{ "best_metric": 0.85, "best_model_checkpoint": "distilhubert-finetuned-gtzan/checkpoint-64", "epoch": 2.986666666666667, "eval_steps": 500, "global_step": 64, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.23, "learning_rate": 9.523809523809525e-06, "loss": 0.0321, "step": 5 }, { "epoch": 0.47, "learning_rate": 1.904761904761905e-05, "loss": 0.0077, "step": 10 }, { "epoch": 0.7, "learning_rate": 2.8571428571428574e-05, "loss": 0.0345, "step": 15 }, { "epoch": 0.93, "learning_rate": 3.80952380952381e-05, "loss": 0.0009, "step": 20 }, { "epoch": 0.98, "eval_accuracy": 0.84, "eval_loss": 1.215333342552185, "eval_runtime": 71.621, "eval_samples_per_second": 1.396, "eval_steps_per_second": 0.237, "step": 21 }, { "epoch": 1.17, "learning_rate": 3.9153439153439155e-05, "loss": 0.0094, "step": 25 }, { "epoch": 1.4, "learning_rate": 3.80952380952381e-05, "loss": 0.0319, "step": 30 }, { "epoch": 1.63, "learning_rate": 3.703703703703704e-05, "loss": 0.0006, "step": 35 }, { "epoch": 1.87, "learning_rate": 3.597883597883598e-05, "loss": 0.0122, "step": 40 }, { "epoch": 1.96, "eval_accuracy": 0.82, "eval_loss": 1.2078099250793457, "eval_runtime": 72.4494, "eval_samples_per_second": 1.38, "eval_steps_per_second": 0.235, "step": 42 }, { "epoch": 2.1, "learning_rate": 3.4920634920634925e-05, "loss": 0.0043, "step": 45 }, { "epoch": 2.33, "learning_rate": 3.386243386243386e-05, "loss": 0.0004, "step": 50 }, { "epoch": 2.57, "learning_rate": 3.280423280423281e-05, "loss": 0.0003, "step": 55 }, { "epoch": 2.8, "learning_rate": 3.1746031746031745e-05, "loss": 0.0204, "step": 60 }, { "epoch": 2.99, "eval_accuracy": 0.85, "eval_loss": 1.3243606090545654, "eval_runtime": 72.2701, "eval_samples_per_second": 1.384, "eval_steps_per_second": 0.235, "step": 64 } ], "logging_steps": 5, "max_steps": 210, "num_train_epochs": 10, "save_steps": 500, "total_flos": 1.8401964823872e+17, "trial_name": null, "trial_params": null }