Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1918 commits
Select commit Hold shift + click to select a range
7e98921
Fix missing mask on tl.atomic_add for constant-index stores (#179833)
shino16 Apr 16, 2026
45af0d6
Codegen mutation epilogue in _create_runtime_wrapper (#179600)
bobrenjc93 Apr 16, 2026
c83ef11
[ROCm] Update scaled_mm DeepSeek error message (#180384)
jagadish-amd Apr 16, 2026
09fca34
[BE] TensorWithFlatten rename to TraceableWrapperSubclass and Improve…
Lucaskabela Apr 16, 2026
d0031a3
[CI] Align vLLM wheel build CUDA versions with PyTorch nightly (#180607)
huydhn Apr 17, 2026
6b7601f
Make strides consistent for _scaled_dot_product_efficient_attention_b…
jansel Apr 16, 2026
f4f124a
Replace isinstance(x, OpaqueBase) with is_opaque_value() (#180530)
aorenste Apr 16, 2026
2005577
[xpu][feature] Support fork-safe device_count by pyzes (#178496)
guangyey Mar 31, 2026
8fdde52
Revert "Codegen mutation epilogue in _create_runtime_wrapper (#179600)"
pytorchmergebot Apr 17, 2026
26ab646
[ROCm] Add gfx1103 to wheel build arch list (#179653)
SiteRelEnby Apr 17, 2026
f07a136
[ROCm] Enable ROCm swizzle check and update scaled_mm swizzle tests (…
jagadish-amd Apr 17, 2026
e11c17b
[autograd] Extract Node class into node.h (#179765)
colesbury Apr 17, 2026
fe11c39
[Inductor][CUTLASS] Use subprocess.list2cmdline to properly quote com…
etaf Apr 16, 2026
9d68039
[dynamo] Refactor GetAttrBuiltinVariable (#179033)
azahed98 Apr 16, 2026
e2294af
Update third_party/kineto submodule to 23b5bb57 (#180606)
scotts Apr 17, 2026
507f793
[ROCm][inductor][UT] Preserve combo kernel HIP compile options (#180277)
naromero77amd Apr 17, 2026
24c273c
[inductor] Fix fp32=>bf16=>fp32 cast being lost (#180575)
jansel Apr 16, 2026
05550c3
Remove unused noqa directives in non-torch/, batch 1 (#180140)
aorenste Apr 17, 2026
34b12dd
Remove unused noqa directives in non-torch/, batch 2 (#180141)
aorenste Apr 17, 2026
9f34b01
[user-streams] Fix cudagraphs compatibility with current stream (#180…
mlazos Apr 17, 2026
91f8173
[xpu][fix] Refine oneDNN stride check (#166861)
guangyey Apr 16, 2026
18c7146
Revert "[ROCm] Add gfx1103 to wheel build arch list (#179653)"
pytorchmergebot Apr 17, 2026
cc3b1bc
Revert "[inductor] Fix fp32=>bf16=>fp32 cast being lost (#180575)"
pytorchmergebot Apr 17, 2026
be5f4f9
Codegen mutation epilogue in _create_runtime_wrapper (#179600)
bobrenjc93 Apr 17, 2026
0585eb9
Revert "Codegen mutation epilogue in _create_runtime_wrapper (#179600)"
pytorchmergebot Apr 17, 2026
3e5f342
Revert "[user-streams] Fix cudagraphs compatibility with current stre…
pytorchmergebot Apr 17, 2026
1dba334
Add cuDNN depthwise convolution backend flag (#176500)
gderossi Apr 17, 2026
10a2f75
[inductor][cutlass] Filter CUTLASS kernels by "fastaccum" only on Hop…
crcrpar Apr 17, 2026
6433e6f
Codegen mutation epilogue in _create_runtime_wrapper (#179600)
bobrenjc93 Apr 17, 2026
93df587
Implement MKLGenerator (#151218)
michalowski-arm Apr 17, 2026
665a875
Accelerate SDPA on Arm CPUs: Unroll exp_sum and max_mul kernels (#177…
fadara01 Apr 17, 2026
d129991
[UT][ROCm][inductor] ROCm-specific XFAILS list for torchinductor_opin…
naromero77amd Apr 17, 2026
3f43a35
Revert "[CI] Migrate B200 operator benchmark to OSDC (#180450)"
pytorchmergebot Apr 17, 2026
f5d1a90
[ROCm] Add gfx1103 to wheel build arch list (#179653)
SiteRelEnby Apr 17, 2026
0125e52
