| { | |
| "best_metric": 0.762063227953411, | |
| "best_model_checkpoint": "finetuned_models/wisesight_sentiment/checkpoint-2400", | |
| "epoch": 5.991124260355029, | |
| "eval_steps": 100, | |
| "global_step": 8100, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.07396449704142012, | |
| "eval_class_f1": { | |
| "neg": 0.018604651162790697, | |
| "neu": 0.6997558991049634, | |
| "pos": 0.0, | |
| "q": 0.0 | |
| }, | |
| "eval_loss": 1.0810712575912476, | |
| "eval_macro_average_f1": 0.17959013756693853, | |
| "eval_micro_average_f1": 0.5391014975041597, | |
| "eval_runtime": 6.8397, | |
| "eval_samples_per_second": 351.476, | |
| "eval_steps_per_second": 22.077, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.14792899408284024, | |
| "eval_class_f1": { | |
| "neg": 0.6223055295220243, | |
| "neu": 0.7497702909647779, | |
| "pos": 0.0, | |
| "q": 0.0 | |
| }, | |
| "eval_loss": 0.8820463418960571, | |
| "eval_macro_average_f1": 0.34301895512170055, | |
| "eval_micro_average_f1": 0.6472545757071547, | |
| "eval_runtime": 6.8712, | |
| "eval_samples_per_second": 349.865, | |
| "eval_steps_per_second": 21.976, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.22189349112426035, | |
| "eval_class_f1": { | |
| "neg": 0.7588757396449703, | |
| "neu": 0.7731384829505916, | |
| "pos": 0.27037037037037037, | |
| "q": 0.0 | |
| }, | |
| "eval_loss": 0.7263810038566589, | |
| "eval_macro_average_f1": 0.4505961482414831, | |
| "eval_micro_average_f1": 0.7059068219633944, | |
| "eval_runtime": 7.0136, | |
| "eval_samples_per_second": 342.761, | |
| "eval_steps_per_second": 21.529, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.2958579881656805, | |
| "eval_class_f1": { | |
| "neg": 0.7806637806637807, | |
| "neu": 0.760541586073501, | |
| "pos": 0.501891551071879, | |
| "q": 0.0909090909090909 | |
| }, | |
| "eval_loss": 0.6896220445632935, | |
| "eval_macro_average_f1": 0.5335015021795629, | |
| "eval_micro_average_f1": 0.7175540765391015, | |
| "eval_runtime": 7.0909, | |
| "eval_samples_per_second": 339.026, | |
| "eval_steps_per_second": 21.295, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.3698224852071006, | |
| "grad_norm": 5.410265922546387, | |
| "learning_rate": 1.828817733990148e-05, | |
| "loss": 0.8994, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.3698224852071006, | |
| "eval_class_f1": { | |
| "neg": 0.7767988252569751, | |
| "neu": 0.7759882869692534, | |
| "pos": 0.41987179487179493, | |
| "q": 0.3777777777777778 | |
| }, | |
| "eval_loss": 0.6673027276992798, | |
| "eval_macro_average_f1": 0.5876091712189503, | |
| "eval_micro_average_f1": 0.7225457570715474, | |
| "eval_runtime": 7.071, | |
| "eval_samples_per_second": 339.978, | |
| "eval_steps_per_second": 21.355, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.4437869822485207, | |
| "eval_class_f1": { | |
| "neg": 0.7704042715484364, | |
| "neu": 0.7415287628053587, | |
| "pos": 0.5426356589147286, | |
| "q": 0.14285714285714285 | |
| }, | |
| "eval_loss": 0.6752218008041382, | |
| "eval_macro_average_f1": 0.5493564590314166, | |
| "eval_micro_average_f1": 0.7050748752079867, | |
| "eval_runtime": 7.2386, | |
| "eval_samples_per_second": 332.107, | |
| "eval_steps_per_second": 20.86, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.5177514792899408, | |
| "eval_class_f1": { | |
| "neg": 0.7780979827089337, | |
| "neu": 0.7535296490520371, | |
| "pos": 0.5532435740514076, | |
| "q": 0.3870967741935484 | |
| }, | |
| "eval_loss": 0.667128324508667, | |
| "eval_macro_average_f1": 0.6179919950014817, | |
| "eval_micro_average_f1": 0.7171381031613977, | |
| "eval_runtime": 7.3636, | |
| "eval_samples_per_second": 326.472, | |
| "eval_steps_per_second": 20.506, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.591715976331361, | |
| "eval_class_f1": { | |
| "neg": 0.7555923777961889, | |
| "neu": 0.7529501332318232, | |
| "pos": 0.5357575757575757, | |
| "q": 0.4161073825503356 | |
| }, | |
| "eval_loss": 0.662220299243927, | |
| "eval_macro_average_f1": 0.6151018673339809, | |
| "eval_micro_average_f1": 0.7059068219633944, | |
| "eval_runtime": 7.3848, | |
| "eval_samples_per_second": 325.535, | |
| "eval_steps_per_second": 20.448, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.665680473372781, | |
| "eval_class_f1": { | |
| "neg": 0.7756714060031595, | |
| "neu": 0.7722698471859858, | |
| "pos": 0.5253807106598984, | |
| "q": 0.2535211267605634 | |
| }, | |
| "eval_loss": 0.6270455121994019, | |
| "eval_macro_average_f1": 0.5817107726524018, | |
| "eval_micro_average_f1": 0.7250415973377704, | |
| "eval_runtime": 7.3625, | |
| "eval_samples_per_second": 326.52, | |
| "eval_steps_per_second": 20.509, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.7396449704142012, | |
| "grad_norm": 7.580224514007568, | |
| "learning_rate": 2.9252054794520548e-05, | |
| "loss": 0.6495, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.7396449704142012, | |
| "eval_class_f1": { | |
| "neg": 0.7862993298585256, | |
| "neu": 0.7885968159940763, | |
| "pos": 0.5219858156028369, | |
| "q": 0.30508474576271183 | |
| }, | |
| "eval_loss": 0.6415818929672241, | |
| "eval_macro_average_f1": 0.6004916768045376, | |
| "eval_micro_average_f1": 0.742928452579035, | |
| "eval_runtime": 7.3324, | |
| "eval_samples_per_second": 327.861, | |
| "eval_steps_per_second": 20.594, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.8136094674556213, | |
| "eval_class_f1": { | |
| "neg": 0.7824267782426777, | |
| "neu": 0.7750972762645915, | |
| "pos": 0.550531914893617, | |
| "q": 0.1923076923076923 | |
| }, | |
| "eval_loss": 0.6599770188331604, | |
| "eval_macro_average_f1": 0.5750909154271446, | |
| "eval_micro_average_f1": 0.7358569051580699, | |
| "eval_runtime": 7.4168, | |
| "eval_samples_per_second": 324.129, | |
| "eval_steps_per_second": 20.359, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.8875739644970414, | |
| "eval_class_f1": { | |
| "neg": 0.7531806615776081, | |
| "neu": 0.7700414000752727, | |
| "pos": 0.5685164212910533, | |
| "q": 0.35955056179775274 | |
| }, | |
| "eval_loss": 0.6348879933357239, | |
| "eval_macro_average_f1": 0.6128222611854217, | |
| "eval_micro_average_f1": 0.721297836938436, | |
| "eval_runtime": 7.4408, | |
| "eval_samples_per_second": 323.084, | |
| "eval_steps_per_second": 20.294, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.9615384615384616, | |
| "eval_class_f1": { | |
| "neg": 0.7920646583394563, | |
| "neu": 0.783076923076923, | |
| "pos": 0.535014005602241, | |
| "q": 0.46616541353383456 | |
| }, | |
| "eval_loss": 0.6110679507255554, | |
| "eval_macro_average_f1": 0.6440802501381137, | |
| "eval_micro_average_f1": 0.7400166389351082, | |
| "eval_runtime": 7.4775, | |
| "eval_samples_per_second": 321.498, | |
| "eval_steps_per_second": 20.194, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.0355029585798816, | |
| "eval_class_f1": { | |
| "neg": 0.8015094339622642, | |
| "neu": 0.8065934065934067, | |
| "pos": 0.5252225519287834, | |
| "q": 0.379746835443038 | |
| }, | |
| "eval_loss": 0.6416576504707336, | |
| "eval_macro_average_f1": 0.628268056981873, | |
| "eval_micro_average_f1": 0.7587354409317804, | |
| "eval_runtime": 7.358, | |
| "eval_samples_per_second": 326.721, | |
| "eval_steps_per_second": 20.522, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.1094674556213018, | |
| "grad_norm": 3.8226146697998047, | |
| "learning_rate": 2.72013698630137e-05, | |
| "loss": 0.6084, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.1094674556213018, | |
| "eval_class_f1": { | |
| "neg": 0.7846277021617293, | |
| "neu": 0.8024917552216929, | |
| "pos": 0.5813953488372093, | |
| "q": 0.2857142857142857 | |
| }, | |
| "eval_loss": 0.6498740315437317, | |
| "eval_macro_average_f1": 0.6135572729837293, | |
| "eval_micro_average_f1": 0.7562396006655574, | |
| "eval_runtime": 7.2975, | |
| "eval_samples_per_second": 329.428, | |
| "eval_steps_per_second": 20.692, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.183431952662722, | |
| "eval_class_f1": { | |
| "neg": 0.7917525773195877, | |
| "neu": 0.7863247863247863, | |
| "pos": 0.5444126074498568, | |
| "q": 0.345679012345679 | |
| }, | |
| "eval_loss": 0.685055673122406, | |
| "eval_macro_average_f1": 0.6170422458599774, | |
| "eval_micro_average_f1": 0.7454242928452579, | |
| "eval_runtime": 7.1996, | |
| "eval_samples_per_second": 333.908, | |
| "eval_steps_per_second": 20.973, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.2573964497041419, | |
| "eval_class_f1": { | |
| "neg": 0.7966231772831925, | |
| "neu": 0.7783018867924529, | |
| "pos": 0.5695216907675196, | |
| "q": 0.25806451612903225 | |
| }, | |
| "eval_loss": 0.6685267090797424, | |
| "eval_macro_average_f1": 0.6006278177430493, | |
| "eval_micro_average_f1": 0.7375207986688852, | |
| "eval_runtime": 7.2849, | |
| "eval_samples_per_second": 329.999, | |
| "eval_steps_per_second": 20.728, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.331360946745562, | |
| "eval_class_f1": { | |
| "neg": 0.8018362662586075, | |
| "neu": 0.7977570093457944, | |
| "pos": 0.5578947368421052, | |
| "q": 0.36363636363636365 | |
| }, | |
| "eval_loss": 0.6347253918647766, | |
| "eval_macro_average_f1": 0.6302810940207177, | |
| "eval_micro_average_f1": 0.7549916805324459, | |
| "eval_runtime": 7.1263, | |
| "eval_samples_per_second": 337.343, | |
| "eval_steps_per_second": 21.189, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.4053254437869822, | |
| "eval_class_f1": { | |
| "neg": 0.7917329093799682, | |
| "neu": 0.7925512104283055, | |
| "pos": 0.5742821473158551, | |
| "q": 0.28125 | |
| }, | |
| "eval_loss": 0.6284430027008057, | |
| "eval_macro_average_f1": 0.6099540667810323, | |
| "eval_micro_average_f1": 0.7491680532445923, | |
| "eval_runtime": 7.2447, | |
| "eval_samples_per_second": 331.831, | |
| "eval_steps_per_second": 20.843, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.4792899408284024, | |
| "grad_norm": 4.046507835388184, | |
| "learning_rate": 2.5146575342465757e-05, | |
| "loss": 0.5135, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.4792899408284024, | |
| "eval_class_f1": { | |
| "neg": 0.7920792079207921, | |
| "neu": 0.7867370007535796, | |
| "pos": 0.5517241379310345, | |
| "q": 0.27586206896551724 | |
| }, | |
| "eval_loss": 0.6431812644004822, | |
| "eval_macro_average_f1": 0.601600603892731, | |
| "eval_micro_average_f1": 0.7437603993344426, | |
| "eval_runtime": 7.2356, | |
| "eval_samples_per_second": 332.247, | |
| "eval_steps_per_second": 20.869, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.5532544378698225, | |
