In this guideline, we see:
attribution = metadata_store_pb2.Attribution()
attribution.artifact_id = model_artifact_id
attribution.context_id = experiment_id
(...)
store.put_attributions_and_associations([attribution], [association])
An Attribution is created to record the relation between the Model Artifact (model after training) and the Context. This is expected because of the definition: "An attribution represents the relationship between artifacts and contexts".
However, I wonder why there isn't an Attribution created to record the relation between Input Artifact and the Context? In the guideline, data_artifact_id (the Input Artifact id) was not used to create an Attribution.
This might be not a bug, but rather an (unexplained) design principle.
