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.

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

Adds wrapper for siglip to allow tiling #1

Merged
merged 1 commit into from
May 30, 2024

Conversation

HuiyingLi
Copy link
Owner

This adds SigLIP tiling and visual token downsampling support. If grid_height/grid_width parameters are larger than 1, a number of SigLip tiles will be added to cover the image (with padding, using black pixels). To reduce number of tokens, feature blocks of size grid_width/grid_height will be also merged, increasing individual SigLIP feature dimension by grid_width*grid_heigth factor.

This choice of downsampling factor is arbitraty, downsampling can be done independent of the tiling. This particular factor keeps the number of visual features same as the original visual encoder.

What does this PR do ?

Add a one line overview of what this PR aims to accomplish.

Collection: [Note which collection this PR will affect]

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

GitHub Actions CI

The Jenkins CI system has been replaced by GitHub Actions self-hosted runners.

The GitHub Actions CI will run automatically when the "Run CICD" label is added to the PR.
To re-run CI remove and add the label again.
To run CI on an untrusted fork, a NeMo user with write access must first click "Approve and run".

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

This adds SigLIP tiling and visual token downsampling support. If grid_height/grid_width
parameters are larger than 1, a number of SigLip tiles will be added to cover the image
(with padding, using black pixels). To reduce number of tokens, feature blocks of size
grid_width/grid_height will be also merged, increasing individual SigLIP feature dimension
by grid_width*grid_heigth factor.

This choice of downsampling factor is arbitraty, downsampling can be done independent
of the tiling. This particular factor keeps the number of visual features same as the
original visual encoder.
@HuiyingLi HuiyingLi merged commit e6a3a53 into HuiyingLi:siglip_merge_llama3 May 30, 2024
4 of 5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants