trainer_state.json

{
  "best_metric": 0.09309405833482742,
  "best_model_checkpoint": "BioM-ELECTRA-Large-SQuAD2-XLM/checkpoint-840",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 840,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_loss": 0.33242297172546387,
      "eval_runtime": 52.3029,
      "eval_samples_per_second": 42.789,
      "eval_steps_per_second": 5.353,
      "step": 280
    },
    {
      "epoch": 1.7857142857142856,
      "grad_norm": 22.950204849243164,
      "learning_rate": 8.095238095238097e-06,
      "loss": 0.5578,
      "step": 500
    },
    {
      "epoch": 2.0,
      "eval_loss": 0.13800670206546783,
      "eval_runtime": 52.4078,
      "eval_samples_per_second": 42.704,
      "eval_steps_per_second": 5.343,
      "step": 560
    },
    {
      "epoch": 3.0,
      "eval_loss": 0.09309405833482742,
      "eval_runtime": 52.4618,
      "eval_samples_per_second": 42.66,
      "eval_steps_per_second": 5.337,
      "step": 840
    }
  ],
  "logging_steps": 500,
  "max_steps": 840,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "total_flos": 4676504058860544.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}