[AC remat] Skip backward regions that don't need recomputation (#180534)
tugsbayasgalan Apr 16, 2026
503340b
[dynamo] object.__getattribute__ to fallback to UDOV generic_getattr …
anijain2305 Apr 17, 2026
168339d
[NCCL][Symmetric Memory] Fix groupName in IntraNodeComm (#180663)
Aidyn-A Apr 17, 2026
cb0287b
[PyTorch] Fix binary_cross_entropy SymInt error with dynamic shapes (…
nandesuka Apr 17, 2026
d7fb73e
[dynamo] Add polyfills for operator.concat and operator.iconcat (#175…
tashapais Apr 17, 2026
a877375
[Inductor] Skip cpp_wrapper for async pipelined autotuning and static…
PaulZhang12 Apr 17, 2026
44327a0
[inductor] Fix fp32=>bf16=>fp32 cast being lost (#180575)
jansel Apr 17, 2026
3f2f731
[OSDC] Enable trunk workflow on OSDC runners (#180537)
huydhn Apr 17, 2026
68a7039
[Inductor] remove skiprocm from combo kernels combokernel tests (#180…
karthickai Apr 17, 2026
702aa84
[ROCm] Fix evaluate_platform_supports_fp8 false-positive (#180518)
jagadish-amd Apr 17, 2026
52a4b71
Updating Assertion to Consider Gil-less Python Environments (#180669)
grantlouisherman Apr 17, 2026
912cc72
[MPS] Support integer inputs to `histc` (#178624)
kurtamohler Apr 17, 2026
8d83b40
[CI] Add TSan job for free-threaded Python 3.14t (#179933)
colesbury Apr 17, 2026
9deaf8d
[CUDA][FP8][CPU][TEST] Properly saturate E4M3 on finite-overflow on C…
eqy Apr 17, 2026
c05e3b9
Move TensorImpl and StorageImpl incref/decref/try_incref wrappers fro…
colesbury Apr 17, 2026
c78ef13
[Inductor] Handle ordering constraints for user-kernel fusion (#179803)
jjvraw Apr 17, 2026
4c910af
Update FBGEMM submodule to v1.6.0 (#180694)
huydhn Apr 17, 2026
0aab3f8
[MPS] Add complex support to `c10/metal/reduction_utils.h` (#180708)
malfet Apr 17, 2026
fa6628f
Revert "[CI] Add TSan job for free-threaded Python 3.14t (#179933)"
pytorchmergebot Apr 18, 2026
077121c
[1/N] Use const_data_ptr (#178234)
cyyever Apr 18, 2026
434c209
[xpu][fix] Fix hard code UT failed on XPU (#180647)
guangyey Apr 17, 2026
dd57412
[xpu][feature] Enable Inductor sycl-tla standalone runner. (#174958)
etaf Apr 16, 2026
e7ee6c0
Refine graph capture check in optimizer (#180453)
guangyey Apr 17, 2026
ee2630e
Skip quantized benchmarks on aarch64 in operator_benchmark (#180532)
aorenste Apr 16, 2026
f13b92d
[CI] Migrate B200 operator benchmark to OSDC (#180450)
huydhn Apr 18, 2026
fe2fa45
[inductor][autotuning] extract _pre_launch/_post_launch from CachingA…
nmacchioni Apr 18, 2026
6f7cfc8
Revert "[ROCm] Handle profiler trace format differences in combo kern…
pytorchmergebot Apr 18, 2026
b4a129a
[MPS] Enable `test_reductions.py` with skips (#179407)
hvaara Apr 18, 2026
2974448
Fix quantized engine registration on aarch64 (#180679)
aorenste Apr 18, 2026
fadf0e3
[dynamo] Constant fold hex, oct, bin, ascii, and format builtins (#18…
anijain2305 Apr 17, 2026
96c92e0
Handle dict in _iterate_exprs for triton kernel wrapper nodes (#18047…
angzhang1 Apr 18, 2026
5275313
Move AllocatorConfig static functions into cpp to fix DSO boundary is…
garfield1997 Apr 18, 2026
50eea6c
Fix GDS smoke test failure on CUDA 13.2 (#180577)
atalman Apr 18, 2026
2f115b7
[2/N] Use const_data_ptr (#180726)
cyyever Apr 18, 2026
b42e51c
[cuDNN][SDPA][TEST] Bump tolerances for TF32 `test_transformerencoder…
eqy Apr 18, 2026
6800075
[shard prop] norm ops (#176992)
pianpwk Apr 16, 2026
81d5c3f
[TEST][CUDA] Fix `test_throw_on_cudamalloc_oom` alllocation size (#18…
eqy Apr 18, 2026
1c5a398
Add batching rule for `aten::view.dtype` (#180728)
ar0ck Apr 18, 2026
a94a176
Add missing FP8 skip decorator to test_scaled_mm_pdl_handles_none_bia…
aorenste Apr 18, 2026
45c9107
Revert "[shard prop] norm ops (#176992)"
pytorchmergebot Apr 18, 2026
fd98962
Disable check_batched_grad in test_sparse_mul gradcheck calls (#180645)
aorenste Apr 18, 2026
92b5ad2
[ROCm] Run test_scaled_mm_deepseek_error_messages on mi350 arch (#180…
jagadish-amd Apr 19, 2026
5269bd2
[caffe2/tensorexpr] Add LLVM 23+ branch for `ObjectLinkingLayerCreato…
yuxuanchen1997 Apr 19, 2026
9c511ab
[functorch] Add vmap batching rule for aten::unbind_copy.int (#178035)
osherlock1 Apr 19, 2026
300d094
[12/12] Enforce C++20 minimum in header guards (#178150) (#178150)
r-barnes Apr 19, 2026
0805164
[1/N] Use PyModule_AddType (#180729)
cyyever Apr 19, 2026
c49639e
[MPS] grid_sampler_3d backward pass (#179388)
mlaves Apr 19, 2026
e2f2dbe
Revert "[12/12] Enforce C++20 minimum in header guards (#178150) (#17…
pytorchmergebot Apr 19, 2026
10a41bb
[MPS] Replace sum/nansum/mean ops wth native Metal kernel (#180709)
malfet Apr 19, 2026
3ce00f4
Revert "[pyrefly] Add type annotations to torch/fx graph and graph_mo…
pytorchmergebot Apr 19, 2026
01fca47
[user-streams] Fix cudagraphs compatibility with current stream (#180…
mlazos Apr 18, 2026
cf14e0a
Use Py_IsNone/Py_IsTrue/Py_IsFalse for singleton identity checks (#18…
cyyever Apr 19, 2026
4d290c0
[BE]: Update pybind11 submodule to 3.0.4 (#175115)
Skylion007 Apr 19, 2026
2cb9a5b
Revert "[MPS] Replace sum/nansum/mean ops wth native Metal kernel (#1…
pytorchmergebot Apr 19, 2026
474a055
[BE] Remove double import in BinaryDivFloorKernel.cu (#179260)
benediktjohannes Apr 19, 2026
2fc2639
Revert "Remove unused noqa directives in non-torch/, batch 2 (#180141)"
pytorchmergebot Apr 19, 2026
88f4e9d
Revert "[inductor] Emit assert_alignment in cpp_wrapper (#177538)"
pytorchmergebot Apr 19, 2026
ad557f0
Revert "[inductor] Emit assert_size_stride in cpp_wrapper (#177535)"
pytorchmergebot Apr 19, 2026
947b712
[3/N] Use const_data_ptr (#180756)
cyyever Apr 19, 2026
fc4271c
Revert "Implement MKLGenerator (#151218)"
pytorchmergebot Apr 19, 2026
513f103
Use = default for trivial default ctors (#180769)
cyyever Apr 19, 2026
3a7e950
Use default member initializer (#180767)
cyyever Apr 19, 2026
a8b370c
Fix tearDown leak in TestOpaqueObject under PYTORCH_TEST_SKIP_FAST (#…
aorenste Apr 19, 2026
77b9eb4
[MPS] Replace sum/nansum/mean ops wth native Metal kernel (#180709)
malfet Apr 19, 2026
fc5e84b
Implement derivatives for nextafter operation (#148820)
baskargopinath Apr 19, 2026
5b23e8c
Refactor AOTAutograd cache key helpers (#180091)
bobrenjc93 Apr 19, 2026
7906593
Fix torch.trapz documentation signature to match torch.trapezoid (#18…
Jah-yee Apr 19, 2026
0951ec0
Remove unused branches for CUDA <12.1 (#171588)
cyyever Apr 20, 2026
3bf5f6f
Remove unused noqa directives in non-torch/, batch 2 (#180141)
aorenste Apr 19, 2026
28791d7
Remove unused pyrefly ignores (#180474)
maggiemoss Apr 20, 2026
aea31e0
[Profiler][PrivateUse1] Make `PrivateUse1ProfilerRegistry::registerWi…
KarhouTam Apr 20, 2026
1cf9d1a
Run clang-tidy with modernize-use-starts-ends-with (#179458)
lakshayg Apr 20, 2026
2545731
[shard prop] norm ops (#176992)
pianpwk Apr 20, 2026
133c64d
[doc][xpu]Update get start xpu for release 2.12 (#180659)
ZhaoqiongZ Apr 20, 2026
0e045b5
[DeviceMesh] Enforce 2-level Layouts (#172089)
