This project aims to create a multi-platform docker image contains nvidia cuda, cudnn, and python pytorch.
We've provided build scripts for you.
Run the following command
./scripts/local_build.sh -j <threads>
where the threads
specifies the number of threads you want to use, just like make -j
.
If the number of threads doesn't specify, i.e.,
./scripts/local_build.sh
then the default value is 4.
First of all, login to your docker registry.
docker login ghcr.io/otischung
Then run the following command
./scripts/local_buildx_push.sh