{ "best_metric": 0.8583291200488784, "best_model_checkpoint": "condenser-bert-base-uncased_mix_0.00001_cls_output_dir", "epoch": 2.0, "global_step": 31252, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "electra_acc": 97.8496, "electra_fix_acc": 0.9447, "electra_rep_acc": 0.0242, "epoch": 0.0, "learning_rate": 6.999776014335083e-06, "loss": 0.0046, "neg_sim": -0.0126, "pos_sim": 0.6661, "step": 1 }, { "epoch": 0.01, "eval_avg_sts": 0.6467281681849183, "eval_sickr_spearman": 0.6493190116334423, "eval_stsb_spearman": 0.6441373247363943, "step": 125 }, { "epoch": 0.02, "eval_avg_sts": 0.6766102197245321, "eval_sickr_spearman": 0.6747404807226937, "eval_stsb_spearman": 0.6784799587263706, "step": 250 }, { "epoch": 0.02, "eval_avg_sts": 0.6927318121134369, "eval_sickr_spearman": 0.6849395969264682, "eval_stsb_spearman": 0.7005240273004057, "step": 375 }, { "electra_acc": 111.3766, "electra_fix_acc": 0.9699, "electra_rep_acc": 0.4695, "epoch": 0.03, "learning_rate": 6.888007167541277e-06, "loss": 0.0027, "neg_sim": -0.0136, "pos_sim": 0.7499, "step": 500 }, { "epoch": 0.03, "eval_avg_sts": 0.70406785938114, "eval_sickr_spearman": 0.6912377231336422, "eval_stsb_spearman": 0.716897995628638, "step": 500 }, { "epoch": 0.04, "eval_avg_sts": 0.7079843162896757, "eval_sickr_spearman": 0.6933601214439895, "eval_stsb_spearman": 0.7226085111353618, "step": 625 }, { "epoch": 0.05, "eval_avg_sts": 0.721992790338854, "eval_sickr_spearman": 0.7041983394834997, "eval_stsb_spearman": 0.7397872411942082, "step": 750 }, { "epoch": 0.06, "eval_avg_sts": 0.7371963659388359, "eval_sickr_spearman": 0.7161540010979675, "eval_stsb_spearman": 0.7582387307797042, "step": 875 }, { "electra_acc": 115.0045, "electra_fix_acc": 0.9686, "electra_rep_acc": 0.6116, "epoch": 0.06, "learning_rate": 6.776014335082554e-06, "loss": 0.002, "neg_sim": -0.0139, "pos_sim": 0.7675, "step": 1000 }, { "epoch": 0.06, "eval_avg_sts": 0.7419491613050363, "eval_sickr_spearman": 0.7199929829426116, "eval_stsb_spearman": 0.763905339667461, "step": 1000 }, { "epoch": 0.07, "eval_avg_sts": 0.7503468807257706, "eval_sickr_spearman": 0.7253389886259777, "eval_stsb_spearman": 0.7753547728255635, "step": 1125 }, { "epoch": 0.08, "eval_avg_sts": 0.755381635152361, "eval_sickr_spearman": 0.7283938147085431, "eval_stsb_spearman": 0.782369455596179, "step": 1250 }, { "epoch": 0.09, "eval_avg_sts": 0.7557951064861299, "eval_sickr_spearman": 0.7286210018183454, "eval_stsb_spearman": 0.7829692111539145, "step": 1375 }, { "electra_acc": 115.5115, "electra_fix_acc": 0.9698, "electra_rep_acc": 0.6302, "epoch": 0.1, "learning_rate": 6.6640215026238325e-06, "loss": 0.0017, "neg_sim": -0.014, "pos_sim": 0.7738, "step": 1500 }, { "epoch": 0.1, "eval_avg_sts": 0.7624381952103951, "eval_sickr_spearman": 0.7329619086730178, "eval_stsb_spearman": 0.7919144817477726, "step": 1500 }, { "epoch": 0.1, "eval_avg_sts": 0.765407531009575, "eval_sickr_spearman": 0.7350173996590618, "eval_stsb_spearman": 0.7957976623600883, "step": 1625 }, { "epoch": 0.11, "eval_avg_sts": 0.7663640946087169, "eval_sickr_spearman": 0.7356288836114704, "eval_stsb_spearman": 0.7970993056059634, "step": 1750 }, { "epoch": 0.12, "eval_avg_sts": 0.7654280480014255, "eval_sickr_spearman": 0.7361625571795493, "eval_stsb_spearman": 0.7946935388233016, "step": 1875 }, { "electra_acc": 115.9922, "electra_fix_acc": 0.971, "electra_rep_acc": 0.641, "epoch": 0.13, "learning_rate": 6.55202867016511e-06, "loss": 0.0016, "neg_sim": -0.014, "pos_sim": 0.7751, "step": 2000 }, { "epoch": 0.13, "eval_avg_sts": 0.7672864050175499, "eval_sickr_spearman": 0.7369789418106929, "eval_stsb_spearman": 0.7975938682244069, "step": 2000 }, { "epoch": 0.14, "eval_avg_sts": 0.7747771973615925, "eval_sickr_spearman": 0.7388607043028352, "eval_stsb_spearman": 0.8106936904203497, "step": 2125 }, { "epoch": 0.14, "eval_avg_sts": 0.7774171758966744, "eval_sickr_spearman": 0.7346681655205053, "eval_stsb_spearman": 0.8201661862728435, "step": 2250 }, { "epoch": 0.15, "eval_avg_sts": 0.7804372234935246, "eval_sickr_spearman": 0.7373037281186174, "eval_stsb_spearman": 0.8235707188684319, "step": 2375 }, { "electra_acc": 116.1787, "electra_fix_acc": 0.971, "electra_rep_acc": 0.6492, "epoch": 0.16, "learning_rate": 6.440035837706387e-06, "loss": 0.0018, "neg_sim": -0.0139, "pos_sim": 0.7692, "step": 2500 }, { "epoch": 0.16, "eval_avg_sts": 0.7810767988630156, "eval_sickr_spearman": 0.7360565525386754, "eval_stsb_spearman": 0.8260970451873557, "step": 2500 }, { "epoch": 0.17, "eval_avg_sts": 0.7810534850927242, "eval_sickr_spearman": 0.737071930023077, "eval_stsb_spearman": 0.8250350401623715, "step": 2625 }, { "epoch": 0.18, "eval_avg_sts": 0.7840637132723827, "eval_sickr_spearman": 0.7369239942306477, "eval_stsb_spearman": 0.8312034323141176, "step": 2750 }, { "epoch": 0.18, "eval_avg_sts": 0.7807650924553761, "eval_sickr_spearman": 0.736852331827302, "eval_stsb_spearman": 0.8246778530834501, "step": 2875 }, { "electra_acc": 116.4021, "electra_fix_acc": 0.9711, "electra_rep_acc": 0.6575, "epoch": 0.19, "learning_rate": 6.328043005247664e-06, "loss": 0.0017, "neg_sim": -0.014, "pos_sim": 0.778, "step": 3000 }, { "epoch": 0.19, "eval_avg_sts": 0.7812090218351604, "eval_sickr_spearman": 0.7374819235049528, "eval_stsb_spearman": 0.824936120165368, "step": 3000 }, { "epoch": 0.2, "eval_avg_sts": 0.7792184022991786, "eval_sickr_spearman": 0.735666443932795, "eval_stsb_spearman": 0.8227703606655621, "step": 3125 }, { "epoch": 0.21, "eval_avg_sts": 0.7868425974055346, "eval_sickr_spearman": 0.7406432345394065, "eval_stsb_spearman": 0.8330419602716627, "step": 3250 }, { "epoch": 0.22, "eval_avg_sts": 0.7879152712258332, "eval_sickr_spearman": 0.74201269730361, "eval_stsb_spearman": 0.8338178451480563, "step": 3375 }, { "electra_acc": 116.4857, "electra_fix_acc": 0.9706, "electra_rep_acc": 0.6652, "epoch": 0.22, "learning_rate": 6.2160501727889415e-06, "loss": 0.0016, "neg_sim": -0.014, "pos_sim": 0.7831, "step": 3500 }, { "epoch": 0.22, "eval_avg_sts": 0.7874505134915601, "eval_sickr_spearman": 0.7415818102926086, "eval_stsb_spearman": 0.8333192166905118, "step": 3500 }, { "epoch": 0.23, "eval_avg_sts": 0.7885881180801638, "eval_sickr_spearman": 0.745498890708191, "eval_stsb_spearman": 0.8316773454521367, "step": 3625 }, { "epoch": 0.24, "eval_avg_sts": 0.786897125097908, "eval_sickr_spearman": 0.7427211560498221, "eval_stsb_spearman": 0.8310730941459938, "step": 3750 }, { "epoch": 0.25, "eval_avg_sts": 0.7884858023873396, "eval_sickr_spearman": 0.7450858232358233, "eval_stsb_spearman": 0.831885781538856, "step": 3875 }, { "electra_acc": 116.6361, "electra_fix_acc": 0.9706, "electra_rep_acc": 0.6699, "epoch": 0.26, "learning_rate": 6.104057340330219e-06, "loss": 0.0016, "neg_sim": -0.014, "pos_sim": 0.7846, "step": 4000 }, { "epoch": 0.26, "eval_avg_sts": 0.7883005639004375, "eval_sickr_spearman": 0.7467304081490642, "eval_stsb_spearman": 0.8298707196518108, "step": 4000 }, { "epoch": 0.26, "eval_avg_sts": 0.7890018606938185, "eval_sickr_spearman": 0.7466418868291138, "eval_stsb_spearman": 0.8313618345585232, "step": 4125 }, { "epoch": 0.27, "eval_avg_sts": 0.794939161048242, "eval_sickr_spearman": 0.7542289276433795, "eval_stsb_spearman": 0.8356493944531046, "step": 4250 }, { "epoch": 0.28, "eval_avg_sts": 0.7950124813232355, "eval_sickr_spearman": 0.7546040025145097, "eval_stsb_spearman": 0.8354209601319615, "step": 4375 }, { "electra_acc": 116.6274, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.6733, "epoch": 0.29, "learning_rate": 5.992064507871497e-06, "loss": 0.0016, "neg_sim": -0.0141, "pos_sim": 0.7871, "step": 4500 }, { "epoch": 0.29, "eval_avg_sts": 0.7938800981934206, "eval_sickr_spearman": 0.7533463081233527, "eval_stsb_spearman": 0.8344138882634886, "step": 4500 }, { "epoch": 0.3, "eval_avg_sts": 0.794615358586916, "eval_sickr_spearman": 0.7546161063520722, "eval_stsb_spearman": 0.8346146108217597, "step": 4625 }, { "epoch": 0.3, "eval_avg_sts": 0.7945912313688637, "eval_sickr_spearman": 0.7546232629861865, "eval_stsb_spearman": 0.8345591997515407, "step": 4750 }, { "epoch": 0.31, "eval_avg_sts": 0.7926504221766928, "eval_sickr_spearman": 0.7513844777851131, "eval_stsb_spearman": 0.8339163665682726, "step": 4875 }, { "electra_acc": 116.73, "electra_fix_acc": 0.9701, "electra_rep_acc": 0.6783, "epoch": 0.32, "learning_rate": 5.880071675412773e-06, "loss": 0.0015, "neg_sim": -0.0141, "pos_sim": 0.7904, "step": 5000 }, { "epoch": 0.32, "eval_avg_sts": 0.7934518878705803, "eval_sickr_spearman": 0.7517016271479089, "eval_stsb_spearman": 0.8352021485932517, "step": 5000 }, { "epoch": 0.33, "eval_avg_sts": 0.7947730330766363, "eval_sickr_spearman": 0.7522364054313209, "eval_stsb_spearman": 0.8373096607219516, "step": 5125 }, { "epoch": 0.34, "eval_avg_sts": 0.7957221976017312, "eval_sickr_spearman": 0.7467118201128077, "eval_stsb_spearman": 0.8447325750906549, "step": 5250 }, { "epoch": 0.34, "eval_avg_sts": 0.7897362921108364, "eval_sickr_spearman": 0.7382444172402828, "eval_stsb_spearman": 0.8412281669813901, "step": 5375 }, { "electra_acc": 116.8512, "electra_fix_acc": 0.9704, "electra_rep_acc": 0.6807, "epoch": 0.35, "learning_rate": 5.7680788429540505e-06, "loss": 0.0016, "neg_sim": -0.014, "pos_sim": 0.7834, "step": 5500 }, { "epoch": 0.35, "eval_avg_sts": 0.791664811568233, "eval_sickr_spearman": 0.7412088968210431, "eval_stsb_spearman": 0.8421207263154228, "step": 5500 }, { "epoch": 0.36, "eval_avg_sts": 0.791979531720516, "eval_sickr_spearman": 0.7422650046794644, "eval_stsb_spearman": 0.8416940587615678, "step": 5625 }, { "epoch": 0.37, "eval_avg_sts": 0.7909414570864208, "eval_sickr_spearman": 0.7407966939085013, "eval_stsb_spearman": 0.8410862202643404, "step": 5750 }, { "epoch": 0.38, "eval_avg_sts": 0.7916033250892136, "eval_sickr_spearman": 0.7420070776647416, "eval_stsb_spearman": 0.8411995725136856, "step": 5875 }, { "electra_acc": 116.9386, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.6839, "epoch": 0.38, "learning_rate": 5.656086010495328e-06, "loss": 0.0014, "neg_sim": -0.014, "pos_sim": 0.7867, "step": 6000 }, { "epoch": 0.38, "eval_avg_sts": 0.7939838516948429, "eval_sickr_spearman": 0.7457015819562598, "eval_stsb_spearman": 0.8422661214334259, "step": 6000 }, { "epoch": 0.39, "eval_avg_sts": 0.7897949803868374, "eval_sickr_spearman": 0.7381373559151772, "eval_stsb_spearman": 0.8414526048584976, "step": 6125 }, { "epoch": 0.4, "eval_avg_sts": 0.792929873578434, "eval_sickr_spearman": 0.7426651998166468, "eval_stsb_spearman": 0.8431945473402213, "step": 6250 }, { "epoch": 0.41, "eval_avg_sts": 0.7941689576028486, "eval_sickr_spearman": 0.7447726604544468, "eval_stsb_spearman": 0.8435652547512503, "step": 6375 }, { "electra_acc": 116.9876, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.6869, "epoch": 0.42, "learning_rate": 5.544093178036605e-06, "loss": 0.0014, "neg_sim": -0.0141, "pos_sim": 0.7923, "step": 6500 }, { "epoch": 0.42, "eval_avg_sts": 0.7943810788058159, "eval_sickr_spearman": 0.7449998475642491, "eval_stsb_spearman": 0.8437623100473826, "step": 6500 }, { "epoch": 0.42, "eval_avg_sts": 0.7943037355857338, "eval_sickr_spearman": 0.7452223756572118, "eval_stsb_spearman": 0.8433850955142557, "step": 6625 }, { "epoch": 0.43, "eval_avg_sts": 0.7941830032512052, "eval_sickr_spearman": 0.7457148385402568, "eval_stsb_spearman": 0.8426511679621534, "step": 6750 }, { "epoch": 0.44, "eval_avg_sts": 0.7945109790018022, "eval_sickr_spearman": 0.7458948591084468, "eval_stsb_spearman": 0.8431270988951577, "step": 6875 }, { "electra_acc": 117.061, "electra_fix_acc": 0.9704, "electra_rep_acc": 0.6883, "epoch": 0.45, "learning_rate": 5.432100345577883e-06, "loss": 0.0014, "neg_sim": -0.0142, "pos_sim": 0.8, "step": 7000 }, { "epoch": 0.45, "eval_avg_sts": 0.7988457805161364, "eval_sickr_spearman": 0.7515535952932767, "eval_stsb_spearman": 0.846137965738996, "step": 7000 }, { "epoch": 0.46, "eval_avg_sts": 0.7981158834848314, "eval_sickr_spearman": 0.75075815222236, "eval_stsb_spearman": 0.845473614747303, "step": 7125 }, { "epoch": 0.46, "eval_avg_sts": 0.7984443557563039, "eval_sickr_spearman": 0.7501957080245197, "eval_stsb_spearman": 0.8466930034880882, "step": 7250 }, { "epoch": 0.47, "eval_avg_sts": 0.7981127107574888, "eval_sickr_spearman": 0.750318523550897, "eval_stsb_spearman": 0.8459068979640807, "step": 7375 }, { "electra_acc": 117.069, "electra_fix_acc": 0.9701, "electra_rep_acc": 0.6888, "epoch": 0.48, "learning_rate": 5.32010751311916e-06, "loss": 0.0015, "neg_sim": -0.0142, "pos_sim": 0.8001, "step": 7500 }, { "epoch": 0.48, "eval_avg_sts": 0.7955868437200273, "eval_sickr_spearman": 0.7487511246176671, "eval_stsb_spearman": 0.8424225628223875, "step": 