[AUTOGENERATED] rocm7.1_internal_testing_IFU_2025-10-03#2694

Closed

jithunnair-amd wants to merge 467 commits intorocm7.1_internal_testingfrom

rocm7.1_internal_testing_IFU_2025-10-03

This pull request is big! We're only showing the most recent 250 commits

Commits on Sep 29, 2025

Consistently use c10_ovrsource in arvr mode everywhere (#164128)

ezyang
authored and
pytorchmergebot
committed
[testing] upload test stats: Add info to the invoking file summary and some other changes (#164016)

clee2000
authored and
pytorchmergebot
committed
Replace setup.py bdist_wheel with python -m build --wheel (#156712)

zklaus
authored and
pytorchmergebot
committed
Revert "Helper to augment graph with additional deps (#163959)"
pytorchmergebot
committed
Revert "refactor bucketing (#163754)"
pytorchmergebot
committed
Revert "[inductor] do comm compute overlap at aten fx level (#163215)"
pytorchmergebot
committed
Bump protobuf from 5.29.4 to 5.29.5 in /.ci/docker (#156157)

dependabot[bot]
and
malfet
authored
[Easy] Add pointwise tag to fma (#164149)

eellison
authored and
pytorchmergebot
committed
[PP] Customize pipeline's submod name (#164037)

kwen2501
authored and
pytorchmergebot
committed

Commits on Sep 30, 2025

[BE] remove manylinuxcxx11-abi-builder:cpu-cxx11-abi docker image (#164187)

atalman
authored and
pytorchmergebot
committed
[3/N] Import Callable from collections.abc in torch/distributed (#164104)

cyyever
authored and
pytorchmergebot
committed
Install `fmtlib` headers. (#164139)

ysiraichi
authored and
pytorchmergebot
committed
[dynamic shapes] unbacked-safe slicing (#161414)

pianpwk
authored and
pytorchmergebot
committed
CUDACachingHostAllocatorImpl skip event query during capture (#164001)

jeffdaily
authored and
pytorchmergebot
committed
[AARCH64][CD][CUDA13][Triton][PTXAS] Turn on BUILD_BUNDLE_PTXAS=1 (#163988)

nWEIdia
authored and
pytorchmergebot
committed
[ROCm][CI] Upgrade ROCm to 7.0 (#163140)

jeffdaily
authored and
pytorchmergebot
committed
[CMake] Fix `USE_FBGEMM_GENAI` option (#164165)

malfet
authored and
pytorchmergebot
committed
[doc] Add AOTInductor intermediate debug printer OSS user manual (#163794)

YUNQIUGUO
authored and
pytorchmergebot
committed
Revert "Enable outer reductions in fbcode (#163884)"
pytorchmergebot
committed
Remove unnecessary list comprehensions (#164103)

cyyever
authored and
pytorchmergebot
committed
[CI] Push `viable/strict/${time}` tags (#164183)

malfet
authored and
pytorchmergebot
committed
Fix invalid f-strings (#164112)

cyyever
authored and
pytorchmergebot
committed
[torchfuzz] Make scalar and tensor distribution configurable (#164034)

bobrenjc93
authored and
pytorchmergebot
committed
[inductor] do comm compute overlap at aten fx level (#163215)

eellison
authored and
pytorchmergebot
committed
refactor bucketing (#163754)

eellison
authored and
pytorchmergebot
committed
Helper to augment graph with additional deps (#163959)

eellison
authored and
pytorchmergebot
committed
Add Comm-Compute Preserving Bucketer (#163960)

eellison
authored and
pytorchmergebot
committed
better error handling for rrelu when lower or upper range is infinite (#160965)

vishalgoyal316
authored and
pytorchmergebot
committed
[dynamo] Special path for cloning of torch dispatch tensors (#164081)

anijain2305
authored and
pytorchmergebot
committed
Back out "Revert D81959389" (#163905)

yyetim
authored and
pytorchmergebot
committed
[dynamo][device_mesh] Support mesh_dim_names (#164200)

anijain2305
authored and
pytorchmergebot
committed
[DTensor][Export] Supporting exporting a model with DTensor params/inputs (#163609)

authored and
pytorchmergebot
committed
distributed/serialization: support zero sized tensors (#164198)

d4l3k
authored and
pytorchmergebot
committed
Add functions to setup PrivateUse1 as a python backend device. (#157859)

