Fixed CUDA-bound error where GPU-enabled run would result in incorrect model loading

[nopgadget]

Fix CUDA Face Garbling Issue

Fixed face garbling issues that occurred when running Deep-Live-Cam with CUDA execution provider by ensuring the correct face swap model is loaded.

Changes Made

• File: modules/processors/frame/face_swapper.py
• Lines: 67
• Change: -'model_name = "inswapper_128_fp16.onnx"', +'model_name = "inswapper_128.onnx"'

Technical Details

The fix ensures the appropriate inswapper_128.onnx model is loaded. This prevents the face garbling that was occurring due to model incompatibility with CUDA-enabled runs.

Testing

• ✅ All existing functionality preserved
• ✅ CUDA execution provider compatibility verified
• ✅ Face swapping quality maintained across different execution backends
• ✅ No regression in performance or accuracy

Impact

This change resolves the face garbling issue that users experienced when running with CUDA acceleration, while maintaining full compatibility with other execution providers.

Summary by Sourcery

Bug Fixes:

• Always use inswapper_128.onnx for CUDA execution provider, removing the fp16 model to prevent face garbling

[sourcery-ai[bot]]

Reviewer's guide (collapsed on small PRs)

Reviewer's Guide

Ensures the full-precision face swap model is used for CUDA execution to eliminate garbled output by replacing the FP16 model reference with the correct ONNX model.

Sequence diagram for face swapper model loading with CUDA provider

sequenceDiagram
    participant User
    participant FaceSwapper
    participant ModelZoo
    User->>FaceSwapper: get_face_swapper()
    FaceSwapper->>FaceSwapper: Check execution_providers
    FaceSwapper->>FaceSwapper: Set model_name = "inswapper_128.onnx"
    FaceSwapper->>ModelZoo: get_model(model_path, providers)
    ModelZoo-->>FaceSwapper: Return loaded model
    FaceSwapper-->>User: Return FACE_SWAPPER

Class diagram for updated face swapper model selection

classDiagram
    class FaceSwapper
    FaceSwapper : +get_face_swapper()
    FaceSwapper : -FACE_SWAPPER
    FaceSwapper : -model_name
    FaceSwapper : -model_path
    FaceSwapper : -models_dir
    FaceSwapper : -execution_providers
    FaceSwapper : Loads "inswapper_128.onnx" for all providers

File-Level Changes

Change	Details	Files
Load correct full-precision model for CUDA backend	Removed reference to the FP16 model when CUDAExecutionProvider is active Set model_name to inswapper_128.onnx by default and under the CUDA condition	modules/processors/frame/face_swapper.py

Possibly linked issues

• #123: PR changes inswapper_128_fp16.onnx to inswapper_128.onnx for CUDA, resolving face garbling issue described.
• #2: PR changes model from inswapper_128_fp16.onnx to inswapper_128.onnx to fix face garbling.

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an issue from a review comment by replying to it. You can also reply to a review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull request title to generate a title at any time. You can also comment @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in the pull request body to generate a PR summary at any time exactly where you want it. You can also comment @sourcery-ai summary on the pull request to (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the pull request to resolve all Sourcery comments. Useful if you've already addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull request to dismiss all existing Sourcery reviews. Especially useful if you want to start fresh with a new review - don't forget to comment @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[Meehey]

Thank you for the fix! But I would appreciate it beeing just 2 changed rows for model name (without recording functionality)

[nopgadget]

Thank you for the fix! But I would appreciate it beeing just 2 changed rows for model name (without recoding functionality)

whoops, that was me committing to the wrong branch just for my personal tinkering. i'll get that fixed now

[nopgadget]

@Meehey Let me know if that didn't fix it and I can figure it out further, but for me looks like I've reverted it back to just the one line change. Thank you!

[Meehey]

This should work. Also I think pull requests need to be made to premain branch first. Main is protected. see https://github.com/hacksider/Deep-Live-Cam/blob/main/CONTRIBUTING.md