| "eval_class_f1": { | |
| "neg": 0.7887550200803214, | |
| "neu": 0.7950581395348836, | |
| "pos": 0.555407209612817, | |
| "q": 0.29032258064516125 | |
| }, | |
| "eval_loss": 0.6327183842658997, | |
| "eval_macro_average_f1": 0.6073857374682958, | |
| "eval_micro_average_f1": 0.7495840266222962, | |
| "eval_runtime": 7.3333, | |
| "eval_samples_per_second": 327.82, | |
| "eval_steps_per_second": 20.591, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 1.6272189349112427, | |
| "eval_class_f1": { | |
| "neg": 0.7658119658119659, | |
| "neu": 0.8002847988608045, | |
| "pos": 0.56951871657754, | |
| "q": 0.4197530864197531 | |
| }, | |
| "eval_loss": 0.6534045338630676, | |
| "eval_macro_average_f1": 0.6388421419175159, | |
| "eval_micro_average_f1": 0.7495840266222962, | |
| "eval_runtime": 7.2578, | |
| "eval_samples_per_second": 331.231, | |
| "eval_steps_per_second": 20.805, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 1.7011834319526629, | |
| "eval_class_f1": { | |
| "neg": 0.7832369942196531, | |
| "neu": 0.7650099403578529, | |
| "pos": 0.5829268292682928, | |
| "q": 0.35955056179775274 | |
| }, | |
| "eval_loss": 0.6581071019172668, | |
| "eval_macro_average_f1": 0.6226810814108878, | |
| "eval_micro_average_f1": 0.7316971713810316, | |
| "eval_runtime": 7.1824, | |
| "eval_samples_per_second": 334.708, | |
| "eval_steps_per_second": 21.024, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 1.7751479289940828, | |
| "eval_class_f1": { | |
| "neg": 0.7951807228915662, | |
| "neu": 0.8109843081312411, | |
| "pos": 0.5441595441595442, | |
| "q": 0.3157894736842105 | |
| }, | |
| "eval_loss": 0.6206311583518982, | |
| "eval_macro_average_f1": 0.6165285122166405, | |
| "eval_micro_average_f1": 0.762063227953411, | |
| "eval_runtime": 7.2501, | |
| "eval_samples_per_second": 331.583, | |
| "eval_steps_per_second": 20.827, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 1.849112426035503, | |
| "grad_norm": 6.195135593414307, | |
| "learning_rate": 2.3091780821917807e-05, | |
| "loss": 0.4995, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.849112426035503, | |
| "eval_class_f1": { | |
| "neg": 0.7932148626817447, | |
| "neu": 0.8, | |
| "pos": 0.5830164765525983, | |
| "q": 0.196078431372549 | |
| }, | |
| "eval_loss": 0.6029447913169861, | |
| "eval_macro_average_f1": 0.5930774426517229, | |
| "eval_micro_average_f1": 0.7562396006655574, | |
| "eval_runtime": 7.1935, | |
| "eval_samples_per_second": 334.192, | |
| "eval_steps_per_second": 20.991, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 1.9230769230769231, | |
| "eval_class_f1": { | |
| "neg": 0.8059236165237724, | |
| "neu": 0.7899159663865546, | |
| "pos": 0.579415501905972, | |
| "q": 0.36666666666666664 | |
| }, | |
| "eval_loss": 0.6066814064979553, | |
| "eval_macro_average_f1": 0.6354804378707414, | |
| "eval_micro_average_f1": 0.7491680532445923, | |
| "eval_runtime": 7.2817, | |
| "eval_samples_per_second": 330.143, | |
| "eval_steps_per_second": 20.737, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 1.997041420118343, | |
| "eval_class_f1": { | |
| "neg": 0.8003157063930545, | |
| "neu": 0.797884397431054, | |
| "pos": 0.5773447015834348, | |
| "q": 0.3835616438356164 | |
| }, | |
| "eval_loss": 0.630171537399292, | |
| "eval_macro_average_f1": 0.63977661231079, | |
| "eval_micro_average_f1": 0.7545757071547421, | |
| "eval_runtime": 7.2376, | |
| "eval_samples_per_second": 332.156, | |
| "eval_steps_per_second": 20.863, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 2.0710059171597632, | |
| "eval_class_f1": { | |
| "neg": 0.7848509266720386, | |
| "neu": 0.7945103857566765, | |
| "pos": 0.5853051058530511, | |
| "q": 0.35294117647058826 | |
| }, | |
| "eval_loss": 0.7064331769943237, | |
| "eval_macro_average_f1": 0.6294018986880886, | |
| "eval_micro_average_f1": 0.7508319467554077, | |
| "eval_runtime": 7.2934, | |
| "eval_samples_per_second": 329.612, | |
| "eval_steps_per_second": 20.704, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 2.1449704142011834, | |
| "eval_class_f1": { | |
| "neg": 0.797752808988764, | |
| "neu": 0.8026412325752018, | |
| "pos": 0.5824742268041238, | |
| "q": 0.26666666666666666 | |
| }, | |
| "eval_loss": 0.7201129794120789, | |
| "eval_macro_average_f1": 0.612383733758689, | |
| "eval_micro_average_f1": 0.7591514143094842, | |
| "eval_runtime": 7.1883, | |
| "eval_samples_per_second": 334.433, | |
| "eval_steps_per_second": 21.006, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 2.2189349112426036, | |
| "grad_norm": 6.065237045288086, | |
| "learning_rate": 2.1036986301369864e-05, | |
| "loss": 0.4003, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.2189349112426036, | |
| "eval_class_f1": { | |
| "neg": 0.7861271676300579, | |
| "neu": 0.800578034682081, | |
| "pos": 0.5670391061452514, | |
| "q": 0.3185840707964602 | |
| }, | |
| "eval_loss": 0.7178497910499573, | |
| "eval_macro_average_f1": 0.6180820948134627, | |
| "eval_micro_average_f1": 0.7508319467554077, | |
| "eval_runtime": 7.2106, | |
| "eval_samples_per_second": 333.398, | |
| "eval_steps_per_second": 20.941, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.2928994082840237, | |
| "eval_class_f1": { | |
| "neg": 0.7955801104972374, | |
