Are Inference Endpoints automatically updated with new pushes?

I pushed changes to our handler.py file and pushed to our HuggingFace repo which is connected to an Inference Endpoint. Does the endpoint automatically update with the latest code or do we need to redeploy the endpoint?

You need to re-deploy. Note that at present you also cannot update the endpoint with a new model version, you will need to deploy a whole new endpoint. Hoping this may change in the future so that the endpoint url doesn’t change each time a new version is deployed (CC @jeffboudier).

Also check out GitHub - MantisAI/hugie: Command Line Interface for Hugging Face Inference Endpoints, it may be useful.

1 Like