Spaces:
Runtime error
Runtime error
| { | |
| "best_metric": 0.5930562407498767, | |
| "best_model_checkpoint": "bert-finetuned-sem_eval-english/checkpoint-49092", | |
| "epoch": 4.0, | |
| "global_step": 49092, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 1.9796300822944678e-05, | |
| "loss": 0.472, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.9592601645889353e-05, | |
| "loss": 0.4302, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.938890246883403e-05, | |
| "loss": 0.4104, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9185203291778702e-05, | |
| "loss": 0.4044, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.8981504114723378e-05, | |
| "loss": 0.3954, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.8777804937668054e-05, | |
| "loss": 0.3921, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.857410576061273e-05, | |
| "loss": 0.393, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.8370406583557402e-05, | |
| "loss": 0.3993, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.816670740650208e-05, | |
| "loss": 0.4029, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7963008229446754e-05, | |
| "loss": 0.4048, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.775930905239143e-05, | |
| "loss": 0.3931, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.7555609875336106e-05, | |
| "loss": 0.3879, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.735191069828078e-05, | |
| "loss": 0.3907, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.7148211521225454e-05, | |
| "loss": 0.3909, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.6944512344170133e-05, | |
| "loss": 0.3854, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.6740813167114806e-05, | |
| "loss": 0.3854, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.6537113990059482e-05, | |
| "loss": 0.3899, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.6333414813004154e-05, | |
| "loss": 0.3855, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.6129715635948834e-05, | |
| "loss": 0.387, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.5926016458893506e-05, | |
| "loss": 0.388, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.5722317281838182e-05, | |
| "loss": 0.3754, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.5518618104782858e-05, | |
| "loss": 0.3844, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.5314918927727534e-05, | |
| "loss": 0.3768, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.5111219750672208e-05, | |
| "loss": 0.3763, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_accuracy": 0.3124490793547336, | |
| "eval_f1": 0.5500132135306554, | |
| "eval_loss": 0.360363245010376, | |
| "eval_roc_auc": 0.7014207843050106, | |
| "eval_runtime": 1041.5255, | |
| "eval_samples_per_second": 11.785, | |
| "eval_steps_per_second": 1.474, | |
| "step": 12273 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.4907520573616886e-05, | |
| "loss": 0.376, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.470382139656156e-05, | |
| "loss": 0.3613, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.4500122219506234e-05, | |
| "loss": 0.3644, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.4296423042450908e-05, | |
| "loss": 0.3535, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.4092723865395586e-05, | |
| "loss": 0.3588, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.388902468834026e-05, | |
| "loss": 0.3505, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.3685325511284934e-05, | |
| "loss": 0.3488, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.3481626334229612e-05, | |
| "loss": 0.3513, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.3277927157174286e-05, | |
| "loss": 0.3536, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.307422798011896e-05, | |
| "loss": 0.3627, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.2870528803063638e-05, | |
| "loss": 0.3439, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.2666829626008312e-05, | |
| "loss": 0.3598, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.2463130448952986e-05, | |
| "loss": 0.3524, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.2259431271897662e-05, | |
| "loss": 0.3533, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.2055732094842338e-05, | |
| "loss": 0.3422, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.1852032917787012e-05, | |
| "loss": 0.3578, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.1648333740731688e-05, | |
| "loss": 0.3554, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.1444634563676364e-05, | |
| "loss": 0.3448, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.124093538662104e-05, | |
| "loss": 0.3539, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.1037236209565714e-05, | |
| "loss": 0.3556, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.083353703251039e-05, | |
| "loss": 0.3545, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.0629837855455066e-05, | |
| "loss": 0.3588, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.042613867839974e-05, | |
| "loss": 0.3519, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0222439501344414e-05, | |
| "loss": 0.3533, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0018740324289092e-05, | |
| "loss": 0.349, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_accuracy": 0.3259736027374939, | |
| "eval_f1": 0.5744935216000517, | |
| "eval_loss": 0.3546680212020874, | |
| "eval_roc_auc": 0.7161739733306818, | |
| "eval_runtime": 1043.4528, | |
| "eval_samples_per_second": 11.763, | |
| "eval_steps_per_second": 1.471, | |
| "step": 24546 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 9.815041147233766e-06, | |
| "loss": 0.3248, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.611341970178442e-06, | |
| "loss": 0.3291, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.407642793123116e-06, | |
| "loss": 0.3279, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.203943616067792e-06, | |
| "loss": 0.3212, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 9.000244439012466e-06, | |
| "loss": 0.3233, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.796545261957142e-06, | |
| "loss": 0.3171, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.592846084901818e-06, | |
| "loss": 0.324, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 8.389146907846492e-06, | |
| "loss": 0.3279, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 8.185447730791168e-06, | |
| "loss": 0.3263, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.981748553735843e-06, | |
| "loss": 0.3171, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 7.778049376680518e-06, | |
| "loss": 0.3259, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 7.574350199625194e-06, | |
| "loss": 0.321, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 7.370651022569869e-06, | |
| "loss": 0.3155, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 7.1669518455145444e-06, | |
| "loss": 0.3317, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 6.96325266845922e-06, | |
| "loss": 0.3195, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 6.759553491403895e-06, | |
| "loss": 0.3215, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 6.555854314348571e-06, | |
| "loss": 0.3188, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 6.3521551372932455e-06, | |
| "loss": 0.3157, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 6.148455960237921e-06, | |
| "loss": 0.318, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 5.944756783182597e-06, | |
| "loss": 0.325, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 5.7410576061272715e-06, | |
| "loss": 0.3213, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 5.537358429071947e-06, | |
| "loss": 0.3306, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 5.333659252016622e-06, | |
| "loss": 0.3254, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 5.1299600749612975e-06, | |
| "loss": 0.3179, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_accuracy": 0.3271956982238879, | |
| "eval_f1": 0.5859622450272393, | |
| "eval_loss": 0.3562159240245819, | |
| "eval_roc_auc": 0.7241606691764836, | |
| "eval_runtime": 1044.5243, | |
| "eval_samples_per_second": 11.751, | |
| "eval_steps_per_second": 1.47, | |
| "step": 36819 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 4.9262608979059726e-06, | |
| "loss": 0.3184, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 4.722561720850648e-06, | |
| "loss": 0.2906, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 4.5188625437953235e-06, | |
| "loss": 0.2902, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 4.3151633667399986e-06, | |
| "loss": 0.3049, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 4.1114641896846745e-06, | |
| "loss": 0.2982, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 3.9077650126293495e-06, | |
| "loss": 0.3011, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 3.7040658355740246e-06, | |
| "loss": 0.2966, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 3.5003666585186996e-06, | |
| "loss": 0.298, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 3.2966674814633755e-06, | |
| "loss": 0.2937, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 3.0929683044080506e-06, | |
| "loss": 0.3008, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 2.8892691273527256e-06, | |
| "loss": 0.3002, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 2.6855699502974007e-06, | |
| "loss": 0.2881, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 2.481870773242076e-06, | |
| "loss": 0.2956, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 2.2781715961867516e-06, | |
| "loss": 0.2945, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 2.074472419131427e-06, | |
| "loss": 0.3016, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 1.8707732420761022e-06, | |
| "loss": 0.2977, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 1.6670740650207777e-06, | |
| "loss": 0.2958, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 1.4633748879654527e-06, | |
| "loss": 0.2957, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 1.259675710910128e-06, | |
| "loss": 0.2996, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.0559765338548034e-06, | |
| "loss": 0.2937, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 8.522773567994785e-07, | |
| "loss": 0.2956, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 6.485781797441539e-07, | |
| "loss": 0.298, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 4.4487900268882914e-07, | |
| "loss": 0.2876, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 2.4117982563350446e-07, | |
| "loss": 0.3056, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 3.748064857817975e-08, | |
| "loss": 0.2951, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_accuracy": 0.3271142251914616, | |
| "eval_f1": 0.5930562407498767, | |
| "eval_loss": 0.3666437268257141, | |
| "eval_roc_auc": 0.7306438427437314, | |
| "eval_runtime": 1044.3283, | |
| "eval_samples_per_second": 11.753, | |
| "eval_steps_per_second": 1.47, | |
| "step": 49092 | |
| } | |
| ], | |
| "max_steps": 49092, | |
| "num_train_epochs": 4, | |
| "total_flos": 2.601467760667853e+16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |