189 Pull requests opened by 114 people

• Extend compute_global_tensor_shape to multi dimension sharding

  #152166 opened Apr 25, 2025

• Generate test reports for pytest when option is given

  #152170 opened Apr 25, 2025

• [c10d] Allow split_group to work with non nccl backends

  #152175 opened Apr 25, 2025

• IGNORE: Testing OIDC

  #152181 opened Apr 25, 2025

• [WIP] New Win Arm64 Runners - User pre installed Visual Studio

  #152184 opened Apr 25, 2025

• xpu: get xpu arch flags at runtime in cpp_extensions

  #152192 opened Apr 25, 2025

• SAC: fix recompute tag propagation for ops with list[tensor] inputs

  #152193 opened Apr 25, 2025

• SAC: fix recompute tag propagation for ops with list[tensor] inputs

  #152194 opened Apr 25, 2025

• SAC: fix recompute tag propagation for ops with list[tensor] inputs

  #152195 opened Apr 25, 2025

• Add detailed triton kernel logging to tlparse

  #152197 opened Apr 25, 2025

• [inductor] propagate shapes in CSEVariable

  #152198 opened Apr 25, 2025

• Synchronize mps backend in the timer

  #152199 opened Apr 25, 2025

• [submodule] Update ONNX to 1.18

  #152200 opened Apr 25, 2025

• Add support for torch.cuda.FloatTensor()

  #152208 opened Apr 25, 2025

• [CI] docker images use tags instead of image name

  #152209 opened Apr 25, 2025

• Move mps_linear forward to use MPS kernels directly instead of MPSGraph

  #152210 opened Apr 25, 2025

• Mini tutorial for provenance tracking

  #152211 opened Apr 25, 2025

• Improve error handling in CachingAutotuner for argument mismatches

  #152215 opened Apr 25, 2025

• [not for land] functionalization hack to try making mutations on graph input slices more efficient

  #152217 opened Apr 25, 2025

• Add `padding="same"` for transposed convolution

  #152228 opened Apr 25, 2025

• Fix: Consider input defined unbacked during inductor codegen for runtime asserts

  #152231 opened Apr 25, 2025

• _get_total_norm should use float64 to avoid rounding errors

  #152234 opened Apr 25, 2025

• At least one of ROCM_HOME or CUDA_HOME must be None

  #152236 opened Apr 26, 2025

• [executorch hash update] update the pinned executorch hash

  #152238 opened Apr 26, 2025

• Updates to build on Noble (Ubuntu24.04) and py3.12

  #152240 opened Apr 26, 2025

• Enable 8byte vector loading for fp16/bf16

  #152242 opened Apr 26, 2025

• Move code out of individual token linters

  #152256 opened Apr 26, 2025

• [BE]: Cleanup traceutils with fmtlib

  #152265 opened Apr 26, 2025

• [ROCm] Maxpool backward NHWC Perf Improvement targeting Resnet scenarios

  #152267 opened Apr 26, 2025

• [Dynamo] Replace `unimplemented` with `unimplemented_v2` in `torch/_dynamo/variables/misc.py` [1/2]

  #152274 opened Apr 27, 2025

• [CI] Add xpu inductor test into periodic workflow

  #152281 opened Apr 27, 2025

• [DTensor] enable SimpleFSDP's composability with Tensor Parallel

  #152286 opened Apr 27, 2025

• [inductor] Skip isinf check for FP8 E4M3 dtype

  #152289 opened Apr 28, 2025

• Enable the AMP precision with freezing for CPU nightly test

  #152298 opened Apr 28, 2025

• [cp] dispatch flex_attention_backward to CP impl in TorchDispatchMode

  #152311 opened Apr 28, 2025

• setuptools.build_meta:__legacy__ backend is deprecated

  #152313 opened Apr 28, 2025

• Fixed RELEASE.md typo

  #152315 opened Apr 28, 2025

• Correct torch.xpu.is_bf16_supported return False if no XPU detected

  #152317 opened Apr 28, 2025

• [dynamo] Use getattr when accessing self.value.__module__ in SkipFunctionVariable

  #152320 opened Apr 28, 2025

• [ROCm] Unskipped test_rnn_dropout_state for ROCm

  #152339 opened Apr 28, 2025

• [ROCm][Inductor][CK] Add ck-tile based universal gemm kernels to torch.mm autotune choices

  #152341 opened Apr 28, 2025

• [Memento] Enable on-demand mode

  #152342 opened Apr 28, 2025

• [WIP] DeadCodeEliminator Mark(block) improvement

  #152348 opened Apr 28, 2025

• [inductor][dynamo] Include operator name in size/stride/alignment assertion

  #152353 opened Apr 28, 2025

• Add codeowner for merge rules

  #152354 opened Apr 28, 2025

• [Will This Work?] Build libgomp (gcc-11) from src on AArch64

  #152361 opened Apr 28, 2025

• Format all headers under ATen/cpu/vec, not just top-level

  #152364 opened Apr 28, 2025

• add is_vec_specialized_for

  #152365 opened Apr 28, 2025

• vec::map: directly process reduced-precision floats when reasonable

  #152366 opened Apr 28, 2025

• [Relandx2] Rewrite the guts of torch::jit::Lexer to speed it up

  #152372 opened Apr 28, 2025

• complex.pow(2) on GPU by replacing with complex * complex to avoid numerical instability

  #152373 opened Apr 28, 2025

• [FP8][CUTLASS] xFail `honor_sm_carveout` on `sm100`

  #152378 opened Apr 28, 2025

• fix: outdated contents in dynamo overview

  #152382 opened Apr 28, 2025

• [inductor][subgraph] Simplify the resulting output code for subgraph

  #152383 opened Apr 28, 2025

• [inductor][invoke_subgraph] Remove assertion checks for outputs of invoke_subgraph

  #152384 opened Apr 28, 2025

• Add vec_reduce_all specialization for std::plus on AArch64

  #152388 opened Apr 28, 2025

• [Hierarchical Compilation] Track node mutations

  #152389 opened Apr 29, 2025

• [Inductor] Fix typing in cuda_template.py

  #152390 opened Apr 29, 2025

• [Inductor] Use `torch._dynamo.utils.same` in block pointer tests, adding atol/rtol kwargs to it.

  #152392 opened Apr 29, 2025

• [Accelerator] Fix Python typing in accelerator

  #152394 opened Apr 29, 2025

• [NFC] [inductor] [compile async] Warn exception if pickler failed

  #152401 opened Apr 29, 2025

• [Do not merge] poke CI with FX IR always on

  #152405 opened Apr 29, 2025

• Call torch.distributed.destroy_process_group() at the end of the example

  #152407 opened Apr 29, 2025

• [Inductor][CPU] bug fix for int8 GEMM compensation epilogue

  #152408 opened Apr 29, 2025

• Cleanup DeviceInterface in triton test

  #152409 opened Apr 29, 2025

• [Hierarchical Compile] Add mutation dependencies to topological sorting

  #152410 opened Apr 29, 2025

• [Quant][X86] add ops to compute uint8 pointwise add/add_relu

  #152411 opened Apr 29, 2025

• Add Vectorized FP8 E4M3

  #152417 opened Apr 29, 2025

• [Inductor][CPP] Enable vectorized fp8 quant dequant

  #152418 opened Apr 29, 2025

• Relax tolerance for test_quick_baddbmm_cpu_complex64

  #152424 opened Apr 29, 2025

• [ROCm] cpp_extension allow user to override default flags

  #152432 opened Apr 29, 2025

• Log aot and idx waitcounters.

  #152444 opened Apr 29, 2025

• [TorchDynamo] Fix failure to realize LazyVariableTracker on stack

  #152446 opened Apr 29, 2025

• Add new profiling events to `DebugAutotuner`

  #152449 opened Apr 29, 2025

• [PT2] Port replace_lce_with_matmul / replace_first_lce_with_fused_matmul_lce to PT2 pre_grad passes

  #152450 opened Apr 29, 2025

• Implement async manifold cache write

  #152452 opened Apr 29, 2025

• Add epoch to fake tensor cache key

  #152453 opened Apr 29, 2025

• [export] add runtime assert messages to python torch checks (#150719)

  #152455 opened Apr 29, 2025

• Fix XLA issue.

  #152456 opened Apr 29, 2025

• fix: Update padding_mode to use Literal for type checking

  #152458 opened Apr 29, 2025

• [IR] Input Adapter refactor prototype

  #152459 opened Apr 29, 2025

• [pytorch][triton] flex attention fwd kernel with TMA loads (#151923)

  #152460 opened Apr 29, 2025

• consolidate guard_or_x and definitely_x

  #152463 opened Apr 29, 2025

• add device generalisation support for distributed tests

  #152471 opened Apr 29, 2025

• [nativert] Move TensorMeta to pytorch core

  #152475 opened Apr 29, 2025

• [DO NOT REVIEW] Attempt a mixed precision fused adam

  #152477 opened Apr 29, 2025

• [ONNX] Suggest users setting dynamo=True when exporting

  #152478 opened Apr 29, 2025

• Fix flaky test in test_custom_ops

  #152484 opened Apr 29, 2025

• Change unsafe_marked_cacheable_functions to a dictionary, so that you can specify a static cache key