| "neu": 0.781854043392505, | |
| "pos": 0.5852585258525853, | |
| "q": 0.28865979381443296 | |
| }, | |
| "eval_loss": 0.7727176547050476, | |
| "eval_macro_average_f1": 0.6128381183891901, | |
| "eval_micro_average_f1": 0.7383527454242929, | |
| "eval_runtime": 7.2299, | |
| "eval_samples_per_second": 332.51, | |
| "eval_steps_per_second": 20.886, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 2.366863905325444, | |
| "eval_class_f1": { | |
| "neg": 0.7893462469733656, | |
| "neu": 0.788983997022702, | |
| "pos": 0.5606060606060607, | |
| "q": 0.2888888888888889 | |
| }, | |
| "eval_loss": 0.7219040393829346, | |
| "eval_macro_average_f1": 0.6069562983727543, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.2669, | |
| "eval_samples_per_second": 330.815, | |
| "eval_steps_per_second": 20.779, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 2.440828402366864, | |
| "eval_class_f1": { | |
| "neg": 0.8073115003808072, | |
| "neu": 0.7814829344841114, | |
| "pos": 0.5855338691159586, | |
| "q": 0.26666666666666666 | |
| }, | |
| "eval_loss": 0.7229210734367371, | |
| "eval_macro_average_f1": 0.610248742661886, | |
| "eval_micro_average_f1": 0.7450083194675541, | |
| "eval_runtime": 7.1283, | |
| "eval_samples_per_second": 337.248, | |
| "eval_steps_per_second": 21.183, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 2.5147928994082838, | |
| "eval_class_f1": { | |
| "neg": 0.7984790874524715, | |
| "neu": 0.7856049004594182, | |
| "pos": 0.5773447015834348, | |
| "q": 0.3 | |
| }, | |
| "eval_loss": 0.7037935853004456, | |
| "eval_macro_average_f1": 0.615357172373831, | |
| "eval_micro_average_f1": 0.747504159733777, | |
| "eval_runtime": 7.2219, | |
| "eval_samples_per_second": 332.876, | |
| "eval_steps_per_second": 20.909, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 2.5887573964497044, | |
| "grad_norm": 3.8475677967071533, | |
| "learning_rate": 1.8982191780821918e-05, | |
| "loss": 0.3579, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 2.5887573964497044, | |
| "eval_class_f1": { | |
| "neg": 0.7871815940838127, | |
| "neu": 0.7871305649083427, | |
| "pos": 0.5738916256157636, | |
| "q": 0.37735849056603776 | |
| }, | |
| "eval_loss": 0.7569752931594849, | |
| "eval_macro_average_f1": 0.6313905687934891, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.3391, | |
| "eval_samples_per_second": 327.56, | |
| "eval_steps_per_second": 20.575, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 2.662721893491124, | |
| "eval_class_f1": { | |
| "neg": 0.8064269319051262, | |
| "neu": 0.7905718701700155, | |
| "pos": 0.5779927448609432, | |
| "q": 0.3255813953488372 | |
| }, | |
| "eval_loss": 0.7201011180877686, | |
| "eval_macro_average_f1": 0.6251432355712305, | |
| "eval_micro_average_f1": 0.75, | |
| "eval_runtime": 7.2188, | |
| "eval_samples_per_second": 333.02, | |
| "eval_steps_per_second": 20.918, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 2.7366863905325443, | |
| "eval_class_f1": { | |
| "neg": 0.7847896440129449, | |
| "neu": 0.7701911822083495, | |
| "pos": 0.5797413793103448, | |
| "q": 0.3703703703703704 | |
| }, | |
| "eval_loss": 0.7302864789962769, | |
| "eval_macro_average_f1": 0.6262731439755023, | |
| "eval_micro_average_f1": 0.7304492512479202, | |
| "eval_runtime": 7.2541, | |
| "eval_samples_per_second": 331.401, | |
| "eval_steps_per_second": 20.816, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 2.8106508875739644, | |
| "eval_class_f1": { | |
| "neg": 0.7971698113207547, | |
| "neu": 0.8014842300556586, | |
| "pos": 0.5839793281653747, | |
| "q": 0.3283582089552239 | |
| }, | |
| "eval_loss": 0.7112248539924622, | |
| "eval_macro_average_f1": 0.627747894624253, | |
| "eval_micro_average_f1": 0.7587354409317804, | |
| "eval_runtime": 7.2905, | |
| "eval_samples_per_second": 329.743, | |
| "eval_steps_per_second": 20.712, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 2.8846153846153846, | |
| "eval_class_f1": { | |
| "neg": 0.7999999999999999, | |
| "neu": 0.7944066515495087, | |
| "pos": 0.5961995249406176, | |
| "q": 0.2545454545454545 | |
| }, | |
| "eval_loss": 0.7105884552001953, | |
| "eval_macro_average_f1": 0.6112879077588952, | |
| "eval_micro_average_f1": 0.7549916805324459, | |
| "eval_runtime": 7.4167, | |
| "eval_samples_per_second": 324.132, | |
| "eval_steps_per_second": 20.359, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 2.9585798816568047, | |
| "grad_norm": 8.97050666809082, | |
| "learning_rate": 1.6927397260273975e-05, | |
| "loss": 0.3409, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 2.9585798816568047, | |
| "eval_class_f1": { | |
| "neg": 0.803088803088803, | |
| "neu": 0.7901328273244782, | |
| "pos": 0.5671641791044775, | |
| "q": 0.3513513513513513 | |
| }, | |
| "eval_loss": 0.7364293932914734, | |
| "eval_macro_average_f1": 0.6279342902172774, | |
| "eval_micro_average_f1": 0.7495840266222962, | |
| "eval_runtime": 7.2488, | |
| "eval_samples_per_second": 331.641, | |
| "eval_steps_per_second": 20.831, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.032544378698225, | |
| "eval_class_f1": { | |
| "neg": 0.7924836601307189, | |
| "neu": 0.7892777364110202, | |
| "pos": 0.5696969696969698, | |
| "q": 0.3287671232876712 | |
| }, | |
| "eval_loss": 0.8425710201263428, | |
