Skip to content

[qwen] refactor attentions for vision/audio#38930

Merged
zucchini-nlp merged 6 commits intohuggingface:mainfrom
zucchini-nlp:qwen-attentions
Jun 24, 2025
Merged

[qwen] refactor attentions for vision/audio#38930
zucchini-nlp merged 6 commits intohuggingface:mainfrom
zucchini-nlp:qwen-attentions

Conversation

@zucchini-nlp
Copy link
Copy Markdown
Member

What does this PR do?

As per title, these modalities were left out when refactoring VLMs earlier

@zucchini-nlp zucchini-nlp requested a review from Cyrilvallez June 20, 2025 08:22
@HuggingFaceDocBuilderDev
Copy link
Copy Markdown

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

Copy link
Copy Markdown
Member

@Cyrilvallez Cyrilvallez left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Very very much welcome!! 🤗 Main concern is about the kwargs here, they are not flowing correctly in this PR, and I'm not sure if we can flow them all the way in general (e.g., some seqlens passed as kwargs for the language model would clash with those explicitly defined in the other modalities), i.e. does it make sense to have them?

Comment thread src/transformers/models/qwen2_5_omni/modular_qwen2_5_omni.py Outdated
Comment thread src/transformers/models/qwen2_vl/modeling_qwen2_vl.py Outdated
Comment thread src/transformers/models/qwen2_5_omni/modular_qwen2_5_omni.py
Comment thread src/transformers/models/qwen2_vl/modeling_qwen2_vl.py Outdated
@zucchini-nlp zucchini-nlp merged commit d3d835d into huggingface:main Jun 24, 2025
14 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants