Can the tensorflow and pytorch container use the cuda container base instead of requiring cuda to be installed on the base os? And if so would this be used in a multi step build process to make a new container?
Hi @jcwscience, I haven’t tried this but in theory it may be possible if you change the base image to be the CUDA base container instead of the l4t-base container. The base images get set in this script: https://github.com/dusty-nv/jetson-containers/blob/master/scripts/docker_base.sh
Thanks I’ll take a look!