276549, "eval_sickr_spearman": 0.5787376820006436, "eval_stsb_spearman": 0.7221443468546662, "step": 5500 }, { "epoch": 0.03, "eval_avg_sts": 0.6517348476964206, "eval_sickr_spearman": 0.578490262749959, "eval_stsb_spearman": 0.7249794326428824, "step": 5750 }, { "epoch": 0.04, "learning_rate": 9.819150731836706e-09, "loss": 0.8965, "step": 6000 }, { "epoch": 0.04, "eval_avg_sts": 0.6534361836672715, "eval_sickr_spearman": 0.5791566213598945, "eval_stsb_spearman": 0.7277157459746485, "step": 6000 }, { "epoch": 0.04, "eval_avg_sts": 0.654605629234608, "eval_sickr_spearman": 0.5794011022579643, "eval_stsb_spearman": 0.7298101562112517, "step": 6250 }, { "epoch": 0.04, "learning_rate": 9.804079959489765e-09, "loss": 0.7796, "step": 6500 }, { "epoch": 0.04, "eval_avg_sts": 0.6556384016928898, "eval_sickr_spearman": 0.5794165410598388, "eval_stsb_spearman": 0.7318602623259409, "step": 6500 }, { "epoch": 0.04, "eval_avg_sts": 0.6565914169527195, "eval_sickr_spearman": 0.5794604669413009, "eval_stsb_spearman": 0.7337223669641381, "step": 6750 }, { "epoch": 0.04, "learning_rate": 9.789009187142823e-09, "loss": 0.6669, "step": 7000 }, { "epoch": 0.04, "eval_avg_sts": 0.6580426081739461, "eval_sickr_spearman": 0.5804354023783784, "eval_stsb_spearman": 0.7356498139695139, "step": 7000 }, { "epoch": 0.04, "eval_avg_sts": 0.659003327547673, "eval_sickr_spearman": 0.5809793462302257, "eval_stsb_spearman": 0.7370273088651204, "step": 7250 }, { "epoch": 0.05, "learning_rate": 9.773938414795883e-09, "loss": 0.5639, "step": 7500 }, { "epoch": 0.05, "eval_avg_sts": 0.6599478368878535, "eval_sickr_spearman": 0.5816024263923263, "eval_stsb_spearman": 0.7382932473833808, "step": 7500 }, { "epoch": 0.05, "eval_avg_sts": 0.6606422753240502, "eval_sickr_spearman": 0.581862097079337, "eval_stsb_spearman": 0.7394224535687633, "step": 7750 }, { "epoch": 0.05, "learning_rate": 9.758867642448941e-09, "loss": 0.4826, "step": 8000 }, { "epoch": 0.05, "eval_avg_sts": 0.6615698341083216, "eval_sickr_spearman": 0.5826618768190197, "eval_stsb_spearman": 0.7404777913976235, "step": 8000 }, { "epoch": 0.05, "eval_avg_sts": 0.6619288475610956, "eval_sickr_spearman": 0.582522877799563, "eval_stsb_spearman": 0.7413348173226283, "step": 8250 }, { "epoch": 0.05, "learning_rate": 9.743796870102e-09, "loss": 0.4145, "step": 8500 }, { "epoch": 0.05, "eval_avg_sts": 0.6623974970827452, "eval_sickr_spearman": 0.5825905595148771, "eval_stsb_spearman": 0.7422044346506133, "step": 8500 }, { "epoch": 0.05, "eval_avg_sts": 0.662889528502383, "eval_sickr_spearman": 0.5827362320809506, "eval_stsb_spearman": 0.7430428249238155, "step": 8750 }, { "epoch": 0.05, "learning_rate": 9.72872609775506e-09, "loss": 0.3629, "step": 9000 }, { "epoch": 0.05, "eval_avg_sts": 0.6636302923347899, "eval_sickr_spearman": 0.5834812289752723, "eval_stsb_spearman": 0.7437793556943073, "step": 9000 }, { "epoch": 0.06, "eval_avg_sts": 0.664159926773715, "eval_sickr_spearman": 0.5837915488929486, "eval_stsb_spearman": 0.7445283046544816, "step": 9250 }, { "epoch": 0.06, "learning_rate": 9.713655325408118e-09, "loss": 0.3202, "step": 9500 }, { "epoch": 0.06, "eval_avg_sts": 0.6648401388341261, "eval_sickr_spearman": 0.5845315157260146, "eval_stsb_spearman": 0.7451487619422374, "step": 9500 }, { "epoch": 0.06, "eval_avg_sts": 0.6654341412381362, "eval_sickr_spearman": 0.5849954766236348, "eval_stsb_spearman": 0.7458728058526375, "step": 9750 }, { "epoch": 0.06, "learning_rate": 9.698584553061174e-09, "loss": 0.2807, "step": 10000 }, { "epoch": 0.06, "eval_avg_sts": 0.665940208652456, "eval_sickr_spearman": 0.5854395559566877, "eval_stsb_spearman": 0.7464408613482243, "step": 10000 }, { "epoch": 0.06, "eval_avg_sts": 0.6665948829175545, "eval_sickr_spearman": 0.5861564745245929, "eval_stsb_spearman": 0.7470332913105162, "step": 10250 }, { "epoch": 0.06, "learning_rate": 9.683513780714234e-09, "loss": 0.2452, "step": 10500 }, { "epoch": 0.06, "eval_avg_sts": 0.6671936027854921, "eval_sickr_spearman": 0.5865762605278176, "eval_stsb_spearman": 0.7478109450431667, "step": 10500 }, { "epoch": 0.06, "eval_avg_sts": 0.6677066615222154, "eval_sickr_spearman": 0.5870991378854943, "eval_stsb_spearman": 0.7483141851589363, "step": 10750 }, { "epoch": 0.07, "learning_rate": 9.668443008367292e-09, "loss": 0.2158, "step": 11000 }, { "epoch": 0.07, "eval_avg_sts": 0.668111446446719, "eval_sickr_spearman": 0.5873266361015022, "eval_stsb_spearman": 0.748896256791936, "step": 11000 }, { "epoch": 0.07, "eval_avg_sts": 0.6684118774641665, "eval_sickr_spearman": 0.5873295246515303, "eval_stsb_spearman": 0.7494942302768027, "step": 11250 }, { "epoch": 0.07, "learning_rate": 9.65337223602035e-09, "loss": 0.191, "step": 11500 }, { "epoch": 0.07, "eval_avg_sts": 0.6685451430874475, "eval_sickr_spearman": 0.5870249320313236, "eval_stsb_spearman": 0.7500653541435714, "step": 11500 }, { "epoch": 0.07, "eval_avg_sts": 0.6688289293567933, "eval_sickr_spearman": 0.5870888785526357, "eval_stsb_spearman": 0.7505689801609509, "step": 11750 }, { "epoch": 0.07, "learning_rate": 9.63830146367341e-09, "loss": 0.1703, "step": 12000 }, { "epoch": 0.07, "eval_avg_sts": 0.6689011931208393, "eval_sickr_spearman": 0.5867304991388019, "eval_stsb_spearman": 0.7510718871028766, "step": 12000 }, { "epoch": 0.07, "eval_avg_sts": 0.6691967252591424, "eval_sickr_spearman": 0.5868564498805454, "eval_stsb_spearman": 0.7515370006377393, "step": 12250 }, { "epoch": 0.08, "learning_rate": 9.623230691326469e-09, "loss": 0.1528, "step": 12500 }, { "epoch": 0.08, "eval_avg_sts": 0.6694502557179467, "eval_sickr_spearman": 0.5868579937607329, "eval_stsb_spearman": 0.7520425176751604, "step": 12500 }, { "epoch": 0.08, "eval_avg_sts": 0.6699632588811146, "eval_sickr_spearman": 0.587389835584014, "eval_stsb_spearman": 0.7525366821782152, "step": 12750 }, { "epoch": 0.08, "learning_rate": 9.608159918979527e-09, "loss": 0.1371, "step": 13000 }, { "epoch": 0.08, "eval_avg_sts": 0.6703483303011183, "eval_sickr_spearman": 0.5876732123022901, "eval_stsb_spearman": 0.7530234482999465, "step": 13000 }, { "epoch": 0.08, "eval_avg_sts": 