| { | |
| "best_metric": 0.838238453177042, | |
| "best_model_checkpoint": "result/noextendneg-202112301439/unsup-53", | |
| "epoch": 1.0, | |
| "global_step": 15626, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "eval_avg_sts": 0.6824317233771873, | |
| "eval_sickr_spearman": 0.6565727646348289, | |
| "eval_stsb_spearman": 0.7082906821195456, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.7088615925204196, | |
| "eval_sickr_spearman": 0.6830176324024222, | |
| "eval_stsb_spearman": 0.7347055526384171, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.748889792906392, | |
| "eval_sickr_spearman": 0.7234262941045092, | |
| "eval_stsb_spearman": 0.7743532917082747, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.9040061436068092e-05, | |
| "loss": 1.4942, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7647833717747232, | |
| "eval_sickr_spearman": 0.736218225226116, | |
| "eval_stsb_spearman": 0.7933485183233304, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.7671547263733642, | |
| "eval_sickr_spearman": 0.7419857518557031, | |
| "eval_stsb_spearman": 0.7923237008910253, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7743718035568281, | |
| "eval_sickr_spearman": 0.7418156256944094, | |
| "eval_stsb_spearman": 0.8069279814192469, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.778288874388263, | |
| "eval_sickr_spearman": 0.7432141953060842, | |
| "eval_stsb_spearman": 0.8133635534704418, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.8080122872136184e-05, | |
| "loss": 1.3761, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.7820337312720824, | |
| "eval_sickr_spearman": 0.7496537250758921, | |
| "eval_stsb_spearman": 0.8144137374682728, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.7843418934590822, | |
| "eval_sickr_spearman": 0.74964997864998, | |
| "eval_stsb_spearman": 0.8190338082681845, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.780707868643423, | |
| "eval_sickr_spearman": 0.7451549399907927, | |
| "eval_stsb_spearman": 0.8162607972960534, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.7822364414457996, | |
| "eval_sickr_spearman": 0.7458706514333221, | |
| "eval_stsb_spearman": 0.818602231458277, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.7120184308204276e-05, | |
| "loss": 1.3752, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.784082685057127, | |
| "eval_sickr_spearman": 0.7478067851322924, | |
| "eval_stsb_spearman": 0.8203585849819615, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.7822808064244355, | |
| "eval_sickr_spearman": 0.7476906459290149, | |
| "eval_stsb_spearman": 0.8168709669198561, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.788703629999421, | |
| "eval_sickr_spearman": 0.7475929026375885, | |
| "eval_stsb_spearman": 0.8298143573612535, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.7896517154212715, | |
| "eval_sickr_spearman": 0.7467137893879666, | |
| "eval_stsb_spearman": 0.8325896414545765, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.6160245744272367e-05, | |
| "loss": 1.3749, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.7848601188664759, | |
| "eval_sickr_spearman": 0.7478141338908124, | |
| "eval_stsb_spearman": 0.8219061038421395, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.791993659168045, | |
| "eval_sickr_spearman": 0.7493434921917033, | |
| "eval_stsb_spearman": 0.8346438261443867, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.7912909275247073, | |
| "eval_sickr_spearman": 0.750289032454614, | |
| "eval_stsb_spearman": 0.8322928225948005, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.789035346084623, | |
| "eval_sickr_spearman": 0.7492525212855795, | |
| "eval_stsb_spearman": 0.8288181708836665, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.5200307180340456e-05, | |
| "loss": 1.3744, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.7856599152275197, | |
| "eval_sickr_spearman": 0.7458348202316492, | |
| "eval_stsb_spearman": 0.8254850102233903, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.7873291440307411, | |
| "eval_sickr_spearman": 0.7489932494000165, | |
| "eval_stsb_spearman": 0.8256650386614655, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7887021861862142, | |
| "eval_sickr_spearman": 0.7506508027106459, | |
| "eval_stsb_spearman": 0.8267535696617826, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.7908330013178573, | |
| "eval_sickr_spearman": 0.7434275494586724, | |
| "eval_stsb_spearman": 0.838238453177042, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.424036861640855e-05, | |
| "loss": 1.3743, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.7864405140650716, | |
| "eval_sickr_spearman": 0.7452477299492377, | |
| "eval_stsb_spearman": 0.8276332981809055, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.7856684955860707, | |
