Refactor QAT to use common fake_quantize_affine primitive #527
Conversation
This file contains 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: Currently there are two QAT quantizers, 8da4w and 4w. Today, these use different autograd functions to represent their fake quantization numerics, but this is not scalable because new QAT quantizers may introduce yet another divergent code path. To address this, this commit refactors both quantizers to use the common fake_quantize_affine QAT primitive. Test Plan: python test/quantization/test_qat.py Reviewers: jerryzh168 Subscribers: jerryzh168, supriyar, msaroufim
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/527
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 8486207 with merge base 6dd82d8 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
jerryzh168
reviewed
Jul 20, 2024
| @@ -25,7 +25,10 @@ | |||
| ZeroPointDomain, | |||
| ) | |||
| from torchao.quantization.unified import TwoStepQuantizer | |||
| from torchao.quantization.utils import get_group_qparams_symmetric | |||
| from torchao.quantization.utils import ( | |||
| _get_per_token_block_size, | |||
There was a problem hiding this comment.
if it's helpful we could have a general util like:
def get_block_size(granularity, **kw_params) -> Callable:
if granularity == Granularity.PER_BLOCK:
...
elif type == Granularity.PER_TOKEN:
...
...
block_size = get_block_size(Granularity.PER_TOKEN)(x)
There was a problem hiding this comment.
Sounds good, let's do that separately
jerryzh168
approved these changes
Jul 20, 2024
Hanxian97
pushed a commit
that referenced
this pull request
Jul 24, 2024
Summary: Currently there are two QAT quantizers, 8da4w and 4w. Today, these use different autograd functions to represent their fake quantization numerics, but this is not scalable because new QAT quantizers may introduce yet another divergent code path. To address this, this commit refactors both quantizers to use the common fake_quantize_affine QAT primitive. Test Plan: python test/quantization/test_qat.py Reviewers: jerryzh168 Subscribers: jerryzh168, supriyar, msaroufim
dbyoung18
pushed a commit
to dbyoung18/ao
that referenced
this pull request
Jul 31, 2024
Summary: Currently there are two QAT quantizers, 8da4w and 4w. Today, these use different autograd functions to represent their fake quantization numerics, but this is not scalable because new QAT quantizers may introduce yet another divergent code path. To address this, this commit refactors both quantizers to use the common fake_quantize_affine QAT primitive. Test Plan: python test/quantization/test_qat.py Reviewers: jerryzh168 Subscribers: jerryzh168, supriyar, msaroufim
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: Currently there are two QAT quantizers, 8da4w and 4w. Today, these use different autograd functions to represent their fake quantization numerics, but this is not scalable because new QAT quantizers may introduce yet another divergent code path. To address this, this commit refactors both quantizers to use the common fake_quantize_affine QAT primitive.
Test Plan:
python test/quantization/test_qat.py
Reviewers: jerryzh168
Subscribers: jerryzh168, supriyar, msaroufim