Hi @Yuki,

Thank you for contacting the Databricks community.

• If you run register_model with the same run twice, you’ll create multiple versions pointing to the same source.

• To avoid that, you can check if the run is already registered before creating a new version:

   

  from mlflow.tracking import MlflowClient
  
  client = MlflowClient()
  existing = [
      v.source for v in client.search_model_versions(f"name='{model_name}'")
  ]
  if f"runs:/{best_run.run_id}/model" not in existing:
      result = mlflow.register_model(f"runs:/{best_run.run_id}/model", model_name)