-
Notifications
You must be signed in to change notification settings - Fork 565
Pull requests: pytorch/ao
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
[test][xpu]Add LD_LIBRARY_PATH ENV for CCL
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.
xpu
Intel XPU related features
#4581
opened Jul 20, 2026 by
zxd1997066
Contributor
Loading…
Fix stale-index pop in _fqn_to_config_handler when multiple FqnToConfig parameters map to None
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4580
opened Jul 17, 2026 by
sohumt123
Loading…
[pat] Low-rank pruning support
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
#4579
opened Jul 17, 2026 by
lisjin
Contributor
Loading…
fix(experimental/mps): raise ImportError so pkgutil.walk_packages can skip the module (#4577)
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4578
opened Jul 16, 2026 by
Anai-Guo
Contributor
Loading…
Relax test_bf16_stochastic_round tolerance
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.
xpu
Intel XPU related features
#4576
opened Jul 16, 2026 by
karol-brejna-i
Loading…
Consolidate tensor subclass boilerplate using flatten/unflatten (fixes #710)
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4575
opened Jul 15, 2026 by
GiGiKoneti
Contributor
Loading…
fix(int4/tile_packed): validate block_size against scale_and_zero to prevent OOB read
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4573
opened Jul 14, 2026 by
Anai-Guo
Contributor
Loading…
Add pass to writeback buffer mutations after QDQ
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4569
opened Jul 13, 2026 by
tom-arm
Contributor
Loading…
[mxfp8] use explicit RCEIL scaling in on-device all-to-all-v
#4568
opened Jul 13, 2026 by
wolfcomos
Loading…
[qat] Add QAT support for Int8DynamicActivationInt8WeightConfig and Int8WeightOnlyConfig
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4567
opened Jul 12, 2026 by
JPL11
Loading…
docs/types(pt2e): annotate and document DuplicateDQPass (#4541)
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4565
opened Jul 11, 2026 by
Anai-Guo
Contributor
Loading…
Fix Int8WeightOnlyConfig crash when in_features=1
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4563
opened Jul 9, 2026 by
egeozkoc
Loading…
5 tasks done
cpu: make ukernel registration idempotent
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4562
opened Jul 7, 2026 by
nascheme
Contributor
Loading…
Implement This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
cat, transpose, mm and addmm ops in Int8Tensor
CLA Signed
Make FP8 blockwise MoE DeepGEMM compile clean
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
#4559
opened Jul 6, 2026 by
iamzainhuda
Contributor
Loading…
kernel: synchronize autotuner cache updates
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4554
opened Jul 3, 2026 by
nascheme
Contributor
Loading…
kernel: avoid partial Triton lazy initialization
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4553
opened Jul 3, 2026 by
nascheme
Contributor
Loading…
Reuse direct grouped FP8 quantizers for blockwise MoE and simplify DeepGEMM path
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
#4551
opened Jul 2, 2026 by
iamzainhuda
Contributor
Loading…
HQQ general
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
[xpu] Route per-block FP8 GEMM through torch._scaled_mm on XPU
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
#4548
opened Jul 2, 2026 by
hoshibara
Loading…
Add MoE blockwise FP8 DTensor and compile coverage
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
#4545
opened Jun 29, 2026 by
iamzainhuda
Contributor
Loading…
fix(_is_mslk_available): guard against missing CUDA/ROCm runtime
ciflow/rocm
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
module: rocm
#4544
opened Jun 29, 2026 by
space192
Loading…
Fix constant_fold cleanup for shared get_attr targets
CLA Signed
This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.
#4542
opened Jun 28, 2026 by
yinli-systems
•
Draft
Previous Next
ProTip!
Filter pull requests by the default branch with base:main.