Docker

Numba dpex now delivers docker support. Dockerfile is capable of building numba-dpex as well as direct dependencies for it: dpctl and dpnp. There are several prebuilt images available: for trying numba_dpex and for building numba-dpex.

Building

Numba dpex ships with multistage Dockerfile, which means there are different targets available for build. The most useful ones:

  • runtime

  • runtime-gpu

  • numba-dpex-builder-runtime

  • numba-dpex-builder-runtime-gpu

To build docker image

docker build --target runtime -t numba-dpex:runtime ./

To run docker image

docker run -it --rm numba-dpex:runtime

Note

If you are building docker image with gpu support it will calls github api to get latest versions of intel gpu drivers. You may face Github API call limits. To avoid this, you can pass your github credentials to increase this limit. You can do it by providing build args GITHUB_USER and GITHUB_PASSWORD. You can use access token <https://docs.github.com/en/authentication/keeping-your-account-and-data-secure/creating-a-personal-access-token> instead of the password:

Note

In case you are building docker image behind firewall and your internet access requires proxy, you can provide proxy build args http_proxy and https_proxy. Please note, that these args must be lowercase.

Dockerfile supports different python versions. To select the one you want, simply set PYTHON_VERSION build arg. By default docker image is based on official python image based on slim debian, so the requested python version must be from the available python docker images. In case you want to build on images on custom image you have to pass BASE_IMAGE environment variable. Be aware that Dockerfile is based on debian so any base image should be debian based, like debian or ubuntu.

Build arguments that could be useful:

  • PYTHON_VERSION

  • CR_TAG

  • IGC_TAG

  • CM_TAG

  • L0_TAG

  • ONEAPI_VERSION

  • DPCTL_GIT_BRANCH

  • DPCTL_GIT_URL

  • DPNP_GIT_BRANCH

  • DPNP_GIT_URL

  • NUMBA_DPEX_GIT_BRANCH

  • NUMBA_DPEX_GIT_URL

  • CMAKE_VERSION

  • CMAKE_VERSION_BUILD

  • INTEL_NUMPY_VERSION

  • INTEL_NUMBA_VERSION

  • CYTHON_VERSION

  • SCIKIT_BUILD_VERSION

  • http_proxy

  • https_proxy

  • GITHUB_USER

  • GITHUB_PASSWORD

  • BASE_IMAGE

Refer to Dockerfile to see all available

Running prebuilt images

An easy way you can try numba_dpex is by using prebuilt images. There are several prebuilt images available:

  • runtime package that provides runtime experience

ghcr.io/intelpython/numba-dpex/runtime:<numba_dpex_version>-py<python_version>[-gpu]

  • builder package that provides building experience

ghcr.io/intelpython/numba-dpex/builder:<numba_dpex_version>-py<python_version>[-gpu]

  • you can also see stages package, but it is used mostly for building stages.

You can use them to build your own docker that is built on top of one of them.

To try out numba dpex simply run:

docker run --rm -it ghcr.io/intelpython/numba-dpex/runtime:0.20.0-py3.10 bash

Within the container you can check available devices by running

sycl-ls

or

python -c "import dpctl; dpctl.lsplatform();".

Note

If you want to enable GPU you need to pass it within container and use *-gpu tag.

For passing GPU into container on linux use arguments --device=/dev/dri. However if you are using WSL you need to pass --device=/dev/dxg -v /usr/lib/wsl:/usr/lib/wsl instead.

So, for example, if you want to run numba dpex container with GPU support on WSL:

docker run --rm -it \
--device=/dev/dxg -v /usr/lib/wsl:/usr/lib/wsl \
ghcr.io/intelpython/numba-dpex/runtime:0.20.0-py3.10-gpu