[AUTOGENERATED] develop_IFU_20251106#2789

Open

pragupta wants to merge 131 commits intodevelopfrom

develop_IFU_20251106

Commits on Nov 4, 2025

Add CUDA MXFP4 scaled mm support via. FBGEMM (#166526)

slayton58
authored and
pytorchmergebot
committed
[CI] Use smaller amx + avx2 runners for inductor test? (#164989)

clee2000
authored and
pytorchmergebot
committed
[ez] Print some more test timing info in the logs (#166447)

clee2000
authored and
pytorchmergebot
committed
Add model code stack trace to cuda.memory._snapshot (#166676)

yushangdi
authored and
pytorchmergebot
committed
[BE] Use `[[maybe_unused]]` (#166865)

malfet
authored and
pytorchmergebot
committed
[Inductor] addmm with bias -> unfuse bias if there is a pointwise/reduction consumer (#166165)

nikitaved
authored and
pytorchmergebot
committed
subproc_pool: Add support for enabling quiesce via a timer (#166467)

c00w
authored and
pytorchmergebot
committed
Delete deprecated fp32 precision warnings (#166956)

zou3519
authored and
pytorchmergebot
committed
Fixed some syntax errors in SECURITY.md file. (#166718)

wenlinchong17-web
authored and
pytorchmergebot
committed
[ROCm][CI] Lower runner check gpu count for distributed jobs (#166961)

amdfaa
authored and
pytorchmergebot
committed
[cuDNN] Smoke-test runtime cuDNN version matches compile time version in CI (#165922)

authored and
pytorchmergebot
committed
More pyrefly local errors (#166976)

oulgen
authored and
pytorchmergebot
committed
[BE] Delete Python-3.9 stdlib definitions from torch.package (#166768)

malfet
authored and
pytorchmergebot
committed
[aotd] Compiled saved tensor hooks context (#166887)

IvanKobzarev
authored and
pytorchmergebot
committed
Revert "[Inductor][Grouped Gemm] Add Blackwell CuTeDSL Kernel (#165036)"
pytorchmergebot
committed
[inductor] coordesc not tune XBLOCK for mix-order-reduction (#166669)

shunting314
authored and
pytorchmergebot
committed
Upload test stats for trunk/sha tag (#166916)

izaitsevfb
authored and
pytorchmergebot
committed
Update triton to 3.5.1 release (#166968)

atalman
authored and
pytorchmergebot
committed
[inductor] runtime estimations disable use_nccl_estimator by default (#166973)

IvanKobzarev
authored and
pytorchmergebot
committed
If USE_CUDA=1 is set, do not fallback to no CUDA (#166982)

oulgen
authored and
pytorchmergebot
committed
inductor: Switch quiesce to use timer based implementation. (#166581)

c00w
authored and
pytorchmergebot
committed
Fix torch.linalg.eig inductor stride mismatch (#162484)

parsshar-RH
authored and
pytorchmergebot
committed
[dynamo] don't use LocalSource for temp variables created by side_effects (#166917)

williamwen42
authored and
pytorchmergebot
committed
Avoid DDE in narrow with unbacked start (#166361)

laithsakka
authored and
pytorchmergebot
committed
Fix quoting in pytest_cache.py invocations (#166955)

Flamefire
authored and
pytorchmergebot
committed
[DebugMode] output, tensor id annotations for DebugMode (#165076)

pianpwk
authored and
pytorchmergebot
committed
Add model code stack trace to torch.profile (#166677)

yushangdi
authored and
pytorchmergebot
committed
[Minor][Inductor] move some combo kernel log from warning to debug (#166993)

BoyuanFeng
authored and
pytorchmergebot
committed
Revert "Add model code stack trace to torch.profile (#166677)"
pytorchmergebot
committed
make narrow_tensor_symint DDE-free (#166379)

laithsakka
authored and
pytorchmergebot
committed
[export] Codemod unittests to use new graph capture API (#166957)

zhxchen17
authored and
pytorchmergebot
committed
Clarify safety of CUDA graph memory pool sharing across graphs that are replayed in arbtirary order. (#166975)

galv
authored and
pytorchmergebot
committed
[inductor][AMD] Filter out invalid Triton Configs for MI350X _scaled_mm (#166442)