qihqi
authored and
pytorchmergebot
committed
[inductor] Small refactor of CachingAutotuner (#162406)

kundaMwiza
authored and
pytorchmergebot
committed
Fix cdist export compute mode validation (#161724)

ahkush
authored and
pytorchmergebot
committed
[OpenReg] Add AMP Integration guide for accelerators (#162050)

authored and
pytorchmergebot
committed
Revert "Add functions to setup PrivateUse1 as a python backend device. (#157859)"
pytorchmergebot
committed
Remove unused PyIntXXX, THPUtils_newReal_BOOL, THPQXXX macros (#164056)

cyyever
authored and
pytorchmergebot
committed
Revert "Add less warps config to inner reductions (#162447)"
pytorchmergebot
committed
Revert "[CI] Push `viable/strict/${time}` tags (#164183)"
pytorchmergebot
committed
Revert "Consistently use c10_ovrsource in arvr mode everywhere (#164128)"
pytorchmergebot
committed
[BE] Remove not existing mnist mirror (#164238)

atalman
authored and
pytorchmergebot
committed
Update LPPool docs to clarify ceil_mode padding semantics when ceil_mode=True (#163186)

Jonahcb
authored and
pytorchmergebot
committed
Update Sphinx theme (#164147)

svekars
authored and
pytorchmergebot
committed
[CUDA][Expandable Segments] Follow-up cleanups for even more expandable segments tests (#163297)

eqy
authored and
pytorchmergebot
committed
[testing] Add upload for test status during test stat uploads (#164189)

authored and
pytorchmergebot
committed
[precompile] Add option to disable guard check on aot-compiled function. (#163432)

zhxchen17
authored and
pytorchmergebot
committed
Fix TestExportOpInfo (#164184)

SherlockNoMad
authored and
pytorchmergebot
committed
[Inductor-FX] Generalize FloorDiv conversion to handle more complex launch grids. Remove python_slow grid mode. (#163828)

blaine-rister
authored and
pytorchmergebot
committed
[ROCm][CI] no longer build almalinux image for ROCm 6.3 (#164201)

jeffdaily
authored and
pytorchmergebot
committed
[submodule] upgrade cutlass version to 4.2.1 and completely resolved python/cutlass name collision (#164156)

henrylhtsang
authored and
pytorchmergebot
committed
Exporting aten.sdpa with cuda under fake mode on a cuda-less machine (#164162)

yiming0416
authored and
pytorchmergebot
committed
[inductor] fx comm overlap: align runtime estimations across dist ranks (#164226)

IvanKobzarev
authored and
pytorchmergebot
committed
[dynamo] format cpython_defs.c (#161838)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] compile actual code in C dynamo (#161555)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] Python dynamo changes to get basic programs working (#161839)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] support some bytecodes, fix CALL_FUNCTION_EX (#163009)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix context managers (#163109)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] support LOAD_CONST on slice, codegen LOAD_CONST slice instead of BINARY/STORE_SLICE (#163110)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix inactive ctx handling in resume functions (#163191)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix WITH_EXCEPT_START (#163292)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix stack ref copy error (#163796)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix BUILD_TUPLE with 0 args (#163818)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] support LOAD_COMMON_CONSTANT (#163919)

williamwen42
authored and
pytorchmergebot
committed
[3.14] make unbacked_sym[int/float]_counter integers (#163920)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix tracing typing.Union (#164004)

williamwen42
authored and
pytorchmergebot
committed
[dynamo, 3.14] fix _detect_and_normalize_assert_statement for 3.14 (#164005)

williamwen42
authored and
pytorchmergebot
committed
[inductor][templates] Template hooks should be finalised inside a kernel context (#164229)

kundaMwiza
authored and
pytorchmergebot
committed
Fix TMA transpose logic to handle 1D shapes + string differences (#163966)

njriasan
authored and
pytorchmergebot
committed
[inductor] use hint_override in kernel benchmark args (#164207)

pianpwk
authored and
pytorchmergebot
committed
fix: inductor non_blocking test - warmup events to make test pass whether it is the first run or not (#164188)

v0i0
authored and
pytorchmergebot
committed
Revert "[BE] Remove HermeticPyObjectTLS and Simplify PythonOpRegistrationTrampoline (#163464)"
pytorchmergebot
committed
Fix the shape check inside gnll loss (#147522)

