Registered models with AML SDK v2 doesn't appear in pipeline

Question

I have successfully been able to register a model in AML workspace, using SDK v2, however, when I navigate to AML workspace UI and the given pipeline, the registered model don't appear:

I have registered the models using both this method:

cloud_model = Model(
    path=model_path,
    name="cloud-path-example",
    type=AssetTypes.CUSTOM_MODEL,
    description="Model created from cloud path.",
)
ml_client.models.create_or_update(cloud_model)

and this method:

mlflow.spacy.save_model(nlp, output_folder)
model_path = os.path.abspath(output_folder)
mlflow.register_model(f"file://{model_path}", "model-test")

Both methods successfully registers the model but the "connection" between the registered model and the pipeline run is missing. What could the reason for this be? My impression is that this "connection" should appear automatically?

Answer 1

Ok, so answering my own question, model path has to be on the form azureml://jobs/<run-id>/outputs/artifacts/paths/model/ or similar in order for the connection to happen.

https://github.com/Azure/azureml-examples/blob/main/sdk/python/assets/model/model.ipynb