8627 perceptual loss errors out after hitting the maximum number of downloads

[virginiafdez]

Fixes #8627. Moves the perceptual loss code to MONAI repository https://github.com/Project-MONAI/perceptual-models and the checkpoints to Huggingface.

Description

This PR changes and simplifies the torch.hub loading process and gets the models from Huggingface lirbary. A few sentences describing the changes proposed in this pull request.

Types of changes

• [x] Non-breaking change (fix or new feature that would not break existing functionality).
• [x] Breaking change (fix or new feature that would cause existing functionality to change).
• [N/A] New tests added to cover the changes.
• [N/A] Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• [x] Quick tests passed locally by running ./runtests.sh --quick --unittests --disttests.
• [x] In-line docstrings updated.
• [x] Documentation updated, tested make html command in the docs/ folder.

[coderabbitai[bot]]

Walkthrough

Perceptual loss backbones now load from Project-MONAI/perceptual-models via torch.hub. Added HF_MONAI_MODELS constant and renamed/expanded the enum to PerceptualNetworkType with new members (radimagenet_resnet50, medicalnet_resnet10_23datasets, medicalnet_resnet50_23datasets, resnet50). PerceptualLoss, MedicalNetPerceptualSimilarity, and RadImageNetPerceptualSimilarity accept a new cache_dir parameter and propagate it to constructors and torch.hub.load calls. MedicalNet models now enforce spatial_dims==3 and is_fake_3d==False; channel_wise is restricted to MedicalNet (non-MedicalNet with channel_wise raises). Model-name inputs are validated against HF_MONAI_MODELS. Axis/slicing bug in channel-wise feature differencing fixed (slice end index corrected). Docstrings and comments updated to reflect new hub sources and cache_dir behavior.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Areas needing extra attention:

• MedicalNet validation: enforcement of spatial_dims==3, is_fake_3d==False, and restriction of channel_wise to MedicalNet
• cache_dir propagation through PerceptualLoss → MedicalNetPerceptualSimilarity / RadImageNetPerceptualSimilarity → torch.hub.load
• torch.hub.load targets changed to Project-MONAI/perceptual-models and model-name validation against HF_MONAI_MODELS
• Channel-wise computation fix: corrected slice indexing for feats_diff
• Public API changes: new cache_dir parameters and renamed enum PerceptualNetworkType (check imports and backward-compatibility)

Pre-merge checks and finishing touches

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 25.00% which is insufficient. The required threshold is 80.00%.	You can run @coderabbitai generate docstrings to improve docstring coverage.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	Title clearly identifies the bug fix (issue #8627) by referencing the specific problem with perceptual loss download limits.
Description check	✅ Passed	Description links to the issue, explains the solution (moving to MONAI repo and HuggingFace), and documents the testing and documentation updates performed.
Linked Issues check	✅ Passed	Changes fully address issue #8627 by migrating models from Google Drive to HuggingFace and updating the torch.hub loading process to prevent download quota failures.
Out of Scope Changes check	✅ Passed	All changes in perceptual.py directly support the migration objective: new model sources, cache_dir propagation, validation enums, and strict MedicalNet handling are all within scope.

✨ Finishing touches

• [ ] 📝 Generate docstrings

🧪 Generate unit tests (beta)

• [ ] Create PR with unit tests
• [ ] Post copyable unit tests in a comment

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}