KohakuBlueleaf
authored and
pytorchmergebot
committed
Update persons of interest for XLA. The previous one is out of date. (#158652)

qihqi
authored and
pytorchmergebot
committed
half support for fused_moving_avg_obs_fake_quant() op (#164175)

jeffdaily
authored and
pytorchmergebot
committed
[1/N] Simplify "in" operation for containers of a single item (#164224)

cyyever
authored and
pytorchmergebot
committed
[CI] Delete binary smoke workflows (#164260)

malfet
authored and
pytorchmergebot
committed
Fix silent incorrectness for bmm/baddmm out_dtype overload (#164095)

PaulZhang12
authored and
pytorchmergebot
committed
Missing lambda in torch._check (#164225)

xadupre
authored and
pytorchmergebot
committed
[FSDP][Replicate] tests replicate gradient accumulation and 1f1b microbatching (#162839)

anshul-si
authored and
pytorchmergebot
committed
[FSDP][Replicate] tests replicate with custom forward method (#162851)

anshul-si
authored and
pytorchmergebot
committed
[FSDP][Replicate] tests replicate is composable with tp (#162853)

anshul-si
authored and
pytorchmergebot
committed
[FSDP][Replicate] tests replicate core functionality with mixed precision (#162855)

anshul-si
authored and
pytorchmergebot
committed
[export] avoid checks during tracing of export verification (#164219)

anijain2305
authored and
pytorchmergebot
committed
[FSDP2] support AC(FSDP) for torchtitan's MOE (#164009)

weifengpy
authored and
pytorchmergebot
committed
[FSDP][Replicate] tests replicate type casting behavior and edge cases in mixed precision (#162861)

anshul-si
authored and
pytorchmergebot
committed
[c10d][BE][ez] Update tensor ptr inside nccl.cpp (#164276)

fduwjj
authored and
pytorchmergebot
committed
[CUDA][CUDAGraph] Reduce capture overhead in CUDA Graph memory reuse (#162186)

eee4017
authored and
pytorchmergebot
committed
[DTensor] Allow redistribute to Partial if src matches (#164253)

SherlockNoMad
authored and
pytorchmergebot
committed
[vllm hash update] update the pinned vllm hash (#164190)

pytorchupdatebot
authored and
pytorchmergebot
committed
Migrate DeviceType to torch/headeronly (#163999)

janeyx99
authored and
pytorchmergebot
committed

Commits on Oct 1, 2025

[ROCm][CD] librocroller.so missing from ROCm 7 wheel (#164244)

jeffdaily
authored and
pytorchmergebot
committed
[PyTorch CCA] Add an API to get expandable segment sizes (#163771)

banitag1
authored and
pytorchmergebot
committed
[aoti] AOTI mingw cross compilation (#163188)

yushangdi
authored and
pytorchmergebot
committed
Split scaled-mm tests into separate file (#164266)

slayton58
authored and
pytorchmergebot
committed
Revert new-test part of #163829 (#164259)

slayton58
authored and
pytorchmergebot
committed
Add smoke tests to verify that stable ABI FA3 wheel runs w/ newer torch (#163782)

janeyx99
authored and
pytorchmergebot
committed
[torchfuzz] remove erroneous can_produce check (#164209)

bobrenjc93
authored and
pytorchmergebot
committed
[torchfuzz] raise if Operator abstract method is not implemented (#164211)

bobrenjc93
authored and
pytorchmergebot
committed
[torchfuzz] add layout operators (#164210)

bobrenjc93
authored and
pytorchmergebot
committed
skip non memory deps in memory estimator (#164294)

eellison
authored and
pytorchmergebot
committed
[dynamo][sac] Support functools partial context_fn for sac (#164308)

anijain2305
authored and
pytorchmergebot
committed
[export] Explicitly passing requires_grad to nn.Parameter() in deserialization (#164290)

yiming0416
authored and
pytorchmergebot
committed
Add algorithm in header (#164295)

ahkush
authored and
pytorchmergebot
committed
Adding check for square matrix for input tensor in matrix_exp backwar… (#163357)

mansiag05
authored and
pytorchmergebot
committed
[BE][Easy]: Add prims common TypeGuard (#164263)

