checkpoint-3090/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.995957962813258,
  "eval_steps": 500,
  "global_step": 3090,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.81,
      "learning_rate": 4.9800000000000004e-05,
      "loss": 0.0634,
      "step": 500
    },
    {
      "epoch": 1.62,
      "learning_rate": 4.557592164542048e-05,
      "loss": 0.0823,
      "step": 1000
    },
    {
      "epoch": 2.43,
      "learning_rate": 3.381275475633684e-05,
      "loss": 0.072,
      "step": 1500
    },
    {
      "epoch": 3.23,
      "learning_rate": 1.8906193524453964e-05,
      "loss": 0.0483,
      "step": 2000
    },
    {
      "epoch": 4.04,
      "learning_rate": 6.193178695564386e-06,
      "loss": 0.0289,
      "step": 2500
    },
    {
      "epoch": 4.85,
      "learning_rate": 1.5889694102974583e-07,
      "loss": 0.012,
      "step": 3000
    }
  ],
  "logging_steps": 500,
  "max_steps": 3090,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 2.618615536631808e+16,
  "trial_name": null,
  "trial_params": null
}