Skip to content

Pull requests: pytorch/ao

New pull request New
477 Open 3,494 Closed
477 Open 3,494 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

Support MXTensor cat for shared-input linear fusion
#4882 opened Sep 11, 2026 by lchen2331 • Draft
2
Fix MoE FP8 Triton autotune key CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: training quantize_ api training flow
#4880 opened Sep 10, 2026 by warrendeng Loading…
4
Make float8 autograd Function allow_in_graph opt-in via config CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4879 opened Sep 10, 2026 by vkuzo Contributor Loading…
2
Fix PT2E Quick Start guide example so it runs as documented CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4878 opened Sep 10, 2026 by Sanjays2402 Loading…
1
1
Add code owners for low-bit CPU kernels CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: not user facing Use this tag if you don't want this PR to show up in release notes
#4874 opened Sep 8, 2026 by JacobSzwejbka Contributor Loading…
1
Fix scalar collapse for all-ones block_size in tinygemm / dont_preserve_zero qparams CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4871 opened Sep 8, 2026 by TusharND12 Loading…
3
fix(float8): handle aten.abs and amax-path ops in Float8TrainingTensor dispatch CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: training quantize_ api training flow
#4870 opened Sep 7, 2026 by somuai Loading…
1
7
Make Int8 and Float8 weight-only reapplication safe CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4869 opened Sep 6, 2026 by yashb98 Loading…
1
Add generalized int1-int8 weight-only MPS quantization with simdgroup GEMM CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4868 opened Sep 4, 2026 by mspinelli Loading…
1
Warn when Int4WeightOnlyConfig uses the PLAIN packing format CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4867 opened Sep 4, 2026 by fletchers16 Loading…
1
Optimize Arm NEON W3A8 decode CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. cpu module: core changes affecting multiple modules, e.g. base config/tensor, observers, quant ops
#4859 opened Aug 31, 2026 by JacobSzwejbka Contributor Loading…
5
Parallelize W3A8 prefill activation packing CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4854 opened Aug 31, 2026 by JacobSzwejbka Contributor • Draft
1
[nvfp4 training][rl] Add the four-over-six grouped GEMM and dispatcher wiring CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4853 opened Aug 31, 2026 by wolfcomos Loading…
1
[nvfp4 training][rl] Add a CuTe DSL fast-path kernel for four-over-six quantize CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4852 opened Aug 31, 2026 by wolfcomos Loading…
1
[nvfp4 training][rl] Add the four-over-six NVFP4 dense training recipe CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4851 opened Aug 31, 2026 by wolfcomos Loading…
@wolfcomos
1
fix(tests/moe_training): float8_dtyep typo makes the fp8 rowwise grouped-mm test TypeError once unskipped CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4850 opened Aug 30, 2026 by Anai-Guo Contributor Loading…
1
Avoid re-registering native Enum pytree constants CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4849 opened Aug 29, 2026 by malaiwah Loading…
1
3
[MXFP8 training] Preserve linear bias in weight wrapper CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4847 opened Aug 29, 2026 by GotFusion Contributor Loading…
1
1
Refactor MXTensor: replace is_swizzled_scales bool with SwizzleType class hierarchy ciflow/xpu label used to trigger xpu CI jobs CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: inference quantize_ api inference flow xpu Intel XPU related features
#4843 opened Aug 28, 2026 by xiaowangintel Collaborator • Draft
1
docs: remove stale MXFP4 training entry CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4842 opened Aug 28, 2026 by GotFusion Contributor Loading…
1
1
[WIP][MX] Add FSDP2 support for inference weights CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4838 opened Aug 27, 2026 by GotFusion Contributor Loading…
1
1
[CI] Run only accelerator PT2E tests on GPU CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. module: pt2e_quant pt2 export quantization (prepare_pt2e, convert_pt2e, quantizer)
#4836 opened Aug 27, 2026 by iamzainhuda Contributor Loading…
7
Limit MXFP8 PyTorch version guard to DTensor usage CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4831 opened Aug 26, 2026 by GotFusion Contributor Loading…
1
1
Refactor ToyConvModel into shared testing architectures CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4829 opened Aug 26, 2026 by AndyYin06 Loading…
2
Fix generate_permute_indices max_len bounds CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4827 opened Aug 25, 2026 by gaoxiaomo Loading…
1
2
ProTip! Adding no:label will show everything without a label.