Revise Linux CPU unittests CI workflow

[samanklesaria]

As we've decided against the approach to solving the CI ffmpeg>4 incompatibility issue by installing our own miniconda and ignoring the pytorch-infra environment (begun in #4093), this draft PR tries a different approach: using a standard github runner, as is used to build the docs for torchcodec. Note that this PR depends on files included in #4095 , so it has that PR as a target.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/audio/4098

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

⏳ 6 Pending, 7 Unrelated Failures

As of commit 84d2a99 with merge base f0a4999 ():

FLAKY - The following job failed but was likely due to flakiness present on trunk:

• .github/workflows/integration-test.yml (gh) (similar failure)

BROKEN TRUNK - The following jobs failed but was present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• Build Aarch64 Linux Wheels / pytorch/audio / build-wheel-py3_10-cuda-aarch6412_6-aarch64 (gh) (trunk failure)
  Process completed with exit code 1.
• Build Aarch64 Linux Wheels / pytorch/audio / build-wheel-py3_10-cuda-aarch6412_8-aarch64 (gh) (trunk failure)
  Process completed with exit code 1.
• Build Aarch64 Linux Wheels / pytorch/audio / build-wheel-py3_10-cuda-aarch6413_0-aarch64 (gh) (trunk failure)
  ImportError: libarm_compute.so: cannot open shared object file: No such file or directory
• Build Aarch64 Linux Wheels / pytorch/audio / upload / upload-wheel-py3_10-cuda-aarch6412_6-aarch64 (gh) (trunk failure)
  Unable to download artifact(s): Artifact not found for name: pytorch_audio__3.10_cu126_aarch64
• Build Aarch64 Linux Wheels / pytorch/audio / upload / upload-wheel-py3_10-cuda-aarch6412_8-aarch64 (gh) (trunk failure)
• Build Aarch64 Linux Wheels / pytorch/audio / upload / upload-wheel-py3_10-cuda-aarch6413_0-aarch64 (gh) (trunk failure)
  Unable to download artifact(s): Artifact not found for name: pytorch_audio__3.10_cu130_aarch64

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[samanklesaria]

Strange- why isn't CI running any more? Perhaps the force pushing confused github. I had to trigger it manually. The run is at https://github.com/pytorch/audio/actions/runs/17588954128

[samanklesaria]

I'm getting the deadly "Could not load libtorchcodec" error from an incorrectly installed ffmpeg. But conda should be installing ffmpeg correctly! Very confusing.

[pearu]

I'm getting the deadly "Could not load libtorchcodec" error from an incorrectly installed ffmpeg. But conda should be installing ffmpeg correctly! Very confusing.

I saw this as well. Adding

export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib:${LD_LIBRARY_PATH}

fixed that.