trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 50,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "grad_norm": 0.4947165846824646,
      "learning_rate": 0.00021314021436425024,
      "loss": 0.3985,
      "step": 20
    },
    {
      "epoch": 4.0,
      "grad_norm": 0.6319659352302551,
      "learning_rate": 3.228009220820085e-05,
      "loss": 0.082,
      "step": 40
    },
    {
      "epoch": 5.0,
      "step": 50,
      "total_flos": 1.326817135982592e+17,
      "train_loss": 0.1957190614938736,
      "train_runtime": 1416.6552,
      "train_samples_per_second": 0.565,
      "train_steps_per_second": 0.035
    }
  ],
  "logging_steps": 20,
  "max_steps": 50,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 6,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.326817135982592e+17,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}