7500 }, { "epoch": 0.49, "eval_avg_sts": 0.7949766204622221, "eval_sickr_spearman": 0.7481150007102207, "eval_stsb_spearman": 0.8418382402142234, "step": 7625 }, { "epoch": 0.5, "eval_avg_sts": 0.7950165968598919, "eval_sickr_spearman": 0.7485309980797762, "eval_stsb_spearman": 0.8415021956400077, "step": 7750 }, { "epoch": 0.5, "eval_avg_sts": 0.7952250496118516, "eval_sickr_spearman": 0.7486211044260741, "eval_stsb_spearman": 0.8418289947976292, "step": 7875 }, { "electra_acc": 117.1619, "electra_fix_acc": 0.9703, "electra_rep_acc": 0.6937, "epoch": 0.51, "learning_rate": 5.2081146806604376e-06, "loss": 0.0014, "neg_sim": -0.0142, "pos_sim": 0.7987, "step": 8000 }, { "epoch": 0.51, "eval_avg_sts": 0.7952021451003002, "eval_sickr_spearman": 0.7491326836874914, "eval_stsb_spearman": 0.8412716065131091, "step": 8000 }, { "epoch": 0.52, "eval_avg_sts": 0.7951490971039081, "eval_sickr_spearman": 0.749455164502547, "eval_stsb_spearman": 0.8408430297052691, "step": 8125 }, { "epoch": 0.53, "eval_avg_sts": 0.7950116620178889, "eval_sickr_spearman": 0.749741093649408, "eval_stsb_spearman": 0.8402822303863696, "step": 8250 }, { "epoch": 0.54, "eval_avg_sts": 0.7951408099684255, "eval_sickr_spearman": 0.7501430179062422, "eval_stsb_spearman": 0.8401386020306089, "step": 8375 }, { "electra_acc": 117.2641, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.6954, "epoch": 0.54, "learning_rate": 5.096121848201715e-06, "loss": 0.0014, "neg_sim": -0.0142, "pos_sim": 0.806, "step": 8500 }, { "epoch": 0.54, "eval_avg_sts": 0.7949691120514141, "eval_sickr_spearman": 0.7501533445930513, "eval_stsb_spearman": 0.8397848795097769, "step": 8500 }, { "epoch": 0.55, "eval_avg_sts": 0.7947062336358759, "eval_sickr_spearman": 0.7503012323543793, "eval_stsb_spearman": 0.8391112349173724, "step": 8625 }, { "epoch": 0.56, "eval_avg_sts": 0.7944181183521635, "eval_sickr_spearman": 0.7493049712483499, "eval_stsb_spearman": 0.8395312654559769, "step": 8750 }, { "epoch": 0.57, "eval_avg_sts": 0.7950044953143297, "eval_sickr_spearman": 0.7496116017999308, "eval_stsb_spearman": 0.8403973888287286, "step": 8875 }, { "electra_acc": 117.2706, "electra_fix_acc": 0.9705, "electra_rep_acc": 0.6956, "epoch": 0.58, "learning_rate": 4.984129015742992e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8093, "step": 9000 }, { "epoch": 0.58, "eval_avg_sts": 0.7950848432290827, "eval_sickr_spearman": 0.7493472386176154, "eval_stsb_spearman": 0.8408224478405498, "step": 9000 }, { "epoch": 0.58, "eval_avg_sts": 0.794847808534162, "eval_sickr_spearman": 0.7489566016696193, "eval_stsb_spearman": 0.8407390153987048, "step": 9125 }, { "epoch": 0.59, "eval_avg_sts": 0.795644348129942, "eval_sickr_spearman": 0.7493907547955183, "eval_stsb_spearman": 0.8418979414643657, "step": 9250 }, { "epoch": 0.6, "eval_avg_sts": 0.7966154624139274, "eval_sickr_spearman": 0.7486175020934662, "eval_stsb_spearman": 0.8446134227343888, "step": 9375 }, { "electra_acc": 117.2471, "electra_fix_acc": 0.9703, "electra_rep_acc": 0.6966, "epoch": 0.61, "learning_rate": 4.872136183284269e-06, "loss": 0.0015, "neg_sim": -0.0143, "pos_sim": 0.8085, "step": 9500 }, { "epoch": 0.61, "eval_avg_sts": 0.7962741237962319, "eval_sickr_spearman": 0.7489400789707246, "eval_stsb_spearman": 0.8436081686217393, "step": 9500 }, { "epoch": 0.62, "eval_avg_sts": 0.7966016709380292, "eval_sickr_spearman": 0.7497793264061527, "eval_stsb_spearman": 0.8434240154699058, "step": 9625 }, { "epoch": 0.62, "eval_avg_sts": 0.7969314966558516, "eval_sickr_spearman": 0.750271933382502, "eval_stsb_spearman": 0.8435910599292011, "step": 9750 }, { "epoch": 0.63, "eval_avg_sts": 0.7965405773738363, "eval_sickr_spearman": 0.7509994124448837, "eval_stsb_spearman": 0.842081742302789, "step": 9875 }, { "electra_acc": 117.3599, "electra_fix_acc": 0.9704, "electra_rep_acc": 0.6998, "epoch": 0.64, "learning_rate": 4.760143350825547e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8077, "step": 10000 }, { "epoch": 0.64, "eval_avg_sts": 0.7964394231656002, "eval_sickr_spearman": 0.7509571931067197, "eval_stsb_spearman": 0.8419216532244808, "step": 10000 }, { "epoch": 0.65, "eval_avg_sts": 0.7967246593169031, "eval_sickr_spearman": 0.7512494143278691, "eval_stsb_spearman": 0.842199904305937, "step": 10125 }, { "epoch": 0.66, "eval_avg_sts": 0.798107636355013, "eval_sickr_spearman": 0.750279041985515, "eval_stsb_spearman": 0.845936230724511, "step": 10250 }, { "epoch": 0.66, "eval_avg_sts": 0.7980710535062169, "eval_sickr_spearman": 0.7505661719099119, "eval_stsb_spearman": 0.8455759351025218, "step": 10375 }, { "electra_acc": 117.3933, "electra_fix_acc": 0.9703, "electra_rep_acc": 0.7006, "epoch": 0.67, "learning_rate": 4.648150518366825e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8081, "step": 10500 }, { "epoch": 0.67, "eval_avg_sts": 0.7985924819917696, "eval_sickr_spearman": 0.7519674312632674, "eval_stsb_spearman": 0.8452175327202718, "step": 10500 }, { "epoch": 0.68, "eval_avg_sts": 0.7991692973398108, "eval_sickr_spearman": 0.7515644683598212, "eval_stsb_spearman": 0.8467741263198004, "step": 10625 }, { "epoch": 0.69, "eval_avg_sts": 0.7987755187970038, "eval_sickr_spearman": 0.751040334943309, "eval_stsb_spearman": 0.8465107026506987, "step": 10750 }, { "epoch": 0.7, "eval_avg_sts": 0.7986197258898549, "eval_sickr_spearman": 0.7514156980010478, "eval_stsb_spearman": 0.8458237537786619, "step": 10875 }, { "electra_acc": 117.4302, "electra_fix_acc": 0.9705, "electra_rep_acc": 0.7012, "epoch": 0.7, "learning_rate": 4.536157685908102e-06, "loss": 0.0013, "neg_sim": -0.0143, "pos_sim": 0.8102, "step": 11000 }, { "epoch": 0.7, "eval_avg_sts": 0.7996251236145968, "eval_sickr_spearman": 0.7529341052108109, "eval_stsb_spearman": 0.8463161420183827, "step": 11000 }, { "epoch": 0.71, "eval_avg_sts": 0.7994727880295167, "eval_sickr_spearman": 0.752963500244891, "eval_stsb_spearman": 0.8459820758141423, "step": 11125 }, { "epoch": 0.72, "eval_avg_sts": 0.7988993639342742, "eval_sickr_spearman": 0.7527381383169434, "eval_stsb_spearman": 0.845060589551605, "step": 11250 }, { "epoch": 0.73, "eval_avg_sts": 0.7949700279790515, "eval_sickr_spearman": 0.7504287549286974, "eval_stsb_spearman": 0.8395113010294055, "step": 11375 }, { "electra_acc": 117.3738, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.6995, "epoch": 0.74, "learning_rate": 4.424164853449379e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8157, "step": 11500 }, { "epoch": 0.74, "eval_avg_sts": 0.7945615778286725, "eval_sickr_spearman": 0.7499855239246266, "eval_stsb_spearman": 0.8391376317327186, "step": 11500 }, { "epoch": 0.74, "eval_avg_sts": 0.7934292933863014, "eval_sickr_spearman": 0.7480520799673368, "eval_stsb_spearman": 0.838806506805266, "step": 11625 }, { "epoch": 0.75, "eval_avg_sts": 0.7937723514656398, "eval_sickr_spearman": 0.7481406973494901, "eval_stsb_spearman": 0.8394040055817897, "step": 11750 }, { "epoch": 0.76, "eval_avg_sts": 0.7945405318685601, "eval_sickr_spearman": 0.7490017989360725, "eval_stsb_spearman": 0.8400792648010476, "step": 11875 }, { "electra_acc": 117.5412, "electra_fix_acc": 0.9707, "electra_rep_acc": 0.7042, "epoch": 0.77, "learning_rate": 4.312172020990656e-06, "loss": 0.0016, "neg_sim": -0.0143, "pos_sim": 0.812, "step": 12000 }, { "epoch": 0.77, "eval_avg_sts": 0.7960514566703415, "eval_sickr_spearman": 0.7499365802322611, "eval_stsb_spearman": 0.8421663331084219, "step": 12000 }, { "epoch": 0.78, "eval_avg_sts": 0.7951893349270129, "eval_sickr_spearman": 0.7490124138094904, "eval_stsb_spearman": 0.8413662560445356, "step": 12125 }, { "epoch": 0.78, "eval_avg_sts": 0.7937236608939515, "eval_sickr_spearman": 0.7449763603556459, "eval_stsb_spearman": 0.8424709614322572, "step": 12250 }, { "epoch": 0.79, "eval_avg_sts": 0.7934627772608434, "eval_sickr_spearman": 0.746114793521932, "eval_stsb_spearman": 0.8408107609997548, "step": 12375 }, { "electra_acc": 117.4273, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.7033, "epoch": 0.8, "learning_rate": 4.2001791885319345e-06, "loss": 0.0016, "neg_sim": -0.0143, "pos_sim": 0.8084, "step": 12500 }, { "epoch": 0.8, "eval_avg_sts": 0.793506856374318, "eval_sickr_spearman": 0.7463023069419462, "eval_stsb_spearman": 0.8407114058066898, "step": 12500 }, { "epoch": 0.81, "eval_avg_sts": 0.7929774718082235, "eval_sickr_spearman": 0.7465257476258366, "eval_stsb_spearman": 0.8394291959906105, "step": 12625 }, { "epoch": 0.82, "eval_avg_sts": 0.7907041469753885, "eval_sickr_spearman": 0.7438328358926072, "eval_stsb_spearman": 0.8375754580581698, "step": 12750 }, { "epoch": 0.82, "eval_avg_sts": 0.7919377238754862, "eval_sickr_spearman": 0.744254116683321, "eval_stsb_spearman": 0.8396213310676515, "step": 12875 }, { "electra_acc": 117.4657, "electra_fix_acc": 0.9701, "electra_rep_acc": 0.7049, "epoch": 0.83, "learning_rate": 4.088186356073212e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8091, "step": 13000 }, { "epoch": 0.83, "eval_avg_sts": 0.7950524625796482, "eval_sickr_spearman": 0.7459233415515996, "eval_stsb_spearman": 0.8441815836076968, "step": 13000 }, { "epoch": 0.84, "eval_avg_sts": 0.7953046984165892, "eval_sickr_spearman": 0.7460563877025831, "eval_stsb_spearman": 0.8445530091305953, "step": 13125 }, { "epoch": 0.85, "eval_avg_sts": 0.7960686224710865, "eval_sickr_spearman": 0.745027801665286, "eval_stsb_spearman": 0.8471094432768869, "step": 13250 }, { "epoch": 0.86, "eval_avg_sts": 0.7964731409284048, "eval_sickr_spearman": 0.7453551336215867, "eval_stsb_spearman": 0.847591148235223, "step": 13375 }, { "electra_acc": 117.5481, "electra_fix_acc": 0.9705, "electra_rep_acc": 0.7049, "epoch": 0.86, "learning_rate": 3.976193523614489e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8115, "step": 13500 }, { "epoch": 0.86, "eval_avg_sts": 0.7966766934403904, "eval_sickr_spearman": 0.7463476483017039, "eval_stsb_spearman": 0.8470057385790768, "step": 13500 }, { "epoch": 0.87, "eval_avg_sts": 0.7966025495961442, "eval_sickr_spearman": 0.7463816062904207, "eval_stsb_spearman": 0.8468234929018676, "step": 13625 }, { "epoch": 0.88, "eval_avg_sts": 0.7967818015292757, "eval_sickr_spearman": 0.746880793527667, "eval_stsb_spearman": 0.8466828095308844, "step": 13750 }, { "epoch": 0.89, "eval_avg_sts": 0.7969535576693652, "eval_sickr_spearman": 0.7471762808637142, "eval_stsb_spearman": 0.8467308344750161, "step": 13875 }, { "electra_acc": 117.5482, "electra_fix_acc": 0.9705, "electra_rep_acc": 0.7068, "epoch": 0.9, "learning_rate": 3.864200691155766e-06, "loss": 0.0013, "neg_sim": -0.0143, "pos_sim": 0.8142, "step": 14000 }, { "epoch": 0.9, "eval_avg_sts": 0.7965629720194758, "eval_sickr_spearman": 0.7468308350906325, "eval_stsb_spearman": 0.8462951089483192, "step": 14000 }, { "epoch": 0.9, "eval_avg_sts": 0.7965266307017043, "eval_sickr_spearman": 0.7447568101909722, "eval_stsb_spearman": 0.8482964512124362, "step": 14125 }, { "epoch": 0.91, "eval_avg_sts": 0.7965480662045883, "eval_sickr_spearman": 0.7448633912050633, "eval_stsb_spearman": 0.8482327412041132, "step": 14250 }, { "epoch": 0.92, "eval_avg_sts": 0.7964806384496487, "eval_sickr_spearman": 0.745111231688484, "eval_stsb_spearman": 0.8478500452108133, "step": 14375 }, { "electra_acc": 117.5208, "electra_fix_acc": 0.9701, "electra_rep_acc": 0.7071, "epoch": 0.93, "learning_rate": 3.7522078586970435e-06, "loss": 0.0013, "neg_sim": -0.0143, "pos_sim": 0.8155, "step": 14500 }, { "epoch": 0.93, "eval_avg_sts": 0.7971949399590785, "eval_sickr_spearman": 0.7457434650767139, "eval_stsb_spearman": 0.8486464148414431, "step": 14500 }, { "epoch": 0.94, "eval_avg_sts": 0.7970029281099968, "eval_sickr_spearman": 0.7458370776934168, "eval_stsb_spearman": 0.8481687785265768, "step": 14625 }, { "epoch": 0.94, "eval_avg_sts": 0.7991675032273975, "eval_sickr_spearman": 0.7483727836316391, "eval_stsb_spearman": 0.8499622228231559, "step": 14750 }, { "epoch": 0.95, "eval_avg_sts": 0.79901349058513, "eval_sickr_spearman": 0.748446895621158, "eval_stsb_spearman": 0.8495800855491022, "step": 14875 }, { "electra_acc": 117.6139, "electra_fix_acc": 0.9705, "electra_rep_acc": 0.7091, "epoch": 0.96, "learning_rate": 3.6402150262383203e-06, "loss": 0.0014, "neg_sim": -0.0143, "pos_sim": 0.8156, "step": 15000 }, { "epoch": 0.96, "eval_avg_sts": 0.79914478688965, "eval_sickr_spearman": 0.7468535118620501, "eval_stsb_spearman": 0.8514360619172497, "step": 15000 }, { "epoch": 0.97, "eval_avg_sts": 0.7991571351120899, "eval_sickr_spearman": 0.7471076924508606, "eval_stsb_spearman": 0.8512065777733191, "step": 15125 }, { "epoch": 0.98, "eval_avg_sts": 0.8012474062999755, "eval_sickr_spearman": 0.7502938355647578, "eval_stsb_spearman": 0.8522009770351932, "step": 