  #152486 opened Apr 29, 2025

• [invoke_subgraph] Simplify output code for subgraph output node

  #152490 opened Apr 29, 2025

• fix tests broken after #152450

  #152493 opened Apr 29, 2025

• [inductor][invoke_subgraph] Free the buffers before the subgraph call

  #152494 opened Apr 29, 2025

• [export] Refactor pt2 save/load

  #152495 opened Apr 30, 2025

• Refactor nested benchmark functions in AlgorithmSelectorCache

  #152502 opened Apr 30, 2025

• [Hierarchical Compilation] Use universal flatten APIs

  #152505 opened Apr 30, 2025

• [Hierarchical Compile] Take into account mutation deps in cycle detection

  #152506 opened Apr 30, 2025

• [inductor] [compile async] Don't compile in eager

  #152507 opened Apr 30, 2025

• [2/N] Deprecate c10::string_view and at::string

  #152509 opened Apr 30, 2025

• [MPS] Migrate mul to TensorIterator

  #152515 opened Apr 30, 2025

• Make torch/csrc/utils.h to be device-agnostic

  #152521 opened Apr 30, 2025

• Remove the unnecessary cuda/Tensor.cpp

  #152522 opened Apr 30, 2025

• [compile async] [cache] testing

  #152523 opened Apr 30, 2025

• elastic: do not shutdown rendezvous on leaving workers

  #152525 opened Apr 30, 2025

• Use std::apply for CPU code

  #152526 opened Apr 30, 2025

• Add methods for checking Triton availability to the device interface

  #152529 opened Apr 30, 2025

• Do not check out nccl when not building it

  #152533 opened Apr 30, 2025

• [CI] Use cmake from pip instead of conda in CI docker images

  #152537 opened Apr 30, 2025

• Disable SLEEF implementation of vec::maximum in vec128_float_neon.h | Accelerate aten::hardtanh_ by 21x

  #152538 opened Apr 30, 2025

• [CUDA] Rest peak memory stats before running `test_set_per_process_memory_fraction`

  #152540 opened Apr 30, 2025

• Add parameters for monitor

  #152541 opened Apr 30, 2025

• strict multidimensional slicing

  #152543 opened Apr 30, 2025

• Migrate perf_test/test_[gc]pu_speed_mnist.sh from conda to venv

  #152544 opened Apr 30, 2025

• ci: Switch benchmark dependency to use pip

  #152545 opened Apr 30, 2025

• Remove Conda Instructions

  #152546 opened Apr 30, 2025

• Implemented `Size.__radd__`

  #152554 opened Apr 30, 2025

• [BE] Update numba versions

  #152557 opened Apr 30, 2025

• xpu: rely on sycl/sycl.hpp to include bfloat16.hpp

  #152562 opened Apr 30, 2025

• [c10d][fr] Make FR vendor neutral so that other backends can use it

  #152563 opened Apr 30, 2025

• [ROCm] Update spack includes

  #152569 opened Apr 30, 2025

• [Hierarchical Compile] Replace tracing alias and mutation check with dynamo impl

  #152570 opened Apr 30, 2025

• [Dynamo] Fix typing in graph_deduplication.py

  #152572 opened May 1, 2025

• Allow decomposeK to fuse

  #152573 opened May 1, 2025

• Added documentation for nonzero_static function (#152347)

  #152574 opened May 1, 2025

• [IR] Input Adapter refactor prototype (#152459)

  #152575 opened May 1, 2025

• [testing] 1

  #152578 opened May 1, 2025

• [aoti] skip codegen for sympy expr when codegening input

  #152579 opened May 1, 2025

• [cutlass backend] cache filtered ops based on layouts

  #152580 opened May 1, 2025

• [invoke_subgraph] rename identifiers to prevent python mangling

  #152581 opened May 1, 2025

• add support for 0 size shardedTensor and recalculate metadata from all_gather

  #152583 opened May 1, 2025

• [c10d][fr] Decouple the core logic of FR with the entry and event type

  #152585 opened May 1, 2025

• [2/N] Use std::filesystem

  #152586 opened May 1, 2025

• [Inductor] Introduce Wrapper IR line for symbolic call args

  #152587 opened May 1, 2025

• [WIP] verbose logging for recompilations

  #152588 opened May 1, 2025

• [Dynamo] Optimize dedupe region ancestor tracking

  #152589 opened May 1, 2025

• Fix #152280: add Literal[…] PaddingMode to Conv modules

  #152590 opened May 1, 2025

• Fix: promote scalar to MPS device in exec_binary_kernel

  #152591 opened May 1, 2025

• [c10d] Add support for ReduceOp::AVG in ProcessGroupMPI for FSDP2

  #152594 opened May 1, 2025

• [wip] base commit

  #152596 opened May 1, 2025

• add backend_specialization kwarg to mark_dynamic

  #152597 opened May 1, 2025

• [testing] 3

  #152599 opened May 1, 2025

• store backend specializations in StatelessSymbolicContext

  #152600 opened May 1, 2025

• use backend specializations in compile_and_call_fx_graph

  #152601 opened May 1, 2025

• [testing] 4

  #152602 opened May 1, 2025

• [BE] Delete `Module_CUDA_fix`

  #152603 opened May 1, 2025

• [Testing] Is FindCUDA.cmake from `Modules_CUDA_fix` called at all?

  #152604 opened May 1, 2025

• [Environment Variable] Use thread-safe getenv functions

  #152609 opened May 1, 2025

• Update padding_mode type annotation to use Literal type (PaddingMode)

  #152610 opened May 1, 2025

• Makefile: refactor build, setup and lint rules

  #152611 opened May 1, 2025

• Revert "Cleanup VS 2019 refs in pytorch (#145863)"

  #152613 opened May 1, 2025

• [WIP] Make FR vendor generic and try to enable it for gloo

  #152614 opened May 1, 2025

• [dynamo] Guard serialization for DUAL LEVEL.

