WIP: Add linux_aarch64 builds

[conda-forge-admin]

Fixes #13

Adds linux_aarch64 builds to dgl.

---

Hi! This is the friendly automated conda-forge-webservice.

I've started rerendering the recipe as instructed in #13.

If I find any needed changes to the recipe, I'll push them to this PR shortly. Thank you for waiting!

Here's a checklist to do before merging.

• Bump the build number if needed.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[jakirkham]

@conda-forge-admin, please re-render

[jakirkham]

Looks like we still need a few more things before this is viable based on this error from CI (also here's the attached log):

The reported errors are:
- Encountered problems while solving:
-   - nothing provides requested pytorch * *cuda*
-   - nothing provides requested nanoflann
-   - nothing provides requested metis 5.1.1.*

[mikemhenry]

metis 5.1.1.*

I can work on getting that to build, last time I tried I got errors that didn't really make any sense

pytorch * cuda

We can get at least CPU support then

nanoflann

Looks like this just needs to be merged? conda-forge/nanoflann-feedstock#15

[jakirkham]

Gotcha ok. Thanks for looking into it. Could you please link the PR once up and can take a look as well?

Yeah think CUDA ARM builds of PyTorch should be possible too. Originally the CUDA arch migrator waited for the arch migration to complete before starting a CUDA arch migration PR. There were issues with the PyTorch arch migration PR ( conda-forge/pytorch-cpu-feedstock#119 ) (particularly on linux_ppc64le), but linux_aarch64 builds were enabled as part of PR ( conda-forge/pytorch-cpu-feedstock#123 ). However as there is not much need for an additional CUDA arch migrator any more, the CUDA arch details were added to the global pinnings in PR ( conda-forge/conda-forge-pinning-feedstock#4733 ), which allows the regular arch migrator to enable CUDA builds. IOW just re-rendering should enable these CUDA arch builds for PyTorch

Thanks for the pointer! 🙏 Asked the team in PR ( conda-forge/nanoflann-feedstock#15 ) if they can take a look

[jakirkham]

@conda-forge-admin , please re-render

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I was trying to look for recipes to lint for you, but it appears we have a merge conflict.
Please try to merge or rebase with the base branch to resolve this conflict.

Please ping the 'conda-forge/core' team (using the @ notation in a comment) if you believe this is a bug.

[conda-forge-webservices]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

[jaimergp]

Synced with main to fetch the .ci_support/*yml stuff to do some debugging for #28

[jaimergp]

Hm, we need PyTorch + CUDA on linux-aarch64 for this to fully work with CUDA, and tensorflow for CPU builds:

• Add ARM + CUDA packages pytorch-cpu-feedstock#149
• Linux-aarch64 support? tensorflow-feedstock#136

[jakirkham]

Yeah couldn't remember the current state. So wanted to retest

Probably still a bit more work before this becomes viable