| "eval_macro_average_f1": 0.6200563723815951, | |
| "eval_micro_average_f1": 0.7454242928452579, | |
| "eval_runtime": 7.1671, | |
| "eval_samples_per_second": 335.422, | |
| "eval_steps_per_second": 21.068, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 3.106508875739645, | |
| "eval_class_f1": { | |
| "neg": 0.7883817427385893, | |
| "neu": 0.7684537684537684, | |
| "pos": 0.5720338983050848, | |
| "q": 0.35294117647058826 | |
| }, | |
| "eval_loss": 0.9264113306999207, | |
| "eval_macro_average_f1": 0.6204526464920077, | |
| "eval_micro_average_f1": 0.7275374376039934, | |
| "eval_runtime": 7.2924, | |
| "eval_samples_per_second": 329.656, | |
| "eval_steps_per_second": 20.706, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 3.1804733727810652, | |
| "eval_class_f1": { | |
| "neg": 0.8064269319051262, | |
| "neu": 0.7787333854573885, | |
| "pos": 0.5774647887323944, | |
| "q": 0.32967032967032966 | |
| }, | |
| "eval_loss": 0.9222328662872314, | |
| "eval_macro_average_f1": 0.6230738589413097, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.2012, | |
| "eval_samples_per_second": 333.833, | |
| "eval_steps_per_second": 20.969, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 3.2544378698224854, | |
| "eval_class_f1": { | |
| "neg": 0.7999999999999999, | |
| "neu": 0.7803557617942769, | |
| "pos": 0.5765124555160142, | |
| "q": 0.35955056179775274 | |
| }, | |
| "eval_loss": 0.9496058821678162, | |
| "eval_macro_average_f1": 0.6291046947770109, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.3334, | |
| "eval_samples_per_second": 327.814, | |
| "eval_steps_per_second": 20.591, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 3.328402366863905, | |
| "grad_norm": 12.435276985168457, | |
| "learning_rate": 1.4872602739726027e-05, | |
| "loss": 0.2249, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 3.328402366863905, | |
| "eval_class_f1": { | |
| "neg": 0.8012718600953895, | |
| "neu": 0.784238714613619, | |
| "pos": 0.5663082437275986, | |
| "q": 0.32323232323232326 | |
| }, | |
| "eval_loss": 0.9026820063591003, | |
| "eval_macro_average_f1": 0.6187627854172325, | |
| "eval_micro_average_f1": 0.7412645590682196, | |
| "eval_runtime": 7.1404, | |
| "eval_samples_per_second": 336.677, | |
| "eval_steps_per_second": 21.147, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 3.4023668639053253, | |
| "eval_class_f1": { | |
| "neg": 0.8043647700701482, | |
| "neu": 0.7884322678843227, | |
| "pos": 0.5676328502415459, | |
| "q": 0.2898550724637681 | |
| }, | |
| "eval_loss": 0.943065345287323, | |
| "eval_macro_average_f1": 0.6125712401649462, | |
| "eval_micro_average_f1": 0.747504159733777, | |
| "eval_runtime": 7.2681, | |
| "eval_samples_per_second": 330.759, | |
| "eval_steps_per_second": 20.776, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 3.4763313609467454, | |
| "eval_class_f1": { | |
| "neg": 0.8018504240555128, | |
| "neu": 0.7930382141505864, | |
| "pos": 0.5692503176620076, | |
| "q": 0.345679012345679 | |
| }, | |
| "eval_loss": 0.9825762510299683, | |
| "eval_macro_average_f1": 0.6274544920534464, | |
| "eval_micro_average_f1": 0.7512479201331115, | |
| "eval_runtime": 7.2921, | |
| "eval_samples_per_second": 329.672, | |
| "eval_steps_per_second": 20.707, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 3.5502958579881656, | |
| "eval_class_f1": { | |
| "neg": 0.7946498819826908, | |
| "neu": 0.7813455657492355, | |
| "pos": 0.5795053003533569, | |
| "q": 0.3055555555555555 | |
| }, | |
| "eval_loss": 0.9374552965164185, | |
| "eval_macro_average_f1": 0.6152640759102097, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.387, | |
| "eval_samples_per_second": 325.436, | |
| "eval_steps_per_second": 20.441, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 3.6242603550295858, | |
| "eval_class_f1": { | |
| "neg": 0.7984790874524715, | |
| "neu": 0.7785547785547785, | |
| "pos": 0.5748218527315915, | |
| "q": 0.3116883116883117 | |
| }, | |
| "eval_loss": 0.9656402468681335, | |
| "eval_macro_average_f1": 0.6158860076067884, | |
| "eval_micro_average_f1": 0.7408485856905158, | |
| "eval_runtime": 7.4821, | |
| "eval_samples_per_second": 321.299, | |
| "eval_steps_per_second": 20.181, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 3.698224852071006, | |
| "grad_norm": 0.6623280644416809, | |
| "learning_rate": 1.2817808219178083e-05, | |
| "loss": 0.2207, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 3.698224852071006, | |
| "eval_class_f1": { | |
| "neg": 0.7987616099071206, | |
| "neu": 0.7862857142857143, | |
| "pos": 0.5647348951911221, | |
| "q": 0.3 | |
| }, | |
| "eval_loss": 0.9422620534896851, | |
| "eval_macro_average_f1": 0.6124455548459892, | |
| "eval_micro_average_f1": 0.7441763727121464, | |
| "eval_runtime": 7.2765, | |
| "eval_samples_per_second": 330.379, | |
| "eval_steps_per_second": 20.752, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 3.772189349112426, | |
| "eval_class_f1": { | |
| "neg": 0.8, | |
| "neu": 0.7671342685370742, | |
| "pos": 0.5714285714285714, | |
| "q": 0.3414634146341463 | |
| }, | |
| "eval_loss": 0.9625053405761719, | |
| "eval_macro_average_f1": 0.620006563649948, | |
| "eval_micro_average_f1": 0.7296173044925125, | |
| "eval_runtime": 7.3445, | |
| "eval_samples_per_second": 327.318, | |
| "eval_steps_per_second": 20.559, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 3.8461538461538463, | |
| "eval_class_f1": { | |
| "neg": 0.8018942383583267, | |
| "neu": 0.7968691762951919, | |
| "pos": 0.5706874189364461, | |
| "q": 0.25287356321839083 | |
| }, | |
| "eval_loss": 0.9822611212730408, | |
| "eval_macro_average_f1": 0.6055810992020889, | |
| "eval_micro_average_f1": 0.7520798668885191, | |
| "eval_runtime": 7.4499, | |
| "eval_samples_per_second": 322.688, | |
| "eval_steps_per_second": 20.269, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 3.9201183431952664, | |
| "eval_class_f1": { | |
| "neg": 0.7891268533772653, | |
| "neu": 0.7875375375375375, | |
| "pos": 0.5810968494749125, | |
| "q": 0.273972602739726 | |
| }, | |
| "eval_loss": 0.9442653656005859, | |
| "eval_macro_average_f1": 0.6079334607823603, | |
| "eval_micro_average_f1": 0.7433444259567388, | |
| "eval_runtime": 7.3731, | |
| "eval_samples_per_second": 326.052, | |
| "eval_steps_per_second": 20.48, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 3.994082840236686, | |
| "eval_class_f1": { | |
| "neg": 0.8046511627906977, | |
| "neu": 0.7945516458569808, | |
| "pos": 0.5775, | |
| "q": 0.32 | |
| }, | |
| "eval_loss": 0.9429491758346558, | |
| "eval_macro_average_f1": 0.6241757021619195, | |
| "eval_micro_average_f1": 0.7537437603993344, | |
| "eval_runtime": 7.3966, | |
| "eval_samples_per_second": 325.013, | |
| "eval_steps_per_second": 20.415, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 4.068047337278107, | |
| "grad_norm": 2.4124114513397217, | |
| "learning_rate": 1.0763013698630138e-05, | |
| "loss": 0.2077, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 4.068047337278107, | |
| "eval_class_f1": { | |
| "neg": 0.8063781321184511, | |
| "neu": 0.7866927592954991, | |
| "pos": 0.5862884160756501, | |
| "q": 0.3333333333333333 | |
| }, | |
| "eval_loss": 1.1077452898025513, | |
| "eval_macro_average_f1": 0.6281731602057334, | |
| "eval_micro_average_f1": 0.7483361064891847, | |
| "eval_runtime": 7.369, | |
| "eval_samples_per_second": 326.23, | |
| "eval_steps_per_second": 20.491, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 4.1420118343195265, | |
| "eval_class_f1": { | |
| "neg": 0.7993920972644377, | |
| "neu": 0.7660256410256411, | |
| "pos": 0.5726775956284154, | |
| "q": 0.29629629629629634 | |
| }, | |
| "eval_loss": 1.1472598314285278, | |
| "eval_macro_average_f1": 0.6085979075536977, | |
| "eval_micro_average_f1": 0.7304492512479202, | |
| "eval_runtime": 7.5033, | |
| "eval_samples_per_second": 320.394, | |
| "eval_steps_per_second": 20.125, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 4.215976331360947, | |
| "eval_class_f1": { | |
| "neg": 0.8024502297090352, | |
| "neu": 0.7868978805394989, | |
| "pos": 0.5731559854897219, | |
| "q": 0.3 | |
| }, | |
| "eval_loss": 1.169406533241272, | |
| "eval_macro_average_f1": 0.615626023934564, | |
| "eval_micro_average_f1": 0.7462562396006656, | |
| "eval_runtime": 7.4833, | |
| "eval_samples_per_second": 321.249, | |
| "eval_steps_per_second": 20.178, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 4.289940828402367, | |
| "eval_class_f1": { | |
| "neg": 0.803951367781155, | |
| "neu": 0.7902550437761706, | |
| "pos": 0.5685019206145967, | |
| "q": 0.30952380952380953 | |
| }, | |
| "eval_loss": 1.1968339681625366, | |
| "eval_macro_average_f1": 0.6180580354239329, | |
| "eval_micro_average_f1": 0.7495840266222962, | |
| "eval_runtime": 7.4382, | |
| "eval_samples_per_second": 323.197, | |
| "eval_steps_per_second": 20.301, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 4.363905325443787, | |
| "eval_class_f1": { | |
| "neg": 0.8024786986831913, | |
| "neu": 0.7916030534351146, | |
| "pos": 0.5773955773955775, | |
| "q": 0.3373493975903615 | |
| }, | |
| "eval_loss": 1.1896393299102783, | |
| "eval_macro_average_f1": 0.6272066817760612, | |
| "eval_micro_average_f1": 0.7504159733777038, | |
| "eval_runtime": 7.423, | |
| "eval_samples_per_second": 323.857, | |
| "eval_steps_per_second": 20.342, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 4.437869822485207, | |
| "grad_norm": 0.9506312608718872, | |
| "learning_rate": 8.708219178082192e-06, | |
| "loss": 0.1324, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 4.437869822485207, | |
| "eval_class_f1": { | |
| "neg": 0.8024316109422492, | |
| "neu": 0.7925840092699884, | |
| "pos": 0.5878048780487805, | |
| "q": 0.28915662650602414 | |
| }, | |
| "eval_loss": 1.2535008192062378, | |
| "eval_macro_average_f1": 0.6179942811917606, | |
| "eval_micro_average_f1": 0.7516638935108153, | |
| "eval_runtime": 7.3808, | |
| "eval_samples_per_second": 325.711, | |
| "eval_steps_per_second": 20.459, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 4.511834319526627, | |
| "eval_class_f1": { | |
| "neg": 0.7901821060965954, | |
| "neu": 0.7819374758780393, | |
| "pos": 0.5821064552661382, | |
| "q": 0.2535211267605634 | |
| }, | |
| "eval_loss": 1.2182434797286987, | |
| "eval_macro_average_f1": 0.601936791000334, | |
| "eval_micro_average_f1": 0.7396006655574043, | |
| "eval_runtime": 7.3498, | |
| "eval_samples_per_second": 327.085, | |
| "eval_steps_per_second": 20.545, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 4.585798816568047, | |