  #152615 opened May 1, 2025

• [dynamo] Guard serialization for FUNCTORCH_STACK_MATCH

  #152616 opened May 1, 2025

• [CUDA][TF32] Account for TF32 in `test_conv2d_same_padding`

  #152618 opened May 1, 2025

• [DO NOT REVIEW] Implement __obj_flatten__ for LinearPackedParamsBase

  #152619 opened May 1, 2025

• Stop proxy-ing autograd.Function.ctx into the graph

  #152621 opened May 1, 2025

• Parameterized CUDA Graph Launch

  #152622 opened May 1, 2025

• [pytree] make `tree_*` functions accept both Python and C++ `PyTreeSpec`

  #152624 opened May 1, 2025

• [Dynamo] Guard serialization for TENSOR_SUBCLASS_METADATA_MATCH

  #152626 opened May 1, 2025

• Make PGO code state not sensitive to file path by hashing file content when the file is available.

  #152628 opened May 1, 2025

• [DCP] Add 30min timeout for IPC communications in async checkpointing

  #152629 opened May 1, 2025

• [ROCm] Initial AITER Integration for mha_bwd asm kernels

  #152630 opened May 1, 2025

• Fix two error messages involving Tensor.dense()

  #152631 opened May 1, 2025

• [ca] wrap flex attention tests with compiled autograd

  #152633 opened May 1, 2025

• Switch to metal kernel for mul

  #152636 opened May 1, 2025

• [export] Add draft-export docs

  #152637 opened May 1, 2025

• [dynamic shapes] use try-catch instead of guard_or_true for reshape_view_helper

  #152638 opened May 1, 2025

• [AOTAutogradCache][Easy] Move `"einops.einops.rearrange"` to `SAFE_NON_TORCH_FUNCTIONS`

  #152640 opened May 1, 2025

• [FlexAttention] explicilty create grad_q w/ strides

  #152641 opened May 1, 2025

• [CUTLASS][WIP] Gate rowwise matmul CUTLASS kernels by compute capability

  #152642 opened May 1, 2025

• [BE]remove vulkan test

  #152643 opened May 1, 2025

• [inductor] Realize bucketize/searchsorted output

  #152644 opened May 1, 2025

• [do-not-land][ca] default on for CI

  #152646 opened May 1, 2025

• [Flight Recorder] Added logging after FR dump completed

  #152648 opened May 2, 2025

• thread through specialization to compile_fx

  #152650 opened May 2, 2025

• Add assert_fp8_close helper for FP8 tensor comparisons

  #152651 opened May 2, 2025

• Refactor some common autotune-related utils into a new file

  #152652 opened May 2, 2025

• Remove incorrect assertion

  #152653 opened May 2, 2025

• Make assertion about pass callable print the bad pass

  #152654 opened May 2, 2025

• cleanup, refactor and add missing self._dde_suppressed checks

  #152657 opened May 2, 2025

• Fix the basic description of torch.min(), torch.max(), torch.all(), torch.any()

  #152658 opened May 2, 2025

• [Inductor] Fix kernel argument ordering when using dynamic shapes with workspace

  #152660 opened May 2, 2025

• Fix evaluate_expr to include suppress_guards_tls in cache key

  #152661 opened May 2, 2025

• Re-enable FakeTensor caching for SymInts

  #152662 opened May 2, 2025

• [MPS][BE] Do not dispatch empty kernels

  #152663 opened May 2, 2025

• Raise error when no record on extra_files

  #152664 opened May 2, 2025

• MXFP8 Fix broken bias support for mxfp8

  #152665 opened May 2, 2025

• [StaticCudaLauncher] Ensure cuda context exists before launching kernels

  #152667 opened May 2, 2025

• Added documentation for nonzero_static function (#152347)

  #152669 opened May 2, 2025

• add codegen layer specialization dispatch

  #152670 opened May 2, 2025

164 Issues closed by 41 people

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int64 (__main__.TestForeachCUDA)

  #150392 closed May 2, 2025

• DISABLED test_nvshmem

  #152649 closed May 2, 2025

• py_limited_api=True in PyTorch2.7 will break the build of extensions

  #152243 closed May 2, 2025

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int32 (__main__.TestForeachCUDA)

  #150350 closed May 2, 2025

• [ONNX] Improve and sort out fallback mechanism

  #151703 closed May 2, 2025

• Should make the doc of `nn.CrossEntropyLoss()` more clear

  #134853 closed May 1, 2025

• torch.compile should not recompiles when `.requires_grad=True` under `torch.no_grad()` context

  #131975 closed May 1, 2025

• compiled autograd + dynamic shapes fails with constraint violation

  #133575 closed May 1, 2025

• [torch.compile] Dynamic shape behavior is different between using torch.compile with and without compiled_autograd.enable

  #113129 closed May 1, 2025

• Export QAT model is not performing as expected when compared to the original model and FX Graph QAT

  #150746 closed May 1, 2025

• DISABLED test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_float32 (__main__.TestForeachCUDA)

  #149409 closed May 1, 2025

• `torch.export` fails on `InstanceNorm1d`

  #152467 closed May 1, 2025

• module.cuda() doesn't work under FakeTensorMode

  #148977 closed May 1, 2025

• [CI] [anaconda] CI Perf Tests

  #148342 closed May 1, 2025

• [Inductor] Dynamo hangs when processing an operator, seemingly depending on a logical argument value

  #151743 closed May 1, 2025

• [export] Warn users when 0/1 specialization happens

  #151582 closed May 1, 2025

• The test 'test_host_memory_stats' is failing in torch2.7.0+cu118

  #152422 closed May 1, 2025

• How does torch.cudagraph capture a hybrid graph?

