[wpyaoxueros]

What does this PR do?

Add isolated regression coverage for the old-TorchAO model-import failure fixed by #15095 (reported in #15085). This changes only tests/others/test_dependencies.py; the merged runtime fix is unchanged.

The test places a minimal TorchAO 0.14.1 package and distribution metadata on a subprocess's PYTHONPATH, verifies Diffusers recognizes the installed older version, and imports the real UNet2DModel and DDPMPipeline classes. The stand-in deliberately lacks the newer FqnToConfig export. It does not monkeypatch Diffusers availability/version helpers, install TorchAO, or download model weights. The subprocess isolates optional-dependency detection from the parent test process; the case skips when PyTorch is disabled.

Validation

All checks ran on the designated remote Linux workstation, Python 3.11.17 / PyTorch 2.6.0+cu124, against main 48cdbaef5ee3d718aaee392b44f2099ae511c4a5:

• python -m pytest -q tests/others/test_dependencies.py: 5 passed.
• The exact same new test with only the Fix model imports when torchao is older than 0.15 #15095 version guard reverted: 1 failed, with cannot import name 'FqnToConfig'; restoring the guard passes. No runtime modification is included in the PR.
• USE_TORCH=0 python -m pytest -q tests/others/test_dependencies.py -k old_torchao: 1 skipped.
• make style and make fix-copies completed; an unrelated generated loader-init change was reverted. Final changed-file Ruff check/format, utils/check_ai.py, utils/check_copies.py, utils/check_dummies.py, and git diff --check pass.

The full Diffusers suite and Windows/macOS were not run.

Self-review notes

Reviewed against .ai/references/review-rules.md, coding-style and testing guidance. No blocking technical findings or dead code found. The fixture represents the optional dependency's package/metadata boundary, and the regression demonstrably fails when the production guard is removed. No new pipeline, model, runtime behavior or documentation surface is added.

AI disclosure: Codex prepared the test, performed the remote validation and self-review, wrote this description, and submitted the PR on behalf of @wpyaoxueros at the contributor's explicit request. No independent human review or personally executed human tests are claimed.

[github-actions]

Hi @wpyaoxueros, thanks for the PR! It does not appear to link an issue it fixes. If this PR addresses an existing issue, please add a closing keyword (e.g. Fixes #1234) to the PR description so the issue is linked. See the contribution guide for more details. If this PR intentionally does not fix a tracked issue, a maintainer can add the no-issue-needed label to silence this reminder.

Please note that PRs without a linked issue are likely to be automatically closed 10 days after this notice.

Once the PR links an issue (or gets the no-issue-needed label), you can ignore this message — it stays here as a comment, but it no longer applies.