Mlflow model serve command Error: Bad object name 'app'

44 Views Asked by At

I am using MlFlow and am trying to serve a model that has been saved in the model registry. I ran the command mlflow model serve -m models:/registered_model_name/1. But I am getting the error: Error: Bad object name 'app'. Also the error contains a message:

There was an exception (AttributeError) importing your module.

It had these arguments: module 'datasets' has no attribute 'Split'

The error trace shows the error is generated while executing the command: waitress-serve --host=127.0.0.1 --port=5000 --ident=mlflow mlflow.pyfunc.scoring_server.wsgi:app'

How do I fix it?

I am trying to understand the error, and have followed the error trace. mlflow.pyfunc.scoring_server defines an app object. But when I try printing something in that file/module nothing is displayed, which is making it harder to debug. I don't think my model_uri is wrong, as I printed an environment variable scoring_server._SERVER_MODEL_PATH, before the command is tried to run, and it displays the correct path to saved artifacts.

0

There are 0 best solutions below