Description
'import_logged_model' function in 'mlflow_export_import/logged_model/import_logged_model.py' does not return any useful information after successfully importing a logged model. This makes it difficult to programatically retrieve the imported model's 'run_id','model_id' and other metadata in automated workflows.
Current behaviour
The function only returns a dictionary in the unsupported MLflow version case:
if not has_logged_model_support():
return {"unsupported": True, "mlflow_version": mlflow.__version__}
Otherwise, it returns None after successful import.
### Expected behaviour
The function should return a dictionary containing metadata about imported model:
return{
"run_id": run_id,
"model_id": logged_model.model_id,
"experiment_id": exp.experiment_id,
"experiment_name": exp.name,
"model_name": src_logged_model_dct["name"],
"status": "success"
}
Current workaround requires querying mlflow after import which is less reliable.
Description
'import_logged_model' function in 'mlflow_export_import/logged_model/import_logged_model.py' does not return any useful information after successfully importing a logged model. This makes it difficult to programatically retrieve the imported model's 'run_id','model_id' and other metadata in automated workflows.
Current behaviour
The function only returns a dictionary in the unsupported MLflow version case:
Current workaround requires querying mlflow after import which is less reliable.