Skylion007
authored and
pytorchmergebot
committed
Use dataclass features in two classes (#164221)

authored and
pytorchmergebot
committed
Skip symmetric memory tests calling `_scaled_mm` on CCC < 8.9 (#164251)

Flamefire
authored and
pytorchmergebot
committed
[BE][CI] Unify requirments (#163396)

malfet
authored and
pytorchmergebot
committed
[PyTorch Pinned Allocator] Pinned memory stats and perf fixes around allocating blocks (#163777)

banitag1
authored and
pytorchmergebot
committed
[CI] Push `viable/strict/${time}` tags (#164183)

malfet
authored and
pytorchmergebot
committed
[BE] Delete stale non-ephemeral runners workarounds (#164285)

malfet
authored and
pytorchmergebot
committed
[dynamo] Use strings instead of modules for fqn info tracking (#164272)

anijain2305
authored and
pytorchmergebot
committed
[2/N] Simplify "in" operation for containers of a single item (#164323)

cyyever
authored and
pytorchmergebot
committed
Improve repeat op to a single copy (#163842)

authored and
pytorchmergebot
committed
Suppress `FutureWarning`s in `torch.distributed.algorithms.ddp_comm_hooks` (#163939)

xuantengh
authored and
pytorchmergebot
committed
[BE][Easy] update CUDA and ROCm sources in nightly tool (#162324)

XuehaiPan
authored and
pytorchmergebot
committed
[Easy] Add notes for setting up dev venv with specific Python version (#164214)

XuehaiPan
authored and
pytorchmergebot
committed
Register MTIA kernel for all_all_out (#164293)

trirpi
authored and
pytorchmergebot
committed
Adds Issue#153109 as a test for CUDAPluggableAllocator (#163575)

syed-ahmed
authored and
pytorchmergebot
committed
Releases multicast object before releasing mapped buffers in CUDASymmetricMemory (#163750)

syed-ahmed
authored and
pytorchmergebot
committed
[dynamo] do not pop from framelocals dict in Python 3.10 (#164316)

williamwen42
authored and
pytorchmergebot
committed
Build and Install Arm Compute Library in manylinux docker image (#159737)

robert-hardwick
authored and
pytorchmergebot
committed
Triton 3.5.x pin update (#164268)

atalman
authored and
pytorchmergebot
committed
Fix non-TMA loads in grouped MM Triton kernel (#163895)

alexsamardzic
authored and
pytorchmergebot
committed
Add a RECORD_FUNCTION for Python fallback so it shows in profile (#160573)

ezyang
authored and
pytorchmergebot
committed
[PyTorch] Pull ARM's box-cox (#164152)

Nicoshev
authored and
pytorchmergebot
committed
Revert "CUDACachingHostAllocatorImpl skip event query during capture (#164001)"
pytorchmergebot
committed
Use TMA loads always for Triton grouped MM kernel (#164256)

alexsamardzic
authored and
pytorchmergebot
committed
Revert nccl upgrade back to 2.27.5 (#164352)

albanD
authored and
pytorchmergebot
committed
Revert "[PP] Customize pipeline's submod name (#164037)"
pytorchmergebot
committed
Revert "[fx] Allow customization of submod name in split graph (#164035)"
pytorchmergebot
committed
Revert "Add num_store to inductor_meta and use it to scale persistent reduction x block (#162446)"
pytorchmergebot
committed
[fx] Allow customization of submod name in split graph (#164035)

kwen2501
authored and
pytorchmergebot
committed
[PP] Customize pipeline's submod name (#164037)

kwen2501
authored and
pytorchmergebot
committed
[1/N] Fix ruff warnings (#164333)

cyyever
authored and
pytorchmergebot
committed
[export] support unbacked stack (#163867)

