trainer_state.json

{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 48,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.625,
      "grad_norm": 6.109557766701803,
      "learning_rate": 9.78800299954203e-06,
      "loss": 1.8311,
      "step": 10
    },
    {
      "epoch": 1.25,
      "grad_norm": 3.414783530388519,
      "learning_rate": 7.604701702439652e-06,
      "loss": 0.4788,
      "step": 20
    },
    {
      "epoch": 1.875,
      "grad_norm": 3.300993172504628,
      "learning_rate": 4.091815745102818e-06,
      "loss": 0.3025,
      "step": 30
    },
    {
      "epoch": 2.5,
      "grad_norm": 2.496541962568354,
      "learning_rate": 1.04251755785373e-06,
      "loss": 0.1666,
      "step": 40
    },
    {
      "epoch": 3.0,
      "step": 48,
      "total_flos": 9109050359808.0,
      "train_loss": 0.597065252562364,
      "train_runtime": 152.3178,
      "train_samples_per_second": 19.696,
      "train_steps_per_second": 0.315
    }
  ],
  "logging_steps": 10,
  "max_steps": 48,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "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": 9109050359808.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}