Seldon Core Microservice does not with Sklearn Pipeline or XGBoost model

[SelvinSelbaraju]

FYI this is my first time ever raising an issue on github for any repo so sorry if it is missing information!

Describe the bug

I have created a Custom Model Class defined here: https://github.com/SelvinSelbaraju/ab-testing-kubernetes/blob/master/services/PrepTimeModel.py. I am trying to test it locally using the seldon-core-microservice command. The server starts successfully, but anytime the model is called for prediction (line 27 in the linked file), the request just hangs and no response is ever received. It doesn't even time out, it hangs for several minutes. I am keen to understand why my model predictions cause the code to freeze.

An example JSON request would look like this: { "data": { "ndarray": [ [ "UK", "Chinese", 5, 6 ] ], "names": [ "country", "type_of_food", "order_value_gbp", "number_of_items" ] } }

It is also worth noting that I edited the source code for the seldon-core package (in accordance with this PR https://github.com/SeldonIO/seldon-core/pull/4114/files) so that I would no longer get issues.

To reproduce

1. Clone my repo
2. Edit the soruce code for the seldon-core package in accordance with the PR
3. Try running the seldon-core-microservice command and try and send requests with the above structure

Expected behaviour

The request should hang and nothing should happen

Environment

Python Version: 3.9.10 OS: MAC OS Version 12.1 (M1 Pro Macbook Pro) Seldon command ran through Poetry

This is all on my local machine, without even deploying to Kubernetes. I am simply trying to use the seldon-core-microservice command (without Docker or s2i).

[ukclivecox]

Can you show the curl you are using? Maybe also add debug mode for your wrapper and add debug in your predict method.

[github-actions[bot]]

This issue is stale because it has been open 10 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[ukclivecox]

closing