| { | |
| "best_metric": 0.866013676946666, | |
| "best_model_checkpoint": "result/esimcse-bert-large-uncased-0.1-bpe-0.32-160-bs64", | |
| "epoch": 1.0, | |
| "global_step": 15626, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.01, | |
| "eval_avg_sts": 0.5655787069891403, | |
| "eval_sickr_spearman": 0.5300969801413821, | |
| "eval_stsb_spearman": 0.6010604338368984, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.6326186877770161, | |
| "eval_sickr_spearman": 0.6116633005413822, | |
| "eval_stsb_spearman": 0.6535740750126502, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "eval_avg_sts": 0.6609133620656575, | |
| "eval_sickr_spearman": 0.6391030845441643, | |
| "eval_stsb_spearman": 0.6827236395871509, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 9.680020478689364e-06, | |
| "loss": 0.0123, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "eval_avg_sts": 0.7221979056484411, | |
| "eval_sickr_spearman": 0.6936728999765546, | |
| "eval_stsb_spearman": 0.7507229113203276, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "eval_avg_sts": 0.7554045800232995, | |
| "eval_sickr_spearman": 0.7318677121511561, | |
| "eval_stsb_spearman": 0.7789414478954428, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "eval_avg_sts": 0.7930988815026507, | |
| "eval_sickr_spearman": 0.761447137629702, | |
| "eval_stsb_spearman": 0.8247506253755994, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.799489871458167, | |
| "eval_sickr_spearman": 0.7621246163154866, | |
| "eval_stsb_spearman": 0.8368551266008474, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.36004095737873e-06, | |
| "loss": 0.001, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.8058576433295244, | |
| "eval_sickr_spearman": 0.7662108142392325, | |
| "eval_stsb_spearman": 0.8455044724198162, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "eval_avg_sts": 0.8085897051320876, | |
| "eval_sickr_spearman": 0.7775263653648347, | |
| "eval_stsb_spearman": 0.8396530448993406, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "eval_avg_sts": 0.812202173171146, | |
| "eval_sickr_spearman": 0.7778328037920097, | |
| "eval_stsb_spearman": 0.8465715425502822, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "eval_avg_sts": 0.8130000077739681, | |
| "eval_sickr_spearman": 0.7843073002036646, | |
| "eval_stsb_spearman": 0.8416927153442717, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 9.040061436068092e-06, | |
| "loss": 0.0007, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.8069831792350133, | |
| "eval_sickr_spearman": 0.784164936019002, | |
| "eval_stsb_spearman": 0.8298014224510247, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "eval_avg_sts": 0.8066347132600319, | |
| "eval_sickr_spearman": 0.7831622867764814, | |
| "eval_stsb_spearman": 0.8301071397435824, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "eval_avg_sts": 0.8020330411815588, | |
| "eval_sickr_spearman": 0.7803842639315037, | |
| "eval_stsb_spearman": 0.8236818184316139, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.8021778407421094, | |
| "eval_sickr_spearman": 0.774247762380669, | |
| "eval_stsb_spearman": 0.83010791910355, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 8.720081914757458e-06, | |
| "loss": 0.0005, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "eval_avg_sts": 0.8065341521992876, | |
| "eval_sickr_spearman": 0.7772098404063574, | |
| "eval_stsb_spearman": 0.8358584639922177, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.794240687869741, | |
| "eval_sickr_spearman": 0.7671064021566476, | |
| "eval_stsb_spearman": 0.8213749735828343, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "eval_avg_sts": 0.8051559486926301, | |
| "eval_sickr_spearman": 0.7794549101497777, | |
| "eval_stsb_spearman": 0.8308569872354824, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "eval_avg_sts": 0.8073887679947801, | |
| "eval_sickr_spearman": 0.7799384392479553, | |
| "eval_stsb_spearman": 0.834839096741605, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 8.400102393446819e-06, | |
| "loss": 0.0005, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "eval_avg_sts": 0.8160314140071933, | |
| "eval_sickr_spearman": 0.7852161447050764, | |
| "eval_stsb_spearman": 0.8468466833093101, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "eval_avg_sts": 0.8233471889455257, | |
| "eval_sickr_spearman": 0.78413381186527, | |
| "eval_stsb_spearman": 0.8625605660257815, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.8227729219450164, | |
| "eval_sickr_spearman": 0.782287832543699, | |
| "eval_stsb_spearman": 0.8632580113463336, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.8256728603497563, | |
| "eval_sickr_spearman": 0.7853320437528465, | |
| "eval_stsb_spearman": 0.866013676946666, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 8.080122872136184e-06, | |
| "loss": 0.0004, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "eval_avg_sts": 0.8234860002979157, | |
| "eval_sickr_spearman": 0.7870213456215266, | |
| "eval_stsb_spearman": 0.8599506549743049, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "eval_avg_sts": 0.8214365447461363, | |
| "eval_sickr_spearman": 0.7847184464319747, | |
| "eval_stsb_spearman": 0.8581546430602978, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "eval_avg_sts": 0.8174673028951771, | |
| "eval_sickr_spearman": 0.7841726690263334, | |
| "eval_stsb_spearman": 0.8507619367640208, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.8164292052372073, | |
| "eval_sickr_spearman": 0.7851938582740091, | |
| "eval_stsb_spearman": 0.8476645522004054, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 7.760143350825547e-06, | |
| "loss": 0.0005, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "eval_avg_sts": 0.8128229716545061, | |
| "eval_sickr_spearman": 0.7825479689890878, | |
| "eval_stsb_spearman": 0.8430979743199243, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_avg_sts": 0.7999160883647138, | |
| "eval_sickr_spearman": 0.7685948379591132, | |
| "eval_stsb_spearman": 0.8312373387703145, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.79444504891346, | |
| "eval_sickr_spearman": 0.7639066662410434, | |
| "eval_stsb_spearman": 0.8249834315858765, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "eval_avg_sts": 0.8041035838235049, | |
| "eval_sickr_spearman": 0.7758875922148677, | |
| "eval_stsb_spearman": 0.8323195754321422, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 7.440163829514912e-06, | |
| "loss": 0.0006, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.8147672333502036, | |
| "eval_sickr_spearman": 0.7849670073819169, | |
| "eval_stsb_spearman": 0.8445674593184904, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "eval_avg_sts": 0.8193365795354237, | |
| "eval_sickr_spearman": 0.786982584522666, | |
| "eval_stsb_spearman": 0.8516905745481813, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "eval_avg_sts": 0.8129220138792039, | |
| "eval_sickr_spearman": 0.7803344076482109, | |
| "eval_stsb_spearman": 0.8455096201101969, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "eval_avg_sts": 0.814173563113507, | |
| "eval_sickr_spearman": 0.7786497647963706, | |
| "eval_stsb_spearman": 0.8496973614306432, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 7.120184308204276e-06, | |
| "loss": 0.0007, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_avg_sts": 0.8204432531176182, | |
| "eval_sickr_spearman": 0.7869734105822913, | |
| "eval_stsb_spearman": 0.853913095652945, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.8175076926619402, | |
| "eval_sickr_spearman": 0.7864348858729672, | |
| "eval_stsb_spearman": 0.8485804994509132, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "eval_avg_sts": 0.8128858173589497, | |
| "eval_sickr_spearman": 0.7823268818291682, | |
| "eval_stsb_spearman": 0.8434447528887313, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.8146504899933802, | |
| "eval_sickr_spearman": 0.7820177056292111, | |
| "eval_stsb_spearman": 0.8472832743575491, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 6.800204786893639e-06, | |
| "loss": 0.0003, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "eval_avg_sts": 0.818213782839917, | |
| "eval_sickr_spearman": 0.7825289006418169, | |
| "eval_stsb_spearman": 0.853898665038017, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "eval_avg_sts": 0.8159978701613437, | |
| "eval_sickr_spearman": 0.7774106584414702, | |
| "eval_stsb_spearman": 0.8545850818812172, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.8016384023594931, | |
| "eval_sickr_spearman": 0.7594139330747252, | |
| "eval_stsb_spearman": 0.8438628716442612, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "eval_avg_sts": 0.801479733175446, | |
| "eval_sickr_spearman": 0.7607142310528577, | |