lw Apr 17, 2026
8ac0b5f
Remove checks for CUDNN 7 (#170506)
cyyever Apr 20, 2026
151c76a
Revert "Run clang-tidy with modernize-use-starts-ends-with (#179458)"
pytorchmergebot Apr 20, 2026
ccb2214
Workaround for nvrtcCompileProgram changing locale in CUDA < 12.6.2 (…
Flamefire Apr 20, 2026
d9e347e
Revert "Remove unused branches for CUDA <12.1 (#171588)"
pytorchmergebot Apr 20, 2026
cc52542
Replace MPSGraph count_nonzero with custom Metal kernel (#180725)
malfet Apr 19, 2026
debaeab
Fix test_hf_logging_logger CI hang by dropping transformers import (#…
aorenste Apr 20, 2026
7da749c
[AOTI] Add FP8 header files in aoti shim.h (#178120)
CaoE Apr 20, 2026
c3878c1
[BE][CD] Always build small wheels for CUDA Linux and remove CUDA 12.…
atalman Apr 20, 2026
59e28df
Enable RUF100 (#180142)
aorenste Apr 20, 2026
938df06
[DeviceMesh] Make the non-overlapping check more strict (#172343)
lw Apr 17, 2026
3493232
[3/N] Use const_data_ptr (#180801)
cyyever Apr 20, 2026
41b494c
Fix DTensor Partial placement lost during autograd layout invariant …
anshul-si Apr 16, 2026
befadd6
[Profiler] Add ryanzhang22 as a profiler owner (#180680)
ryanzhang22 Apr 20, 2026
34d67bd
Use shared constants for cache key prefixes (#180093)
bobrenjc93 Apr 20, 2026
9daaaa2
Refactor duplicated AOTAutograd cache-save checks (#180094)
bobrenjc93 Apr 20, 2026
ee20fa0
[torchelastic] Start health check server before remote_pre_launch in …
sourabhrohilla Apr 20, 2026
3a89337
[CI] Remove runtime Chocolatey installs from Windows CI
ZainRizvi Apr 20, 2026
dfd48e3
[inductor][estimations] profile guided estimation option for runtime …
IvanKobzarev Apr 20, 2026
26e558b
Revert "[CI] Remove runtime Chocolatey installs from Windows CI" (#18…
ZainRizvi Apr 20, 2026
0c40acf
[MPS] fix lstm train/eval error (#180873)
Isalia20 Apr 20, 2026
105bdd1
Correct the Repo Name in allowlist.yaml (#180794)
fffrog Apr 20, 2026
c00aa0a
[CI] Fix target-determination-indexer pytorch checkout path (#180711)
huydhn Apr 20, 2026
493ee4b
[profiler] Add profiler chrome trace validator with rules (#177947)
jiannanWang Apr 20, 2026
bdec947
[inductor] Fix torch.compile crash on cumsum with broadcast input whe…
liqiangxl Apr 20, 2026
3fa6c02
Make Generator an Opaque Object (#179661)
yushangdi Apr 18, 2026
5c61b13
Revert "Make Generator an Opaque Object (#179661)"
pytorchmergebot Apr 20, 2026
05e702f
[ROCm][UT] Remove previously retained Triton 3.7 skip for torchinduct…
naromero77amd Apr 20, 2026
5e90d15
[Profiler] Remove references to _KinetoProfile in public docs (#180672)
ryanzhang22 Apr 20, 2026
a2e8fa3
[cuDNN][SDPA] Don't route to cuDNN SDPA for batch size or head dim > …
eqy Apr 20, 2026
292a779
[DTensor] Register sharding strategies for upsample/interpolation bac…
shuhuayu Apr 20, 2026
18357d9
Make Generator an Opaque Object (#179661)
yushangdi Apr 20, 2026
6ed4d6e
Fix FakeProcessGroup collectives leaving output uninitialized (#178078)
ydwu4 Apr 17, 2026
bbe4a1e
[dynamo] Fix duplicated-word typos in comments and docstrings (#180609)
frgossen Apr 20, 2026
e361471
[CI][Build] Goodbye Bazel (#180883)
malfet Apr 20, 2026
24b14fd
[ROCm][CI] Modify permissions in nightly workflow (#180877)
amdfaa Apr 20, 2026
0876a27
[dynamo][refactor] Reparent DefaultDictVariable to UserDefinedDictVar…
anijain2305 Apr 20, 2026
f926772
[BE]: Enable clang-tidy on c10::SmallVector reserves to reduce reallo…
Skylion007 Apr 20, 2026