| "eval_class_f1": { | |
| "neg": 0.7945425361155697, | |
| "neu": 0.7956989247311828, | |
| "pos": 0.5761006289308177, | |
| "q": 0.22857142857142854 | |
| }, | |
| "eval_loss": 1.2836171388626099, | |
| "eval_macro_average_f1": 0.5987283795872497, | |
| "eval_micro_average_f1": 0.7508319467554077, | |
| "eval_runtime": 7.4332, | |
| "eval_samples_per_second": 323.412, | |
| "eval_steps_per_second": 20.314, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 4.659763313609467, | |
| "eval_class_f1": { | |
| "neg": 0.8003025718608169, | |
| "neu": 0.7733970529669454, | |
| "pos": 0.5691609977324262, | |
| "q": 0.3225806451612903 | |
| }, | |
| "eval_loss": 1.2842472791671753, | |
| "eval_macro_average_f1": 0.6163603169303697, | |
| "eval_micro_average_f1": 0.7346089850249584, | |
| "eval_runtime": 7.4941, | |
| "eval_samples_per_second": 320.786, | |
| "eval_steps_per_second": 20.149, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 4.733727810650888, | |
| "eval_class_f1": { | |
| "neg": 0.799375487900078, | |
| "neu": 0.7936865839909809, | |
| "pos": 0.5614489003880984, | |
| "q": 0.3225806451612903 | |
| }, | |
| "eval_loss": 1.3067219257354736, | |
| "eval_macro_average_f1": 0.619272904360112, | |
| "eval_micro_average_f1": 0.7487520798668885, | |
| "eval_runtime": 7.3805, | |
| "eval_samples_per_second": 325.723, | |
| "eval_steps_per_second": 20.459, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 4.8076923076923075, | |
| "grad_norm": 0.2732117772102356, | |
| "learning_rate": 6.653424657534246e-06, | |
| "loss": 0.1441, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 4.8076923076923075, | |
| "eval_class_f1": { | |
| "neg": 0.8027628549501151, | |
| "neu": 0.7751572327044026, | |
| "pos": 0.5694760820045559, | |
| "q": 0.3132530120481927 | |
| }, | |
| "eval_loss": 1.2718240022659302, | |
| "eval_macro_average_f1": 0.6151622954268166, | |
| "eval_micro_average_f1": 0.7371048252911814, | |
| "eval_runtime": 7.3822, | |
| "eval_samples_per_second": 325.649, | |
| "eval_steps_per_second": 20.455, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 4.881656804733728, | |
| "eval_class_f1": { | |
| "neg": 0.796875, | |
| "neu": 0.7868601986249045, | |
| "pos": 0.5735115431348725, | |
| "q": 0.29885057471264365 | |
| }, | |
| "eval_loss": 1.261472225189209, | |
| "eval_macro_average_f1": 0.6140243291181051, | |
| "eval_micro_average_f1": 0.7441763727121464, | |
| "eval_runtime": 7.5114, | |
| "eval_samples_per_second": 320.048, | |
| "eval_steps_per_second": 20.103, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 4.955621301775148, | |
| "eval_class_f1": { | |
| "neg": 0.7930763178599529, | |
| "neu": 0.7766536964980545, | |
| "pos": 0.5714285714285715, | |
| "q": 0.35294117647058826 | |
| }, | |
| "eval_loss": 1.2753080129623413, | |
| "eval_macro_average_f1": 0.6235249405642919, | |
| "eval_micro_average_f1": 0.7358569051580699, | |
| "eval_runtime": 7.3656, | |
| "eval_samples_per_second": 326.381, | |
| "eval_steps_per_second": 20.501, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 5.029585798816568, | |
| "eval_class_f1": { | |
| "neg": 0.7962962962962963, | |
| "neu": 0.7754943776657619, | |
| "pos": 0.5657276995305165, | |
| "q": 0.345679012345679 | |
| }, | |
| "eval_loss": 1.3079357147216797, | |
| "eval_macro_average_f1": 0.6207993464595634, | |
| "eval_micro_average_f1": 0.7366888519134775, | |
| "eval_runtime": 7.4026, | |
| "eval_samples_per_second": 324.753, | |
| "eval_steps_per_second": 20.398, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 5.103550295857988, | |
| "eval_class_f1": { | |
| "neg": 0.7972136222910216, | |
| "neu": 0.7786790266512167, | |
| "pos": 0.5721040189125295, | |
| "q": 0.3703703703703704 | |
| }, | |
| "eval_loss": 1.3499900102615356, | |
| "eval_macro_average_f1": 0.6295917595562845, | |
| "eval_micro_average_f1": 0.740432612312812, | |
| "eval_runtime": 7.3777, | |
| "eval_samples_per_second": 325.846, | |
| "eval_steps_per_second": 20.467, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 5.177514792899408, | |
| "grad_norm": 11.024497985839844, | |
| "learning_rate": 4.598630136986302e-06, | |
| "loss": 0.1111, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 5.177514792899408, | |
| "eval_class_f1": { | |
| "neg": 0.7956147220046985, | |
| "neu": 0.7807853602744949, | |
| "pos": 0.5693606755126658, | |
| "q": 0.3544303797468354 | |
| }, | |
| "eval_loss": 1.4051584005355835, | |
| "eval_macro_average_f1": 0.6250477843846737, | |
| "eval_micro_average_f1": 0.7412645590682196, | |
| "eval_runtime": 7.2531, | |
| "eval_samples_per_second": 331.444, | |
| "eval_steps_per_second": 20.819, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 5.2514792899408285, | |
| "eval_class_f1": { | |
| "neg": 0.7925407925407926, | |
| "neu": 0.7769230769230769, | |
| "pos": 0.5737898465171192, | |
| "q": 0.27027027027027023 | |
| }, | |
| "eval_loss": 1.4020917415618896, | |
| "eval_macro_average_f1": 0.6033809965628147, | |
| "eval_micro_average_f1": 0.7375207986688852, | |
| "eval_runtime": 7.404, | |
| "eval_samples_per_second": 324.69, | |
| "eval_steps_per_second": 20.394, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 5.325443786982248, | |
| "eval_class_f1": { | |
| "neg": 0.7949326999208235, | |
| "neu": 0.7753846153846154, | |
| "pos": 0.5727482678983833, | |
| "q": 0.27848101265822783 | |
| }, | |
| "eval_loss": 1.4238033294677734, | |
| "eval_macro_average_f1": 0.6053866489655125, | |
| "eval_micro_average_f1": 0.7358569051580699, | |
| "eval_runtime": 7.4328, | |
| "eval_samples_per_second": 323.433, | |
| "eval_steps_per_second": 20.315, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 5.399408284023669, | |
| "eval_class_f1": { | |
| "neg": 0.7969348659003831, | |
| "neu": 0.7798306389530408, | |
| "pos": 0.5721212121212121, | |
| "q": 0.3 | |
| }, | |
| "eval_loss": 1.4431192874908447, | |
| "eval_macro_average_f1": 0.612221679243659, | |
| "eval_micro_average_f1": 0.7408485856905158, | |
| "eval_runtime": 7.3682, | |
| "eval_samples_per_second": 326.266, | |
| "eval_steps_per_second": 20.493, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 5.4733727810650885, | |
| "eval_class_f1": { | |
| "neg": 0.7940717628705148, | |
| "neu": 0.783072817384674, | |
| "pos": 0.5703883495145632, | |
| "q": 0.3037974683544304 | |
| }, | |
| "eval_loss": 1.4316595792770386, | |
| "eval_macro_average_f1": 0.6128325995310456, | |
| "eval_micro_average_f1": 0.7416805324459235, | |
| "eval_runtime": 7.3736, | |
| "eval_samples_per_second": 326.03, | |
| "eval_steps_per_second": 20.479, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 5.547337278106509, | |
| "grad_norm": 0.4265735149383545, | |
| "learning_rate": 2.543835616438356e-06, | |
| "loss": 0.0933, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 5.547337278106509, | |
| "eval_class_f1": { | |
| "neg": 0.7978311386522074, | |
| "neu": 0.7788089713843775, | |
| "pos": 0.567409144196952, | |
| "q": 0.30769230769230765 | |
| }, | |
| "eval_loss": 1.4399964809417725, | |
| "eval_macro_average_f1": 0.6129353904814612, | |
| "eval_micro_average_f1": 0.7387687188019967, | |
| "eval_runtime": 7.2697, | |
| "eval_samples_per_second": 330.689, | |
| "eval_steps_per_second": 20.771, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 5.621301775147929, | |
| "eval_class_f1": { | |
| "neg": 0.7984375, | |
| "neu": 0.7812379853902346, | |
| "pos": 0.5714285714285714, | |
| "q": 0.32500000000000007 | |
| }, | |
| "eval_loss": 1.4240373373031616, | |
| "eval_macro_average_f1": 0.6190260142047015, | |
| "eval_micro_average_f1": 0.7412645590682196, | |
| "eval_runtime": 7.4341, | |
| "eval_samples_per_second": 323.375, | |
| "eval_steps_per_second": 20.312, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 5.695266272189349, | |
| "eval_class_f1": { | |
| "neg": 0.7987470634299139, | |
| "neu": 0.7843286420692278, | |
| "pos": 0.5703883495145632, | |
| "q": 0.30769230769230765 | |
| }, | |
| "eval_loss": 1.4332064390182495, | |
| "eval_macro_average_f1": 0.6152890906765031, | |
| "eval_micro_average_f1": 0.7437603993344426, | |
| "eval_runtime": 7.4434, | |
| "eval_samples_per_second": 322.969, | |
| "eval_steps_per_second": 20.286, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 5.769230769230769, | |
| "eval_class_f1": { | |
| "neg": 0.7981220657276996, | |
| "neu": 0.781874039938556, | |
| "pos": 0.5731132075471698, | |
| "q": 0.30769230769230765 | |
| }, | |
| "eval_loss": 1.4344979524612427, | |
| "eval_macro_average_f1": 0.6152004052264332, | |
| "eval_micro_average_f1": 0.7416805324459235, | |
| "eval_runtime": 7.3808, | |
| "eval_samples_per_second": 325.708, | |
| "eval_steps_per_second": 20.458, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 5.84319526627219, | |
| "eval_class_f1": { | |
| "neg": 0.7990654205607477, | |
| "neu": 0.7815384615384616, | |
| "pos": 0.5724465558194775, | |
| "q": 0.3414634146341463 | |
| }, | |
| "eval_loss": 1.4412455558776855, | |
| "eval_macro_average_f1": 0.6236284631382082, | |
| "eval_micro_average_f1": 0.7420965058236273, | |
| "eval_runtime": 7.3915, | |
| "eval_samples_per_second": 325.237, | |
| "eval_steps_per_second": 20.429, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 5.9171597633136095, | |
| "grad_norm": 16.41318702697754, | |
| "learning_rate": 4.89041095890411e-07, | |
| "loss": 0.1006, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 5.9171597633136095, | |
| "eval_class_f1": { | |
| "neg": 0.7987519500780033, | |
| "neu": 0.7813098429720413, | |
| "pos": 0.5700598802395208, | |
| "q": 0.32500000000000007 | |
| }, | |
| "eval_loss": 1.4469937086105347, | |
| "eval_macro_average_f1": 0.6187804183223914, | |
| "eval_micro_average_f1": 0.7416805324459235, | |
| "eval_runtime": 7.3689, | |
| "eval_samples_per_second": 326.236, | |
| "eval_steps_per_second": 20.492, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 5.991124260355029, | |
| "eval_class_f1": { | |
| "neg": 0.7990654205607477, | |
| "neu": 0.781441717791411, | |
| "pos": 0.569377990430622, | |
| "q": 0.32500000000000007 | |
| }, | |
| "eval_loss": 1.4454258680343628, | |
| "eval_macro_average_f1": 0.6187212821956952, | |
| "eval_micro_average_f1": 0.7416805324459235, | |
| "eval_runtime": 7.448, | |
| "eval_samples_per_second": 322.77, | |
| "eval_steps_per_second": 20.274, | |
| "step": 8100 | |
| } | |
| ], | |
| "logging_steps": 500, | |
| "max_steps": 8112, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 6, | |
| "save_steps": 100, | |
| "total_flos": 1.0485727069042368e+16, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |