trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "global_step": 2500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 4.4552e-05,
      "loss": 0.0239,
      "step": 500
    },
    {
      "epoch": 0.4,
      "learning_rate": 3.3413999999999996e-05,
      "loss": 0.0265,
      "step": 1000
    },
    {
      "epoch": 0.6,
      "learning_rate": 2.2276e-05,
      "loss": 0.0204,
      "step": 1500
    },
    {
      "epoch": 0.8,
      "learning_rate": 1.1138e-05,
      "loss": 0.0139,
      "step": 2000
    },
    {
      "epoch": 1.0,
      "learning_rate": 2.2276e-08,
      "loss": 0.0116,
      "step": 2500
    }
  ],
  "max_steps": 2500,
  "num_train_epochs": 1,
  "total_flos": 2467722643636224.0,
  "trial_name": null,
  "trial_params": null
}