  #152584 closed May 1, 2025

• Add switch to disable truncation to long list print

  #152427 closed May 1, 2025

• `torch.randint` can't handle large `high` argument (and in general high range of `torch.uint64`)

  #152564 closed Apr 30, 2025

• torch.randint should accept high=2**63

  #81446 closed Apr 30, 2025

• pytorch index_select is too slow

  #111247 closed Apr 30, 2025

• cuda graphs produce two additional kernel calls

  #143572 closed Apr 30, 2025

• [regression] Not getting `CUDA error: device-side assert triggered` on main for CUDA_KERNEL_ASSERT2

  #107396 closed Apr 30, 2025

• [CI] [anaconda] Benchmarks anaconda removal

  #152123 closed Apr 30, 2025

• More logs to show why fx graph cache isn't hit / created?

  #152065 closed Apr 30, 2025

• Mr

  #152549 closed Apr 30, 2025

• Add Description of `validate_args` in `torch.distributions.`

  #152165 closed Apr 30, 2025

• [ROCm] "No available kernel" when running EFFICIENT_ATTENTION sdpa

  #138864 closed Apr 30, 2025

• difficulty creating magma tarball when new rocm or cuda versions are deployed

  #151707 closed Apr 30, 2025

• [CUDA Graph tree] Cannot capture buffer allocation on side CUDA Streams

  #151199 closed Apr 30, 2025

• Unify nccl versions for x86 and aarch64 builds

  #149554 closed Apr 30, 2025

• Release torch with CUDA12.1 for 2.6 and even latest version

  #152524 closed Apr 30, 2025

• Failed to create Gloo new group after initialized with NCCL

  #68726 closed Apr 30, 2025

• Cudnn 9.2 is out!

  #119400 closed Apr 30, 2025

• [ONNX] scatter_reduce with max reduction not correctly converted to ONNX for 2d input

  #152419 closed Apr 30, 2025

• DISABLED test_fn_grad_grid_sampler_2d_cuda_float64 (__main__.TestBwdGradientsCUDA)

  #131079 closed Apr 30, 2025

• NotImplementedError: Operator aten.view.dtype does not have a sharding strategy registered.

  #152530 closed Apr 30, 2025

• DISABLED test_inductor_debug (__main__.LoggingTests)

  #152511 closed Apr 30, 2025

• DISABLED test_einsum_cpu (__main__.TestUnbackedSymintsCPU)

  #151380 closed Apr 30, 2025

• [export] export doesn't save custom meta for constant tensors

  #151476 closed Apr 30, 2025

• Regression: Multiple OpenMP runtimes linked to libtorch_cpu.so

  #146603 closed Apr 29, 2025

• Floating point exception (core dumped) in `native_channel_shuffle`

  #142453 closed Apr 29, 2025

• UNSTABLE pull / linux-jammy-py3-clang12-executorch / test (executorch)

  #144480 closed Apr 29, 2025

• DISABLED test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_complex64 (__main__.TestForeachCUDA)

  #149199 closed Apr 29, 2025

• [CI] Remove conda usage from lint related jobs

  #148110 closed Apr 29, 2025

• `make pdflatex` Sphinx error: Builder name pdflatex not registered or available through entry point

  #147027 closed Apr 29, 2025

• [CI] [anaconda] Utility scripts and workflows

  #152124 closed Apr 29, 2025

• cd: There's no way to test changes to container images for binary builds

  #149679 closed Apr 29, 2025

• [RFC] PyTorch next wheel build platform: manylinux-2.28

  #123649 closed Apr 29, 2025

• [Performance] Simple arithemtic operations are slower using MPS than Metal

  #143874 closed Apr 29, 2025

• DISABLED test_parity__foreach_add_fastpath_inplace_cuda_complex64 (__main__.TestForeachCUDA)

  #151313 closed Apr 29, 2025

• [BUG] when invoking torch::manul_seed() program crashed in libtorch 2.2.1

  #121658 closed Apr 29, 2025

• Loading weights using `torch.distributed.checkpoint` leads to large loss values

  #145378 closed Apr 29, 2025

• FSDP OOM during initialization

  #152263 closed Apr 29, 2025

• [Inductor] Different results with Conv2d and BN2d not in `eval mode`

  #141317 closed Apr 29, 2025

• DISABLED test_parity__foreach_add_fastpath_inplace_cuda_bool (__main__.TestForeachCUDA)

  #151268 closed Apr 29, 2025

• DISABLED test_guard_failure_fn2 (__main__.MiscTests)

  #148217 closed Apr 29, 2025

• DISABLED test_inlined_functions_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148544 closed Apr 29, 2025

• DISABLED test_nested_tuple_output_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148606 closed Apr 29, 2025

