fix cif_predictor.py bug in mac GPU

[yileld]

When the result of fire >= threshold is an 0-dim tensor, torch.where will crash in Mac mps GPU.

File "/asr_test/lib/python3.9/site-packages/funasr/models/bicif_paraformer/cif_predictor.py", line 65, in cif
    l = torch.index_select(frames[b, :, :], 0, torch.nonzero(fire >= threshold).squeeze())
IndexError: Dimension specified as -1 but tensor has no dimensions

This step should be run in cpu.

[gemini-code-assist]

Summary of Changes

Hello @yileld, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request addresses a specific runtime error encountered by users leveraging Mac MPS GPUs with the cif_predictor.py module. The core problem was an IndexError triggered by torch.index_select when it received a 0-dimensional tensor on the MPS backend. The solution intelligently routes this particular tensor operation to the CPU under the identified conditions, ensuring the module functions correctly on Apple Silicon without compromising performance on other devices.

Highlights

• Mac MPS GPU Compatibility Fix: Resolved an IndexError occurring in cif_predictor.py when running on Mac systems with MPS (Metal Performance Shaders) GPUs. The issue arose when torch.nonzero(...).squeeze() produced a 0-dimensional tensor, causing torch.index_select to crash on MPS.
• Conditional CPU Offload: Implemented a conditional check within the cif function to detect if the problematic 0-dimensional tensor is encountered on an MPS device. If so, the torch.index_select operation is temporarily executed on the CPU before moving the result back to the original device, preventing the crash.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request addresses a crash on MPS-powered Macs that occurs when torch.index_select receives a 0-dimensional tensor as an index. The proposed solution is a workaround that moves the operation to the CPU for this specific case.

My review provides a more robust and efficient alternative. Instead of a device-specific check and data transfer, I suggest modifying the tensor manipulation to ensure the index is always 1-dimensional by using .squeeze(-1) instead of .squeeze(). This is a cleaner solution that avoids performance penalties.

Additionally, the same pattern of code exists in funasr/models/paraformer/cif_predictor.py and could lead to the same bug. It would be beneficial to apply the recommended fix there as well.