15250 }, { "epoch": 0.98, "eval_avg_sts": 0.7993104516704408, "eval_sickr_spearman": 0.7512376467080167, "eval_stsb_spearman": 0.8473832566328647, "step": 15375 }, { "electra_acc": 117.5407, "electra_fix_acc": 0.9702, "electra_rep_acc": 0.7085, "epoch": 0.99, "learning_rate": 3.5282221937795976e-06, "loss": 0.0015, "neg_sim": -0.0143, "pos_sim": 0.8109, "step": 15500 }, { "epoch": 0.99, "eval_avg_sts": 0.7991830405974262, "eval_sickr_spearman": 0.7510609883169274, "eval_stsb_spearman": 0.847305092877925, "step": 15500 }, { "epoch": 1.0, "eval_avg_sts": 0.7993142209919086, "eval_sickr_spearman": 0.750953302587503, "eval_stsb_spearman": 0.8476751393963143, "step": 15625 }, { "epoch": 1.01, "eval_avg_sts": 0.79998042307191, "eval_sickr_spearman": 0.7521459148362114, "eval_stsb_spearman": 0.8478149313076085, "step": 15750 }, { "epoch": 1.02, "eval_avg_sts": 0.8000408218838944, "eval_sickr_spearman": 0.7522030718469228, "eval_stsb_spearman": 0.847878571920866, "step": 15875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.02, "learning_rate": 3.4162293613208757e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8075, "step": 16000 }, { "epoch": 1.02, "eval_avg_sts": 0.7998575815186055, "eval_sickr_spearman": 0.7520637336216508, "eval_stsb_spearman": 0.8476514294155602, "step": 16000 }, { "epoch": 1.03, "eval_avg_sts": 0.7998976758754202, "eval_sickr_spearman": 0.752151678568384, "eval_stsb_spearman": 0.8476436731824563, "step": 16125 }, { "epoch": 1.04, "eval_avg_sts": 0.7993375115090717, "eval_sickr_spearman": 0.7512065225542848, "eval_stsb_spearman": 0.8474685004638588, "step": 16250 }, { "epoch": 1.05, "eval_avg_sts": 0.8005700206739319, "eval_sickr_spearman": 0.7500328285448524, "eval_stsb_spearman": 0.8511072128030113, "step": 16375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.06, "learning_rate": 3.304236528862153e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8139, "step": 16500 }, { "epoch": 1.06, "eval_avg_sts": 0.800932566485014, "eval_sickr_spearman": 0.7492046823085472, "eval_stsb_spearman": 0.8526604506614809, "step": 16500 }, { "epoch": 1.06, "eval_avg_sts": 0.8020790163363096, "eval_sickr_spearman": 0.7523164732774181, "eval_stsb_spearman": 0.851841559395201, "step": 16625 }, { "epoch": 1.07, "eval_avg_sts": 0.8034376489964099, "eval_sickr_spearman": 0.7546485273455429, "eval_stsb_spearman": 0.8522267706472768, "step": 16750 }, { "epoch": 1.08, "eval_avg_sts": 0.8035121251749653, "eval_sickr_spearman": 0.7551173589266803, "eval_stsb_spearman": 0.8519068914232504, "step": 16875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.09, "learning_rate": 3.19224369640343e-06, "loss": 0.0015, "neg_sim": NaN, "pos_sim": 0.8123, "step": 17000 }, { "epoch": 1.09, "eval_avg_sts": 0.8034498780822419, "eval_sickr_spearman": 0.755224660407293, "eval_stsb_spearman": 0.8516750957571909, "step": 17000 }, { "epoch": 1.1, "eval_avg_sts": 0.8032831913725695, "eval_sickr_spearman": 0.7549221124993344, "eval_stsb_spearman": 0.8516442702458048, "step": 17125 }, { "epoch": 1.1, "eval_avg_sts": 0.8025478170596181, "eval_sickr_spearman": 0.7531311768200115, "eval_stsb_spearman": 0.8519644572992248, "step": 17250 }, { "epoch": 1.11, "eval_avg_sts": 0.8023474610208348, "eval_sickr_spearman": 0.7529551428332407, "eval_stsb_spearman": 0.8517397792084288, "step": 17375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.12, "learning_rate": 3.0802508639447074e-06, "loss": 0.0015, "neg_sim": NaN, "pos_sim": 0.8124, "step": 17500 }, { "epoch": 1.12, "eval_avg_sts": 0.8036377678015303, "eval_sickr_spearman": 0.7540948728392658, "eval_stsb_spearman": 0.8531806627637948, "step": 17500 }, { "epoch": 1.13, "eval_avg_sts": 0.803436785093961, "eval_sickr_spearman": 0.7537820462755993, "eval_stsb_spearman": 0.8530915239123227, "step": 17625 }, { "epoch": 1.14, "eval_avg_sts": 0.8034543332861956, "eval_sickr_spearman": 0.753641363179487, "eval_stsb_spearman": 0.8532673033929044, "step": 17750 }, { "epoch": 1.14, "eval_avg_sts": 0.8031959700775069, "eval_sickr_spearman": 0.7533919857008204, "eval_stsb_spearman": 0.8529999544541933, "step": 17875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.15, "learning_rate": 2.9682580314859847e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8138, "step": 18000 }, { "epoch": 1.15, "eval_avg_sts": 0.8033449509317736, "eval_sickr_spearman": 0.7535802676184579, "eval_stsb_spearman": 0.8531096342450895, "step": 18000 }, { "epoch": 1.16, "eval_avg_sts": 0.803034323911963, "eval_sickr_spearman": 0.7532389586116386, "eval_stsb_spearman": 0.8528296892122876, "step": 18125 }, { "epoch": 1.17, "eval_avg_sts": 0.8030336070807917, "eval_sickr_spearman": 0.753408940679628, "eval_stsb_spearman": 0.8526582734819556, "step": 18250 }, { "epoch": 1.18, "eval_avg_sts": 0.8042466753465274, "eval_sickr_spearman": 0.754791275779017, "eval_stsb_spearman": 0.8537020749140378, "step": 18375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.18, "learning_rate": 2.8562651990272623e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8154, "step": 18500 }, { "epoch": 1.18, "eval_avg_sts": 0.8041866669158593, "eval_sickr_spearman": 0.7544020317329624, "eval_stsb_spearman": 0.8539713020987562, "step": 18500 }, { "epoch": 1.19, "eval_avg_sts": 0.8040941868331957, "eval_sickr_spearman": 0.7544721091099607, "eval_stsb_spearman": 0.8537162645564307, "step": 18625 }, { "epoch": 1.2, "eval_avg_sts": 0.8041380482719174, "eval_sickr_spearman": 0.7546320526777496, "eval_stsb_spearman": 0.8536440438660853, "step": 18750 }, { "epoch": 1.21, "eval_avg_sts": 0.8038765751251531, "eval_sickr_spearman": 0.7544602934590069, "eval_stsb_spearman": 0.8532928567912992, "step": 18875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.22, "learning_rate": 2.7442723665685396e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8165, "step": 19000 }, { "epoch": 1.22, "eval_avg_sts": 0.8066267695432612, "eval_sickr_spearman": 0.7576009030887407, "eval_stsb_spearman": 0.8556526359977817, "step": 19000 }, { "epoch": 1.22, "eval_avg_sts": 0.8079773606909346, "eval_sickr_spearman": 0.7579651709620474, "eval_stsb_spearman": 0.8579895504198217, "step": 19125 }, { "epoch": 1.23, "eval_avg_sts": 0.8076160011490612, "eval_sickr_spearman": 0.7573030141976216, "eval_stsb_spearman": 0.8579289881005008, "step": 19250 }, { "epoch": 1.24, "eval_avg_sts": 