cc5f2ca
Revert "[CI][Build] Goodbye Bazel (#180883)"
pytorchmergebot Apr 20, 2026
7a59cc2
[Inductor] Add batch-invariant accuracy mode for benchmark perf tests…
karthickai Apr 20, 2026
c48c691
[torch.compile] Add torch.compiler.set_default_backend (#178944)
angelayi Apr 20, 2026
100e540
[CI][Build] Goodbye Bazel (#180883)
malfet Apr 20, 2026
83ba944
[dynamo] Fix cuda_stream pointer extraction for generic torch.Stream …
GSumanth109 Apr 20, 2026
9df7f79
[ROCm] Fix FlexAttention fp16 default num_warps (8 -> 4) on AMD GPUs …
nithinsubbiah Apr 20, 2026
f65be27
[Pytorch][Kineto] Reland comm_id generation changes (#180615) (#180615)
ycui1984 Apr 20, 2026
092226f
[FakeTensor] Preserve grad_dtype when converting real tensors to fake…
ydwu4 Apr 16, 2026
b2e1f52
expose num_splits in FA2 + bump flash attention submodule commit (#17…
liangel-02 Apr 20, 2026
cb242af
Revert "Make Generator an Opaque Object (#179661)"
pytorchmergebot Apr 21, 2026
f710883
[pyrefly] Add type annotations to torch/fx graph and graph_module (#1…
Lucaskabela Apr 21, 2026
2cef5ba
[Dynamo]Add python_type() to TritonKernelVariable (#180882)
fxdawnn Apr 20, 2026
cc5d8df
Reorder Fake dispatch key before Python (#178428)
liangel-02 Apr 9, 2026
f63d04e
[Inductor] Fix flaky test_interim_result_with_future_pattern race con…
karthickai Apr 15, 2026
80fa7f9
Allow pytorch-bot to trigger Claude Autorevert Advisor workflow (#180…
izaitsevfb Apr 21, 2026
c4ec73b
[inductor] Re-enable split reductions when cooperative reductions won…
benjaminglass1 Apr 20, 2026
7943f32
Revert "Fix FakeProcessGroup collectives leaving output uninitialized…
pytorchmergebot Apr 21, 2026
0cfb7e2
Revert "[inductor][estimations] profile guided estimation option for …
pytorchmergebot Apr 21, 2026
0309973
[autograd] Convert shared_ptr<Node> to intrusive_ptr<Node> (#179766)
colesbury Apr 17, 2026
d4b9bd5
enable software pipelining for mix-order reduction (#180892)
shunting314 Apr 20, 2026
d4791a1
[dynamo] Filter aliased intermediates in autograd.Function forward tr…
anijain2305 Apr 20, 2026
7b8a833
[inductor] Fix test_extension_backend race condition (#180667)
choijon5 Apr 17, 2026
7a197d9
[mem viz] use segment size for envelope to show fragmentation (#180515)
yushangdi Apr 20, 2026
95e6cf8
[mem viz] Add summary blocks to pools (#180613)
yushangdi Apr 20, 2026
1b421fa
[inductor] Fix MSVC const pointer emission in cpp wrapper temporary a…
yuchengliu1 Apr 21, 2026
30884b8
[FSDP2] add unit test and invariant comments for HSDP AR buffer lifet…
weifengpy Apr 20, 2026
0bdf1e5
[FSDP2] Fix post_accumulate_grad_hook staleness under CPUOffloadPolic…
weifengpy Apr 21, 2026
d5347ac
[dynamo] Update test_model_output for transformers v5 and Make decomp…
mansiag05 Apr 21, 2026
5eac661
[dynamo] Lint Fx graph before sending to backend (#180922)
anijain2305 Apr 21, 2026
71d3bd2
[DTensor] Prevent squeeze from redistributing with strict_view (#175798)
stmcgovern Apr 21, 2026
a11cc39
[CI] Add TSan job for free-threaded Python 3.14t (#179933)
colesbury Apr 21, 2026
244c581
Use C++20 std::atomic_ref to simplify cpu_atomic_add_float (#180784)
cyyever Apr 21, 2026
84ac583
Fix pad_tensor/unpad_tensor creating unnecessary guards on symbolic p…
fmassa Apr 21, 2026
9947816
Don't redefine `TORCH_USE_CUDA_DSA` (#175146)
Flamefire Apr 21, 2026
d08a9ae
Revert "[pyrefly] Add type annotations to torch/fx graph and graph_mo…
pytorchmergebot Apr 21, 2026
781db14
[inductor][estimations] profile guided estimation option for runtime …
IvanKobzarev Apr 21, 2026
7231f9e
[inductor] Fix cpp_wrapper lazy compile for combo kernels with per-su…
desertfire Apr 16, 2026
e2dc222
Revert "expose num_splits in FA2 + bump flash attention submodule com…
pytorchmergebot Apr 21, 2026
a66567f
Move some pre-build steps from setup.py to CMake (#177641)
zklaus Apr 20, 2026
4aef62c
Make Generator an Opaque Object (#179661)
yushangdi Apr 21, 2026
dbae54c
Add ghstack workflow notes to CLAUDE.md (#180991)
ezyang Apr 21, 2026
07dc5b8
[ROCm] Symmetric Memory - Add rocSHMEM Triton integration (#178658)
drprajap Apr 21, 2026
22615cb
Revert "[inductor] Fix cpp_wrapper lazy compile for combo kernels wit…
pytorchmergebot Apr 21, 2026
0ad0d42
[inductor][cutlass evt] skip some on SM 12.0 (#180817)
crcrpar Apr 21, 2026
17e6275
Revert "[ROCm] Symmetric Memory - Add rocSHMEM Triton integration (#1…
pytorchmergebot Apr 21, 2026
446033e
[RISCV] disable cuda-bingings on riscv64 CI (#173663)
yuzibo Apr 21, 2026
300d114
[inductor] Fix mix_order_reduction over-fusion via load count check (…
abaybektursun Apr 21, 2026
5842b0e
[PyTorch] Fix broadcast_shapes op missing in selective builds (#18086…
nandesuka Apr 21, 2026
b72c0ac
Revert "Workaround for nvrtcCompileProgram changing locale in CUDA < …
pytorchmergebot Apr 21, 2026
3cfc333
[BE] suggest running with CUDA_LOG_FILE in cuda error msg (#181011)
ngimel Apr 21, 2026
9644178
[nativert][sigmoid] Fix None input values having incorrect producer p…
georgiaphillips Apr 21, 2026
ba36784
[inductor] Fix duplicated-word typos in comments and docstrings (#180…
frgossen Apr 20, 2026
fd43b72
Fix folded constant offset indexing in AOTI constant buffer updateFix…
FFXIVYYDS Apr 21, 2026
f305166
[pyrefly] Add type annotations to torch/fx graph and graph_module (#…
Lucaskabela Apr 21, 2026
add9804
[CI] Use portable jemalloc path lookup in build.sh (#180983)
subinz1 Apr 21, 2026
eafb5d1
[ROCm] Build with rocshmem support only if rocshmem ver>=3.3.0 (#180901)
pragupta Apr 21, 2026
ef44be6
[vllm hash update] update the pinned vllm hash (#180516)
pytorchupdatebot Apr 21, 2026
5b4d57d
Unify Inductor cache key strategies (#180799)
bobrenjc93 Apr 21, 2026
fc92844
[MPS] Fix sliced channels_last tensors handling (#180992)
malfet Apr 21, 2026
fcc7e46
[ROCm][CI] Use env vars and clean up docker-cache-rocm.yml (#180710)
jithunnair-amd Apr 21, 2026
eb06bf6
[inductor] Fix stride mismatch for user-visible reductions (#180197)
liqiangxl Apr 21, 2026
320c41a
[inductor] Raise split reduction threshold from 8K to 524K on Blackwe…
liqiangxl Apr 21, 2026
c194562
Codegen output alias regeneration in _create_runtime_wrapper (#179601)
bobrenjc93 Apr 21, 2026
46c0c35
Add torch._dynamo.override_optimization_hint API for unbacked symbols…
laithsakka Apr 21, 2026
420f50f
[inductor] Fix bug with contiguous checks and comprehensive_padding (…
jansel Apr 21, 2026
eec2645
[misc] Fix duplicated-word typos in comments and docstrings (#181028)
frgossen Apr 21, 2026
dab4e83
use reserve, move, and simplify vector construction in inductor (#180…
cyyever Apr 22, 2026
5e344fd
[MTIA][Inductor] Eliminate redundant copy kernel for concat graph out…
nandesuka Apr 22, 2026
aaf9265
[Inductor] Make persistent-reduction config pick batch-invariant unde…
karthickai Apr 21, 2026
0274ad6
[Inductor] Forward optimize_mem to combo kernel inductor_meta (#180790)
karthickai Apr 21, 2026
6a6780b
Fix undefined symbolic variable in fx_graph_runnable repro scripts (#…
yushangdi Apr 22, 2026
bca0ce3
Skip llms-full.txt during Sphinx build, generate in nightly push (#18…
svekars Apr 22, 2026
17b1573
[s390x CI] Disable scheduled s390x-periodic runs (#181005)
ZainRizvi Apr 22, 2026
8e58f30
[BE] Remove pyobj_interpreter_ from PyObjectSlot (#179631)
colesbury Apr 22, 2026
50294ed
[MPS] Flatten 5D tensors to 4D in batch_norm for performance (#180335)
nuri-yoo Apr 22, 2026
1aa5e36
Fix fbgemm check from #180679 (#180958)
aorenste Apr 21, 2026
a39ec69
[Inductor] Use triton_meta_common() in combo kernel jit_line for disa…
karthickai Apr 21, 2026
c161f2c
Add non-strict backward patching helper (#181012)
tugsbayasgalan Apr 21, 2026
b627bfb
[dynamo] Taint filtered aliased intermediates with clear error messag…
anijain2305 Apr 22, 2026
ba15482
Add TLS stack_bounds so we dont pay expensive reads (#181018)
drisspg Apr 21, 2026
a54d6de
Revert "[Inductor] Use triton_meta_common() in combo kernel jit_line …
pytorchmergebot Apr 22, 2026
c6d36a4
Fix FakeProcessGroup collectives leaving output uninitialized (#178078)
ydwu4 Apr 21, 2026
4c9b9c3
[vllm hash update] update the pinned vllm hash (#181048)
pytorchupdatebot Apr 22, 2026
3d1e839
[Profiler][PrivateUse1] Expose backend name as alias in ProfilerActiv…
KarhouTam Apr 22, 2026
c56d764
[inductor] Fix CPU CPP fusion crash on GroupNorm + SDPA + bmm (#181015)
jansel Apr 22, 2026
12c7b75
[Benchmark] Fix xpu benchmark workflow issue (#180825)
chuanqi129 Apr 22, 2026
89ed986
[Inductor] Use triton_meta_common() in combo kernel jit_line for disa…
karthickai Apr 22, 2026
69daaed
Formalize out= operators and custom operators definition (#180851)
zou3519 Apr 22, 2026
b139074
Support out-tagged custom operators in torch.compile (#180852)
zou3519 Apr 22, 2026
9c406e3
Auto-generate fake kernels for Tag.out custom operators (#180987)
zou3519 Apr 22, 2026
2db14fe
Add doc for CRCR (#180976)
can-gaa-hou Apr 22, 2026
b5784b9
[Triton 3.7] Update Triton hash (#181001)
atalman Apr 22, 2026
2b5ff12
Revert "[Inductor] Use triton_meta_common() in combo kernel jit_line …
pytorchmergebot Apr 22, 2026
138d202
expose num_splits in FA2 + bump flash attention submodule commit (#17…
liangel-02 Apr 22, 2026
db8c0b8
FakeTensor C++ Migration: Modifying TensorImpl (#178429)
liangel-02 Apr 22, 2026
8908772
Revert "[BE] Remove pyobj_interpreter_ from PyObjectSlot (#179631)"
pytorchmergebot Apr 22, 2026
633d8e1
custom_op: handle omitted optional mutated defaults (#180621)
bobrenjc93 Apr 22, 2026
a9cd064
Revert "[DeviceMesh] Make the non-overlapping check more strict (#172…
pytorchmergebot Apr 22, 2026
529b891
Revert "[DeviceMesh] Enforce 2-level Layouts (#172089)"
pytorchmergebot Apr 22, 2026
9372b02
Revert "[autograd] Convert shared_ptr<Node> to intrusive_ptr<Node> (#…
pytorchmergebot Apr 22, 2026
76e3699
Revert "Fix FakeProcessGroup collectives leaving output uninitialized…
pytorchmergebot Apr 22, 2026
b5dc529
Revert "enable software pipelining for mix-order reduction (#180892)"
pytorchmergebot Apr 22, 2026
83de8b8
[PT2] Make functorch JVP operator torch.exporatble (#179686) (#179686)
malaybag Apr 22, 2026
6ecd5d4
Merge remote-tracking branch 'upstream/main' into develop_IFU_20260422
github-actions[bot] Apr 22, 2026
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
The table of contents is too big for display.
Diff view
Diff view
  •  
  •  
  •  
114 changes: 0 additions & 114 deletions .bazelrc

This file was deleted.

1 change: 0 additions & 1 deletion .bazelversion

This file was deleted.

1 change: 0 additions & 1 deletion .ci/docker/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,6 @@ See `build.sh` for valid build environments (it's the giant switch).

* `conda` - Dockerfile and build.sh to build Docker images used in nightly conda builds
* `manywheel` - Dockerfile and build.sh to build Docker images used in nightly manywheel builds
* `libtorch` - Dockerfile and build.sh to build Docker images used in nightly libtorch builds

## Usage

Expand Down
10 changes: 4 additions & 6 deletions .ci/docker/almalinux/Dockerfile
Original file line number Diff line number Diff line change
Expand Up @@ -19,15 +19,12 @@ RUN git config --global --add safe.directory '*'
ENV PATH=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/bin:$PATH

# cmake-3.18.4 from pip
# NS: Apr 1 2026 3.18.4 is gone, reported here https://github.com/scikit-build/cmake-python-distributions/issues/693
RUN yum install -y python3-pip && \
python3 -mpip install cmake==3.18.4 && \
python3 -mpip install cmake==3.18.4.post1 && \
ln -s /usr/local/bin/cmake /usr/bin/cmake3
RUN rm -rf /usr/local/cuda-*

FROM base as openssl
ADD ./common/install_openssl.sh install_openssl.sh
RUN bash ./install_openssl.sh && rm install_openssl.sh

FROM base as patchelf
# Install patchelf
ADD ./common/install_patchelf.sh install_patchelf.sh
Expand Down Expand Up @@ -84,6 +81,8 @@ RUN yum -y update && \
yum -y install glibc-langpack-en && \
yum install -y sudo wget curl perl util-linux xz bzip2 git patch which perl zlib-devel openssl-devel yum-utils autoconf automake make gcc-toolset-${DEVTOOLSET_VERSION}-gcc gcc-toolset-${DEVTOOLSET_VERSION}-gcc-c++ gcc-toolset-${DEVTOOLSET_VERSION}-gcc-gfortran gcc-toolset-${DEVTOOLSET_VERSION}-gdb
RUN git config --global --add safe.directory '*'
# All rocm clang cfg files load the same rocm.cfg, make sure it points to the right toolchain.
RUN echo "--gcc-toolchain=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr" >> /opt/rocm/llvm/bin/rocm.cfg
ENV PATH=/opt/rh/gcc-toolset-${DEVTOOLSET_VERSION}/root/usr/bin:$PATH

FROM rocm_base as rocm
Expand All @@ -109,7 +108,6 @@ COPY --from=cuda13.2 /usr/local/cuda-13.2 /usr/local/cuda-13.2
# Final step
FROM ${BASE_TARGET} as final
ARG DEVTOOLSET_VERSION=13
COPY --from=openssl /opt/openssl /opt/openssl
COPY --from=patchelf /patchelf /usr/local/bin/patchelf
COPY --from=conda /opt/conda /opt/conda

Expand Down
2 changes: 1 addition & 1 deletion .ci/docker/almalinux/build.sh
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ case ${DOCKER_TAG_PREFIX} in
;;
rocm*)
BASE_TARGET=rocm
PYTORCH_ROCM_ARCH="gfx900;gfx906;gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1200;gfx1201;gfx950;gfx1150;gfx1151"
PYTORCH_ROCM_ARCH="gfx900;gfx906;gfx908;gfx90a;gfx942;gfx1030;gfx1100;gfx1101;gfx1102;gfx1103;gfx1200;gfx1201;gfx950;gfx1150;gfx1151"
EXTRA_BUILD_ARGS="${EXTRA_BUILD_ARGS} --build-arg PYTORCH_ROCM_ARCH=${PYTORCH_ROCM_ARCH}"
;;
*)
Expand Down
Loading