| "eval_stsb_spearman": 0.8422452352980342, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 6.480225265583003e-06, | |
| "loss": 0.0004, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_avg_sts": 0.8042683855550379, | |
| "eval_sickr_spearman": 0.7645871228551171, | |
| "eval_stsb_spearman": 0.8439496482549587, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "eval_avg_sts": 0.8012942192582326, | |
| "eval_sickr_spearman": 0.7631563243743772, | |
| "eval_stsb_spearman": 0.839432114142088, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.8032207475650094, | |
| "eval_sickr_spearman": 0.767318363407294, | |
| "eval_stsb_spearman": 0.8391231317227248, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.8026675474584128, | |
| "eval_sickr_spearman": 0.769569437038394, | |
| "eval_stsb_spearman": 0.8357656578784318, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 6.1602457442723675e-06, | |
| "loss": 0.0004, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "eval_avg_sts": 0.8027527543269901, | |
| "eval_sickr_spearman": 0.7721073524072823, | |
| "eval_stsb_spearman": 0.833398156246698, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "eval_avg_sts": 0.8080890156965373, | |
| "eval_sickr_spearman": 0.7777212755744705, | |
| "eval_stsb_spearman": 0.8384567558186042, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "eval_avg_sts": 0.8099187475300627, | |
| "eval_sickr_spearman": 0.7792697982848352, | |
| "eval_stsb_spearman": 0.8405676967752902, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_avg_sts": 0.8109755257224806, | |
| "eval_sickr_spearman": 0.779351547219483, | |
| "eval_stsb_spearman": 0.8425995042254781, | |
| "step": 6375 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 5.840266222961732e-06, | |
| "loss": 0.0004, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.807811991532841, | |
| "eval_sickr_spearman": 0.7731304628990157, | |
| "eval_stsb_spearman": 0.8424935201666665, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "eval_avg_sts": 0.8015312813914837, | |
| "eval_sickr_spearman": 0.7648701701358919, | |
| "eval_stsb_spearman": 0.8381923926470756, | |
| "step": 6625 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.8023491493628093, | |
| "eval_sickr_spearman": 0.7662689799030741, | |
| "eval_stsb_spearman": 0.8384293188225446, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "eval_avg_sts": 0.8032060815338051, | |
| "eval_sickr_spearman": 0.7652516331435133, | |
| "eval_stsb_spearman": 0.8411605299240971, | |
| "step": 6875 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 5.520286701651095e-06, | |
| "loss": 0.0003, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "eval_avg_sts": 0.8060138329601017, | |
| "eval_sickr_spearman": 0.7675326781819108, | |
| "eval_stsb_spearman": 0.8444949877382923, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.8073011071530847, | |
| "eval_sickr_spearman": 0.7807883495879026, | |
| "eval_stsb_spearman": 0.8338138647182667, | |
| "step": 7125 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "eval_avg_sts": 0.8026818710000232, | |
| "eval_sickr_spearman": 0.774955644753664, | |
| "eval_stsb_spearman": 0.8304080972463824, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_avg_sts": 0.8035394838977947, | |
| "eval_sickr_spearman": 0.7752477218815091, | |
| "eval_stsb_spearman": 0.8318312459140802, | |
| "step": 7375 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 5.200307180340458e-06, | |
| "loss": 0.0005, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "eval_avg_sts": 0.8041748456008513, | |
| "eval_sickr_spearman": 0.7710755002550874, | |
| "eval_stsb_spearman": 0.8372741909466153, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.8031490475132864, | |
| "eval_sickr_spearman": 0.7735819072214325, | |
| "eval_stsb_spearman": 0.8327161878051403, | |
| "step": 7625 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.796365553979756, | |
| "eval_sickr_spearman": 0.7677243703077503, | |
| "eval_stsb_spearman": 0.8250067376517616, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "eval_avg_sts": 0.7985580022948444, | |
| "eval_sickr_spearman": 0.7685888821025348, | |
| "eval_stsb_spearman": 0.8285271224871539, | |
| "step": 7875 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 4.8803276590298225e-06, | |
| "loss": 0.0004, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "eval_avg_sts": 0.8005935975506946, | |
| "eval_sickr_spearman": 0.771644332589419, | |
