trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 10,
  "global_step": 5,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.8,
      "learning_rate": 1.0000000000000002e-06,
      "loss": 32.25,
      "step": 1
    },
    {
      "epoch": 1.6,
      "learning_rate": 2.0000000000000003e-06,
      "loss": 19.375,
      "step": 2
    },
    {
      "epoch": 2.4,
      "learning_rate": 3e-06,
      "loss": 0.0,
      "step": 3
    },
    {
      "epoch": 3.2,
      "learning_rate": 4.000000000000001e-06,
      "loss": 7.625,
      "step": 4
    },
    {
      "epoch": 4.0,
      "learning_rate": 5e-06,
      "loss": 53.75,
      "step": 5
    },
    {
      "epoch": 4.0,
      "step": 5,
      "total_flos": 4.694388885356544e+16,
      "train_loss": 22.6,
      "train_runtime": 346.8366,
      "train_samples_per_second": 0.144,
      "train_steps_per_second": 0.014
    }
  ],
  "logging_steps": 1,
  "max_steps": 5,
  "num_train_epochs": 5,
  "save_steps": 98,
  "total_flos": 4.694388885356544e+16,
  "trial_name": null,
  "trial_params": null
}