0.6707168479183498, "eval_sickr_spearman": 0.5879633123697694, "eval_stsb_spearman": 0.7534703834669302, "step": 13250 }, { "epoch": 0.08, "learning_rate": 9.593089146632587e-09, "loss": 0.1252, "step": 13500 }, { "epoch": 0.08, "eval_avg_sts": 0.6712428310314182, "eval_sickr_spearman": 0.5885554651255342, "eval_stsb_spearman": 0.7539301969373022, "step": 13500 }, { "epoch": 0.08, "eval_avg_sts": 0.6715811760730475, "eval_sickr_spearman": 0.5886914759897894, "eval_stsb_spearman": 0.7544708761563054, "step": 13750 }, { "epoch": 0.08, "learning_rate": 9.578018374285645e-09, "loss": 0.1124, "step": 14000 }, { "epoch": 0.08, "eval_avg_sts": 0.6720556074865778, "eval_sickr_spearman": 0.5892022513188999, "eval_stsb_spearman": 0.7549089636542557, "step": 14000 }, { "epoch": 0.09, "eval_avg_sts": 0.6723349901448149, "eval_sickr_spearman": 0.5893755145180006, "eval_stsb_spearman": 0.7552944657716292, "step": 14250 }, { "epoch": 0.09, "learning_rate": 9.562947601938704e-09, "loss": 0.1028, "step": 14500 }, { "epoch": 0.09, "eval_avg_sts": 0.6726527822394905, "eval_sickr_spearman": 0.589669150569135, "eval_stsb_spearman": 0.755636413909846, "step": 14500 }, { "epoch": 0.09, "eval_avg_sts": 0.6730171959193298, "eval_sickr_spearman": 0.5900167228216568, "eval_stsb_spearman": 0.756017669017003, "step": 14750 }, { "epoch": 0.09, "learning_rate": 9.547876829591762e-09, "loss": 0.094, "step": 15000 }, { "epoch": 0.09, "eval_avg_sts": 0.6735421112570777, "eval_sickr_spearman": 0.5906376116699428, "eval_stsb_spearman": 0.7564466108442127, "step": 15000 }, { "epoch": 0.09, "eval_avg_sts": 0.6735606956141224, "eval_sickr_spearman": 0.5903632990224448, "eval_stsb_spearman": 0.7567580922057999, "step": 15250 }, { "epoch": 0.09, "learning_rate": 9.532806057244822e-09, "loss": 0.0868, "step": 15500 }, { "epoch": 0.09, "eval_avg_sts": 0.6737977477369647, "eval_sickr_spearman": 0.5904703745838319, "eval_stsb_spearman": 0.7571251208900974, "step": 15500 }, { "epoch": 0.09, "eval_avg_sts": 0.6742547600358272, "eval_sickr_spearman": 0.591002216407113, "eval_stsb_spearman": 0.7575073036645413, "step": 15750 }, { "epoch": 0.1, "learning_rate": 9.51773528489788e-09, "loss": 0.08, "step": 16000 }, { "epoch": 0.1, "eval_avg_sts": 0.6744646822989027, "eval_sickr_spearman": 0.5910811535070195, "eval_stsb_spearman": 0.7578482110907857, "step": 16000 }, { "epoch": 0.1, "eval_avg_sts": 0.6748799475614093, "eval_sickr_spearman": 0.5915160794965986, "eval_stsb_spearman": 0.7582438156262201, "step": 16250 }, { "epoch": 0.1, "learning_rate": 9.502664512550939e-09, "loss": 0.0728, "step": 16500 }, { "epoch": 0.1, "eval_avg_sts": 0.6751155830172513, "eval_sickr_spearman": 0.5916266412390543, "eval_stsb_spearman": 0.7586045247954482, "step": 16500 }, { "epoch": 0.1, "eval_avg_sts": 0.6754758089125048, "eval_sickr_spearman": 0.5920287971265901, "eval_stsb_spearman": 0.7589228206984197, "step": 16750 }, { "epoch": 0.1, "learning_rate": 9.487593740203998e-09, "loss": 0.0679, "step": 17000 }, { "epoch": 0.1, "eval_avg_sts": 0.6760131658651831, "eval_sickr_spearman": 0.5928007372203121, "eval_stsb_spearman": 0.7592255945100542, "step": 17000 }, { "epoch": 0.1, "eval_avg_sts": 0.6762108045214816, "eval_sickr_spearman": 0.5929441686699845, "eval_stsb_spearman": 0.7594774403729787, "step": 17250 }, { "epoch": 0.11, "learning_rate": 9.472522967857057e-09, "loss": 0.0627, "step": 17500 }, { "epoch": 0.11, "eval_avg_sts": 0.6766050515480309, "eval_sickr_spearman": 0.5934244648160396, "eval_stsb_spearman": 0.7597856382800222, "step": 17500 }, { "epoch": 0.11, "eval_avg_sts": 0.6767814381948531, "eval_sickr_spearman": 0.5934582807724034, "eval_stsb_spearman": 0.7601045956173028, "step": 17750 }, { "epoch": 0.11, "learning_rate": 9.457452195510115e-09, "loss": 0.0584, "step": 18000 }, { "epoch": 0.11, "eval_avg_sts": 0.6769896075076911, "eval_sickr_spearman": 0.5936528096760212, "eval_stsb_spearman": 0.7603264053393611, "step": 18000 }, { "epoch": 0.11, "eval_avg_sts": 0.677209761262622, "eval_sickr_spearman": 0.593758341357221, "eval_stsb_spearman": 0.7606611811680228, "step": 18250 }, { "epoch": 0.11, "learning_rate": 9.442381423163175e-09, "loss": 0.0534, "step": 18500 }, { "epoch": 0.11, "eval_avg_sts": 0.6775168218365983, "eval_sickr_spearman": 0.5941294702332478, "eval_stsb_spearman": 0.7609041734399488, "step": 18500 }, { "epoch": 0.11, "eval_avg_sts": 0.6777132242888597, "eval_sickr_spearman": 0.5942964085038387, "eval_stsb_spearman": 0.7611300400738806, "step": 18750 }, { "epoch": 0.11, "learning_rate": 9.427310650816233e-09, "loss": 0.0507, "step": 19000 }, { "epoch": 0.11, "eval_avg_sts": 0.6778013660288258, "eval_sickr_spearman": 0.594225439817803, "eval_stsb_spearman": 0.7613772922398485, "step": 19000 }, { "epoch": 0.12, "eval_avg_sts": 0.6781823053409229, "eval_sickr_spearman": 0.594783129183577, "eval_stsb_spearman": 0.7615814814982689, "step": 19250 }, { "epoch": 0.12, "learning_rate": 9.412239878469292e-09, "loss": 0.0472, "step": 19500 }, { "epoch": 0.12, "eval_avg_sts": 0.6783265737471122, "eval_sickr_spearman": 0.5948400037375791, "eval_stsb_spearman": 0.7618131437566455, "step": 19500 }, { "epoch": 0.12, "eval_avg_sts": 0.6784263694342607, "eval_sickr_spearman": 0.5948344258478697, "eval_stsb_spearman": 0.7620183130206516, "step": 19750 }, { "epoch": 0.12, "learning_rate": 9.397169106122352e-09, "loss": 0.0433, "step": 20000 }, { "epoch": 0.12, "eval_avg_sts": 0.6786123514092111, "eval_sickr_spearman": 0.5949314412867451, "eval_stsb_spearman": 0.7622932615316771, "step": 20000 }, { "epoch": 0.12, "eval_avg_sts": 0.6787803024178876, "eval_sickr_spearman": 0.5951198444722, "eval_stsb_spearman": 0.762440760363575, "step": 20250 }, { "epoch": 0.12, "learning_rate": 9.38209833377541e-09, "loss": 0.0414, "step": 20500 }, { "epoch": 0.12, "eval_avg_sts": 0.679031627015686, "eval_sickr_spearman": 0.5953973942878331, "eval_stsb_spearman": 0.7626658597435387, "step": 20500 }, { "epoch": 0.13, "eval_avg_sts": 0.6793044898299303, "eval_sickr_spearman": 0.5957000944097455, "eval_stsb_spearman": 0.7629088852501152, "step": 20750 }, { "epoch": 