| "eval_stsb_spearman": 0.8295428625119703, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "eval_avg_sts": 0.7989269976081819, | |
| "eval_sickr_spearman": 0.7673743676715707, | |
| "eval_stsb_spearman": 0.830479627544793, | |
| "step": 8125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_avg_sts": 0.8016572067178073, | |
| "eval_sickr_spearman": 0.7698353852470569, | |
| "eval_stsb_spearman": 0.8334790281885577, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.8076839071350832, | |
| "eval_sickr_spearman": 0.7736355099306375, | |
| "eval_stsb_spearman": 0.8417323043395288, | |
| "step": 8375 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 4.560348137719187e-06, | |
| "loss": 0.0003, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "eval_avg_sts": 0.808135722801363, | |
| "eval_sickr_spearman": 0.7724129262746318, | |
| "eval_stsb_spearman": 0.8438585193280942, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.8053092989350472, | |
| "eval_sickr_spearman": 0.770703355281145, | |
| "eval_stsb_spearman": 0.8399152425889496, | |
| "step": 8625 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "eval_avg_sts": 0.8084270733769378, | |
| "eval_sickr_spearman": 0.7738568852771656, | |
| "eval_stsb_spearman": 0.84299726147671, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "eval_avg_sts": 0.8026266771856059, | |
| "eval_sickr_spearman": 0.7651564355004631, | |
| "eval_stsb_spearman": 0.8400969188707486, | |
| "step": 8875 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 4.24036861640855e-06, | |
| "loss": 0.0004, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.8051260109076794, | |
| "eval_sickr_spearman": 0.7693196272798144, | |
| "eval_stsb_spearman": 0.8409323945355446, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "eval_avg_sts": 0.8086459894042075, | |
| "eval_sickr_spearman": 0.7755324502508343, | |
| "eval_stsb_spearman": 0.8417595285575807, | |
| "step": 9125 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_avg_sts": 0.8042063963931729, | |
| "eval_sickr_spearman": 0.7749119844824567, | |
| "eval_stsb_spearman": 0.8335008083038891, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "eval_avg_sts": 0.7931093622523872, | |
| "eval_sickr_spearman": 0.7642648821955689, | |
| "eval_stsb_spearman": 0.8219538423092054, | |
| "step": 9375 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 3.920389095097914e-06, | |
| "loss": 0.0005, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.790343225626717, | |
| "eval_sickr_spearman": 0.7611612084828409, | |
| "eval_stsb_spearman": 0.8195252427705931, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7899767423661693, | |
| "eval_sickr_spearman": 0.7618361895513507, | |
| "eval_stsb_spearman": 0.818117295180988, | |
| "step": 9625 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "eval_avg_sts": 0.7900723613141147, | |
| "eval_sickr_spearman": 0.7621474310886697, | |
| "eval_stsb_spearman": 0.8179972915395599, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "eval_avg_sts": 0.7944895699959831, | |
| "eval_sickr_spearman": 0.7671481411837973, | |
| "eval_stsb_spearman": 0.8218309988081689, | |
| "step": 9875 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 3.600409573787278e-06, | |
| "loss": 0.0003, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_avg_sts": 0.7947303624992417, | |
| "eval_sickr_spearman": 0.7661302660821208, | |
| "eval_stsb_spearman": 0.8233304589163626, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "eval_avg_sts": 0.7953635000035459, | |
| "eval_sickr_spearman": 0.7651970217811783, | |
| "eval_stsb_spearman": 0.8255299782259135, | |
| "step": 10125 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7981958931316387, | |
| "eval_sickr_spearman": 0.7667117305961306, | |
| "eval_stsb_spearman": 0.8296800556671469, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "eval_avg_sts": 0.7955967572281076, | |
| "eval_sickr_spearman": 0.7647861637394766, | |
| "eval_stsb_spearman": 0.8264073507167387, | |
| "step": 10375 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 3.280430052476642e-06, | |
| "loss": 0.0005, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.7904469016234321, | |
| "eval_sickr_spearman": 0.7560659251489815, | |
| "eval_stsb_spearman": 0.8248278780978826, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "eval_avg_sts": 0.7955565138560285, | |
| "eval_sickr_spearman": 0.7594871324733168, | |