• DISABLED test_make_closure_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148889 closed Apr 29, 2025

• DISABLED test_capture_tracked_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148427 closed Apr 29, 2025

• DISABLED test_return_captured_var_used_multiple_times_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148624 closed Apr 29, 2025

• DISABLED test_export_defaults_ok_dynamic_shapes (__main__.DynamicShapesExportTests)

  #148331 closed Apr 29, 2025

• DISABLED test_internal_nonlocal_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148558 closed Apr 29, 2025

• DISABLED test_int_shape_binops (__main__.MiscTests)

  #148296 closed Apr 29, 2025

• DISABLED test_user_defined_binop (__main__.MiscTests)

  #148443 closed Apr 29, 2025

• DISABLED test_empty_graph_nested_calls_fullgraph_True_dynamic_shapes (__main__.DynamicShapesReproTests)

  #148311 closed Apr 29, 2025

• DISABLED test_wrap_kwarg_default_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148951 closed Apr 29, 2025

• DISABLED test_capture_untracked_global_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148442 closed Apr 29, 2025

• DISABLED test_freevars_as_inputs_to_wrap_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148514 closed Apr 29, 2025

• DISABLED test_donated_buffer1_dynamic_shapes (__main__.DynamicShapesAotAutogradFallbackTests)

  #149101 closed Apr 29, 2025

• DISABLED test_sys_modules_dynamic_shapes (__main__.DynamicShapesMiscTests)

  #148330 closed Apr 29, 2025

• DISABLED test_int_shape_inplace_binops (__main__.MiscTests)

  #148312 closed Apr 29, 2025

• DISABLED test_guard_failure_fn_shape_control_dynamic_shapes (__main__.DynamicShapesMiscTests)

  #148216 closed Apr 29, 2025

• DISABLED test_lift_tensors_with_shared_symbols_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148870 closed Apr 29, 2025

• DISABLED test_wrap_kwarg_only_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #149024 closed Apr 29, 2025

• DISABLED test_symint_in_slice_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148643 closed Apr 29, 2025

• DISABLED test_capture_tracked_nested_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148851 closed Apr 29, 2025

• DISABLED test_wrap_kwarg_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #149000 closed Apr 29, 2025

• DISABLED test_param_shape_binops (__main__.MiscTests)

  #148369 closed Apr 29, 2025

• DISABLED test_wrap_kwarg_default_if_branch_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148961 closed Apr 29, 2025

• DISABLED test_dont_aggressively_write_assert_dynamic_shapes (__main__.DynamicShapesReproTests)

  #148295 closed Apr 29, 2025

• DISABLED test_side_effect_local_list_append_no_graph_break_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148935 closed Apr 29, 2025

• DISABLED test_export_with_cond_dynamic_shape_pred_dynamic_shapes (__main__.DynamicShapesExportTests)

  #148368 closed Apr 29, 2025

• DISABLED test_empty_graph_nested_calls_fullgraph_False_dynamic_shapes (__main__.DynamicShapesReproTests)

  #148426 closed Apr 29, 2025

• DISABLED test_shape_int_inplace_binops (__main__.MiscTests)

  #148392 closed Apr 29, 2025

• DISABLED test_capture_untracked_nonlocal_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148464 closed Apr 29, 2025

• DISABLED test_sys_modules (__main__.MiscTests)

  #148428 closed Apr 29, 2025

• DISABLED test_wrap_pytree_kwargs_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #149079 closed Apr 29, 2025

• DISABLED test_dynamic_sources_dynamic_override (__main__.MiscTests)

  #148218 closed Apr 29, 2025

• DISABLED test_parity__foreach_add_fastpath_inplace_cuda_complex128 (__main__.TestForeachCUDA)

  #151300 closed Apr 29, 2025

• DISABLED test_mark_unbacked_strict_dynamic_shapes (__main__.DynamicShapesMiscTests)

  #148215 closed Apr 29, 2025

• DISABLED test_nested_wrap_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148914 closed Apr 29, 2025

• DISABLED test_mark_unbacked_strict (__main__.MiscTests)

  #148332 closed Apr 29, 2025

• DISABLED test_wrap_all_kwarg_dynamic_shapes (__main__.DynamicShapesHigherOrderOpTests)

  #148665 closed Apr 29, 2025

• float' object is not callable when using scheduler.step() with MultiplicativeLR

  #81554 closed Apr 29, 2025

• MPS SDPA `float32` memory leak

  #152344 closed Apr 29, 2025

• [Break XPU] chunk_cat accuracy failed on XPU Inductor UT.

  #152296 closed Apr 29, 2025

• [XPU] The updated torch-xpu-ops caused interpolate_bilinear accuracy error.

