{ "best_metric": 0.6080508828163147, "best_model_checkpoint": "BERT-WMM/run-2/checkpoint-533", "epoch": 5.0, "eval_steps": 500, "global_step": 2665, "is_hyper_param_search": true, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.94, "grad_norm": 9.333391189575195, "learning_rate": 1.2018217372024236e-05, "loss": 0.7056, "step": 500 }, { "epoch": 1.0, "eval_accuracy": 0.7427230046948357, "eval_loss": 0.6080508828163147, "eval_runtime": 1.9882, "eval_samples_per_second": 1071.329, "eval_steps_per_second": 67.398, "step": 533 }, { "epoch": 1.88, "grad_norm": 10.452563285827637, "learning_rate": 9.790971750818855e-06, "loss": 0.4566, "step": 1000 }, { "epoch": 2.0, "eval_accuracy": 0.7647887323943662, "eval_loss": 0.6108278632164001, "eval_runtime": 2.0426, "eval_samples_per_second": 1042.799, "eval_steps_per_second": 65.603, "step": 1066 }, { "epoch": 2.81, "grad_norm": 15.208818435668945, "learning_rate": 7.563726129613473e-06, "loss": 0.2952, "step": 1500 }, { "epoch": 3.0, "eval_accuracy": 0.7723004694835681, "eval_loss": 0.7061524391174316, "eval_runtime": 2.0864, "eval_samples_per_second": 1020.91, "eval_steps_per_second": 64.226, "step": 1599 }, { "epoch": 3.75, "grad_norm": 71.34471893310547, "learning_rate": 5.3364805084080925e-06, "loss": 0.1924, "step": 2000 }, { "epoch": 4.0, "eval_accuracy": 0.7746478873239436, "eval_loss": 0.9748370051383972, "eval_runtime": 1.9954, "eval_samples_per_second": 1067.453, "eval_steps_per_second": 67.154, "step": 2132 }, { "epoch": 4.69, "grad_norm": 0.35381996631622314, "learning_rate": 3.1092348872027118e-06, "loss": 0.1118, "step": 2500 }, { "epoch": 5.0, "eval_accuracy": 0.7718309859154929, "eval_loss": 1.1978907585144043, "eval_runtime": 2.0749, "eval_samples_per_second": 1026.553, "eval_steps_per_second": 64.581, "step": 2665 } ], "logging_steps": 500, "max_steps": 3198, "num_input_tokens_seen": 0, "num_train_epochs": 6, "save_steps": 500, "total_flos": 846139862236032.0, "train_batch_size": 16, "trial_name": null, "trial_params": { "learning_rate": 1.4245462993229616e-05, "num_train_epochs": 6, "per_device_train_batch_size": 16, "weight_decay": 0.002285356379677195 } }