[model] fix qwen2_5_vl fps #5306
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @Jintao-Huang, 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!
I've implemented a fix for the Qwen2.5-VL model to correctly handle video frames per second (FPS). Previously, the model might have been using a static or incorrect FPS value for video processing, leading to potential inaccuracies. This change ensures that the actual FPS of the input video is used, improving the precision of video-related computations within the model.
Highlights
- Video Metadata Handling: I've updated the replace_tag method in qwen.py to ensure that when video data is fetched, associated metadata, specifically frames per second (FPS), is also captured and stored alongside the video tensor.
- FPS Extraction: In the _encode method, I've modified how video inputs are processed. Instead of just extracting the video tensor, I now explicitly separate the video tensor from its corresponding FPS value for each video.
- Dynamic FPS Calculation: The second_per_grid_ts calculation, crucial for temporal processing in the Qwen2.5-VL model, has been updated to use the dynamically extracted FPS values for each video, replacing a previously fixed or assumed FPS value.
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 in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| 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 issue 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 or fill out our survey to provide feedback.
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
-
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. ↩
There was a problem hiding this comment.
Code Review
This pull request correctly fixes an issue with FPS (frames per second) handling for qwen2_5_vl models by fetching and using per-video FPS values instead of a global constant. The changes are well-implemented to propagate this information through the encoding process. My review includes one suggestion to refactor a small piece of code for better efficiency and conciseness.
| videos = [video[0] for video in inputs.videos] | ||
| fps = [video[1] for video in inputs.videos] |
There was a problem hiding this comment.
These two separate list comprehensions iterate over inputs.videos twice to unpack video data and FPS values. This can be made more efficient and Pythonic by using zip(*...) to perform the unpacking in a single pass.
| videos = [video[0] for video in inputs.videos] | |
| fps = [video[1] for video in inputs.videos] | |
| videos, fps = zip(*inputs.videos) if inputs.videos else ([], []) | |
| videos, fps = list(videos), list(fps) |
No description provided.