Slim Image

[noslouch]

My colleague trained a model that we'd like to host internally. I'm on a corporate deployment platform that sets a max size for docker images. In our case it's 3gb uncompressed. The ludwig base image is 4ish gb.

I guess I'm wondering if there's any interest in putting out a serve-only image that's only got what's needed for the /predict endpoint. Maybe that's already the case in the base image, but if there's stuff that can be dropped to helped shrink the image a bit, it would help us out!