Move packing format used by int4 to int4_packing_format.py #2946
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/2946
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 715ec74 with merge base e7b310b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
jerryzh168
added
the
topic: improvement
jerryzh168
requested review from
kimishpatel,
mergennachin,
metascroy and
vkuzo
jerryzh168
force-pushed
the
int4-packing-format-refactor
branch
from
c11aef3 to
e55001b
Compare
Summary: We found that there is not much reuse of packing format, so we now plan to define packing format for each of the dtype (int4, float8, intx), instead of having a global packing_format that's used by all the tensors this reduces the interference between different dtype configs. This doesn't change tensor subclass, so no BC changes for tensor subclass. For v2 of Int4WeightOnlyConfig, it breaks BC, but we don't have any official models saved with this config yet, so it's fine. We also didn't add bc testing for this since it's not finalized yet. We'll add that later. Test Plan: Regression tests: python test/quantization/quantize_/workflows/int4/test_int4_marlin_sparse_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_opaque_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_preshuffled_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_tile_packed_to_4d_tensor.py python test/core/test_config.py python test/integration/test_load_and_run_checkpoint.py Reviewers: Subscribers: Tasks: Tags:
jerryzh168
force-pushed
the
int4-packing-format-refactor
branch
from
e55001b to
715ec74
Compare
metascroy
approved these changes
Sep 5, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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:
We found that there is not much reuse of packing format, so we now plan to define packing format for each of the dtype (int4, float8, intx), instead of having a global packing_format that's used by all the tensors. this reduces the interference between different dtype configs.
Changes
Int4PackingFormatenumpacking_formatarg in Int4WeightOnlyConfig and Float8DynamicActivationInt4WeightConfig to beint4_packing_formatThis doesn't change tensor subclass, so no BC changes for tensor subclass. For v2 of Int4WeightOnlyConfig, it breaks BC, but we don't have any official models saved with this config yet, so it's fine. We also didn't add bc testing for this since it's not finalized yet. We'll add that later.
Test Plan:
Regression tests:
python test/quantization/quantize_/workflows/int4/test_int4_marlin_sparse_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_opaque_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_preshuffled_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_tensor.py python test/quantization/quantize_/workflows/int4/test_int4_tile_packed_to_4d_tensor.py python test/core/test_config.py
python test/integration/test_load_and_run_checkpoint.py
Reviewers:
Subscribers:
Tasks:
Tags: