Support Qwen3-VL-4B with hidden states extraction and compression script - #2
Open
WhXcjm wants to merge 1 commit into
Open
Support Qwen3-VL-4B with hidden states extraction and compression script#2WhXcjm wants to merge 1 commit into
WhXcjm wants to merge 1 commit into
Conversation
There was a problem hiding this comment.
Pull request overview
This PR adds support for extracting decoder hidden states across different decoder-layer output formats (needed for Qwen3-VL-4B), and introduces a new convenience script for compressing/evaluating Qwen3-VL-4B.
Changes:
- Add a shared helper to consistently extract hidden states from decoder layer outputs during calibration/block search.
- Update block forward paths to use the new hidden-state extraction helper instead of assuming tuple outputs.
- Add a Qwen3-VL-4B compression shell script and ignore a new
submissions/output folder.
Reviewed changes
Copilot reviewed 3 out of 4 changed files in this pull request and generated 1 comment.
| File | Description |
|---|---|
scripts/compress_qwen3vl4b_jsq_hessian.sh |
Adds a runnable script for compressing Qwen3-VL-4B with JSQ settings and running lmms-eval tasks. |
jsq/compression/collector.py |
Adds _extract_hidden_states() and uses it when running decoder blocks to get correct hidden states across output formats. |
jsq/compression/block_search.py |
Uses _extract_hidden_states() in the lightweight forward path during MA-JSQ block search. |
.gitignore |
Ignores submissions/ directory. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Author
|
@copilot review current codes |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Introduce functionality to extract hidden states from decoder layer outputs and add a new compression script for Qwen3-VL-4B.