client.create_model_version(MODEL_NAME, source, run_id, description) giving AttributeError: 'str' object has no attribute 'items'

Question

I have a function to handle deployment of a Model from an Experiment to the Model Registry. I copied code from the example in the documentation and run it on a docker container on a Google Compute Engine Instance but I am getting an error:

Traceback (most recent call last):
  File "/app/train.py", line 214, in <module>
    deploy_model_to_staging(client=client, run=training, run_date=args.run_date)
  File "/app/train.py", line 162, in deploy_model_to_staging
    client.create_model_version(MODEL_NAME, source, run_id, description)
  File "/usr/local/lib/python3.9/site-packages/mlflow/tracking/client.py", line 2658, in create_model_version
    return self._create_model_version(
  File "/usr/local/lib/python3.9/site-packages/mlflow/tracking/client.py", line 2575, in _create_model_version
    return self._get_registry_client().create_model_version(
  File "/usr/local/lib/python3.9/site-packages/mlflow/tracking/_model_registry/client.py", line 193, in create_model_version
    tags = [ModelVersionTag(key, str(value)) for key, value in tags.items()]
AttributeError: 'str' object has no attribute 'items'

The function being called is:

def deploy_model_to_staging(
    client: mlflow.client.MlflowClient, run: mlflow.entities.Run, run_date: str
) -> None:
    """
    Deploy model to Model Registry with stage "Staging".

    Args:
      client (mlflow.client.MlflowClient): MLflow Client.
      run (mlflow.entities.Run): MLflow Run for training.
    """
    description = f"Model trained on date {run_date}"
    run_id = run.info.run_id

    print(f"Creating Model Version for Run ID: {run_id}")
    logging.info(f"Creating Model Version for Run ID: {run_id}")
    runs_uri = f"runs:/{run_id}/model"  # URI for the training run.
    source = RunsArtifactRepository.get_underlying_uri(runs_uri)
    client.create_model_version(MODEL_NAME, source, run_id, description)

    version = client.search_model_version(filter_string=f"run_id='{run_id}'")[0].version
    print(f"Model Version {version} created.")
    logging.info(f"Model Version {version} created.")

    client.transition_model_version_stage(
        name=MODEL_NAME, version=version, stage="Staging"
    )
    print(f"Model Version {version} transitioned to Staging.")
    logging.info(f"Model Version {version} transitioned to Staging.")

Does anyone know what is causing the Attribute Error here?

Answer 1

tags is the 4-th arg, description is the 6-th

https://mlflow.org/docs/latest/python_api/mlflow.client.html#mlflow.client.MlflowClient.create_model_version