trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 10.0,
  "eval_steps": 500.0,
  "global_step": 2040,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.450980392156863,
      "grad_norm": 32.19416046142578,
      "learning_rate": 1.5e-05,
      "loss": 1.0988,
      "step": 500
    },
    {
      "epoch": 4.901960784313726,
      "grad_norm": 32.042449951171875,
      "learning_rate": 3e-05,
      "loss": 0.4359,
      "step": 1000
    },
    {
      "epoch": 7.352941176470588,
      "grad_norm": 16.227807998657227,
      "learning_rate": 1.557692307692308e-05,
      "loss": 0.2055,
      "step": 1500
    },
    {
      "epoch": 9.803921568627452,
      "grad_norm": 7.223657131195068,
      "learning_rate": 1.153846153846154e-06,
      "loss": 0.0488,
      "step": 2000
    },
    {
      "epoch": 10.0,
      "step": 2040,
      "total_flos": 3.031314932000124e+16,
      "train_loss": 0.4388976242028031,
      "train_runtime": 38108.8591,
      "train_samples_per_second": 13.675,
      "train_steps_per_second": 0.054
    }
  ],
  "logging_steps": 500,
  "max_steps": 2040,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 10,
  "save_steps": 1000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 3.031314932000124e+16,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}