  #152020 closed Apr 29, 2025

• Compilation Issues with sm_129 (RTX 5070 Ti) on WSL - Seeking Advice

  #152400 closed Apr 29, 2025

• Proposal: Beautify torch.distributed.tensor.debug.visualize_sharding

  #151857 closed Apr 29, 2025

• [CI] [anaconda] Utilities

  #152126 closed Apr 29, 2025

• torch.bucketize works incorrectly on uint input with negative boundaries after torch.compile-gpu

  #145929 closed Apr 28, 2025

• Potential bug in torch.optim.lr_scheduler.CosineAnnealingWarmRestarts

  #88791 closed Apr 28, 2025

• [inductor] Precompilation start time is the time when a config is added to the queue, not when executor starts compiling the config

  #148777 closed Apr 28, 2025

• [inductor] [dtype] `ReplicationPad` raise dtype error on eager but pass the check on indcutor

  #143779 closed Apr 28, 2025

• Error building pytorch from source

  #138315 closed Apr 28, 2025

• Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable upper limit

  #152351 closed Apr 28, 2025

• torch.arange bf16 results are not accurate

  #137774 closed Apr 28, 2025

• DISABLED test_parity__foreach_add_fastpath_inplace_cuda_bfloat16 (__main__.TestForeachCUDA)

  #151228 closed Apr 28, 2025

• DISABLED test_foreach_reduce_large_input__foreach_max_w_empty_False_cuda_bool (__main__.TestForeachCUDA)

  #151229 closed Apr 28, 2025

• [discussion] Consolidation of audio-visual I/O in a new package

  #81102 closed Apr 28, 2025

• [PTD][RPC] Verify RPC Tutorials contents and scripts

  #138832 closed Apr 28, 2025

• Error in DTensor uneven shard view op

  #143372 closed Apr 28, 2025

• Incorrect Gradient Computation in `torch.log1p`

  #152088 closed Apr 28, 2025

• AWS A100 runners reliability issue

  #140332 closed Apr 28, 2025

• [CI] [anaconda] CI Build and Test scripts MacOS

  #152113 closed Apr 28, 2025

• peak memory is lower for subsequent fresh runs compared to the first run of a torch.compiled model

  #151995 closed Apr 28, 2025

• DISABLED test_dynamic_sources_dynamic_override_dynamic_shapes (__main__.DynamicShapesMiscTests)

  #148214 closed Apr 28, 2025

• DISABLED test_parity__foreach_acos_fastpath_outplace_cuda_float64 (__main__.TestForeachCUDA)

  #151214 closed Apr 28, 2025

• DISABLED test_setting_meta_device_model_broadcasting_and_memory (__main__.TestStateDict)

  #143994 closed Apr 28, 2025

• torch._dynamo.exc.BackendCompilerFailed: backend='inductor' raised: ModuleNotFoundError: No module named 'expecttest'

  #152225 closed Apr 28, 2025

• Pytorch aten::col2im not currently supported on the MPS backend

  #151820 closed Apr 28, 2025

• prod_cpu not implemented for 'BFloat16'

  #89372 closed Apr 28, 2025

• Windows CUDA Build Failure: Ambiguous std in cuda_vectorized_test.cu (CUDA 12.6/MSVC 2019)

  #152291 closed Apr 28, 2025

• `torch._inductor.exc.InductorError: CppCompileError: C++ compile error` after Torch 2.7 Release

  #152172 closed Apr 28, 2025

• Aborted (core dumped) in torch.flipud

  #152253 closed Apr 27, 2025

• Aborted (core dumped) in torch.fliplr

  #152085 closed Apr 27, 2025

• Less Check on the triangular tensor of `L` in `torch.cholesky_solve()`

  #152164 closed Apr 27, 2025

• [inductor] `.to_sparse()-.to_dense()` throws `LoweringException: NotImplementedError:`

  #151522 closed Apr 26, 2025

• "TypeError: unhashable type: non-nested SymInt" with `torch.compile`

  #135099 closed Apr 26, 2025

• Pytorch 2.7.0 with XPU (silently) crashing

  #152255 closed Apr 26, 2025

• [Inductor] weird reordering behavior with `wait_tensor`

  #152252 closed Apr 26, 2025

• memoryview support for `torch._C.import_ir_module_from_buffer`

  #107099 closed Apr 26, 2025

• Compute Capability Misrecognition on NVIDIA Force RTX 50Ge70 Ti (Blackwell Architecture)

  #152223 closed Apr 25, 2025

• [MPS/Inductor] polygamma is miscompiled for some inputs

  #152205 closed Apr 25, 2025

• Lint rule for always using std::optional?

  #150313 closed Apr 25, 2025

• DISABLED test_parity__foreach_acos_fastpath_outplace_cuda_float32 (__main__.TestForeachCUDA)

  #151136 closed Apr 25, 2025

• DISABLED test_foreach_check_stride_ignore_dims_of_one_cuda_float32 (__main__.TestForeachCUDA)

  #150026 closed Apr 25, 2025

• DISABLED test_parity__foreach_acos_fastpath_outplace_cuda_float16 (__main__.TestForeachCUDA)

  #151114 closed Apr 25, 2025

• [AOTI] aoti_compile_and_package + use_runtime_constant_folding gives "Error: CUDA driver error: file not found"

  #152067 closed Apr 25, 2025

• [`Torch 2.7.0 x Py 3.9`] Incompatible dep versions with networkx

  #152191 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_ScheduleClass3 (__main__.ComposabilityTest)

  #151089 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_MP_ScheduleClass0 (__main__.ComposabilityTest)

