trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.0,
  "global_step": 5000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2,
      "learning_rate": 8e-05,
      "loss": 0.4177,
      "step": 1000
    },
    {
      "epoch": 0.4,
      "learning_rate": 6e-05,
      "loss": 0.1062,
      "step": 2000
    },
    {
      "epoch": 0.6,
      "learning_rate": 4e-05,
      "loss": 0.0974,
      "step": 3000
    },
    {
      "epoch": 0.8,
      "learning_rate": 2e-05,
      "loss": 0.0928,
      "step": 4000
    },
    {
      "epoch": 1.0,
      "learning_rate": 0.0,
      "loss": 0.0887,
      "step": 5000
    },
    {
      "epoch": 1.0,
      "step": 5000,
      "total_flos": 5.6319417604353024e+17,
      "train_runtime": 34435.0316,
      "train_samples_per_second": 0.145
    }
  ],
  "max_steps": 5000,
  "num_train_epochs": 1,
  "total_flos": 5.6319417604353024e+17,
  "trial_name": null,
  "trial_params": null
}