trainer_state.json

{
  "best_metric": 3.6615889072418213,
  "best_model_checkpoint": "flan-t5-small-flan_qa\\checkpoint-25",
  "epoch": 5.0,
  "global_step": 25,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "eval_gen_len": 10.222222222222221,
      "eval_loss": 5.466943740844727,
      "eval_rouge1": 3.1746,
      "eval_rouge2": 0.0,
      "eval_rougeL": 3.1746,
      "eval_rougeLsum": 3.1746,
      "eval_runtime": 6.5462,
      "eval_samples_per_second": 1.375,
      "eval_steps_per_second": 0.306,
      "step": 5
    },
    {
      "epoch": 2.0,
      "eval_gen_len": 10.444444444444445,
      "eval_loss": 4.63550329208374,
      "eval_rouge1": 3.1746,
      "eval_rouge2": 0.0,
      "eval_rougeL": 3.1746,
      "eval_rougeLsum": 3.1746,
      "eval_runtime": 6.5259,
      "eval_samples_per_second": 1.379,
      "eval_steps_per_second": 0.306,
      "step": 10
    },
    {
      "epoch": 3.0,
      "eval_gen_len": 7.222222222222222,
      "eval_loss": 4.082330703735352,
      "eval_rouge1": 7.4074,
      "eval_rouge2": 0.0,
      "eval_rougeL": 7.4074,
      "eval_rougeLsum": 7.4074,
      "eval_runtime": 6.106,
      "eval_samples_per_second": 1.474,
      "eval_steps_per_second": 0.328,
      "step": 15
    },
    {
      "epoch": 4.0,
      "eval_gen_len": 5.555555555555555,
      "eval_loss": 3.769094705581665,
      "eval_rouge1": 7.4074,
      "eval_rouge2": 0.0,
      "eval_rougeL": 7.4074,
      "eval_rougeLsum": 7.4074,
      "eval_runtime": 6.3722,
      "eval_samples_per_second": 1.412,
      "eval_steps_per_second": 0.314,
      "step": 20
    },
    {
      "epoch": 5.0,
      "eval_gen_len": 5.333333333333333,
      "eval_loss": 3.6615889072418213,
      "eval_rouge1": 7.4074,
      "eval_rouge2": 0.0,
      "eval_rougeL": 7.4074,
      "eval_rougeLsum": 7.4074,
      "eval_runtime": 5.3258,
      "eval_samples_per_second": 1.69,
      "eval_steps_per_second": 0.376,
      "step": 25
    }
  ],
  "max_steps": 25,
  "num_train_epochs": 5,
  "total_flos": 1916995829760.0,
  "trial_name": null,
  "trial_params": null
}