0.8076237482649742, "eval_sickr_spearman": 0.7578992242597729, "eval_stsb_spearman": 0.8573482722701754, "step": 19375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.25, "learning_rate": 2.632279534109817e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.817, "step": 19500 }, { "epoch": 1.25, "eval_avg_sts": 0.8074421455549705, "eval_sickr_spearman": 0.7572703530486437, "eval_stsb_spearman": 0.8576139380612975, "step": 19500 }, { "epoch": 1.26, "eval_avg_sts": 0.8072772138963991, "eval_sickr_spearman": 0.7571532532233376, "eval_stsb_spearman": 0.8574011745694605, "step": 19625 }, { "epoch": 1.26, "eval_avg_sts": 0.8073302425810744, "eval_sickr_spearman": 0.7571811592932731, "eval_stsb_spearman": 0.8574793258688757, "step": 19750 }, { "epoch": 1.27, "eval_avg_sts": 0.8072437293168806, "eval_sickr_spearman": 0.7571653090297986, "eval_stsb_spearman": 0.8573221496039627, "step": 19875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.28, "learning_rate": 2.5202867016510945e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.819, "step": 20000 }, { "epoch": 1.28, "eval_avg_sts": 0.8077588380658864, "eval_sickr_spearman": 0.7571885560828946, "eval_stsb_spearman": 0.8583291200488784, "step": 20000 }, { "epoch": 1.29, "eval_avg_sts": 0.8073697057890559, "eval_sickr_spearman": 0.7569122331563212, "eval_stsb_spearman": 0.8578271784217905, "step": 20125 }, { "epoch": 1.3, "eval_avg_sts": 0.8072112242255804, "eval_sickr_spearman": 0.7568430683702502, "eval_stsb_spearman": 0.8575793800809108, "step": 20250 }, { "epoch": 1.3, "eval_avg_sts": 0.8070283089413652, "eval_sickr_spearman": 0.7565875429105995, "eval_stsb_spearman": 0.8574690749721309, "step": 20375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.31, "learning_rate": 2.4082938691923718e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8202, "step": 20500 }, { "epoch": 1.31, "eval_avg_sts": 0.806707361091437, "eval_sickr_spearman": 0.7562538228178075, "eval_stsb_spearman": 0.8571608993650666, "step": 20500 }, { "epoch": 1.32, "eval_avg_sts": 0.8065118723390399, "eval_sickr_spearman": 0.7561714014477396, "eval_stsb_spearman": 0.85685234323034, "step": 20625 }, { "epoch": 1.33, "eval_avg_sts": 0.8067084472431667, "eval_sickr_spearman": 0.756075627431472, "eval_stsb_spearman": 0.8573412670548614, "step": 20750 }, { "epoch": 1.34, "eval_avg_sts": 0.8066874969794837, "eval_sickr_spearman": 0.7560982500802493, "eval_stsb_spearman": 0.857276743878718, "step": 20875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.34, "learning_rate": 2.296301036733649e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8217, "step": 21000 }, { "epoch": 1.34, "eval_avg_sts": 0.80597633722049, "eval_sickr_spearman": 0.7559074225142358, "eval_stsb_spearman": 0.8560452519267441, "step": 21000 }, { "epoch": 1.35, "eval_avg_sts": 0.8058007667522757, "eval_sickr_spearman": 0.7558859045807915, "eval_stsb_spearman": 0.8557156289237601, "step": 21125 }, { "epoch": 1.36, "eval_avg_sts": 0.8052139487149964, "eval_sickr_spearman": 0.7551075125508855, "eval_stsb_spearman": 0.8553203848791073, "step": 21250 }, { "epoch": 1.37, "eval_avg_sts": 0.8050337032021436, "eval_sickr_spearman": 0.7547609681540094, "eval_stsb_spearman": 0.8553064382502779, "step": 21375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.38, "learning_rate": 2.1843082042749267e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8233, "step": 21500 }, { "epoch": 1.38, "eval_avg_sts": 0.8048704441406708, "eval_sickr_spearman": 0.7537258018558153, "eval_stsb_spearman": 0.8560150864255263, "step": 21500 }, { "epoch": 1.38, "eval_avg_sts": 0.8048149155952563, "eval_sickr_spearman": 0.7537628338350241, "eval_stsb_spearman": 0.8558669973554885, "step": 21625 }, { "epoch": 1.39, "eval_avg_sts": 0.8049705872600981, "eval_sickr_spearman": 0.7530663348330701, "eval_stsb_spearman": 0.8568748396871261, "step": 21750 }, { "epoch": 1.4, "eval_avg_sts": 0.8048803781742058, "eval_sickr_spearman": 0.7529998117575781, "eval_stsb_spearman": 0.8567609445908335, "step": 21875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.41, "learning_rate": 2.072315371816204e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8232, "step": 22000 }, { "epoch": 1.41, "eval_avg_sts": 0.8048396134318049, "eval_sickr_spearman": 0.7529523570293574, "eval_stsb_spearman": 0.8567268698342524, "step": 22000 }, { "epoch": 1.42, "eval_avg_sts": 0.8052282014222374, "eval_sickr_spearman": 0.754343433789208, "eval_stsb_spearman": 0.8561129690552667, "step": 22125 }, { "epoch": 1.42, "eval_avg_sts": 0.8052272637457882, "eval_sickr_spearman": 0.7546001119952933, "eval_stsb_spearman": 0.855854415496283, "step": 22250 }, { "epoch": 1.43, "eval_avg_sts": 0.8050641056280918, "eval_sickr_spearman": 0.7546412266181242, "eval_stsb_spearman": 0.8554869846380595, "step": 22375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.44, "learning_rate": 1.9603225393574807e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8233, "step": 22500 }, { "epoch": 1.44, "eval_avg_sts": 0.8049566807526001, "eval_sickr_spearman": 0.7546373841300092, "eval_stsb_spearman": 0.8552759773751909, "step": 22500 }, { "epoch": 1.45, "eval_avg_sts": 0.8022531756326772, "eval_sickr_spearman": 0.7524902498024213, "eval_stsb_spearman": 0.8520161014629332, "step": 22625 }, { "epoch": 1.46, "eval_avg_sts": 0.80224163626539, "eval_sickr_spearman": 0.7525507689902333, "eval_stsb_spearman": 0.8519325035405467, "step": 22750 }, { "epoch": 1.46, "eval_avg_sts": 0.8021754313869522, "eval_sickr_spearman": 0.7525084055587649, "eval_stsb_spearman": 0.8518424572151397, "step": 22875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.47, "learning_rate": 1.8483297068987584e-06, "loss": 0.0015, "neg_sim": NaN, "pos_sim": 0.8241, "step": 23000 }, { "epoch": 1.47, "eval_avg_sts": 0.8024969817513419, "eval_sickr_spearman": 0.752065078492491, "eval_stsb_spearman": 0.8529288850101928, "step": 23000 }, { "epoch": 1.48, "eval_avg_sts": 0.8024605472785751, "eval_sickr_spearman": 0.7519645974282826, "eval_stsb_spearman": 0.8529564971288677, "step": 23125 }, { "epoch": 1.49, "eval_avg_sts": 0.8024320134196474, "eval_sickr_spearman": 0.751913444305251, "eval_stsb_spearman": 0.8529505825340439, "step": 23250 }, { "epoch": 1.5, "eval_avg_sts": 0.8024048624411102, "eval_sickr_spearman": 0.7518156049516216, "eval_stsb_spearman": 