ColinPeppler
authored and
pytorchmergebot
committed
Enable keep-going for trunk tags (#164307)

izaitsevfb
authored and
pytorchmergebot
committed
[FSDP][Replicate] created ReplicateModule and changed replicate to use it instead of FSDPModule (#163897)

anshul-si
authored and
pytorchmergebot
committed
Revert "[dynamo, 3.14] fix _detect_and_normalize_assert_statement for 3.14 (#164005)"
pytorchmergebot
committed
Add option to FakeProcessGroup to raise error if comms are invoked. (#162841)

ezyang
authored and
pytorchmergebot
committed
[SymmMem] Add get_nbi the nonblocking version (#163540)

kwen2501
authored and
pytorchmergebot
committed
[Replicate][Pipeline Parallelism] integration of new replicate function with pipeline parallelism (#164031)

anshul-si
authored and
pytorchmergebot
committed
[export] Preserve nn_module_stack for aliased nn modules (#164311)

anijain2305
authored and
pytorchmergebot
committed
Make viable/strict updatable again (#164374)

malfet
authored and
pytorchmergebot
committed
[inductor] Fix constant shape for float constants (#164241)

isuruf
authored and
pytorchmergebot
committed
[inductor] add a runtime assert for triton shapes (#164242)

isuruf
authored and
pytorchmergebot
committed
[4/N] Apply ruff UP035 rule to python code (#164206)

cyyever
authored and
pytorchmergebot
committed
[ROCm] fix carveout feature (#164303)

jeffdaily
authored and
pytorchmergebot
committed
[testing] Better short job name during upload additional stats (#164287)

clee2000
authored and
pytorchmergebot
committed
Skip windows unittest in fbcode (#164363)

yushangdi
authored and
pytorchmergebot
committed
[CD][CUDA13][NCCL] Fix nccl version typo for cu13 (#164383)

nWEIdia
authored and
pytorchmergebot
committed
Add functions to setup PrivateUse1 as a python backend device. (#157859)

qihqi
authored and
pytorchmergebot
committed
Enable torch.nn.functional.batch_norm in test_export_opinfo (#164261)

yiming0416
authored and
pytorchmergebot
committed
Add xfailing test case for inplace mutation of local DTensor (#164355)

ezyang
authored and
pytorchmergebot
committed
Work Around exposing statically linked libstdc++ CXX11 ABI strong symbols (#163980)

authored and
pytorchmergebot
committed
Add B200 to Operator Microbenchmark CI (#164288)

jainapurva
authored and
pytorchmergebot
committed

Commits on Oct 2, 2025

[inductor] log kernel autotuning result to a csv (#164191)

shunting314
authored and
pytorchmergebot
committed
Revert "[DCP] Decrease checkpoint background process Gloo pg init timeout (#162760)"
pytorchmergebot
committed
Speed up FP precision lookup (#164044)

lakshayg
authored and
pytorchmergebot
committed
Better error handling in torch/csrc/jit/codegen/* (#163948)

licy666
authored and
pytorchmergebot
committed
[CI][CUDA] Fix distributed tests for b200 (#164345)

Aidyn-A
authored and
pytorchmergebot
committed
[cutass backend] remove cutlass presets (#164380)

henrylhtsang
authored and
pytorchmergebot
committed
C++-accessible Placements via pybind11 (#163030)

swolchok
authored and
pytorchmergebot
committed
Fix unbacked replacement where LHS is purely backed expr and RHS is unbacked expr (#164013)

ColinPeppler
authored and
pytorchmergebot
committed
[torchfuzz] make fuzzer deterministic (#164397)

bobrenjc93
authored and
pytorchmergebot
committed
[torchfuzz] add matmuls (#164284)

bobrenjc93
authored and
pytorchmergebot
committed
[torchfuzz] keep track of operator stats (#164334)

bobrenjc93
authored and
pytorchmergebot
committed
[MTIA] Enable deserialization for FP8 checkpoint loading (#163559)

PatriceVignola
authored and
pytorchmergebot
committed
[inductor] teach bisector to look at pre_grad passes (#164250)

avikchaudhuri
authored and
pytorchmergebot
committed
[inductor] separate preamble from main work in compile_fx (#164169)

avikchaudhuri
authored and
pytorchmergebot
committed
unbacked reshape_copy (#164336)

laithsakka
authored and
pytorchmergebot
committed
[vision hash update] update the pinned vision hash (#154694)

authored and
pytorchmergebot
committed
[5/N] Apply ruff UP035 rule (#164423)

cyyever
authored and
pytorchmergebot
committed
DebugMode add ignore_compile_internals (#164205)

SherlockNoMad
authored and
pytorchmergebot
committed
Better path handling for nightly setup tool (#164215)