| "eval_stsb_spearman": 0.8316258952387402, | |
| "step": 10625 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "eval_avg_sts": 0.794608835583937, | |
| "eval_sickr_spearman": 0.7577217973710605, | |
| "eval_stsb_spearman": 0.8314958737968136, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.7984620732119101, | |
| "eval_sickr_spearman": 0.7611833988517053, | |
| "eval_stsb_spearman": 0.8357407475721149, | |
| "step": 10875 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 2.960450531166006e-06, | |
| "loss": 0.0004, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_avg_sts": 0.8001810968684413, | |
| "eval_sickr_spearman": 0.7626987801020777, | |
| "eval_stsb_spearman": 0.8376634136348049, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "eval_avg_sts": 0.8009359528786534, | |
| "eval_sickr_spearman": 0.7635318795565218, | |
| "eval_stsb_spearman": 0.838340026200785, | |
| "step": 11125 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "eval_avg_sts": 0.8018165626143542, | |
| "eval_sickr_spearman": 0.761889167856237, | |
| "eval_stsb_spearman": 0.8417439573724714, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.8019628942005317, | |
| "eval_sickr_spearman": 0.7621596790195363, | |
| "eval_stsb_spearman": 0.8417661093815271, | |
| "step": 11375 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 2.640471009855369e-06, | |
| "loss": 0.0004, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.8023474777679851, | |
| "eval_sickr_spearman": 0.7618140952446891, | |
| "eval_stsb_spearman": 0.8428808602912811, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "eval_avg_sts": 0.8007664692491112, | |
| "eval_sickr_spearman": 0.7609206206957373, | |
| "eval_stsb_spearman": 0.8406123178024852, | |
| "step": 11625 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "eval_avg_sts": 0.8013536551112785, | |
| "eval_sickr_spearman": 0.763201473609729, | |
| "eval_stsb_spearman": 0.8395058366128281, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_avg_sts": 0.8016071248722578, | |
| "eval_sickr_spearman": 0.7642604633342366, | |
| "eval_stsb_spearman": 0.838953786410279, | |
| "step": 11875 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 2.3204914885447333e-06, | |
| "loss": 0.0006, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "eval_avg_sts": 0.8045261324324948, | |
| "eval_sickr_spearman": 0.7673434836733461, | |
| "eval_stsb_spearman": 0.8417087811916436, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.8040455198533101, | |
| "eval_sickr_spearman": 0.7673006399308633, | |
| "eval_stsb_spearman": 0.840790399775757, | |
| "step": 12125 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "eval_avg_sts": 0.8055135790258787, | |
| "eval_sickr_spearman": 0.7688476256459821, | |
| "eval_stsb_spearman": 0.8421795324057754, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "eval_avg_sts": 0.8056301407246984, | |
| "eval_sickr_spearman": 0.7684332613338755, | |
| "eval_stsb_spearman": 0.8428270201155214, | |
| "step": 12375 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.000511967234097e-06, | |
| "loss": 0.0005, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.8073712082258814, | |
| "eval_sickr_spearman": 0.7699142042845167, | |
| "eval_stsb_spearman": 0.844828212167246, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "eval_avg_sts": 0.8056455420795055, | |
| "eval_sickr_spearman": 0.769138694120697, | |
| "eval_stsb_spearman": 0.8421523900383138, | |
| "step": 12625 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.8035991703759889, | |
| "eval_sickr_spearman": 0.7667966976145747, | |
| "eval_stsb_spearman": 0.8404016431374032, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_avg_sts": 0.8002085597346069, | |
| "eval_sickr_spearman": 0.7639235731887497, | |
| "eval_stsb_spearman": 0.8364935462804641, | |
| "step": 12875 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.6805324459234608e-06, | |
| "loss": 0.0005, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "eval_avg_sts": 0.8015160073860743, | |
| "eval_sickr_spearman": 0.7661993348059888, | |
| "eval_stsb_spearman": 0.8368326799661597, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "eval_avg_sts": 0.7993655095996954, | |
| "eval_sickr_spearman": 0.7635688154735277, | |
| "eval_stsb_spearman": 0.8351622037258633, | |
| "step": 13125 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "eval_avg_sts": 0.8001347328783066, | |
| "eval_sickr_spearman": 0.7649028312848698, | |
| "eval_stsb_spearman": 0.8353666344717434, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.8014690463126566, | |
| "eval_sickr_spearman": 0.7669867566829652, | |
| "eval_stsb_spearman": 0.835951335942348, | |
| "step": 13375 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3605529246128248e-06, | |
| "loss": 0.0006, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.801102067772123, | |
| "eval_sickr_spearman": 0.7658747406224701, | |
| "eval_stsb_spearman": 0.8363293949217758, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "eval_avg_sts": 0.8007868057264769, | |
| "eval_sickr_spearman": 0.7664748892349392, | |
| "eval_stsb_spearman": 0.8350987222180144, | |
| "step": 13625 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_avg_sts": 0.8005208288353287, | |
| "eval_sickr_spearman": 0.7666100487543861, | |
| "eval_stsb_spearman": 0.8344316089162714, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "eval_avg_sts": 0.8032898804251193, | |
| "eval_sickr_spearman": 0.7685208220317972, | |
| "eval_stsb_spearman": 0.8380589388184414, | |
| "step": 13875 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0405734033021888e-06, | |
| "loss": 0.0005, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.8039134264299834, | |
| "eval_sickr_spearman": 0.7689818725745016, | |
| "eval_stsb_spearman": 0.8388449802854653, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "eval_avg_sts": 0.8045052668715511, | |
| "eval_sickr_spearman": 0.7691046881008788, | |
| "eval_stsb_spearman": 0.8399058456422233, | |
| "step": 14125 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "eval_avg_sts": 0.806178393426559, | |
| "eval_sickr_spearman": 0.7702171844723884, | |
| "eval_stsb_spearman": 0.8421396023807296, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.8065048225721209, | |
| "eval_sickr_spearman": 0.7704268882612672, | |
| "eval_stsb_spearman": 0.8425827568829746, | |
| "step": 14375 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 7.205938819915525e-07, | |
| "loss": 0.0003, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "eval_avg_sts": 0.8069821583081658, | |
| "eval_sickr_spearman": 0.7708896198925219, | |
| "eval_stsb_spearman": 0.8430746967238096, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.8077968468485142, | |
| "eval_sickr_spearman": 0.7718276192725069, | |
| "eval_stsb_spearman": 0.8437660744245216, | |
| "step": 14625 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_avg_sts": 0.8079862371583421, | |
| "eval_sickr_spearman": 0.772007111498581, | |
| "eval_stsb_spearman": 0.8439653628181031, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "eval_avg_sts": 0.8081309498929534, | |
| "eval_sickr_spearman": 0.7723302647490567, | |
| "eval_stsb_spearman": 0.8439316350368501, | |
| "step": 14875 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.0061436068091647e-07, | |
| "loss": 0.0006, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "eval_avg_sts": 0.8084974176098827, | |
| "eval_sickr_spearman": 0.7726533699684309, | |
| "eval_stsb_spearman": 0.8443414652513344, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "eval_avg_sts": 0.8085496480083677, | |
| "eval_sickr_spearman": 0.7727022175985934, | |
| "eval_stsb_spearman": 0.844397078418142, | |
| "step": 15125 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.8085722899568727, | |
| "eval_sickr_spearman": 0.7729186457416736, | |
| "eval_stsb_spearman": 0.8442259341720718, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.8085289743109576, | |
| "eval_sickr_spearman": 0.77284338100572, | |
| "eval_stsb_spearman": 0.8442145676161951, | |
| "step": 15375 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.06348393702803e-08, | |
| "loss": 0.0005, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_avg_sts": 0.8085319187853082, | |
| "eval_sickr_spearman": 0.7728611044821507, | |
| "eval_stsb_spearman": 0.8442027330884655, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_avg_sts": 0.8084971429010983, | |
| "eval_sickr_spearman": 0.7727954939975863, | |
| "eval_stsb_spearman": 0.8441987918046104, | |
| "step": 15625 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 15626, | |
| "train_runtime": 11240.3589, | |
| "train_samples_per_second": 1.39 | |
| } | |
| ], | |
| "max_steps": 15626, | |
| "num_train_epochs": 1, | |
| "total_flos": 329269573742764032, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |