{ "best_metric": 90.00032091396297, "best_model_checkpoint": "bert-base-multilingual-cased-finetuned-jsquad2/checkpoint-4989", "epoch": 4.0, "global_step": 6652, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.3, "learning_rate": 2.819603126879134e-05, "loss": 0.7946, "step": 500 }, { "epoch": 0.6, "learning_rate": 2.6392062537582684e-05, "loss": 0.5012, "step": 1000 }, { "epoch": 0.9, "learning_rate": 2.458809380637402e-05, "loss": 0.4683, "step": 1500 }, { "epoch": 1.0, "eval_HasAns_exact": 75.28479445269936, "eval_HasAns_f1": 75.4044906719498, "eval_HasAns_total": 4038, "eval_NoAns_exact": 98.09418806111199, "eval_NoAns_f1": 98.09418806111199, "eval_NoAns_total": 6349, "eval_best_exact": 89.22691826321363, "eval_best_exact_thresh": 0.0, "eval_best_f1": 89.27345078784379, "eval_best_f1_thresh": 0.0, "eval_exact": 89.22691826321363, "eval_f1": 89.27345078784377, "eval_total": 10387, "step": 1663 }, { "epoch": 1.2, "learning_rate": 2.2784125075165364e-05, "loss": 0.3759, "step": 2000 }, { "epoch": 1.5, "learning_rate": 2.0980156343956704e-05, "loss": 0.3514, "step": 2500 }, { "epoch": 1.8, "learning_rate": 1.9176187612748048e-05, "loss": 0.3417, "step": 3000 }, { "epoch": 2.0, "eval_HasAns_exact": 76.91926696384348, "eval_HasAns_f1": 77.08023774145617, "eval_HasAns_total": 4038, "eval_NoAns_exact": 97.43266656166325, "eval_NoAns_f1": 97.43266656166325, "eval_NoAns_total": 6349, "eval_best_exact": 89.45797631654953, "eval_best_exact_thresh": 0.0, "eval_best_f1": 89.520554539328, "eval_best_f1_thresh": 0.0, "eval_exact": 89.45797631654953, "eval_f1": 89.520554539328, "eval_total": 10387, "step": 3326 }, { "epoch": 2.1, "learning_rate": 1.7372218881539384e-05, "loss": 0.309, "step": 3500 }, { "epoch": 2.41, "learning_rate": 1.5568250150330728e-05, "loss": 0.2527, "step": 4000 }, { "epoch": 2.71, "learning_rate": 1.376428141912207e-05, "loss": 0.2641, "step": 4500 }, { "epoch": 3.0, "eval_HasAns_exact": 77.46409113422486, "eval_HasAns_f1": 77.57140498596667, "eval_HasAns_total": 4038, "eval_NoAns_exact": 97.90518191841235, "eval_NoAns_f1": 97.90518191841235, "eval_NoAns_total": 6349, "eval_best_exact": 89.95860209877732, "eval_best_exact_thresh": 0.0, "eval_best_f1": 90.00032091396297, "eval_best_f1_thresh": 0.0, "eval_exact": 89.95860209877732, "eval_f1": 90.00032091396297, "eval_total": 10387, "step": 4989 }, { "epoch": 3.01, "learning_rate": 1.1960312687913411e-05, "loss": 0.2521, "step": 5000 }, { "epoch": 3.31, "learning_rate": 1.0156343956704751e-05, "loss": 0.1924, "step": 5500 }, { "epoch": 3.61, "learning_rate": 8.352375225496093e-06, "loss": 0.1895, "step": 6000 }, { "epoch": 3.91, "learning_rate": 6.548406494287432e-06, "loss": 0.1969, "step": 6500 }, { "epoch": 4.0, "eval_HasAns_exact": 77.73650321941555, "eval_HasAns_f1": 77.82730724781244, "eval_HasAns_total": 4038, "eval_NoAns_exact": 97.57442116868798, "eval_NoAns_f1": 97.57442116868798, "eval_NoAns_total": 6349, "eval_best_exact": 89.86232790988736, "eval_best_exact_thresh": 0.0, "eval_best_f1": 89.89762844581368, "eval_best_f1_thresh": 0.0, "eval_exact": 89.86232790988736, "eval_f1": 89.89762844581368, "eval_total": 10387, "step": 6652 } ], "max_steps": 8315, "num_train_epochs": 5, "total_flos": 1.3905168206462976e+16, "trial_name": null, "trial_params": null }