With the new multiarch Docker infrastructure, you can run, say, python:3.6 and it will automatically choose the right image you based on the host architecture. Now, how can I override this and run a variant of the image built for another architecture, assuming that I have the appropriate qemu and binfmt support in place?
I'm not sure if this is consistent across all images, but for official images all it takes is prefixing an image name with the architecture:
docker run s390x/python:3.7-alpine3.8 python --version
docker run arm64v8/alpine uname -a
You can check the above repo for supported architectures for each release.