JChunX
authored and
pytorchmergebot
committed
use_cpp_bmm_template supports more use cases (#165469)

helloguo
authored and
pytorchmergebot
committed
Add default `.github/copilot-instructions.md` and item in `.gitignore` for allowing local changes (#166864)

authored and
pytorchmergebot
committed
[BE][Typing][Dynamo] Type torch/_dynamo/variables/ctx_manager.py (#166878)

Lucaskabela
authored and
pytorchmergebot
committed

Commits on Nov 5, 2025

don't produce invalid grid configs (#166974)

ngimel
authored and
pytorchmergebot
committed
[RFC] Add experimental Pallas TorchInductor backend (#166822)

oulgen
authored and
pytorchmergebot
committed
shrink_group implementation to expose ncclCommShrink API (#164518)

brchang24
authored and
pytorchmergebot
committed
[CD] Upload XPU inductor benchmark test reports to s3 (#166954)

chuanqi129
authored and
pytorchmergebot
committed
[HOP][print] Add HOP subclass for printing (#166660)

authored and
pytorchmergebot
committed
Fix the vmap op fallback bug (#166032)

haifeng-jin
authored and
pytorchmergebot
committed
Send / recv support in local tensor (#166595)

dzmitry-huba
authored and
pytorchmergebot
committed
Annotation should be deepcopied (#167017)

yushangdi
authored and
pytorchmergebot
committed
Revert "make narrow_tensor_symint DDE-free (#166379)"
pytorchmergebot
committed
Revert "Avoid DDE in narrow with unbacked start (#166361)"
pytorchmergebot
committed
[12/N] Apply ruff UP035 rule (#166929)

cyyever
authored and
pytorchmergebot
committed
Fix typos in complex numbers docs (#166671)

klamike
authored and
pytorchmergebot
committed
[unified v2][apple] Clean up `APPLETVOS` from caffe2 (#166953)

mzlee
authored and
pytorchmergebot
committed
Fix typo in gloo_hip library name (#166502)

AngryLoki
authored and
pytorchmergebot
committed
[ROCm][CI] revert ROCm magma commit hash to last known good (#167044)

jeffdaily
authored and
pytorchmergebot
committed
Add min/max support for barebones uint types (#166813)

ezyang
authored and
pytorchmergebot
committed
Add model code stack trace to torch.profile (#166677)

yushangdi
authored and
pytorchmergebot
committed
[dynamo] extend `collections.defaultdict` support with `*args`, `**kwargs` and custom `default_factory` (#166793)

XuehaiPan
authored and
pytorchmergebot
committed
[fx] Add strict argument validation to Interpreter.boxed_run (#166784)

meghendra6
authored and
pytorchmergebot
committed
[Inductor][Grouped Gemm] Add Blackwell CuTeDSL Kernel (#167003)

NikhilAPatel
authored and
pytorchmergebot
committed
fixes keyerror when loading parameter with unsaved optimizer state (#165228)

arkadip-maitra
authored and
pytorchmergebot
committed
[dynamo] Introduce _set_lru_cache (#167038)

xmfan
authored and
pytorchmergebot
committed
Revert "[Inductor][Grouped Gemm] Add Blackwell CuTeDSL Kernel (#167003)"
pytorchmergebot
committed
Refactor out headeronly ArrayRef (#164991)

janeyx99
authored and
pytorchmergebot
committed
Widen ops support to take in IntHOArrayRef vs only std::vec (#165152)

janeyx99
authored and
pytorchmergebot
committed
Invert unary read and write for fusion (#161404)

eellison
authored and
pytorchmergebot
committed
Fix clang-21 warnings (#166859)

Sqvid
authored and
pytorchmergebot
committed
[pytree][dynamo] add test to ensure `tree_map` preserves `dict` order (#166236)

