feat: support zero-point decompression for asymmetric quantization (packed) #463
Conversation
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
- Fix decompress_weight method in PackedQuantizationCompressor to support unpacking zero-points - Add comprehensive tests for zero-point packing/unpacking with GROUP and CHANNEL strategies - Add end-to-end integration tests for asymmetric quantization workflow - Ensure packed tensors are contiguous for safetensors compatibility Resolves issue referenced in vllm-project/llm-compressor#1704
…move manual creation
…v similarity; cleanup temp usage
Etelis
force-pushed
the
fix/zero-point-decompression
branch
from
2cf7124 to
126fc89
Compare
|
I've addressed all your concerns. |
dsikka
reviewed
Sep 13, 2025
| } | ||
|
|
||
| state_dict = model.state_dict() | ||
| compressed_state_dict = compressor.compress( |
There was a problem hiding this comment.
It may be simpler to use compress_model / decompress_model, both of which will work on the model in memory
compress / decompress are primarily meant for disk
| if hasattr(module, param_name): | ||
| getattr(module, param_name).data = value.clone() | ||
| else: | ||
| module.register_parameter( |
There was a problem hiding this comment.
Can you explain why we need to separately register parameters here?
| if hasattr(module, param_name): | ||
| getattr(module, param_name).data = value.clone() | ||
| else: | ||
| module.register_parameter( |
There was a problem hiding this comment.
This looks good pending @dsikka 's comments, thanks for updating!
rahul-tuli
reviewed
Sep 17, 2025
There was a problem hiding this comment.
LGTM pending @dsikka 's comments! Thank you for this contribution
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.
Follow-up to PR #459 (which was closed due to an accidental history rewrite). The branch has been restored to the pre-rewrite tip (2cf7124).
This PR:
Refs vllm-project/llm-compressor#1704.