| "eval_sickr_spearman": 0.7435629971647278, | |
| "eval_stsb_spearman": 0.8277739940074136, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.7859315388383803, | |
| "eval_sickr_spearman": 0.7420500174694273, | |
| "eval_stsb_spearman": 0.8298130602073334, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7859735969509316, | |
| "eval_sickr_spearman": 0.7443584402356445, | |
| "eval_stsb_spearman": 0.8275887536662188, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3280430052476642e-05, | |
| "loss": 1.3743, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.7842904145859011, | |
| "eval_sickr_spearman": 0.745185343678003, | |
| "eval_stsb_spearman": 0.823395485493799, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7839231555379245, | |
| "eval_sickr_spearman": 0.744413964188907, | |
| "eval_stsb_spearman": 0.8234323468869418, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.7854707830375165, | |
| "eval_sickr_spearman": 0.745022758399635, | |
| "eval_stsb_spearman": 0.8259188076753979, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.7850613325633844, | |
| "eval_sickr_spearman": 0.7463711294077354, | |
| "eval_stsb_spearman": 0.8237515357190333, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2320491488544734e-05, | |
| "loss": 1.3742, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7830541412616541, | |
| "eval_sickr_spearman": 0.7444240507202089, | |
| "eval_stsb_spearman": 0.8216842318030994, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.7813602214602952, | |
| "eval_sickr_spearman": 0.741438773672526, | |
| "eval_stsb_spearman": 0.8212816692480643, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.7789870537555355, | |
| "eval_sickr_spearman": 0.7424057838377794, | |
| "eval_stsb_spearman": 0.8155683236732915, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.7813460307361557, | |
| "eval_sickr_spearman": 0.7435493563319195, | |
| "eval_stsb_spearman": 0.8191427051403918, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1360552924612826e-05, | |
| "loss": 1.3743, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.776201517183297, | |
| "eval_sickr_spearman": 0.7414799843575598, | |
| "eval_stsb_spearman": 0.8109230500090343, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.779063792601691, | |
| "eval_sickr_spearman": 0.7449220371799193, | |
| "eval_stsb_spearman": 0.8132055480234627, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.7771260741272044, | |
| "eval_sickr_spearman": 0.7445626204478579, | |
| "eval_stsb_spearman": 0.8096895278065507, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.7792767945158564, | |
| "eval_sickr_spearman": 0.7424322970057733, | |
| "eval_stsb_spearman": 0.8161212920259395, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0400614360680917e-05, | |
| "loss": 1.3741, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.7801777270254462, | |
| "eval_sickr_spearman": 0.745189954663741, | |
| "eval_stsb_spearman": 0.8151654993871514, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.7790507403010363, | |
| "eval_sickr_spearman": 0.7436056968139063, | |
| "eval_stsb_spearman": 0.8144957837881661, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7754481116510958, | |
| "eval_sickr_spearman": 0.7431048765192112, | |
| "eval_stsb_spearman": 0.8077913467829804, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.7759560400168242, | |
| "eval_sickr_spearman": 0.7460634002433932, | |
| "eval_stsb_spearman": 0.805848679790255, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.944067579674901e-05, | |
| "loss": 1.374, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.7779457482540455, | |
| "eval_sickr_spearman": 0.7447529677028571, | |
| "eval_stsb_spearman": 0.8111385288052337, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.7812654214930765, | |
| "eval_sickr_spearman": 0.7431360487040445, | |
| "eval_stsb_spearman": 0.8193947942821087, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.7791932156518167, | |
| "eval_sickr_spearman": 0.7427627029525661, | |
| "eval_stsb_spearman": 0.8156237283510671, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.781330663934428, | |
| "eval_sickr_spearman": 0.740891939582653, | |
| "eval_stsb_spearman": 0.8217693882862028, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.84807372328171e-05, | |
| "loss": 1.3743, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.7792436520419758, | |
| "eval_sickr_spearman": 0.7425573219628167, | |
| "eval_stsb_spearman": 0.8159299821211349, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.778491260905172, | |
| "eval_sickr_spearman": 0.7429476707242042, | |
| "eval_stsb_spearman": 0.8140348510861396, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.7781813540457969, | |
| "eval_sickr_spearman": 0.7442114650652438, | |
| "eval_stsb_spearman": 0.8121512430263501, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.7746842721533463, | |
| "eval_sickr_spearman": 0.7427369582821953, | |
| "eval_stsb_spearman": 0.8066315860244974, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.7520798668885192e-05, | |
| "loss": 1.3741, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.774494584762071, | |
| "eval_sickr_spearman": 0.7398474072196785, | |
| "eval_stsb_spearman": 0.8091417623044636, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.7766158847039688, | |
| "eval_sickr_spearman": 0.7414669679290701, | |
| "eval_stsb_spearman": 0.8117648014788674, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.7785386253067408, | |
| "eval_sickr_spearman": 0.7400865540737388, | |
| "eval_stsb_spearman": 0.8169906965397429, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.7773462713510697, | |
| "eval_sickr_spearman": 0.7418193240892201, | |
| "eval_stsb_spearman": 0.8128732186129194, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.6560860104953284e-05, | |
| "loss": 1.3739, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.7731374701085579, | |
| "eval_sickr_spearman": 0.7378116089852239, | |
| "eval_stsb_spearman": 0.808463331231892, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7740007626157097, | |
| "eval_sickr_spearman": 0.7376639133483018, | |
| "eval_stsb_spearman": 0.8103376118831176, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.7765981231805895, | |
| "eval_sickr_spearman": 0.7397111429848985, | |
| "eval_stsb_spearman": 0.8134851033762805, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.7743899061192407, | |
| "eval_sickr_spearman": 0.7414097148561558, | |
| "eval_stsb_spearman": 0.8073700973823258, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.5600921541021372e-05, | |
| "loss": 1.3744, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.7788123677105294, | |
| "eval_sickr_spearman": 0.7435954661893001, | |
| "eval_stsb_spearman": 0.8140292692317587, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.7772272696959476, | |
| "eval_sickr_spearman": 0.7404882381750656, | |
| "eval_stsb_spearman": 0.8139663012168296, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7794965841665269, | |
| "eval_sickr_spearman": 0.742402998033896, | |
| "eval_stsb_spearman": 0.8165901702991577, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7720147134204476, | |
| "eval_sickr_spearman": 0.739454224623306, | |
| "eval_stsb_spearman": 0.8045752022175892, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.4640982977089467e-05, | |
| "loss": 1.3744, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.7772065286971521, | |
| "eval_sickr_spearman": 0.7399364088506433, | |
| "eval_stsb_spearman": 0.814476648543661, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7732682543973268, | |
| "eval_sickr_spearman": 0.7381198245631523, | |
| "eval_stsb_spearman": 0.8084166842315015, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.7754551024695455, | |
| "eval_sickr_spearman": 0.739559796984267, | |
| "eval_stsb_spearman": 0.8113504079548242, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7752465045338841, | |
| "eval_sickr_spearman": 0.7382731878700443, | |
| "eval_stsb_spearman": 0.8122198211977241, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.368104441315756e-05, | |
| "loss": 1.374, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.7738482342971208, | |
| "eval_sickr_spearman": 0.7392923117803581, | |
| "eval_stsb_spearman": 0.8084041568138834, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.7798108521674825, | |
| "eval_sickr_spearman": 0.739234530365328, | |
| "eval_stsb_spearman": 0.820387173969637, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.7757746119221507, | |
| "eval_sickr_spearman": 0.7358846972577298, | |
| "eval_stsb_spearman": 0.8156645265865717, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.7754440529084836, | |
| "eval_sickr_spearman": 0.7365315801318982, | |
| "eval_stsb_spearman": 0.8143565256850689, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.2721105849225649e-05, | |
| "loss": 1.3738, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7752604709709456, | |
| "eval_sickr_spearman": 0.7375815400093355, | |
| "eval_stsb_spearman": 0.8129394019325558, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.7719041040932904, | |
| "eval_sickr_spearman": 0.7360112528588094, | |
| "eval_stsb_spearman": 0.8077969553277713, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.7740838243178477, | |
| "eval_sickr_spearman": 0.737044840481866, | |
| "eval_stsb_spearman": 0.8111228081538294, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7744233573614463, | |
| "eval_sickr_spearman": 0.7352095720959158, | |
| "eval_stsb_spearman": 0.8136371426269768, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.176116728529374e-05, | |
| "loss": 1.3739, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.7728152704473774, | |
| "eval_sickr_spearman": 0.7345371366757821, | |
| "eval_stsb_spearman": 0.8110934042189727, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7748003786183549, | |
| "eval_sickr_spearman": 0.736938499623282, | |
| "eval_stsb_spearman": 0.8126622576134279, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7736787434429957, | |
| "eval_sickr_spearman": 0.736373461745964, | |
| "eval_stsb_spearman": 0.8109840251400274, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.7738638985267913, | |
| "eval_sickr_spearman": 0.7368789890886002, | |
| "eval_stsb_spearman": 0.8108488079649824, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.0801228721361832e-05, | |
| "loss": 1.3741, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7745000944024674, | |
| "eval_sickr_spearman": 0.7357973767153152, | |
| "eval_stsb_spearman": 0.8132028120896195, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7755728981675435, | |
| "eval_sickr_spearman": 0.7363922419066262, | |
| "eval_stsb_spearman": 0.8147535544284608, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7759902317249895, | |
| "eval_sickr_spearman": 0.7375042579671216, | |
| "eval_stsb_spearman": 0.8144762054828574, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7738194667098928, | |
| "eval_sickr_spearman": 0.7379243379802992, | |
| "eval_stsb_spearman": 0.8097145954394862, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 9.841290157429926e-06, | |
| "loss": 1.3739, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7752023601085498, | |
| "eval_sickr_spearman": 0.7386433155377263, | |
| "eval_stsb_spearman": 0.8117614046793732, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.7748475183160914, | |
| "eval_sickr_spearman": 0.7387013371082636, | |
| "eval_stsb_spearman": 0.8109936995239194, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.7728917936913275, | |
| "eval_sickr_spearman": 0.736409533103144, | |
| "eval_stsb_spearman": 0.8093740542795109, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7725897625926305, | |
| "eval_sickr_spearman": 0.7348871393119617, | |
| "eval_stsb_spearman": 0.8102923858732992, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.881351593498018e-06, | |
| "loss": 1.3738, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7728014259850515, | |
| "eval_sickr_spearman": 0.733524785150771, | |
| "eval_stsb_spearman": 0.8120780668193319, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.7713921235193044, | |
| "eval_sickr_spearman": 0.7339205133955196, | |
| "eval_stsb_spearman": 0.8088637336430892, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.7728581212092439, | |
| "eval_sickr_spearman": 0.7340948182626384, | |
| "eval_stsb_spearman": 0.8116214241558494, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.7720358887305934, | |
| "eval_sickr_spearman": 0.7342980378528232, | |
| "eval_stsb_spearman": 0.8097737396083635, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.921413029566108e-06, | |
| "loss": 1.3737, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7695547454198304, | |
| "eval_sickr_spearman": 0.7319195377096078, | |
| "eval_stsb_spearman": 0.8071899531300528, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.7706681719236664, | |
| "eval_sickr_spearman": 0.732619896782625, | |
| "eval_stsb_spearman": 0.8087164470647079, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.7696699833641414, | |
| "eval_sickr_spearman": 0.7322283296607535, | |
| "eval_stsb_spearman": 0.8071116370675294, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.7655661802332044, | |
| "eval_sickr_spearman": 0.7308695473420087, | |
| "eval_stsb_spearman": 0.8002628131244002, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 6.961474465634199e-06, | |
| "loss": 1.3738, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.7683310753929219, | |
| "eval_sickr_spearman": 0.733969889367798, | |
| "eval_stsb_spearman": 0.8026922614180458, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7689113009272721, | |
| "eval_sickr_spearman": 0.7331248782001972, | |
| "eval_stsb_spearman": 0.8046977236543469, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.7727908372056179, | |
| "eval_sickr_spearman": 0.7345837748752785, | |
| "eval_stsb_spearman": 0.8109978995359574, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.7712377661783789, | |
| "eval_sickr_spearman": 0.7337775248065384, | |
| "eval_stsb_spearman": 0.8086980075502196, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 6.001535901702292e-06, | |
| "loss": 1.3737, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.772200684127089, | |
| "eval_sickr_spearman": 0.7341291605001669, | |
| "eval_stsb_spearman": 0.810272207754011, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.774531532169478, | |
| "eval_sickr_spearman": 0.7340682090324417, | |
| "eval_stsb_spearman": 0.8149948553065143, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7725061483445825, | |
| "eval_sickr_spearman": 0.7340744530756287, | |
| "eval_stsb_spearman": 0.8109378436135363, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.7714926649368181, | |
| "eval_sickr_spearman": 0.7341153275429525, | |
| "eval_stsb_spearman": 0.8088700023306838, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 5.0415973377703825e-06, | |
| "loss": 1.3738, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.7722886101322686, | |
| "eval_sickr_spearman": 0.7347338720672726, | |
| "eval_stsb_spearman": 0.8098433481972646, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7724511969333603, | |
| "eval_sickr_spearman": 0.7351566418221308, | |
| "eval_stsb_spearman": 0.8097457520445899, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.771434894841144, | |
| "eval_sickr_spearman": 0.7347215761053044, | |
| "eval_stsb_spearman": 0.8081482135769836, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7714665080668874, | |
| "eval_sickr_spearman": 0.7342130228032777, | |
| "eval_stsb_spearman": 0.8087199933304972, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.081658773838474e-06, | |
| "loss": 1.3737, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.7716889452949547, | |
| "eval_sickr_spearman": 0.733557350237546, | |
| "eval_stsb_spearman": 0.8098205403523633, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.7708524022909773, | |
| "eval_sickr_spearman": 0.7333273773238603, | |
| "eval_stsb_spearman": 0.8083774272580941, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.7695297711312417, | |
| "eval_sickr_spearman": 0.7321257832591831, | |
| "eval_stsb_spearman": 0.8069337590033002, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.7723285421418242, | |
| "eval_sickr_spearman": 0.7334592226973079, | |
| "eval_stsb_spearman": 0.8111978615863407, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 3.121720209906566e-06, | |
| "loss": 1.3739, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7712955547799614, | |
| "eval_sickr_spearman": 0.7343331966190758, | |
| "eval_stsb_spearman": 0.8082579129408469, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.7718610397493488, | |
| "eval_sickr_spearman": 0.7346661482142449, | |
| "eval_stsb_spearman": 0.8090559312844529, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.7723370584228957, | |
| "eval_sickr_spearman": 0.7339858837245768, | |
| "eval_stsb_spearman": 0.8106882331212146, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.7716753925020436, | |
| "eval_sickr_spearman": 0.7339269975942136, | |
| "eval_stsb_spearman": 0.8094237874098735, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.1617816459746575e-06, | |
| "loss": 1.3738, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.7714165422777011, | |
| "eval_sickr_spearman": 0.7332472614466615, | |
| "eval_stsb_spearman": 0.8095858231087406, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7723213232875585, | |
| "eval_sickr_spearman": 0.7332828524928272, | |
| "eval_stsb_spearman": 0.8113597940822898, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.7719368310828905, | |
| "eval_sickr_spearman": 0.7333558597670131, | |
| "eval_stsb_spearman": 0.8105178023987679, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.7712586994141075, | |
| "eval_sickr_spearman": 0.7330949067928999, | |
| "eval_stsb_spearman": 0.8094224920353151, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2018430820427492e-06, | |
| "loss": 1.3737, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.7710566218066379, | |
| "eval_sickr_spearman": 0.732867767714199, | |
| "eval_stsb_spearman": 0.8092454758990769, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.7710765276458553, | |
| "eval_sickr_spearman": 0.7330723321752239, | |
| "eval_stsb_spearman": 0.8090807231164866, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.771355472096863, | |
| "eval_sickr_spearman": 0.7332387119106056, | |
| "eval_stsb_spearman": 0.8094722322831204, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.7716261263484865, | |
| "eval_sickr_spearman": 0.7334210860027661, | |
| "eval_stsb_spearman": 0.8098311666942067, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.4190451811084087e-07, | |
| "loss": 1.3738, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.7715659350580353, | |
| "eval_sickr_spearman": 0.7333172123300392, | |
| "eval_stsb_spearman": 0.8098146577860313, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.7714923724726137, | |
| "eval_sickr_spearman": 0.7332129192091333, | |
| "eval_stsb_spearman": 0.8097718257360942, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 8616.861, | |
| "train_samples_per_second": 1.813 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 420412116908851200, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |