trainer_state.json

{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 30.0,
  "eval_steps": 500,
  "global_step": 120,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 12.5,
      "grad_norm": 309.19952392578125,
      "learning_rate": 3.2916666666666668e-06,
      "loss": 5.09,
      "step": 50
    },
    {
      "epoch": 25.0,
      "grad_norm": 290.5226745605469,
      "learning_rate": 1.2083333333333333e-06,
      "loss": 3.4821,
      "step": 100
    },
    {
      "epoch": 30.0,
      "step": 120,
      "total_flos": 2.006776622592e+17,
      "train_loss": 4.1240334192911785,
      "train_runtime": 64.9739,
      "train_samples_per_second": 6.464,
      "train_steps_per_second": 1.847
    }
  ],
  "logging_steps": 50,
  "max_steps": 120,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 30,
  "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": 2.006776622592e+17,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}