Error message when trying to send a JSON input data request to a model endpoint

Pass one more input parameter from the list of input columns in the endpoint’s model signature.

Written by Shyamprasad Miryala

Last published at: October 23rd, 2024

Problem

After loading your model using the Pyfunc interface in a notebook and passing it in as a Pandas DataFrame, you then try to send a JSON input data request to the model’s endpoint. In response you receive an error message. 

{"error_code": "BAD_REQUEST", "message": "Encountered an unexpected error while evaluating the model. Verify that the input is compatible with the model for inference. Error ''<' not supported between instances of 'NoneType' and 'NoneType'"}.

 

The issue occurs when the model is a feature_store flavor with a scikit-learn sub-flavor. 

 

Cause

There is a discrepancy between the endpoint and the loaded model in a notebook to get predictions, where the feature column is NULL for all rows. This causes the entire column to be None.  

 

Solution

Pass one more input parameter from the list of input columns in the endpoint’s model signature. This adjustment allows the endpoint to process the input correctly.