0.13, "learning_rate": 9.367027561428468e-09, "loss": 0.0391, "step": 21000 }, { "epoch": 0.13, "eval_avg_sts": 0.6796013133058254, "eval_sickr_spearman": 0.5961489651235983, "eval_stsb_spearman": 0.7630536614880524, "step": 21000 }, { "epoch": 0.13, "eval_avg_sts": 0.6797676755235869, "eval_sickr_spearman": 0.5963239216106464, "eval_stsb_spearman": 0.7632114294365272, "step": 21250 }, { "epoch": 0.13, "learning_rate": 9.351956789081526e-09, "loss": 0.0368, "step": 21500 }, { "epoch": 0.13, "eval_avg_sts": 0.6799738806786508, "eval_sickr_spearman": 0.5966217410790631, "eval_stsb_spearman": 0.7633260202782385, "step": 21500 }, { "epoch": 0.13, "eval_avg_sts": 0.6802068401411112, "eval_sickr_spearman": 0.5968975976067485, "eval_stsb_spearman": 0.763516082675474, "step": 21750 }, { "epoch": 0.13, "learning_rate": 9.336886016734586e-09, "loss": 0.0347, "step": 22000 }, { "epoch": 0.13, "eval_avg_sts": 0.6802982007395906, "eval_sickr_spearman": 0.5969305669191386, "eval_stsb_spearman": 0.7636658345600424, "step": 22000 }, { "epoch": 0.13, "eval_avg_sts": 0.680360206773362, "eval_sickr_spearman": 0.5969050679947523, "eval_stsb_spearman": 0.7638153455519716, "step": 22250 }, { "epoch": 0.14, "learning_rate": 9.321815244387645e-09, "loss": 0.0329, "step": 22500 }, { "epoch": 0.14, "eval_avg_sts": 0.6804424688694628, "eval_sickr_spearman": 0.596870206184068, "eval_stsb_spearman": 0.7640147315548577, "step": 22500 }, { "epoch": 0.14, "eval_avg_sts": 0.6804860386341411, "eval_sickr_spearman": 0.5967837488935713, "eval_stsb_spearman": 0.764188328374711, "step": 22750 }, { "epoch": 0.14, "learning_rate": 9.306744472040703e-09, "loss": 0.0311, "step": 23000 }, { "epoch": 0.14, "eval_avg_sts": 0.6805827014319539, "eval_sickr_spearman": 0.5968390795673858, "eval_stsb_spearman": 0.7643263232965221, "step": 23000 }, { "epoch": 0.14, "eval_avg_sts": 0.6806972200568748, "eval_sickr_spearman": 0.5969168712077983, "eval_stsb_spearman": 0.7644775689059513, "step": 23250 }, { "epoch": 0.14, "learning_rate": 9.291673699693763e-09, "loss": 0.0297, "step": 23500 }, { "epoch": 0.14, "eval_avg_sts": 0.6808736496467026, "eval_sickr_spearman": 0.5970506409556523, "eval_stsb_spearman": 0.7646966583377529, "step": 23500 }, { "epoch": 0.14, "eval_avg_sts": 0.6810096396199108, "eval_sickr_spearman": 0.5971779363672365, "eval_stsb_spearman": 0.7648413428725852, "step": 23750 }, { "epoch": 0.14, "learning_rate": 9.276602927346821e-09, "loss": 0.0284, "step": 24000 }, { "epoch": 0.14, "eval_avg_sts": 0.6811351307194636, "eval_sickr_spearman": 0.5972490693237311, "eval_stsb_spearman": 0.7650211921151961, "step": 24000 }, { "epoch": 0.15, "eval_avg_sts": 0.6813719782830303, "eval_sickr_spearman": 0.5975800424521857, "eval_stsb_spearman": 0.7651639141138749, "step": 24250 }, { "epoch": 0.15, "learning_rate": 9.26153215499988e-09, "loss": 0.027, "step": 24500 }, { "epoch": 0.15, "eval_avg_sts": 0.6815488144179666, "eval_sickr_spearman": 0.5977527578228325, "eval_stsb_spearman": 0.7653448710131006, "step": 24500 }, { "epoch": 0.15, "eval_avg_sts": 0.6816865586733576, "eval_sickr_spearman": 0.5979984339829829, "eval_stsb_spearman": 0.7653746833637323, "step": 24750 }, { "epoch": 0.15, "learning_rate": 9.24646138265294e-09, "loss": 0.026, "step": 25000 }, { "epoch": 0.15, "eval_avg_sts": 0.681900742705164, "eval_sickr_spearman": 0.5982624872976225, "eval_stsb_spearman": 0.7655389981127053, "step": 25000 }, { "epoch": 0.15, "eval_avg_sts": 0.6819519221717609, "eval_sickr_spearman": 0.5982854462900874, "eval_stsb_spearman": 0.7656183980534346, "step": 25250 }, { "epoch": 0.15, "learning_rate": 9.231390610305998e-09, "loss": 0.0247, "step": 25500 }, { "epoch": 0.15, "eval_avg_sts": 0.6819903776778287, "eval_sickr_spearman": 0.5982351456775288, "eval_stsb_spearman": 0.7657456096781285, "step": 25500 }, { "epoch": 0.16, "eval_avg_sts": 0.6821595590267062, "eval_sickr_spearman": 0.5985327659355987, "eval_stsb_spearman": 0.7657863521178138, "step": 25750 }, { "epoch": 0.16, "learning_rate": 9.216319837959056e-09, "loss": 0.0238, "step": 26000 }, { "epoch": 0.16, "eval_avg_sts": 0.6822216826747098, "eval_sickr_spearman": 0.5984852542678948, "eval_stsb_spearman": 0.7659581110815249, "step": 26000 }, { "epoch": 0.16, "eval_avg_sts": 0.6822324678023577, "eval_sickr_spearman": 0.5984695664530869, "eval_stsb_spearman": 0.7659953691516284, "step": 26250 }, { "epoch": 0.16, "learning_rate": 9.201249065612116e-09, "loss": 0.0226, "step": 26500 }, { "epoch": 0.16, "eval_avg_sts": 0.6824471355452602, "eval_sickr_spearman": 0.598771668943959, "eval_stsb_spearman": 0.7661226021465615, "step": 26500 }, { "epoch": 0.16, "eval_avg_sts": 0.6825452027323136, "eval_sickr_spearman": 0.5989038948116256, "eval_stsb_spearman": 0.7661865106530016, "step": 26750 }, { "epoch": 0.16, "learning_rate": 9.186178293265174e-09, "loss": 0.0219, "step": 27000 }, { "epoch": 0.16, "eval_avg_sts": 0.682652241260723, "eval_sickr_spearman": 0.5990028027487954, "eval_stsb_spearman": 0.7663016797726506, "step": 27000 }, { "epoch": 0.16, "eval_avg_sts": 0.6827966424642413, "eval_sickr_spearman": 0.5991921521833974, "eval_stsb_spearman": 0.7664011327450853, "step": 27250 }, { "epoch": 0.17, "learning_rate": 9.171107520918232e-09, "loss": 0.0208, "step": 27500 }, { "epoch": 0.17, "eval_avg_sts": 0.6828478001704832, "eval_sickr_spearman": 0.5991956881670526, "eval_stsb_spearman": 0.7664999121739138, "step": 27500 }, { "epoch": 0.17, "eval_avg_sts": 0.6830724849298578, "eval_sickr_spearman": 0.5994928103992555, "eval_stsb_spearman": 0.7666521594604602, "step": 27750 }, { "epoch": 0.17, "learning_rate": 9.15603674857129e-09, "loss": 0.02, "step": 28000 }, { "epoch": 0.17, "eval_avg_sts": 0.6830958603489119, "eval_sickr_spearman": 0.5994919637552817, "eval_stsb_spearman": 0.7666997569425422, "step": 28000 }, { "epoch": 0.17, "eval_avg_sts": 0.6830795409925201, "eval_sickr_spearman": 0.5993960439733135, "eval_stsb_spearman": 0.7667630380117267, "step": 28250 }, { "epoch": 0.17, "learning_rate": 9.14096597622435e-09, "loss": 0.0195, "step": 28500 }, { "epoch": 0.17, "eval_avg_sts": 