last-checkpoint/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0,
  "global_step": 8540,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "learning_rate": 9.825795424664739e-05,
      "loss": 7.898,
      "step": 4270
    },
    {
      "epoch": 1.0,
      "eval_cer": 0.18354978354978355,
      "eval_loss": 0.892418622970581,
      "eval_new_wer": 0.1,
      "eval_old_wer": 0.4714285714285714,
      "eval_runtime": 8.5895,
      "eval_samples_per_second": 24.448,
      "eval_steps_per_second": 3.143,
      "step": 4270
    },
    {
      "epoch": 2.0,
      "learning_rate": 9.124046805153827e-05,
      "loss": 0.9517,
      "step": 8540
    },
    {
      "epoch": 2.0,
      "eval_cer": 0.14545454545454545,
      "eval_loss": 0.8618147373199463,
      "eval_new_wer": 0.12857142857142856,
      "eval_old_wer": 0.30952380952380953,
      "eval_runtime": 8.4071,
      "eval_samples_per_second": 24.979,
      "eval_steps_per_second": 3.212,
      "step": 8540
    }
  ],
  "max_steps": 64050,
  "num_train_epochs": 15,
  "total_flos": 6.465255166176154e+18,
  "trial_name": null,
  "trial_params": null
}