SquishyQuokka
SquishyQuokka

How do you all deploy ML models for production?

Mostly, I have seen people in my team creating an API endpoint for it and deploy on a VM via uvicorn. Anything better?

12mo ago
Find out if you are being paid fairly.Download Grapevine
FloatingWaffle
FloatingWaffle

On a VM? Seems too manual.

First yes API end point is mandatory.. Something should have it.

  1. Kubernetes deployment with CI-CD enabled
  2. Use a model host and push model there. Your kubernetes service should always call the latest model from the model host.
SquishyQuokka
SquishyQuokka
Gojek12mo

What kind of a model host would you recommend? Huggingface?

Discover more
Curated from across