Add Float8ActInt4WeightQATQuantizer #2289
Merged
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/2289
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 656d17d with merge base 4610850 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
andrewor14
commented
Jun 2, 2025
andrewor14
added
the
topic: new feature
vkuzo
reviewed
Jun 3, 2025
vkuzo
reviewed
Jun 3, 2025
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
2 times, most recently
from
452c147 to
620f676
Compare
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
2 times, most recently
from
c0b808c to
5e08eca
Compare
drisspg
reviewed
Jun 3, 2025
| @@ -17,6 +17,9 @@ | |||
| from torch.ao.quantization.fx._decomposed import quantized_decomposed_lib # noqa: F401 | |||
|
|
|||
| from torchao import quantize_ | |||
| from torchao.float8.config import ScalingGranularity | |||
There was a problem hiding this comment.
I kinda hate that we have ScalingGranuliarty and Ganularity of the other FP8 inference APIs
There was a problem hiding this comment.
I think this is worth fixing before landing. @andrewor14 , how about just using rowwise scaling (since I assume that the one you want) and removing the option to confugure it? That will at least keep this problem away from the BC surface of QAT in a way that we can more easily fix later.
drisspg
approved these changes
Jun 3, 2025
vkuzo
requested changes
Jun 3, 2025
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
from
5e08eca to
59ca3ca
Compare
|
Removed ScalingGranularity from the public API |
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
2 times, most recently
from
cf45f47 to
cfead5c
Compare
vkuzo
reviewed
Jun 4, 2025
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
from
cfead5c to
8269247
Compare
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
from
8269247 to
2a371fb
Compare
**Summary:** This commit adds a QAT quantizer that performs float8 dynamic activation + int4 symmetric per channel weight fake quantization. Note that there is no corresponding config for float8 QAT yet. This will be added in a future PR. **Test Plan:** python test/quantization/test_qat.py -k test_float8_fake_quantize python test/quantization/test_qat.py -k test_qat_fp8a4w_quantizer
andrewor14
force-pushed
the
fp8-int4-qat-quantizer
branch
from
2a371fb to
656d17d
Compare
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: This commit adds a QAT quantizer that performs float8 dynamic activation + int4 symmetric per channel weight fake quantization. Note that there is no corresponding config for float8 QAT yet. This will be added in a future PR.
Test Plan:
python test/quantization/test_qat.py -k test_float8_fake_quantize
python test/quantization/test_qat.py -k test_qat_fp8a4w_quantizer