Correct the inaccurate docstring for int4_row_quantize() #4904
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
Summary: X-link: facebookresearch/FBGEMM#1930 When I was reading the code in `quantize.py`, the docstring in int4_row_quantize() seems inaccurate. The `int4_row_quantize()` function returns `wq` with shape `[N, K]`, not `[N, K // 2]`: ``` # Line 101: Concatenate chunks back to original K dimension out = torch.cat(out, dim=-1) # Line 104: Convert to int8 dtype out = out.to(dtype=torch.int8)` ``` So `wq` is `[N, K]` stored as `int8` elements, where each `int8` element contains a single int4 value. And after it, the actual packing could be done via: https://www.internalfb.com/code/fbsource/[c12bfdd174f7897f4615f98b630f2ac612c471fa]/fbcode/deeplearning/fbgemm/fbgemm_gpu/experimental/gen_ai/gen_ai/quantize.py?lines=18 Differential Revision: D82919480 Privacy Context Container: 151967047006994
✅ Deploy Preview for pytorch-fbgemm-docs ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
|
@geng-meta has exported this pull request. If you are a Meta employee, you can view the originating diff in D82919480. |
|
This pull request has been merged in 8ec3635. |
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Outdated suggestions cannot be applied.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.
Summary:
X-link: https://github.com/facebookresearch/FBGEMM/pull/1930
When I was reading the code in
quantize.py, the docstring in int4_row_quantize() seems inaccurate.The
int4_row_quantize()function returnswqwith shape[N, K], not[N, K // 2]:So
wqis[N, K]stored asint8elements, where eachint8element contains a single int4 value. And after it, the actual packing could be done via:https://www.internalfb.com/code/fbsource/[c12bfdd174f7897f4615f98b630f2ac612c471fa]/fbcode/deeplearning/fbgemm/fbgemm_gpu/experimental/gen_ai/gen_ai/quantize.py?lines=18
Differential Revision:
D82919480
Privacy Context Container: 151967047006994