| { | |
| "best_metric": 0.8754395915831782, | |
| "best_model_checkpoint": "v3v2dqwsyn8103.san-roberta-large", | |
| "epoch": 3.0, | |
| "global_step": 6126, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "eval_avg_sts": 0.8149644786494956, | |
| "eval_sickr_spearman": 0.7907241592935935, | |
| "eval_stsb_spearman": 0.8392047980053978, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_avg_sts": 0.8388577709591086, | |
| "eval_sickr_spearman": 0.8233415039069986, | |
| "eval_stsb_spearman": 0.8543740380112187, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_avg_sts": 0.8396589199542435, | |
| "eval_sickr_spearman": 0.8129205800767828, | |
| "eval_stsb_spearman": 0.8663972598317042, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 9.183806725432584e-06, | |
| "loss": 0.5341, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "eval_avg_sts": 0.8363342200594388, | |
| "eval_sickr_spearman": 0.8093077766888108, | |
| "eval_stsb_spearman": 0.8633606634300668, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "eval_avg_sts": 0.8338329341909468, | |
| "eval_sickr_spearman": 0.8055060189477813, | |
| "eval_stsb_spearman": 0.8621598494341124, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "eval_avg_sts": 0.839609299977649, | |
| "eval_sickr_spearman": 0.8169147023790718, | |
| "eval_stsb_spearman": 0.8623038975762263, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "eval_avg_sts": 0.846074071178897, | |
| "eval_sickr_spearman": 0.8259940214839182, | |
| "eval_stsb_spearman": 0.8661541208738758, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 8.367613450865165e-06, | |
| "loss": 0.1279, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "eval_avg_sts": 0.8453144553836196, | |
| "eval_sickr_spearman": 0.8210699209955844, | |
| "eval_stsb_spearman": 0.8695589897716549, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "eval_avg_sts": 0.845320179475549, | |
| "eval_sickr_spearman": 0.8282974490155858, | |
| "eval_stsb_spearman": 0.8623429099355123, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_avg_sts": 0.8460474498271838, | |
| "eval_sickr_spearman": 0.826672508822833, | |
| "eval_stsb_spearman": 0.8654223908315346, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "eval_avg_sts": 0.8426410233010573, | |
| "eval_sickr_spearman": 0.8190030466384995, | |
| "eval_stsb_spearman": 0.8662789999636149, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.551420176297748e-06, | |
| "loss": 0.1108, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "eval_avg_sts": 0.8549386812398281, | |
| "eval_sickr_spearman": 0.8434020296403363, | |
| "eval_stsb_spearman": 0.8664753328393199, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "eval_avg_sts": 0.8501103601570379, | |
| "eval_sickr_spearman": 0.834787891752918, | |
| "eval_stsb_spearman": 0.8654328285611577, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "eval_avg_sts": 0.849271502533093, | |
| "eval_sickr_spearman": 0.8330568988881898, | |
| "eval_stsb_spearman": 0.8654861061779964, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "eval_avg_sts": 0.8536431760480087, | |
| "eval_sickr_spearman": 0.8382815780093236, | |
| "eval_stsb_spearman": 0.8690047740866937, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 6.735226901730331e-06, | |
| "loss": 0.0995, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "eval_avg_sts": 0.859307976655676, | |
| "eval_sickr_spearman": 0.8446237487365161, | |
| "eval_stsb_spearman": 0.873992204574836, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "eval_avg_sts": 0.8539432294379701, | |
| "eval_sickr_spearman": 0.839737448725014, | |
| "eval_stsb_spearman": 0.868149010150926, | |
| "step": 2125 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "eval_avg_sts": 0.8550783980122398, | |
| "eval_sickr_spearman": 0.8398799570029809, | |
| "eval_stsb_spearman": 0.8702768390214988, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "eval_avg_sts": 0.8513744867439872, | |
| "eval_sickr_spearman": 0.8382596758270676, | |
| "eval_stsb_spearman": 0.8644892976609067, | |
| "step": 2375 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 5.919033627162912e-06, | |
| "loss": 0.0755, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "eval_avg_sts": 0.852586624774087, | |
| "eval_sickr_spearman": 0.8348869799151848, | |
| "eval_stsb_spearman": 0.8702862696329892, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "eval_avg_sts": 0.8536153767856529, | |
| "eval_sickr_spearman": 0.8408841432407476, | |
| "eval_stsb_spearman": 0.8663466103305583, | |
| "step": 2625 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "eval_avg_sts": 0.8516843556326315, | |
| "eval_sickr_spearman": 0.8374268645592322, | |
| "eval_stsb_spearman": 0.8659418467060308, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "eval_avg_sts": 0.8415420172889546, | |
| "eval_sickr_spearman": 0.819215344106856, | |
| "eval_stsb_spearman": 0.8638686904710532, | |
| "step": 2875 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 5.102840352595495e-06, | |
| "loss": 0.0701, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "eval_avg_sts": 0.8561622122931456, | |
| "eval_sickr_spearman": 0.8410006666928366, | |
| "eval_stsb_spearman": 0.8713237578934547, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "eval_avg_sts": 0.8482745103445485, | |
| "eval_sickr_spearman": 0.8300929476495453, | |
| "eval_stsb_spearman": 0.8664560730395516, | |
| "step": 3125 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "eval_avg_sts": 0.8509009129630769, | |
| "eval_sickr_spearman": 0.8311688923528606, | |
| "eval_stsb_spearman": 0.8706329335732933, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "eval_avg_sts": 0.8546193575409455, | |
| "eval_sickr_spearman": 0.833799123498713, | |
| "eval_stsb_spearman": 0.8754395915831782, | |
| "step": 3375 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 4.286647078028077e-06, | |
| "loss": 0.0691, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "eval_avg_sts": 0.8480763616470857, | |
| "eval_sickr_spearman": 0.8280656989511468, | |
| "eval_stsb_spearman": 0.8680870243430245, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "eval_avg_sts": 0.8510804038712448, | |
| "eval_sickr_spearman": 0.8315828243850542, | |
| "eval_stsb_spearman": 0.8705779833574355, | |
| "step": 3625 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "eval_avg_sts": 0.8481254894982377, | |
| "eval_sickr_spearman": 0.8284941363759749, | |
| "eval_stsb_spearman": 0.8677568426205006, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "eval_avg_sts": 0.8518486456802068, | |
| "eval_sickr_spearman": 0.8313096234800742, | |
| "eval_stsb_spearman": 0.8723876678803395, | |
| "step": 3875 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 3.47045380346066e-06, | |
| "loss": 0.0639, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "eval_avg_sts": 0.8494848749262629, | |
| "eval_sickr_spearman": 0.8275968673700096, | |
| "eval_stsb_spearman": 0.8713728824825163, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "eval_avg_sts": 0.8478683507440115, | |
| "eval_sickr_spearman": 0.8257009357029429, | |
| "eval_stsb_spearman": 0.87003576578508, | |
| "step": 4125 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "eval_avg_sts": 0.8511953137011732, | |
| "eval_sickr_spearman": 0.8320414253415853, | |
| "eval_stsb_spearman": 0.870349202060761, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "eval_avg_sts": 0.8501760957050806, | |
| "eval_sickr_spearman": 0.8315510758270036, | |
| "eval_stsb_spearman": 0.8688011155831574, | |
| "step": 4375 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 2.654260528893242e-06, | |
| "loss": 0.0525, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "eval_avg_sts": 0.8496234541718779, | |
| "eval_sickr_spearman": 0.8301971271085646, | |
| "eval_stsb_spearman": 0.8690497812351913, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "eval_avg_sts": 0.8437167774343648, | |
| "eval_sickr_spearman": 0.820443979681643, | |
| "eval_stsb_spearman": 0.8669895751870865, | |
| "step": 4625 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "eval_avg_sts": 0.8439814114677218, | |
| "eval_sickr_spearman": 0.8205907627276378, | |
| "eval_stsb_spearman": 0.8673720602078059, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "eval_avg_sts": 0.848080574777744, | |
| "eval_sickr_spearman": 0.8265265903366641, | |
| "eval_stsb_spearman": 0.869634559218824, | |
| "step": 4875 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 1.8380672543258246e-06, | |
| "loss": 0.0462, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "eval_avg_sts": 0.8452623360298752, | |
| "eval_sickr_spearman": 0.8238323337325946, | |
| "eval_stsb_spearman": 0.866692338327156, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "eval_avg_sts": 0.847977497396065, | |
| "eval_sickr_spearman": 0.8265478681146011, | |
| "eval_stsb_spearman": 0.8694071266775288, | |
| "step": 5125 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "eval_avg_sts": 0.8479098375907992, | |
| "eval_sickr_spearman": 0.826889897587942, | |
| "eval_stsb_spearman": 0.8689297775936565, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "eval_avg_sts": 0.8457732999493977, | |
| "eval_sickr_spearman": 0.8235308425088674, | |
| "eval_stsb_spearman": 0.8680157573899278, | |
| "step": 5375 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 1.021873979758407e-06, | |
| "loss": 0.0454, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "eval_avg_sts": 0.8463330711965237, | |
| "eval_sickr_spearman": 0.824173882894921, | |
| "eval_stsb_spearman": 0.8684922594981264, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "eval_avg_sts": 0.8475412183069994, | |
| "eval_sickr_spearman": 0.8259262015686877, | |
| "eval_stsb_spearman": 0.869156235045311, | |
| "step": 5625 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "eval_avg_sts": 0.8474056652640968, | |
| "eval_sickr_spearman": 0.8258597265242973, | |
| "eval_stsb_spearman": 0.8689516040038963, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "eval_avg_sts": 0.846709994814922, | |
| "eval_sickr_spearman": 0.8247918990771251, | |
| "eval_stsb_spearman": 0.868628090552719, | |
| "step": 5875 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 2.0568070519098922e-07, | |
| "loss": 0.0456, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "eval_avg_sts": 0.8476613552908148, | |
| "eval_sickr_spearman": 0.8262576161686107, | |
| "eval_stsb_spearman": 0.8690650944130188, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_avg_sts": 0.8477936113942104, | |
| "eval_sickr_spearman": 0.8264356674616418, | |
| "eval_stsb_spearman": 0.8691515553267789, | |
| "step": 6125 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 6126, | |
| "train_runtime": 4478.0897, | |
| "train_samples_per_second": 1.368 | |
| } | |
| ], | |
| "max_steps": 6126, | |
| "num_train_epochs": 3, | |
| "total_flos": 213963637016494080, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |