Repository to host the Docker images files used in conda-forge
It can be useful to build and test these images locally. For many images
this can be done by running docker build from the root-level of this
repo and passing in the path to the Dockerfile that you wish to build.
For example:
docker build --rm --build-arg DISTRO_NAME=centos --build-arg DISTRO_VER=6 -f linux-anvil-comp7/Dockerfile .However, certain images, such as those building CUDA, will need to have environment variables passed in to be able to build. In this case, you will want to use a command similar to the following:
docker build --rm --build-arg DISTRO_NAME=centos --build-arg DISTRO_VER=6 --build-arg CUDA_VER=11.8.0 -f linux-anvil-cuda/Dockerfile .CUDA_VER: This is the cuda & cudatoolkit version that will be used. The value of this variable should be in major-minor-patch format, e.g.11.8.0.DISTRO_ARCH: This is the Linux architecture that the image should use. Should match the upstream Docker image, e.g.amd64.DISTRO_NAME: This is the Linux distro image name that should be built with. Should match the upstream Docker image, e.g.centos.DISTRO_VER: This is version of Linux distro (typical CentOS) that the image should be built with. This is the major-only version number, e.g.6or7. You'll usually want to build with the lowest working value for maximum compatibility.