XuehaiPan
authored and
pytorchmergebot
committed
Revert "Update triton to 3.5.1 release (#166968)"
pytorchmergebot
committed
[CI] Parse xml and upload json while running (#166988)

clee2000
authored and
pytorchmergebot
committed
[Inductor] Fix unbacked float symbol handling in kernel codegen (#166890)

karthickai
authored and
pytorchmergebot
committed
[Code Clean] Replace `assert` with if statement and raise `AssertionError` (#166935)

KarhouTam
authored and
pytorchmergebot
committed
[ROCm] Enable group gemm through CK (#166334)

jagadish-amd
authored and
pytorchmergebot
committed
Revert "Add model code stack trace to torch.profile (#166677)"
pytorchmergebot
committed
Revert "[cuDNN] Smoke-test runtime cuDNN version matches compile time version in CI (#165922)"
pytorchmergebot
committed
Add some code for exploring the space of accessible size/stride configs via plain views (#167076)

ezyang
authored and
pytorchmergebot
committed
Update triton to 3.5.1 release (#166968)

atalman
authored and
pytorchmergebot
committed
ci: Add aarch64 docker builds for modern clang (#166416)

seemethere
authored and
pytorchmergebot
committed
include DTensor metadata when pretty-printing fx.Graphs (#166750)

bdhirsh
authored and
pytorchmergebot
committed
[BE][Typing][Dynamo] Type torch/_dynamo/variables/dicts.py (#167022)

Lucaskabela
authored and
pytorchmergebot
committed
[Inductor] Naive foreach autotune support (#162053)

authored and
pytorchmergebot
committed
Update typing docs to reference pyrefly (#166883)

maggiemoss
authored and
pytorchmergebot
committed
Revert "[Inductor] Fix unbacked float symbol handling in kernel codegen (#166890)"
pytorchmergebot
committed
Revert "[12/N] Apply ruff UP035 rule (#166929)"
pytorchmergebot
committed
Don't hardcode double argument for reduction base (#166951)

ezyang
authored and
pytorchmergebot
committed
[13/N] Apply ruff UP035 rule (#167048)

cyyever
authored and
pytorchmergebot
committed
Revert "[DebugMode] output, tensor id annotations for DebugMode (#165076)"
pytorchmergebot
committed
Add torch::stable::Tensor sizes and strides (#165153)

janeyx99
authored and
pytorchmergebot
committed
Remove python workaround for ContextDecorator (#167049)

cyyever
authored and
pytorchmergebot
committed
[user-streams] Add graph annotation checks (#167019)

mlazos
authored and
pytorchmergebot
committed
[user-streams] Add backward test (#167021)

mlazos
authored and
pytorchmergebot
committed
[DebugMode] output, tensor id annotations for DebugMode (#165076)

pianpwk
authored and
pytorchmergebot
committed
fix nccl estimations (#167093)

IvanKobzarev
authored and
pytorchmergebot
committed
[12/N] Apply ruff UP035 rule (#166929)

cyyever
authored and
pytorchmergebot
committed
[CP][BE][3/N] Add _templated_ring_attention to the backward compatility stub (#166991)

fegin
authored and
pytorchmergebot
committed
[ATEN][CUDA] Reduce register pressure in radix_sort_pairs to improve torch.sort performance (#167094)

YyWangCS
authored and
pytorchmergebot
committed
fix fr reset api (#166970)

tushar00jain
authored and
pytorchmergebot
committed
[ROCm] Skip nvfp4 tests on ROCm (#167066)

jagadish-amd
authored and
pytorchmergebot
committed
Add missing skipIf(not PLATFORM_SUPPORTS_MEM_EFF_ATTENTION) to test/test_transformers.py (#166969)

xinyazhang
authored and
pytorchmergebot
committed

Commits on Nov 6, 2025

Move almalinux docker image to DEVTOOLSET 13 (#167018)

atalman
authored and
pytorchmergebot
committed
[cuBLAS] Force tensor-core-no-reduction algo in `cuBLASLt` for `n=1` cases (#166735)

eqy
authored and
pytorchmergebot
committed
Avoid DDE in narrow with unbacked start (#166361)

laithsakka
authored and
pytorchmergebot
committed
Reland "Add model code stack trace to torch.profile (#166677)" (#167110)

