8

While trying to run the docker image provided by fastai --> fastpages-jekyll I am running into several issues.

This is a linux/amd64 image. I suspect that if there was a linux/arm64 image then it would run smoothly on my mackbook air M1 having apple silicon chip. I could not find such a pre-existing image on docker hub.

Also I cannot find Dockerfile associated with fastpages-jekyll image.

Is it possible to build an arm64 image from the amd64 image somehow ?
What are my options ?

pX0r
  • 1,472
  • 12
  • 14
  • I have basically the same question; my impression is that there is no such docker image yet. I managed to set up an (amazingly fast) tensorflow with the metal GPU plugin, but could not find a docker image yet. (And I found out that Docker for Mac would not allow GPU use, so my motivation is reduced anyhow now, but I would still be interested.) – hans_meine May 06 '22 at 11:56

1 Answers1

0

This github is the source for pre-built arm64 (Apple Silicon) docker images with PyTorch. That said, you'll still need to install fastai inside this image.

I am currently working on a new version of my "native" fastai image (which uses the above image as a base image) that will not require sonisa's base image and will include Ruby (so you can run a Ruby on Rails server that can access pickle files created by fastai). However, that's still a work in progress. Watch this repo as I hope to publish the new docker file in the coming days.

Ted Stresen-Reuter
  • 392
  • 2
  • 13