0.6831755132481567, "eval_sickr_spearman": 0.5994744332447662, "eval_stsb_spearman": 0.7668765932515471, "step": 28500 }, { "epoch": 0.17, "eval_avg_sts": 0.6833569004212535, "eval_sickr_spearman": 0.5997113441496589, "eval_stsb_spearman": 0.7670024566928482, "step": 28750 }, { "epoch": 0.17, "learning_rate": 9.125895203877409e-09, "loss": 0.0187, "step": 29000 }, { "epoch": 0.17, "eval_avg_sts": 0.6834568554940637, "eval_sickr_spearman": 0.5998296750956386, "eval_stsb_spearman": 0.7670840358924887, "step": 29000 }, { "epoch": 0.18, "eval_avg_sts": 0.6834650699517616, "eval_sickr_spearman": 0.5997757388942514, "eval_stsb_spearman": 0.7671544010092718, "step": 29250 }, { "epoch": 0.18, "learning_rate": 9.110824431530467e-09, "loss": 0.0179, "step": 29500 }, { "epoch": 0.18, "eval_avg_sts": 0.6835179845459645, "eval_sickr_spearman": 0.5998240474033423, "eval_stsb_spearman": 0.7672119216885865, "step": 29500 }, { "epoch": 0.18, "eval_avg_sts": 0.6835631347255009, "eval_sickr_spearman": 0.5997912773012991, "eval_stsb_spearman": 0.7673349921497027, "step": 29750 }, { "epoch": 0.18, "learning_rate": 9.095753659183525e-09, "loss": 0.0176, "step": 30000 }, { "epoch": 0.18, "eval_avg_sts": 0.6836041595446583, "eval_sickr_spearman": 0.5998075627471473, "eval_stsb_spearman": 0.7674007563421691, "step": 30000 }, { "epoch": 0.18, "eval_avg_sts": 0.6837299737321467, "eval_sickr_spearman": 0.6000278395880849, "eval_stsb_spearman": 0.7674321078762084, "step": 30250 }, { "epoch": 0.18, "learning_rate": 9.080682886836584e-09, "loss": 0.017, "step": 30500 }, { "epoch": 0.18, "eval_avg_sts": 0.6837571024354734, "eval_sickr_spearman": 0.5999885951497719, "eval_stsb_spearman": 0.7675256097211749, "step": 30500 }, { "epoch": 0.19, "eval_avg_sts": 0.6838406104965742, "eval_sickr_spearman": 0.6001280423925087, "eval_stsb_spearman": 0.7675531786006398, "step": 30750 }, { "epoch": 0.19, "learning_rate": 9.065612114489644e-09, "loss": 0.0164, "step": 31000 }, { "epoch": 0.19, "eval_avg_sts": 0.6839781957977387, "eval_sickr_spearman": 0.6003198321538586, "eval_stsb_spearman": 0.7676365594416187, "step": 31000 }, { "epoch": 0.19, "eval_avg_sts": 0.6841738441816796, "eval_sickr_spearman": 0.6006274129292668, "eval_stsb_spearman": 0.7677202754340923, "step": 31250 }, { "epoch": 0.19, "learning_rate": 9.050541342142702e-09, "loss": 0.0159, "step": 31500 }, { "epoch": 0.19, "eval_avg_sts": 0.6842156611206922, "eval_sickr_spearman": 0.600646138701863, "eval_stsb_spearman": 0.7677851835395214, "step": 31500 }, { "epoch": 0.19, "eval_avg_sts": 0.6841708520756637, "eval_sickr_spearman": 0.600490057395171, "eval_stsb_spearman": 0.7678516467561564, "step": 31750 }, { "epoch": 0.19, "learning_rate": 9.03547056979576e-09, "loss": 0.0153, "step": 32000 }, { "epoch": 0.19, "eval_avg_sts": 0.6840981881424868, "eval_sickr_spearman": 0.6003198321538586, "eval_stsb_spearman": 0.7678765441311148, "step": 32000 }, { "epoch": 0.19, "eval_avg_sts": 0.6842101456597175, "eval_sickr_spearman": 0.6004448366464549, "eval_stsb_spearman": 0.7679754546729801, "step": 32250 }, { "epoch": 0.2, "learning_rate": 9.02039979744882e-09, "loss": 0.015, "step": 32500 }, { "epoch": 0.2, "eval_avg_sts": 0.684201669305099, "eval_sickr_spearman": 0.6004205329841494, "eval_stsb_spearman": 0.7679828056260486, "step": 32500 }, { "epoch": 0.2, "eval_avg_sts": 0.6843570539386391, "eval_sickr_spearman": 0.600673928545237, "eval_stsb_spearman": 0.7680401793320412, "step": 32750 }, { "epoch": 0.2, "learning_rate": 9.005329025101879e-09, "loss": 0.0143, "step": 33000 }, { "epoch": 0.2, "eval_avg_sts": 0.6844358440678995, "eval_sickr_spearman": 0.6008083955293048, "eval_stsb_spearman": 0.7680632926064943, "step": 33000 }, { "epoch": 0.2, "eval_avg_sts": 0.6845585283651245, "eval_sickr_spearman": 0.6010039702872432, "eval_stsb_spearman": 0.7681130864430059, "step": 33250 }, { "epoch": 0.2, "learning_rate": 8.990258252754937e-09, "loss": 0.014, "step": 33500 }, { "epoch": 0.2, "eval_avg_sts": 0.6845940476732861, "eval_sickr_spearman": 0.6010599483946848, "eval_stsb_spearman": 0.7681281469518876, "step": 33500 }, { "epoch": 0.2, "eval_avg_sts": 0.6846217305035208, "eval_sickr_spearman": 0.6010828575845629, "eval_stsb_spearman": 0.7681606034224787, "step": 33750 }, { "epoch": 0.2, "learning_rate": 8.975187480407995e-09, "loss": 0.0135, "step": 34000 }, { "epoch": 0.2, "eval_avg_sts": 0.6847057396162426, "eval_sickr_spearman": 0.6011810184829323, "eval_stsb_spearman": 0.7682304607495529, "step": 34000 }, { "epoch": 0.21, "eval_avg_sts": 0.684865672317778, "eval_sickr_spearman": 0.6015010001024268, "eval_stsb_spearman": 0.768230344533129, "step": 34250 }, { "epoch": 0.21, "learning_rate": 8.960116708061055e-09, "loss": 0.013, "step": 34500 }, { "epoch": 0.21, "eval_avg_sts": 0.6849459219282323, "eval_sickr_spearman": 0.6016653984410962, "eval_stsb_spearman": 0.7682264454153683, "step": 34500 }, { "epoch": 0.21, "eval_avg_sts": 0.684890754096638, "eval_sickr_spearman": 0.6015818297006276, "eval_stsb_spearman": 0.7681996784926486, "step": 34750 }, { "epoch": 0.21, "learning_rate": 8.945045935714113e-09, "loss": 0.0127, "step": 35000 }, { "epoch": 0.21, "eval_avg_sts": 0.6848642701791275, "eval_sickr_spearman": 0.6015226642276378, "eval_stsb_spearman": 0.7682058761306173, "step": 35000 }, { "epoch": 0.21, "eval_avg_sts": 0.6848120976692862, "eval_sickr_spearman": 0.6013494010285371, "eval_stsb_spearman": 0.7682747943100354, "step": 35250 }, { "epoch": 0.21, "learning_rate": 8.929975163367172e-09, "loss": 0.0126, "step": 35500 }, { "epoch": 0.21, "eval_avg_sts": 0.6848348708799592, "eval_sickr_spearman": 0.6014064249902991, "eval_stsb_spearman": 0.7682633167696193, "step": 35500 }, { "epoch": 0.22, "eval_avg_sts": 0.6849041036803378, "eval_sickr_spearman": 0.6014944261609835, "eval_stsb_spearman": 0.7683137811996923, "step": 35750 }, { "epoch": 0.22, "learning_rate": 8.914904391020232e-09, "loss": 0.0123, "step": 36000 }, { "epoch": 0.22, "eval_avg_sts": 0.6850199516781598, "eval_sickr_spearman": 0.6016458260245264, "eval_stsb_spearman": 0.7683940773317931, "step": 36000 }, { "epoch": 0.22, "eval_avg_sts": 0.6850955218108025, "eval_sickr_spearman": 0.6017597245402905, "eval_stsb_spearman": 0.7684313190813145, "step": 36250 }, { "epoch": 0.22, "learning_rate": 8.89983361867329e-09, "loss": 0.0117, "step": 36500 }, { "epoch": 0.22, "eval_avg_sts": 0.6850673538769948, "eval_sickr_spearman": 0.6016572308168788, "eval_stsb_spearman": 0.7684774769371108, "step": 36500 }, { "epoch": 0.22, "eval_avg_sts": 0.6851541104973866, "eval_sickr_spearman": 0.6018085310752482, "eval_stsb_spearman": 0.768499689919525, "step": 36750 }, { "epoch": 0.22, "learning_rate": 8.884762846326348e-09, "loss": 0.0115, "step": 37000 }, { "epoch": 0.22, "eval_avg_sts": 0.685221784647517, "eval_sickr_spearman": 0.6018979267183601, "eval_stsb_spearman": 0.7685456425766739, "step": 37000 }, { "epoch": 0.22, "eval_avg_sts": 0.6852791986160438, "eval_sickr_spearman": 0.6019803001967482, "eval_stsb_spearman": 0.7685780970353394, "step": 37250 }, { "epoch": 0.23, "learning_rate": 8.869692073979408e-09, "loss": 0.0111, "step": 37500 }, { "epoch": 0.23, "eval_avg_sts": 0.6853061325467835, "eval_sickr_spearman": 0.602020391279035, "eval_stsb_spearman": 0.7685918738145322, "step": 37500 }, { "epoch": 0.23, "eval_avg_sts": 0.6853268783841169, "eval_sickr_spearman": 0.6020406609318185, "eval_stsb_spearman": 0.7686130958364152, "step": 37750 }, { "epoch": 0.23, "learning_rate": 8.854621301632466e-09, "loss": 0.011, "step": 38000 }, { "epoch": 0.23, "eval_avg_sts": 0.6853295441786447, "eval_sickr_spearman": 0.6019849318373105, "eval_stsb_spearman": 0.768674156519979, "step": 38000 }, { "epoch": 0.23, "eval_avg_sts": 0.6853341365400603, "eval_sickr_spearman": 0.6020266664049582, "eval_stsb_spearman": 0.7686416066751625, "step": 38250 }, { "epoch": 0.23, "learning_rate": 8.839550529285525e-09, "loss": 0.0109, "step": 38500 }, { "epoch": 0.23, "eval_avg_sts": 0.6853246034683043, "eval_sickr_spearman": 0.6020385692231776, "eval_stsb_spearman": 0.7686106377134309, "step": 38500 }, { "epoch": 0.23, "eval_avg_sts": 0.6853568834070937, "eval_sickr_spearman": 0.602098581340141, "eval_stsb_spearman": 0.7686151854740463, "step": 38750 }, { "epoch": 0.24, "learning_rate": 8.824479756938583e-09, "loss": 0.0105, "step": 39000 }, { "epoch": 0.24, "eval_avg_sts": 0.6854495878694604, "eval_sickr_spearman": 0.6022639757305444, "eval_stsb_spearman": 0.7686352000083764, "step": 39000 }, { "epoch": 0.24, "eval_avg_sts": 0.685494545030269, "eval_sickr_spearman": 0.6023270258052962, "eval_stsb_spearman": 0.7686620642552416, "step": 39250 }, { "epoch": 0.24, "learning_rate": 8.809408984591643e-09, "loss": 0.0104, "step": 39500 }, { "epoch": 0.24, "eval_avg_sts": 0.6855596476767631, "eval_sickr_spearman": 0.6024400776770863, "eval_stsb_spearman": 0.76867921767644, "step": 39500 }, { "epoch": 0.24, "eval_avg_sts": 0.6855377092658718, "eval_sickr_spearman": 0.6024064012182297, "eval_stsb_spearman": 0.7686690173135138, "step": 39750 }, { "epoch": 0.24, "learning_rate": 8.794338212244701e-09, "loss": 0.0101, "step": 40000 }, { "epoch": 0.24, "eval_avg_sts": 0.6855094541954295, "eval_sickr_spearman": 0.6023502338106945, "eval_stsb_spearman": 0.7686686745801645, "step": 40000 }, { "epoch": 0.24, "eval_avg_sts": 0.6855909496558752, "eval_sickr_spearman": 0.6024469504340499, "eval_stsb_spearman": 0.7687349488777003, "step": 40250 }, { "epoch": 0.24, "learning_rate": 8.77926743989776e-09, "loss": 0.0098, "step": 40500 }, { "epoch": 0.24, "eval_avg_sts": 0.6855538581482595, "eval_sickr_spearman": 0.6023648757711819, "eval_stsb_spearman": 0.7687428405253371, "step": 40500 }, { "epoch": 0.25, "eval_avg_sts": 0.6855329227615412, "eval_sickr_spearman": 0.6022896240626906, "eval_stsb_spearman": 0.7687762214603917, "step": 40750 }, { "epoch": 0.25, "learning_rate": 8.76419666755082e-09, "loss": 0.0097, "step": 41000 }, { "epoch": 0.25, "eval_avg_sts": 0.6856315964474966, "eval_sickr_spearman": 0.6024412729391669, "eval_stsb_spearman": 0.7688219199558264, "step": 41000 }, { "epoch": 0.25, "eval_avg_sts": 0.6856635284536272, "eval_sickr_spearman": 0.6024548690453339, "eval_stsb_spearman": 0.7688721878619207, "step": 41250 }, { "epoch": 0.25, "learning_rate": 8.749125895203878e-09, "loss": 0.0094, "step": 41500 }, { "epoch": 0.25, "eval_avg_sts": 0.6857199355051676, "eval_sickr_spearman": 0.6025655303929629, "eval_stsb_spearman": 0.7688743406173724, "step": 41500 }, { "epoch": 0.25, "eval_avg_sts": 0.68571207819909, "eval_sickr_spearman": 0.6025172716864586, "eval_stsb_spearman": 0.7689068847117213, "step": 41750 }, { "epoch": 0.25, "learning_rate": 8.734055122856936e-09, "loss": 0.0092, "step": 42000 }, { "epoch": 0.25, "eval_avg_sts": 0.685660615312776, "eval_sickr_spearman": 0.6024261329528127, "eval_stsb_spearman": 0.7688950976727392, "step": 42000 }, { "epoch": 0.25, "eval_avg_sts": 0.6857670926985571, "eval_sickr_spearman": 0.602601687070901, "eval_stsb_spearman": 0.7689324983262132, "step": 42250 }, { "epoch": 0.26, "learning_rate": 8.718984350509996e-09, "loss": 0.009, "step": 42500 }, { "epoch": 0.26, "eval_avg_sts": 0.6857819219395811, "eval_sickr_spearman": 0.602576984987902, "eval_stsb_spearman": 0.7689868588912602, "step": 42500 }, { "epoch": 0.26, "eval_avg_sts": 0.6858294973016628, "eval_sickr_spearman": 0.602676988581979, "eval_stsb_spearman": 0.7689820060213465, "step": 42750 }, { "epoch": 0.26, "learning_rate": 8.703913578163054e-09, "loss": 0.0089, "step": 43000 }, { "epoch": 0.26, "eval_avg_sts": 0.6858174137793278, "eval_sickr_spearman": 0.6026706636534691, "eval_stsb_spearman": 0.7689641639051866, "step": 43000 }, { "epoch": 0.26, "eval_avg_sts": 0.6858729153589367, "eval_sickr_spearman": 0.6027516924620167, "eval_stsb_spearman": 0.7689941382558566, "step": 43250 }, { "epoch": 0.26, "learning_rate": 8.688842805816112e-09, "loss": 0.0087, "step": 43500 }, { "epoch": 0.26, "eval_avg_sts": 0.6858947608283459, "eval_sickr_spearman": 0.6027528379215106, "eval_stsb_spearman": 0.7690366837351812, "step": 43500 }, { "epoch": 0.26, "eval_avg_sts": 0.6858911001750951, "eval_sickr_spearman": 0.6027214622918947, "eval_stsb_spearman": 0.7690607380582957, "step": 43750 }, { "epoch": 0.27, "learning_rate": 8.673772033469172e-09, "loss": 0.0085, "step": 44000 }, { "epoch": 0.27, "eval_avg_sts": 0.6859079265589377, "eval_sickr_spearman": 0.602721561897068, "eval_stsb_spearman": 0.7690942912208074, "step": 44000 }, { "epoch": 0.27, "eval_avg_sts": 0.6860282186997773, "eval_sickr_spearman": 0.6029633036528699, "eval_stsb_spearman": 0.7690931337466848, "step": 44250 }, { "epoch": 0.27, "learning_rate": 8.65870126112223e-09, "loss": 0.0082, "step": 44500 }, { "epoch": 0.27, "eval_avg_sts": 0.6861163399063477, "eval_sickr_spearman": 0.6030905492618673, "eval_stsb_spearman": 0.769142130550828, "step": 44500 }, { "epoch": 0.27, "eval_avg_sts": 0.6861892434287493, "eval_sickr_spearman": 0.603183331480874, "eval_stsb_spearman": 0.7691951553766246, "step": 44750 }, { "epoch": 0.27, "learning_rate": 8.643630488775289e-09, "loss": 0.0082, "step": 45000 }, { "epoch": 0.27, "eval_avg_sts": 0.6861477662185661, "eval_sickr_spearman": 0.603116446606947, "eval_stsb_spearman": 0.7691790858301852, "step": 45000 }, { "epoch": 0.27, "eval_avg_sts": 0.6862196383572703, "eval_sickr_spearman": 0.603254100956563, "eval_stsb_spearman": 0.7691851757579776, "step": 45250 }, { "epoch": 0.27, "learning_rate": 8.628559716428347e-09, "loss": 0.0081, "step": 45500 }, { "epoch": 0.27, "eval_avg_sts": 0.686198388321166, "eval_sickr_spearman": 0.6032430447823174, "eval_stsb_spearman": 0.7691537318600145, "step": 45500 }, { "epoch": 0.28, "eval_avg_sts": 0.6861370624006556, "eval_sickr_spearman": 0.6031246640337512, "eval_stsb_spearman": 0.7691494607675601, "step": 45750 }, { "epoch": 0.28, "learning_rate": 8.613488944081407e-09, "loss": 0.0079, "step": 46000 }, { "epoch": 0.28, "eval_avg_sts": 0.6861745072632264, "eval_sickr_spearman": 0.6032205340131327, "eval_stsb_spearman": 0.7691284805133201, "step": 46000 }, { "epoch": 0.28, "eval_avg_sts": 0.6862384540266784, "eval_sickr_spearman": 0.6033426499557009, "eval_stsb_spearman": 0.7691342580976559, "step": 46250 }, { "epoch": 0.28, "learning_rate": 8.598418171734465e-09, "loss": 0.0078, "step": 46500 }, { "epoch": 0.28, "eval_avg_sts": 0.6862749736889266, "eval_sickr_spearman": 0.6034227325151013, "eval_stsb_spearman": 0.769127214862752, "step": 46500 }, { "epoch": 0.28, "eval_avg_sts": 0.6863208881531138, "eval_sickr_spearman": 0.6035210428212308, "eval_stsb_spearman": 0.7691207334849967, "step": 46750 }, { "epoch": 0.28, "learning_rate": 8.583347399387524e-09, "loss": 0.0077, "step": 47000 }, { "epoch": 0.28, "eval_avg_sts": 0.6863572468626877, "eval_sickr_spearman": 0.6035845411192627, "eval_stsb_spearman": 0.7691299526061127, "step": 47000 }, { "epoch": 0.28, "eval_avg_sts": 0.6863364230688664, "eval_sickr_spearman": 0.603581104740781, "eval_stsb_spearman": 0.7690917413969519, "step": 47250 }, { "epoch": 0.29, "learning_rate": 8.568276627040584e-09, "loss": 0.0075, "step": 47500 }, { "epoch": 0.29, "eval_avg_sts": 0.6863403563629373, "eval_sickr_spearman": 0.6035747300096844, "eval_stsb_spearman": 0.7691059827161902, "step": 47500 }, { "epoch": 0.29, "eval_avg_sts": 0.6863889077611471, "eval_sickr_spearman": 0.6036478900095347, "eval_stsb_spearman": 0.7691299255127595, "step": 47750 }, { "epoch": 0.29, "learning_rate": 8.553205854693642e-09, "loss": 0.0075, "step": 48000 }, { "epoch": 0.29, "eval_avg_sts": 0.6864311113667037, "eval_sickr_spearman": 0.6037579936520918, "eval_stsb_spearman": 0.7691042290813154, "step": 48000 }, { "epoch": 0.29, "eval_avg_sts": 0.686500195230545, "eval_sickr_spearman": 0.6039076603017187, "eval_stsb_spearman": 0.7690927301593713, "step": 48250 }, { "epoch": 0.29, "learning_rate": 8.5381350823467e-09, "loss": 0.0072, "step": 48500 }, { "epoch": 0.29, "eval_avg_sts": 0.6864608843930222, "eval_sickr_spearman": 0.6038096488111093, "eval_stsb_spearman": 0.769112119974935, "step": 48500 }, { "epoch": 0.29, "eval_avg_sts": 0.6865355929074337, "eval_sickr_spearman": 0.603904373330997, "eval_stsb_spearman": 0.7691668124838703, "step": 48750 }, { "epoch": 0.3, "learning_rate": 8.52306430999976e-09, "loss": 0.0071, "step": 49000 }, { "epoch": 0.3, "eval_avg_sts": 0.6866365816613527, "eval_sickr_spearman": 0.6040533328677922, "eval_stsb_spearman": 0.7692198304549132, "step": 49000 }, { "epoch": 0.3, "eval_avg_sts": 0.6866187505580584, "eval_sickr_spearman": 0.6040197659243618, "eval_stsb_spearman": 0.7692177351917551, "step": 49250 }, { "epoch": 0.3, "learning_rate": 8.507993537652819e-09, "loss": 0.0071, "step": 49500 }, { "epoch": 0.3, "eval_avg_sts": 0.6866502129431028, "eval_sickr_spearman": 0.6040536316833123, "eval_stsb_spearman": 0.7692467942028933, "step": 49500 }, { "epoch": 0.3, "eval_avg_sts": 0.686725828130758, "eval_sickr_spearman": 0.6041687752637436, "eval_stsb_spearman": 0.7692828809977725, "step": 49750 }, { "epoch": 0.3, "learning_rate": 8.492922765305877e-09, "loss": 0.0069, "step": 50000 }, { "epoch": 0.3, "eval_avg_sts": 0.6867152080403021, "eval_sickr_spearman": 0.6041749009819067, "eval_stsb_spearman": 0.7692555150986975, "step": 50000 }, { "epoch": 0.3, "eval_avg_sts": 0.6867052144023214, "eval_sickr_spearman": 0.6041375490418878, "eval_stsb_spearman": 0.7692728797627548, "step": 50250 }, { "epoch": 0.3, "learning_rate": 8.477851992958935e-09, "loss": 0.0067, "step": 50500 }, { "epoch": 0.3, "eval_avg_sts": 0.6867264422199005, "eval_sickr_spearman": 0.6041753990077735, "eval_stsb_spearman": 0.7692774854320276, "step": 50500 }, { "epoch": 0.31, "eval_avg_sts": 0.6867642770376152, "eval_sickr_spearman": 0.6042159383133408, "eval_stsb_spearman": 0.7693126157618897, "step": 50750 }, { "epoch": 0.31, "learning_rate": 8.462781220611993e-09, "loss": 0.0067, "step": 51000 }, { "epoch": 0.31, "eval_avg_sts": 0.6867016135707726, "eval_sickr_spearman": 0.6041016911794699, "eval_stsb_spearman": 0.7693015359620754, "step": 51000 }, { "epoch": 0.31, "eval_avg_sts": 0.6867455519331296, "eval_sickr_spearman": 0.6041852101173518, "eval_stsb_spearman": 0.7693058937489072, "step": 51250 }, { "epoch": 0.31, "learning_rate": 8.447710448265052e-09, "loss": 0.0066, "step": 51500 }, { "epoch": 0.31, "eval_avg_sts": 