Support quantize_ for convolution on XNNPACK
Open
@GregoryComer is already working on this.
Since Feb 23, 2026.
module: xnnpack
- Dominant language
- Python
- Stars
- 5k
- Forks
- 1.2k
- Avg merge
- 2d 10h
- Merged PRs (30d)
- 581
Description
XNNPACK's partitioner and lowering logic doesn't recognize the quantize_ (de)quantize_affine representation in the IR for convolution nodes. We need to wire this up.
Example quantization code:
conv_config = IntxWeightOnlyConfig(
weight_dtype=torch.int8,
granularity=PerAxis(0),
)
quantize_(
model,
conv_config,
lambda m, fqn: isinstance(m, torch.nn.Conv2d),
)
cc @digantdesai @cbilgin
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.