This repository was archived by the owner on Sep 13, 2023. It is now read-only.

Description
It seems that, to deploy to Heroku and Fly.io, you don’t have to build and upload the docker image. But those don’t offer GPU instances. So, for a model using GPUs, I have to build and upload a docker image (which are around 8GB). This can take up to 1 hour, with Mexico’s slow internet (when there is no internet interruption). This negates the advantage of having a mlem folder of small size (<1GB)