XuehaiPan
authored and
pytorchmergebot
committed
[PyTorch / Sigrid GPU] Fixes in pinned stats collection and add new ODS pinned memory stats (#164412)

banitag1
authored and
pytorchmergebot
committed
[torchfuzz] add test suite of fuzzer repros that we xfail (#164430)

bobrenjc93
authored and
pytorchmergebot
committed
[inductor] Handle patterns where input/output nodes are the same (#163994)

angelayi
authored and
pytorchmergebot
committed
[dynamo][export] Do not graph break on torch.autograd._profiler_enabled for export (#164418)

anijain2305
authored and
pytorchmergebot
committed
Fix SAC + Flex issue (#164421)

drisspg
authored and
pytorchmergebot
committed
Add failing bitwise equivalence UT for aot_eager on rms_norm (#164280)

ezyang
authored and
pytorchmergebot
committed
Add methods to access data and unpack_hook on SavedVariable (#164358)

soulitzer
authored and
pytorchmergebot
committed
Revert "Improve repeat op to a single copy (#163842)"
pytorchmergebot
committed
[DeviceMesh] Simplifying internal bookkeeping with CuTe layout (#163213)

fduwjj
authored and
pytorchmergebot
committed
Revert "Add magic TORCH_MAKE_PYBIND_ENUM_FASTER macro (#163527)"
pytorchmergebot
committed
Stop parsing command line arguments every time common_utils is imported. (#156703)

AnthonyBarbier
authored and
pytorchmergebot
committed
Make sure Windows CUDA 12.8 build follow same arches as Linux builds (#164470)

atalman
authored and
pytorchmergebot
committed
Revert "Stop parsing command line arguments every time common_utils is imported. (#156703)"
pytorchmergebot
committed
Add CUDA release architecture matrix (#164471)

atalman
authored and
pytorchmergebot
committed
Revert "[vision hash update] update the pinned vision hash (#154694)"
pytorchmergebot
committed
Add provenance to inductor IR nodes created after graph.run (#164255)

yushangdi
authored and
pytorchmergebot
committed
Revert "Use TMA loads always for Triton grouped MM kernel (#164256)"
pytorchmergebot
committed
Update torch.rst (#164408)

parthava-adabala
authored and
pytorchmergebot
committed
Revert "C++-accessible Placements via pybind11 (#163030)"
pytorchmergebot
committed
[AOTI win] Add ABI stable method for updating constant buffer (#163819)

yushangdi
authored and
pytorchmergebot
committed
Add Aidyn-A to CUDA codeowners (#164436)

Aidyn-A
authored and
pytorchmergebot
committed
Fix NestedTensor max/min operations for integer dtypes. (#162273)

adabeyta
authored and
pytorchmergebot
committed
[RELAND v2] Close some sources of fake tensors (#164372)

tugsbayasgalan
authored and
pytorchmergebot
committed
Use posix_fallocate() to reserve disk space for shared memory (#161910)

wenjianhn
authored and
pytorchmergebot
committed
Remove old CUDA version checks (#164199)

cyyever
authored and
pytorchmergebot
committed
typo corrected in ivalue.cpp's comment (#164485)

RajeshvShiyal
authored and
pytorchmergebot
committed
add tensor subclass printing support in fx/graph.py (#164403)

bobrenjc93
authored and
pytorchmergebot
committed
Fix THP_PyObject_VirtualFree return type (#163763)

guangyey
authored and
pytorchmergebot
committed
Fix readibility checks in TIDY and apply them (#164475)

authored and
pytorchmergebot
committed
[ROCm][CI] fix test_cudnn_convolution_relu_cuda (#164466)

jeffdaily
authored and
pytorchmergebot
committed
Add initial suppressions for pyrefly (#164177)

maggiemoss
authored and
pytorchmergebot
committed
Revert "Speed up FP precision lookup (#164044)"
pytorchmergebot
committed
[DTensor] raise error if the local_tensor argument passed to DTensor.from_local is a DTensor (#164496)

XilunWu
authored and
pytorchmergebot
committed
Fix type hints in PrepareModuleInput and PrepareModuleInputOutput (#164482)

RohitRathore1
authored and
pytorchmergebot
committed
[inductor] fix TestTemplateRender in select_algorithm (#164158)

isuruf
authored and
pytorchmergebot
committed
[inductor] require shape in TritonCSEVariable (#162275)

isuruf
authored and
pytorchmergebot
committed
Revert "Add provenance to inductor IR nodes created after graph.run (#164255)"
pytorchmergebot
committed
Fake process group Direct construction error (#163665)

ahkush
authored and
pytorchmergebot
committed
Stop parsing command line arguments every time common_utils is imported. (#156703)

