Re-enable CUDA 11.8 on Linux ARM

[conda-forge-admin]

Fixes #28

Thanks to the recent Docker image updates in the global pinnings ( conda-forge/conda-forge-pinning-feedstock#6626 ). Not to mention fixes on the cuDNN GLIBC constraint side ( conda-forge/cudnn-feedstock#85 ). It is now possible to re-enable CUDA 11.8 builds on Linux ARM

So this PR...

1. Drops the skip
2. Re-renders
3. Bumps the build/number

Which should be all that is needed to add these builds back

[conda-forge-admin]

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/meta.yaml) and found it was in an excellent condition.

[jakirkham]

@conda-forge-admin , please re-render

[jakirkham]

Could a maintainer please push a commit to start cirun?

[h-vetinari]

Just a note that after this PR, a single CI run monopolizes the entirety of the open-gpu server for >24h, so please be judicious about triggering CI, and be proactive about cancelling runs you don't need (e.g. in this PR, all the jobs that didn't change or get added could be cancelled, and then run CI for a aarch+CUDA 11.8 for a single python version to see if it works).

I kind of doubt that supporting 11.8 at all here is still worth the use of this very limited & shared resource, but if you think so, fine.

[h-vetinari]

While you debug the CUDA 11.8 + aarch builds, please skip everything else here to conserve resources.

build:
  # debug
  skip: true  # [not (aarch64 and cuda_compiler_version == "11.8")]

since all those builds failed, I cancelled the rest of the jobs.

[carterbox]

The builds failed because the solver could not find a libtorch package for CUDA 11 and arm. If pytorch has already dropped support for this combo, then we do not need it here.

[jakirkham]

Yeah was noticing the same thing. Agreed think we can close this as won't fix

Sorry for the noise