0.8529941199305989, "step": 23375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.5, "learning_rate": 1.7363368744400357e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8244, "step": 23500 }, { "epoch": 1.5, "eval_avg_sts": 0.802379657071767, "eval_sickr_spearman": 0.7517114254926023, "eval_stsb_spearman": 0.8530478886509316, "step": 23500 }, { "epoch": 1.51, "eval_avg_sts": 0.8022866209294395, "eval_sickr_spearman": 0.7517222805215273, "eval_stsb_spearman": 0.8528509613373516, "step": 23625 }, { "epoch": 1.52, "eval_avg_sts": 0.8029982809362839, "eval_sickr_spearman": 0.7519777099189752, "eval_stsb_spearman": 0.8540188519535926, "step": 23750 }, { "epoch": 1.53, "eval_avg_sts": 0.8029407406070306, "eval_sickr_spearman": 0.7520159907068213, "eval_stsb_spearman": 0.8538654905072398, "step": 23875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.54, "learning_rate": 1.6243440419813131e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8241, "step": 24000 }, { "epoch": 1.54, "eval_avg_sts": 0.8026597102203472, "eval_sickr_spearman": 0.7517222805215273, "eval_stsb_spearman": 0.8535971399191672, "step": 24000 }, { "epoch": 1.54, "eval_avg_sts": 0.8025235294672179, "eval_sickr_spearman": 0.7517699273741539, "eval_stsb_spearman": 0.8532771315602818, "step": 24125 }, { "epoch": 1.55, "eval_avg_sts": 0.8024910901785833, "eval_sickr_spearman": 0.7517942311114816, "eval_stsb_spearman": 0.8531879492456851, "step": 24250 }, { "epoch": 1.56, "eval_avg_sts": 0.8022646880892426, "eval_sickr_spearman": 0.7516456709147336, "eval_stsb_spearman": 0.8528837052637515, "step": 24375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.57, "learning_rate": 1.5123512095225906e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8242, "step": 24500 }, { "epoch": 1.57, "eval_avg_sts": 0.8021439174552474, "eval_sickr_spearman": 0.7515729518271562, "eval_stsb_spearman": 0.8527148830833385, "step": 24500 }, { "epoch": 1.58, "eval_avg_sts": 0.8023944530487281, "eval_sickr_spearman": 0.7522279999885693, "eval_stsb_spearman": 0.8525609061088869, "step": 24625 }, { "epoch": 1.58, "eval_avg_sts": 0.8023338277994898, "eval_sickr_spearman": 0.7522989419253933, "eval_stsb_spearman": 0.8523687136735865, "step": 24750 }, { "epoch": 1.59, "eval_avg_sts": 0.8021448888713669, "eval_sickr_spearman": 0.7524111906094542, "eval_stsb_spearman": 0.8518785871332797, "step": 24875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.6, "learning_rate": 1.4003583770638678e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.825, "step": 25000 }, { "epoch": 1.6, "eval_avg_sts": 0.8020245919486174, "eval_sickr_spearman": 0.7524232944470166, "eval_stsb_spearman": 0.8516258894502182, "step": 25000 }, { "epoch": 1.61, "eval_avg_sts": 0.8027255384898491, "eval_sickr_spearman": 0.753412110732323, "eval_stsb_spearman": 0.8520389662473753, "step": 25125 }, { "epoch": 1.62, "eval_avg_sts": 0.8025519191789285, "eval_sickr_spearman": 0.7524277133083489, "eval_stsb_spearman": 0.8526761250495083, "step": 25250 }, { "epoch": 1.62, "eval_avg_sts": 0.8020490496082332, "eval_sickr_spearman": 0.7522735334727325, "eval_stsb_spearman": 0.8518245657437339, "step": 25375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.63, "learning_rate": 1.2883655446051453e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8247, "step": 25500 }, { "epoch": 1.63, "eval_avg_sts": 0.8019816352585948, "eval_sickr_spearman": 0.7522344361561619, "eval_stsb_spearman": 0.8517288343610276, "step": 25500 }, { "epoch": 1.64, "eval_avg_sts": 0.8013792896117395, "eval_sickr_spearman": 0.7513900974239813, "eval_stsb_spearman": 0.8513684817994976, "step": 25625 }, { "epoch": 1.65, "eval_avg_sts": 0.8012957662635658, "eval_sickr_spearman": 0.7513326522266613, "eval_stsb_spearman": 0.8512588803004703, "step": 25750 }, { "epoch": 1.66, "eval_avg_sts": 0.8012601775924122, "eval_sickr_spearman": 0.751396101311661, "eval_stsb_spearman": 0.8511242538731633, "step": 25875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.66, "learning_rate": 1.1763727121464225e-06, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.825, "step": 26000 }, { "epoch": 1.66, "eval_avg_sts": 0.80119947875736, "eval_sickr_spearman": 0.7513846699095188, "eval_stsb_spearman": 0.8510142876052013, "step": 26000 }, { "epoch": 1.67, "eval_avg_sts": 0.8011593246270685, "eval_sickr_spearman": 0.751327464867706, "eval_stsb_spearman": 0.8509911843864311, "step": 26125 }, { "epoch": 1.68, "eval_avg_sts": 0.8011629421191884, "eval_sickr_spearman": 0.7513375033679065, "eval_stsb_spearman": 0.8509883808704702, "step": 26250 }, { "epoch": 1.69, "eval_avg_sts": 0.801390135441459, "eval_sickr_spearman": 0.7513261680279671, "eval_stsb_spearman": 0.8514541028549507, "step": 26375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.7, "learning_rate": 1.0643798796877e-06, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8251, "step": 26500 }, { "epoch": 1.7, "eval_avg_sts": 0.8011107242729139, "eval_sickr_spearman": 0.7510607481614202, "eval_stsb_spearman": 0.8511607003844076, "step": 26500 }, { "epoch": 1.7, "eval_avg_sts": 0.8017416851794723, "eval_sickr_spearman": 0.7518608022180748, "eval_stsb_spearman": 0.8516225681408698, "step": 26625 }, { "epoch": 1.71, "eval_avg_sts": 0.8012848966551385, "eval_sickr_spearman": 0.7518499471891499, "eval_stsb_spearman": 0.8507198461211269, "step": 26750 }, { "epoch": 1.72, "eval_avg_sts": 0.8018531908928693, "eval_sickr_spearman": 0.7526851600120571, "eval_stsb_spearman": 0.8510212217736816, "step": 26875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.73, "learning_rate": 9.523870472289774e-07, "loss": 0.0015, "neg_sim": NaN, "pos_sim": 0.8252, "step": 27000 }, { "epoch": 1.73, "eval_avg_sts": 0.8022264378652083, "eval_sickr_spearman": 0.7527636908629084, "eval_stsb_spearman": 0.8516891848675082, "step": 27000 }, { "epoch": 1.74, "eval_avg_sts": 0.8022941628228328, "eval_sickr_spearman": 0.7528165731055918, "eval_stsb_spearman": 0.851771752540074, "step": 27125 }, { "epoch": 1.74, "eval_avg_sts": 0.802330897941351, "eval_sickr_spearman": 0.7527780521622385, "eval_stsb_spearman": 0.8518837437204637, "step": 27250 }, { "epoch": 1.75, "eval_avg_sts": 0.8023123636818033, "eval_sickr_spearman": 0.75282886906756, "eval_stsb_spearman": 0.8517958582960464, "step": 27375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.76, "learning_rate": 