last-checkpoint/trainer_state.json

{
  "best_metric": 2.020920515060425,
  "best_model_checkpoint": "dq158/morbius/checkpoint-790",
  "epoch": 0.9993674889310563,
  "eval_steps": 500,
  "global_step": 790,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.63,
      "learning_rate": 7e-06,
      "loss": 2.15,
      "step": 500
    },
    {
      "epoch": 1.0,
      "eval_bleu": 1.0,
      "eval_brevity_penalty": 1.0,
      "eval_length_ratio": 1.0,
      "eval_loss": 2.020920515060425,
      "eval_precisions": [
        1.0,
        1.0,
        1.0,
        1.0
      ],
      "eval_reference_length": 51925,
      "eval_runtime": 586.8865,
      "eval_samples_per_second": 4.79,
      "eval_steps_per_second": 0.6,
      "eval_translation_length": 51925,
      "step": 790
    }
  ],
  "logging_steps": 500,
  "max_steps": 15800,
  "num_train_epochs": 20,
  "save_steps": 500,
  "total_flos": 1.7318198200762368e+16,
  "trial_name": null,
  "trial_params": null
}