Install torchvision-extra-decoders from nightly channel

[NicolasHug]

No description provided.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/8963

• 📄 Preview Python docs built from this PR

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

❌ 6 New Failures, 7 Pending, 4 Unrelated Failures

As of commit 726eec8 with merge base de14de0 ():

NEW FAILURES - The following jobs have failed:

• Build Linux Wheels / pytorch/vision / upload / manywheel-py3_9-cuda12_6 (gh)
  Unable to download artifact(s): Artifact not found for name: pytorch_vision__3.9_cu126_x86_64
• Tests / unittests-linux (3.10, linux.12xlarge, cpu) / linux-job (gh)
  RuntimeError: Command docker exec -t 75cfe2788573d11b91d6ddd1b6ba9b0bda988dee8ad347e0a335506efd32a76e /exec failed with exit code 1
• Tests / unittests-linux (3.11, linux.12xlarge, cpu) / linux-job (gh)
  RuntimeError: Command docker exec -t fe03f3488d2bfc094a09b64a101c8ad00ad224790825c5a5b6f9b1506ba7dc43 /exec failed with exit code 1
• Tests / unittests-linux (3.12, linux.12xlarge, cpu) / linux-job (gh)
  RuntimeError: Command docker exec -t 0a8548d33ea76c32b0ffdf3268c64ff7f3f388328ad525e6e52da2b257a68703 /exec failed with exit code 1
• Tests / unittests-linux (3.9, linux.12xlarge, cpu) / linux-job (gh)
  RuntimeError: Command docker exec -t 4b237beb10819b07990e36ef80d3fb4d2565db154aee1bc9169ca6b8b7da5e11 /exec failed with exit code 1
• Tests / unittests-macos (3.11, macos-m1-stable) / macos-job (gh)
  RuntimeError: Command bash /Users/ec2-user/runner/_work/_temp/exec_script failed with exit code 1

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

• Build Linux Wheels / pytorch/vision / upload / manywheel-py3_9-cuda11_8 (gh) (detected as infra flaky with no log or failing log classifier)

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/vision / upload / wheel-py3_9-cuda-aarch6412_8-aarch64 (gh) (trunk failure)
  Unable to download artifact(s): Artifact not found for name: pytorch_vision__3.9_cu128_aarch64
• Build Linux Wheels / pytorch/vision / upload / manywheel-py3_9-cuda12_8 (gh) (trunk failure)
  Unable to download artifact(s): Artifact not found for name: pytorch_vision__3.9_cu128_x86_64
• Tests / unittests-windows (3.9, windows.g5.4xlarge.nvidia.gpu, cuda, 11.8) / windows-job (gh) (trunk failure)
  AttributeError: '_OpNamespace' 'image' object has no attribute '_jpeg_version'

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

[NicolasHug]

We're going from:

2025-03-17T12:57:12.1247843Z �[31mFAILED�[0m test/test_image.py::�[1mtest_decode_bad_encoded_data[decode_heic-Invalid input: No 'ftyp' box]�[0m - OSError: /opt/conda/envs/ci/lib/python3.10/site-packages/torchvision_extra_decoders/extra_decoders_lib.so: undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSsb

To

2025-03-17T12:56:28.1186348Z �[31mFAILED�[0m test/test_image.py::�[1mtest_decode_avif[decode_avif]�[0m - OSError: /opt/conda/envs/ci/bin/../lib/libstdc++.so.6: version `GLIBCXX_3.4.31' not found (required by /opt/conda/envs/ci/lib/python3.10/site-packages/torchvision_extra_decoders/../torchvision_extra_decoders.libs/libheif-2b91cd5e.so.1)

So this is still progress. Before we had a torch symbol issue, now we have a glibcxx symbol issue. But it's mostly because the machines we're testing on are older than what torchvision-extra-decoder was built on. Sigh. I'll merge this and follow-up. As a first step I'll deactivate the tests but I'll also try to look into building torchvision-extra-decoder` with older dependencies (libheif, libavif) which hopefully won't require recent version symbols.

[github-actions]

Hey @NicolasHug!

You merged this PR, but no labels were added.
The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py