yushangdi
authored and
pytorchmergebot
committed
[MPS] Fix crash in BCELoss backwards with reduction="none" and inputs with trailing 1s in shape (#166786)

inventshah
authored and
pytorchmergebot
committed
Bugfix to forward autodiff causing different datatype 2 (#165784)

skpark-rh
authored and
pytorchmergebot
committed
Change template 'Release highlight for proposed Feature'->'New Feature for Release' (#167145)

atalman
authored and
pytorchmergebot
committed
[c10d] Fix split_group bug by having the parent pg option deep copied (#167125)

fduwjj
authored and
pytorchmergebot
committed
[1/N] Use `key in dict` for existence checks (#167035)

cyyever
authored and
pytorchmergebot
committed
[ci][cpu] Update compiler to GCC-13 in jammy-aarch64 (#166849)

fadara01
authored and
pytorchmergebot
committed
Fix duplicate benchmarking entries for addmm (#166652)

jainapurva
authored and
pytorchmergebot
committed
[7/N] Fix unused loop variables in tests (#167043)

cyyever
authored and
pytorchmergebot
committed
Refactor: extract OperatorArgsKwargsView from parseIValuesToPyArgsKwargs (#166368)

swolchok
authored and
pytorchmergebot
committed
[PP] make runtime dbg log print custom actions (#167113)

wconstab
authored and
pytorchmergebot
committed
[1/N] Add return types of Python functions (#167162)

cyyever
authored and
pytorchmergebot
committed
torch.fx: add debug-level logging to Interpreter.run_node (#117351) (#166622)

mdbarnesUCSD
authored and
pytorchmergebot
committed
[dynamo, 3.14] disable dynamo cpython tests in 3.14 (again) (#167000)

williamwen42
authored and
pytorchmergebot
committed
[audio hash update] update the pinned audio hash (#167031)

pytorchupdatebot
authored and
pytorchmergebot
committed
[user-streams] Enable stream ops to work in eager (#167141)

mlazos
authored and
pytorchmergebot
committed
[user-streams] Add record/wait ops (#167151)

mlazos
authored and
pytorchmergebot
committed
[user-streams] Mark stream ops as side effectful (#167152)

mlazos
authored and
pytorchmergebot
committed
Expose torch.compiler.config.force_disable_caches as a public API (#166699)

gmagogsfm
authored and
pytorchmergebot
committed
Update `tensorpipe` submodule (#167108)

malfet
authored and
pytorchmergebot
committed
make narrow_tensor_symint DDE-free (#166379)

laithsakka
authored and
pytorchmergebot
committed
CustomOp Inline Fusion (#165952)

tianrengao
authored and
pytorchmergebot
committed
Check cluster_dims attribute exists before access (#167187)

yf225
authored and
pytorchmergebot
committed
[2/N] Use `key in dict` for existence checks (#167174)

cyyever
authored and
pytorchmergebot
committed
[8/N] Fix unused loop variables in tests (#166921)

cyyever
authored and
pytorchmergebot
committed
Revert "Don't hardcode double argument for reduction base (#166951)"
pytorchmergebot
committed
[inductor] Use runtime estimations in iterative reorder collectives pass (#167080)

IvanKobzarev
authored and
pytorchmergebot
committed
[MTIA][PyTorch] Add mtia as native device for PyTorch tests (#167089)

jvandebon
authored and
pytorchmergebot
committed
Add per_process_memory_fraction to PYTORCH_CUDA_ALLOC_CONF (#161035)

lakshayg
authored and
pytorchmergebot
committed
[inductor] Use runtime estimations in iterative sink waits pass (#167081)

IvanKobzarev
authored and
pytorchmergebot
committed
Make CUDA preload logic more straightforward (#167046)

malfet
authored and
pytorchmergebot
committed
`nn.Linear`: nD contiguous input + bias -- dispatch to addmm also when weight is sparse (#166071)

nikitaved
authored and
pytorchmergebot
committed
Merge remote-tracking branch 'upstream/main' into develop_IFU_20251106
github-actions[bot]
committed