8.403942147702547e-07, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8254, "step": 27500 }, { "epoch": 1.76, "eval_avg_sts": 0.8022805236271966, "eval_sickr_spearman": 0.7531723875050453, "eval_stsb_spearman": 0.8513886597493479, "step": 27500 }, { "epoch": 1.77, "eval_avg_sts": 0.8021820813348178, "eval_sickr_spearman": 0.7531067289893795, "eval_stsb_spearman": 0.8512574336802562, "step": 27625 }, { "epoch": 1.78, "eval_avg_sts": 0.8021045835263574, "eval_sickr_spearman": 0.7530147494301255, "eval_stsb_spearman": 0.8511944176225893, "step": 27750 }, { "epoch": 1.78, "eval_avg_sts": 0.8021931793671637, "eval_sickr_spearman": 0.7531494766696594, "eval_stsb_spearman": 0.8512368820646681, "step": 27875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.79, "learning_rate": 7.284013823115321e-07, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8249, "step": 28000 }, { "epoch": 1.79, "eval_avg_sts": 0.8019519952437764, "eval_sickr_spearman": 0.752846640575092, "eval_stsb_spearman": 0.8510573499124608, "step": 28000 }, { "epoch": 1.8, "eval_avg_sts": 0.8020597161190861, "eval_sickr_spearman": 0.7531391019517488, "eval_stsb_spearman": 0.8509803302864235, "step": 28125 }, { "epoch": 1.81, "eval_avg_sts": 0.8017160897849254, "eval_sickr_spearman": 0.7524687799000784, "eval_stsb_spearman": 0.8509633996697724, "step": 28250 }, { "epoch": 1.82, "eval_avg_sts": 0.8019391531942626, "eval_sickr_spearman": 0.7527104243714136, "eval_stsb_spearman": 0.8511678820171117, "step": 28375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.82, "learning_rate": 6.164085498528094e-07, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8248, "step": 28500 }, { "epoch": 1.82, "eval_avg_sts": 0.8019393193841213, "eval_sickr_spearman": 0.7527119133355582, "eval_stsb_spearman": 0.8511667254326846, "step": 28500 }, { "epoch": 1.83, "eval_avg_sts": 0.8019492124313456, "eval_sickr_spearman": 0.7531468829901817, "eval_stsb_spearman": 0.8507515418725097, "step": 28625 }, { "epoch": 1.84, "eval_avg_sts": 0.8018476761350394, "eval_sickr_spearman": 0.753012059688445, "eval_stsb_spearman": 0.8506832925816338, "step": 28750 }, { "epoch": 1.85, "eval_avg_sts": 0.8021637081346856, "eval_sickr_spearman": 0.7534332924480571, "eval_stsb_spearman": 0.8508941238213141, "step": 28875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.86, "learning_rate": 5.044157173940868e-07, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8253, "step": 29000 }, { "epoch": 1.86, "eval_avg_sts": 0.8021424201726441, "eval_sickr_spearman": 0.7534233980411609, "eval_stsb_spearman": 0.8508614423041275, "step": 29000 }, { "epoch": 1.86, "eval_avg_sts": 0.8021059378722697, "eval_sickr_spearman": 0.753409276897338, "eval_stsb_spearman": 0.8508025988472012, "step": 29125 }, { "epoch": 1.87, "eval_avg_sts": 0.802109297825347, "eval_sickr_spearman": 0.7534563954078488, "eval_stsb_spearman": 0.8507622002428452, "step": 29250 }, { "epoch": 1.88, "eval_avg_sts": 0.802140899314784, "eval_sickr_spearman": 0.7535362711295405, "eval_stsb_spearman": 0.8507455275000276, "step": 29375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.89, "learning_rate": 3.9242288493536413e-07, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8247, "step": 29500 }, { "epoch": 1.89, "eval_avg_sts": 0.8021494286648734, "eval_sickr_spearman": 0.7535442923234806, "eval_stsb_spearman": 0.850754565006266, "step": 29500 }, { "epoch": 1.9, "eval_avg_sts": 0.802139809482882, "eval_sickr_spearman": 0.7535292585887305, "eval_stsb_spearman": 0.8507503603770336, "step": 29625 }, { "epoch": 1.9, "eval_avg_sts": 0.8021667358363417, "eval_sickr_spearman": 0.7535602866802595, "eval_stsb_spearman": 0.8507731849924238, "step": 29750 }, { "epoch": 1.91, "eval_avg_sts": 0.802187610188689, "eval_sickr_spearman": 0.7535883368434994, "eval_stsb_spearman": 0.8507868835338787, "step": 29875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.92, "learning_rate": 2.804300524766415e-07, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8251, "step": 30000 }, { "epoch": 1.92, "eval_avg_sts": 0.802156902456721, "eval_sickr_spearman": 0.7535403537731628, "eval_stsb_spearman": 0.8507734511402791, "step": 30000 }, { "epoch": 1.93, "eval_avg_sts": 0.8024228445195005, "eval_sickr_spearman": 0.7538071665416515, "eval_stsb_spearman": 0.8510385224973495, "step": 30125 }, { "epoch": 1.94, "eval_avg_sts": 0.802384678105279, "eval_sickr_spearman": 0.753726522322337, "eval_stsb_spearman": 0.8510428338882212, "step": 30250 }, { "epoch": 1.94, "eval_avg_sts": 0.8023422930470364, "eval_sickr_spearman": 0.7536587504382077, "eval_stsb_spearman": 0.8510258356558651, "step": 30375 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.95, "learning_rate": 1.6843722001791884e-07, "loss": 0.0014, "neg_sim": NaN, "pos_sim": 0.8246, "step": 30500 }, { "epoch": 1.95, "eval_avg_sts": 0.8023481067748586, "eval_sickr_spearman": 0.7536835344865497, "eval_stsb_spearman": 0.8510126790631675, "step": 30500 }, { "epoch": 1.96, "eval_avg_sts": 0.8023432918069697, "eval_sickr_spearman": 0.7536910273383741, "eval_stsb_spearman": 0.8509955562755652, "step": 30625 }, { "epoch": 1.97, "eval_avg_sts": 0.8023505057059448, "eval_sickr_spearman": 0.7537099996234421, "eval_stsb_spearman": 0.8509910117884474, "step": 30750 }, { "epoch": 1.98, "eval_avg_sts": 0.8023412928143316, "eval_sickr_spearman": 0.7536968391016481, "eval_stsb_spearman": 0.8509857465270153, "step": 30875 }, { "electra_acc": NaN, "electra_fix_acc": NaN, "electra_rep_acc": NaN, "epoch": 1.98, "learning_rate": 5.64443875591962e-08, "loss": 0.0013, "neg_sim": NaN, "pos_sim": 0.8246, "step": 31000 }, { "epoch": 1.98, "eval_avg_sts": 0.8023353086079636, "eval_sickr_spearman": 0.7536697975915385, "eval_stsb_spearman": 0.8510008196243888, "step": 31000 }, { "epoch": 1.99, "eval_avg_sts": 0.8023565101430139, "eval_sickr_spearman": 0.7536807486826663, "eval_stsb_spearman": 0.8510322716033614, "step": 31125 }, { "epoch": 2.0, "eval_avg_sts": 0.802324478822219, "eval_sickr_spearman": 0.7536380010023865, "eval_stsb_spearman": 0.8510109566420515, "step": 31250 }, { "epoch": 2.0, "step": 31252, "train_runtime": 14615.2615, "train_samples_per_second": 2.138 } ], "max_steps": 31252, "num_train_epochs": 2, "total_flos": 603571104573277440, "trial_name": null, "trial_params": null }