0.6866994936909452, "eval_sickr_spearman": 0.604103284862244, "eval_stsb_spearman": 0.7692957025196464, "step": 51500 }, { "epoch": 0.31, "eval_avg_sts": 0.6866555978626698, "eval_sickr_spearman": 0.6040234513157771, "eval_stsb_spearman": 0.7692877444095627, "step": 51750 }, { "epoch": 0.31, "learning_rate": 8.432639675918112e-09, "loss": 0.0064, "step": 52000 }, { "epoch": 0.31, "eval_avg_sts": 0.6866565885255093, "eval_sickr_spearman": 0.6039624431470797, "eval_stsb_spearman": 0.7693507339039388, "step": 52000 }, { "epoch": 0.31, "eval_avg_sts": 0.6866521898462536, "eval_sickr_spearman": 0.6039828622076233, "eval_stsb_spearman": 0.7693215174848839, "step": 52250 }, { "epoch": 0.32, "learning_rate": 8.41756890357117e-09, "loss": 0.0064, "step": 52500 }, { "epoch": 0.32, "eval_avg_sts": 0.6866535588513245, "eval_sickr_spearman": 0.6039745451756456, "eval_stsb_spearman": 0.7693325725270033, "step": 52500 }, { "epoch": 0.32, "eval_avg_sts": 0.6865157750150845, "eval_sickr_spearman": 0.6037358413776323, "eval_stsb_spearman": 0.7692957086525368, "step": 52750 }, { "epoch": 0.32, "learning_rate": 8.402498131224228e-09, "loss": 0.0063, "step": 53000 }, { "epoch": 0.32, "eval_avg_sts": 0.6865347142126632, "eval_sickr_spearman": 0.6037773767349331, "eval_stsb_spearman": 0.7692920516903932, "step": 53000 }, { "epoch": 0.32, "eval_avg_sts": 0.6865489888709764, "eval_sickr_spearman": 0.6038000369118779, "eval_stsb_spearman": 0.7692979408300747, "step": 53250 }, { "epoch": 0.32, "learning_rate": 8.387427358877288e-09, "loss": 0.0063, "step": 53500 }, { "epoch": 0.32, "eval_avg_sts": 0.686531075517989, "eval_sickr_spearman": 0.6037771775245864, "eval_stsb_spearman": 0.7692849735113917, "step": 53500 }, { "epoch": 0.32, "eval_avg_sts": 0.6865282711445495, "eval_sickr_spearman": 0.6037651751011937, "eval_stsb_spearman": 0.7692913671879054, "step": 53750 }, { "epoch": 0.33, "learning_rate": 8.372356586530346e-09, "loss": 0.0061, "step": 54000 }, { "epoch": 0.33, "eval_avg_sts": 0.6865362702221531, "eval_sickr_spearman": 0.6037866898186445, "eval_stsb_spearman": 0.7692858506256617, "step": 54000 }, { "epoch": 0.33, "eval_avg_sts": 0.6865545509742186, "eval_sickr_spearman": 0.6038044195395067, "eval_stsb_spearman": 0.7693046824089305, "step": 54250 }, { "epoch": 0.33, "learning_rate": 8.357285814183405e-09, "loss": 0.0061, "step": 54500 }, { "epoch": 0.33, "eval_avg_sts": 0.6865219027035194, "eval_sickr_spearman": 0.6037482422217184, "eval_stsb_spearman": 0.7692955631853203, "step": 54500 }, { "epoch": 0.33, "eval_avg_sts": 0.686487675439571, "eval_sickr_spearman": 0.6036708490019995, "eval_stsb_spearman": 0.7693045018771424, "step": 54750 }, { "epoch": 0.33, "learning_rate": 8.342215041836465e-09, "loss": 0.0059, "step": 55000 }, { "epoch": 0.33, "eval_avg_sts": 0.6864760791160216, "eval_sickr_spearman": 0.6036339452852609, "eval_stsb_spearman": 0.7693182129467823, "step": 55000 }, { "epoch": 0.33, "eval_avg_sts": 0.6864571435422872, "eval_sickr_spearman": 0.6036292140395252, "eval_stsb_spearman": 0.7692850730450491, "step": 55250 }, { "epoch": 0.33, "learning_rate": 8.327144269489523e-09, "loss": 0.0059, "step": 55500 }, { "epoch": 0.33, "eval_avg_sts": 0.6863885216776884, "eval_sickr_spearman": 0.6034630726103215, "eval_stsb_spearman": 0.7693139707450554, "step": 55500 }, { "epoch": 0.34, "eval_avg_sts": 0.6862557558724565, "eval_sickr_spearman": 0.6032416005073034, "eval_stsb_spearman": 0.7692699112376095, "step": 55750 }, { "epoch": 0.34, "learning_rate": 8.312073497142581e-09, "loss": 0.0058, "step": 56000 }, { "epoch": 0.34, "eval_avg_sts": 0.6862266731200726, "eval_sickr_spearman": 0.603156338478887, "eval_stsb_spearman": 0.7692970077612581, "step": 56000 }, { "epoch": 0.34, "eval_avg_sts": 0.6861787937303312, "eval_sickr_spearman": 0.6030421411476029, "eval_stsb_spearman": 0.7693154463130595, "step": 56250 }, { "epoch": 0.34, "learning_rate": 8.297002724795641e-09, "loss": 0.0057, "step": 56500 }, { "epoch": 0.34, "eval_avg_sts": 0.6862001677803937, "eval_sickr_spearman": 0.6030790448643415, "eval_stsb_spearman": 0.7693212906964461, "step": 56500 }, { "epoch": 0.34, "eval_avg_sts": 0.6861681437241836, "eval_sickr_spearman": 0.6030312343811175, "eval_stsb_spearman": 0.7693050530672497, "step": 56750 }, { "epoch": 0.34, "learning_rate": 8.2819319524487e-09, "loss": 0.0056, "step": 57000 }, { "epoch": 0.34, "eval_avg_sts": 0.6861963542675861, "eval_sickr_spearman": 0.6030919437342946, "eval_stsb_spearman": 0.7693007648008775, "step": 57000 }, { "epoch": 0.35, "eval_avg_sts": 0.6861640128431876, "eval_sickr_spearman": 0.603011114136094, "eval_stsb_spearman": 0.7693169115502814, "step": 57250 }, { "epoch": 0.35, "learning_rate": 8.266861180101758e-09, "loss": 0.0056, "step": 57500 }, { "epoch": 0.35, "eval_avg_sts": 0.686116276282335, "eval_sickr_spearman": 0.6029430340000863, "eval_stsb_spearman": 0.7692895185645835, "step": 57500 }, { "epoch": 0.35, "eval_avg_sts": 0.6861118072711813, "eval_sickr_spearman": 0.6029420379483524, "eval_stsb_spearman": 0.7692815765940101, "step": 57750 }, { "epoch": 0.35, "learning_rate": 8.251790407754816e-09, "loss": 0.0055, "step": 58000 }, { "epoch": 0.35, "eval_avg_sts": 0.6860992615916662, "eval_sickr_spearman": 0.6029261011206111, "eval_stsb_spearman": 0.7692724220627214, "step": 58000 }, { "epoch": 0.35, "eval_avg_sts": 0.6860843714724788, "eval_sickr_spearman": 0.6028835697115763, "eval_stsb_spearman": 0.7692851732333813, "step": 58250 }, { "epoch": 0.35, "learning_rate": 8.236719635407876e-09, "loss": 0.0054, "step": 58500 }, { "epoch": 0.35, "eval_avg_sts": 0.6860407186846093, "eval_sickr_spearman": 0.602817232666103, "eval_stsb_spearman": 0.7692642047031155, "step": 58500 }, { "epoch": 0.35, "eval_avg_sts": 0.6860687869250233, "eval_sickr_spearman": 0.6028696747898894, "eval_stsb_spearman": 0.7692678990601571, "step": 58750 }, { "epoch": 0.36, "learning_rate": 8.221648863060934e-09, "loss": 0.0054, "step": 59000 }, { "epoch": 0.36, "eval_avg_sts": 0.6859999949935256, "eval_sickr_spearman": 0.6027299287316323, "eval_stsb_spearman": 0.769270061255419, "step": 