trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 16.0,
  "eval_steps": 500,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 2.0,
      "learning_rate": 0.0,
      "loss": 1.2573,
      "step": 25
    },
    {
      "epoch": 4.0,
      "learning_rate": 0.0002,
      "loss": 0.5919,
      "step": 50
    },
    {
      "epoch": 6.0,
      "learning_rate": 0.0,
      "loss": 0.273,
      "step": 75
    },
    {
      "epoch": 8.0,
      "learning_rate": 0.0002,
      "loss": 0.1305,
      "step": 100
    },
    {
      "epoch": 10.0,
      "learning_rate": 0.0,
      "loss": 0.0698,
      "step": 125
    },
    {
      "epoch": 12.0,
      "learning_rate": 0.0002,
      "loss": 0.0456,
      "step": 150
    },
    {
      "epoch": 14.0,
      "learning_rate": 0.0,
      "loss": 0.0342,
      "step": 175
    },
    {
      "epoch": 16.0,
      "learning_rate": 0.0002,
      "loss": 0.0276,
      "step": 200
    }
  ],
  "logging_steps": 25,
  "max_steps": 200,
  "num_train_epochs": 17,
  "save_steps": 200,
  "total_flos": 1.4083057682703974e+17,
  "trial_name": null,
  "trial_params": null
}