Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Faster / more memory efficient Qwen VL #114

Merged
merged 1 commit into from
Nov 14, 2024
Merged

Faster / more memory efficient Qwen VL #114

merged 1 commit into from
Nov 14, 2024

Conversation

awni
Copy link
Contributor

@awni awni commented Nov 14, 2024

  • Remove the specialized RoPE in the language model as it is equivalent to a regular RoPE in that case. Using the MLX fast RoPE is also faster
  • Avoid accidentally upcasting everything to fp32. This can make a big difference in memory use, especially for lots of images or a videos.

Blaizzy
Blaizzy approved these changes Nov 14, 2024
View reviewed changes
Copy link
Owner

@Blaizzy Blaizzy left a comment

Choose a reason for hiding this comment

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

LGTM!

@Blaizzy
Copy link
Owner

Blaizzy commented Nov 14, 2024

Thanks a lot @awni!

This brings another significant boost in performance.

I will review other models as soon as I finish Molmo #112

@Blaizzy Blaizzy merged commit 95f0102 into Blaizzy:main Nov 14, 2024
1 check passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Blaizzy Blaizzy Blaizzy approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@awni @Blaizzy