59000 }, { "epoch": 0.36, "eval_avg_sts": 0.6860607016043287, "eval_sickr_spearman": 0.6028618557837788, "eval_stsb_spearman": 0.7692595474248788, "step": 59250 }, { "epoch": 0.36, "learning_rate": 8.206578090713993e-09, "loss": 0.0053, "step": 59500 }, { "epoch": 0.36, "eval_avg_sts": 0.6860587113413112, "eval_sickr_spearman": 0.6028526423052408, "eval_stsb_spearman": 0.7692647803773817, "step": 59500 }, { "epoch": 0.36, "eval_avg_sts": 0.6860611200235003, "eval_sickr_spearman": 0.6028758503106392, "eval_stsb_spearman": 0.7692463897363614, "step": 59750 }, { "epoch": 0.36, "learning_rate": 8.191507318367052e-09, "loss": 0.0053, "step": 60000 }, { "epoch": 0.36, "eval_avg_sts": 0.6859463597837794, "eval_sickr_spearman": 0.6026531331429537, "eval_stsb_spearman": 0.7692395864246052, "step": 60000 }, { "epoch": 0.36, "eval_avg_sts": 0.6859239543312032, "eval_sickr_spearman": 0.6025836585345187, "eval_stsb_spearman": 0.7692642501278879, "step": 60250 }, { "epoch": 0.36, "learning_rate": 8.17643654602011e-09, "loss": 0.0052, "step": 60500 }, { "epoch": 0.36, "eval_avg_sts": 0.6859466504629157, "eval_sickr_spearman": 0.6025991969415665, "eval_stsb_spearman": 0.7692941039842648, "step": 60500 }, { "epoch": 0.37, "eval_avg_sts": 0.6858718275078098, "eval_sickr_spearman": 0.6024521797056525, "eval_stsb_spearman": 0.7692914753099672, "step": 60750 }, { "epoch": 0.37, "learning_rate": 8.161365773673169e-09, "loss": 0.0052, "step": 61000 }, { "epoch": 0.37, "eval_avg_sts": 0.6858825814276086, "eval_sickr_spearman": 0.6024696106109946, "eval_stsb_spearman": 0.7692955522442226, "step": 61000 }, { "epoch": 0.37, "eval_avg_sts": 0.6859028515979515, "eval_sickr_spearman": 0.6024987451242092, "eval_stsb_spearman": 0.7693069580716939, "step": 61250 }, { "epoch": 0.37, "learning_rate": 8.146295001326229e-09, "loss": 0.005, "step": 61500 }, { "epoch": 0.37, "eval_avg_sts": 0.6859036333129487, "eval_sickr_spearman": 0.6025038747906385, "eval_stsb_spearman": 0.7693033918352589, "step": 61500 }, { "epoch": 0.37, "eval_avg_sts": 0.6859029568341515, "eval_sickr_spearman": 0.6024901292767116, "eval_stsb_spearman": 0.7693157843915914, "step": 61750 }, { "epoch": 0.37, "learning_rate": 8.131224228979287e-09, "loss": 0.0051, "step": 62000 }, { "epoch": 0.37, "eval_avg_sts": 0.6858457296550131, "eval_sickr_spearman": 0.6024017296853338, "eval_stsb_spearman": 0.7692897296246924, "step": 62000 }, { "epoch": 0.38, "eval_avg_sts": 0.6859309086152754, "eval_sickr_spearman": 0.6025292243072646, "eval_stsb_spearman": 0.769332592923286, "step": 62250 }, { "epoch": 0.38, "learning_rate": 8.116153456632346e-09, "loss": 0.0049, "step": 62500 }, { "epoch": 0.38, "eval_avg_sts": 0.6859456802235347, "eval_sickr_spearman": 0.6025707596645655, "eval_stsb_spearman": 0.769320600782504, "step": 62500 }, { "epoch": 0.38, "eval_avg_sts": 0.6860315979765899, "eval_sickr_spearman": 0.6026728051646969, "eval_stsb_spearman": 0.7693903907884827, "step": 62750 }, { "epoch": 0.38, "learning_rate": 8.101082684285404e-09, "loss": 0.0049, "step": 63000 }, { "epoch": 0.38, "eval_avg_sts": 0.686043555768491, "eval_sickr_spearman": 0.602662296818905, "eval_stsb_spearman": 0.7694248147180769, "step": 63000 }, { "epoch": 0.38, "eval_avg_sts": 0.6861302651703853, "eval_sickr_spearman": 0.6028059274789239, "eval_stsb_spearman": 0.7694546028618466, "step": 63250 }, { "epoch": 0.38, "learning_rate": 8.086011911938464e-09, "loss": 0.0049, "step": 63500 }, { "epoch": 0.38, "eval_avg_sts": 0.6861418997023286, "eval_sickr_spearman": 0.6028308287722698, "eval_stsb_spearman": 0.7694529706323875, "step": 63500 }, { "epoch": 0.38, "eval_avg_sts": 0.6861436270503518, "eval_sickr_spearman": 0.6028189757566372, "eval_stsb_spearman": 0.7694682783440664, "step": 63750 }, { "epoch": 0.39, "learning_rate": 8.070941139591522e-09, "loss": 0.0048, "step": 64000 }, { "epoch": 0.39, "eval_avg_sts": 0.6861273202059766, "eval_sickr_spearman": 0.6027953693305453, "eval_stsb_spearman": 0.7694592710814079, "step": 64000 }, { "epoch": 0.39, "eval_avg_sts": 0.6860970441745926, "eval_sickr_spearman": 0.6027587146267401, "eval_stsb_spearman": 0.7694353737224451, "step": 64250 }, { "epoch": 0.39, "learning_rate": 8.05587036724458e-09, "loss": 0.0048, "step": 64500 }, { "epoch": 0.39, "eval_avg_sts": 0.6861758328776337, "eval_sickr_spearman": 0.6029076243609485, "eval_stsb_spearman": 0.7694440413943191, "step": 64500 }, { "epoch": 0.39, "eval_avg_sts": 0.6861925925169595, "eval_sickr_spearman": 0.602933073482748, "eval_stsb_spearman": 0.769452111551171, "step": 64750 }, { "epoch": 0.39, "learning_rate": 8.04079959489764e-09, "loss": 0.0047, "step": 65000 }, { "epoch": 0.39, "eval_avg_sts": 0.6862022830505737, "eval_sickr_spearman": 0.6029402450552316, "eval_stsb_spearman": 0.7694643210459159, "step": 65000 }, { "epoch": 0.39, "eval_avg_sts": 0.6862314740498379, "eval_sickr_spearman": 0.6030238635982871, "eval_stsb_spearman": 0.7694390845013885, "step": 65250 }, { "epoch": 0.39, "learning_rate": 8.025728822550699e-09, "loss": 0.0046, "step": 65500 }, { "epoch": 0.39, "eval_avg_sts": 0.6861844875947325, "eval_sickr_spearman": 0.6029598174718015, "eval_stsb_spearman": 0.7694091577176634, "step": 65500 }, { "epoch": 0.4, "eval_avg_sts": 0.6862009373727546, "eval_sickr_spearman": 0.6030062832851849, "eval_stsb_spearman": 0.7693955914603242, "step": 65750 }, { "epoch": 0.4, "learning_rate": 8.010658050203757e-09, "loss": 0.0046, "step": 66000 }, { "epoch": 0.4, "eval_avg_sts": 0.6862719488998004, "eval_sickr_spearman": 0.6031286482406865, "eval_stsb_spearman": 0.7694152495589143, "step": 66000 }, { "epoch": 0.4, "eval_avg_sts": 0.686297897917143, "eval_sickr_spearman": 0.6031807915489527, "eval_stsb_spearman": 0.7694150042853335, "step": 66250 }, { "epoch": 0.4, "learning_rate": 7.995587277856817e-09, "loss": 0.0047, "step": 66500 }, { "epoch": 0.4, "eval_avg_sts": 0.6862834571479077, "eval_sickr_spearman": 0.6031002109636855, "eval_stsb_spearman": 0.7694667033321299, "step": 66500 } ], "max_steps": 331768, "num_train_epochs": 2, "total_flos": 0, "trial_name": null, "trial_params": null }