AnthonyBarbier
authored and
pytorchmergebot
committed
Revert "[DeviceMesh] Simplifying internal bookkeeping with CuTe layout (#163213)"
pytorchmergebot
committed
Fix FloorDiv should not generate non integer rationals (due to sympy bug) (#164398)

laithsakka
authored and
pytorchmergebot
committed
Improved support for autotuning in wrapper_fxir (#164132)

nandesuka
authored and
pytorchmergebot
committed
Support setting grad_dtype on leaf tensors (#162815)

soulitzer
authored and
pytorchmergebot
committed
MX: Remove redundant PLATFORM_SUPPORTS_MX_GEMM constant (#164320)

jagadish-amd
authored and
pytorchmergebot
committed
Fix vllm build issue (#164361)

authored and
pytorchmergebot
committed

Commits on Oct 3, 2025

[cuBLAS] update cuBLAS determinism docs, remove workspace requirement checks (#161749)

eqy
authored and
pytorchmergebot
committed
[6/N] Apply ruff UP035 rule (#164438)

cyyever
authored and
pytorchmergebot
committed
Update mask dtype (#164472)

eellison
authored and
pytorchmergebot
committed
[MPS] Add backward pass for `embedding_bag` (#163931)

kurtamohler
authored and
pytorchmergebot
committed
remove unnecessary registration (#164481)

eellison
authored and
pytorchmergebot
committed
fix flex attention eager: dont round down scores to low-precision (closes #163588) (#163986)

v0i0
authored and
pytorchmergebot
committed
multimem reduce (#164517)

kwen2501
authored and
pytorchmergebot
committed
Add pyrefly suppressions 2/n (#164513)

maggiemoss
authored and
pytorchmergebot
committed
[cutlass-4][take 2] upgrade to cutlass 4.2.1 (#164159)

henrylhtsang
authored and
pytorchmergebot
committed
[DeviceMesh] Simplifying internal bookkeeping with CuTe layout (#163213)

fduwjj
authored and
pytorchmergebot
committed
[torchfuzz] Support EagerVsFullGraphDynamicCompileWithNumericsCheck (#164432)

bobrenjc93
authored and
pytorchmergebot
committed
Add Rocm to Operator Microbenchmark CI (#164173)

jainapurva
authored and
pytorchmergebot
committed
Move call to output generated code in inductor (#161615)

CWOA
authored and
pytorchmergebot
committed
config for dcache + unit tests (#164512)

nmacchioni
authored and
pytorchmergebot
committed
[vllm hash update] update the pinned vllm hash (#164319)

authored and
pytorchmergebot
committed
Use torch.testing.test_close instead of torch.testing.test_allclose (#164539)

cyyever
authored and
pytorchmergebot
committed
[torchfuzz] add nn functional ops (#164434)

bobrenjc93
authored and
pytorchmergebot
committed
[torchfuzz] add norm operators (#164514)

bobrenjc93
authored and
pytorchmergebot
committed
Pin conda version for Docker builds (#164575)

atalman
authored and
pytorchmergebot
committed
forward fix #164481 (#164578)

jeffdaily
authored and
pytorchmergebot
committed
Support partial _DynamoCacheEntries when not all backends available (#163521)

jamesjwu
authored and
pytorchmergebot
committed
Change default device to current acclerator (#164399)

drisspg
authored and
pytorchmergebot
committed
Support calling torch.compile inside non-strict export (#164171)

tugsbayasgalan
authored and
pytorchmergebot
committed
[dtensor] avoid shape recompilations on DTensorSpec (#163820)

pianpwk
authored and
pytorchmergebot
committed
Remove old workaround in launch_logcumsumexp_cuda_kernel (#164567)

cyyever
authored and
pytorchmergebot
committed
[PyTorch Pinned Allocator] Add support of reserved pinned memory segment to avoid slow paths (#164501)

banitag1
authored and
pytorchmergebot
committed
Merge remote-tracking branch 'upstream/main' into rocm7.1_internal_testing_IFU_2025-10-03
github-actions[bot]
committed
Fix merge conflicts
pragupta
committed