  #151083 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_MP_ScheduleClass1 (__main__.ComposabilityTest)

  #151084 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_MP_ScheduleClass3 (__main__.ComposabilityTest)

  #151090 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_MP_ScheduleClass2 (__main__.ComposabilityTest)

  #151085 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_ScheduleClass1 (__main__.ComposabilityTest)

  #151087 closed Apr 25, 2025

• DISABLED test_pp_ddp_ScheduleClass2 (__main__.ComposabilityTest)

  #151082 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_ScheduleClass2 (__main__.ComposabilityTest)

  #151088 closed Apr 25, 2025

• DISABLED test_pp_ddp_ScheduleClass1 (__main__.ComposabilityTest)

  #151081 closed Apr 25, 2025

• DISABLED test_pp_ddp_ScheduleClass0 (__main__.ComposabilityTest)

  #151078 closed Apr 25, 2025

• DISABLED test_pp_fsdp_dp_type_FSDP_ScheduleClass0 (__main__.ComposabilityTest)

  #151086 closed Apr 25, 2025

• UserWarning: There is a performance drop because we have not yet implemented the batching rule for aten::scatter_reduce.two

  #152141 closed Apr 25, 2025

• DISABLED test_parity__foreach_acos_fastpath_outplace_cuda_complex64 (__main__.TestForeachCUDA)

  #151099 closed Apr 25, 2025

• ENH: Publish full-fledged tarballs also for release candidates

  #150649 closed Apr 25, 2025

• [ONNX] Exporting with `dynamo=True` and `Dim.DYNAMIC` in `dynamic_shapes` passes for `scaled_dot_product_attention`, but doesn't do anything

  #152018 closed Apr 25, 2025

• Fix the Inconsistency and Description of `device_type` in `torch.random.fork_rng()`

  #151784 closed Apr 25, 2025

• `out` should exist as an instance variable out of the func itself

  #146676 closed Apr 25, 2025

• Size of `tau` can mismatch with the context in `torch.ormqr()`

  #150674 closed Apr 25, 2025

• Whether `x` and `dx` can be used together in `torch.trapezoid()`?

  #151105 closed Apr 25, 2025

• [ONNX] Dynamic shapes: support `torch.sym_not`

  #136572 closed Apr 25, 2025

• What is the difference between normal_tensor.storage().use_count() and viewed_tensor's?

  #152100 closed Apr 25, 2025

121 Issues opened by 76 people

• Torch BF16 group gemm hangs in backward pass - core issue isolated, needs proper resolution.

  #152668 opened May 2, 2025

• DISABLED test_comprehensive_nansum_cuda_int32 (__main__.TestInductorOpInfoCUDA)

  #152666 opened May 2, 2025

• Add explicit error message for def infer_size(a, b): that specificy that non broadcast path was picked due to unbacked existing in both inputs.

  #152656 opened May 2, 2025

• UNSTABLE docker-cache-mi300 / docker-cache

  #152655 opened May 2, 2025

• Check for if two tensors are overall similar instead of bitwise similar?

  #152647 opened May 2, 2025

• ProcessGroupGloo.allgather_into_tensor_coalesced crashes with CUDA tensors

  #152645 opened May 1, 2025

• static cuda launcher causes `RuntimeError: CUDA driver error: invalid device context` in torchtitan CI

  #152639 opened May 1, 2025

• TestFlexAttentionCUDA.test_GQA_score_mod7_cuda_float16 fails on h100

  #152635 opened May 1, 2025

• Incorrect strides for `nonzero_static` compilation

  #152634 opened May 1, 2025

• DISABLED test_torchvision_models_efficientnet_v2_l (__main__.TestVisionTracing)

  #152632 opened May 1, 2025

• [v2.7.1] Release Tracker

  #152627 opened May 1, 2025

• modded-nanogpt flaky NCCL hang starting 3/30 nightly

  #152623 opened May 1, 2025

• Pytorch Profiler crashes while using it with Pytorch Lightning module

  #152617 opened May 1, 2025

• Enable AOTI for Metal inductor

  #152612 opened May 1, 2025

• [triton pin update] Run Inductor CI on pin updates for Triton and the PyTorch nightly branch

  #152608 opened May 1, 2025

• Loops impacting output when utilizing hooks

  #152607 opened May 1, 2025

• AOTI regression on SAM and tts-angular

  #152606 opened May 1, 2025

• ROCm, 7900 XTX: Pytorch SDPA is 2.5x slower than manual implementation with non-continuous v

  #152595 opened May 1, 2025

• Flex Attention doesn't scale with custom bias

  #152593 opened May 1, 2025

• [ratter-build] Cannot detect CUDA when build from source

  #152592 opened May 1, 2025

• [MPS] Binary kernels produce incorrect results when one of the tensor arguments is from a wrapped scalar

  #152582 opened May 1, 2025

• [Benchmark] High compilation time variance on benchmark dashboards

  #152566 opened Apr 30, 2025

• DISABLED test_graph_partition_reorder_cpu_and_gpu_interleave (__main__.CudaGraphTreeTests)

  #152561 opened Apr 30, 2025

• DISABLED test_pending_fusion_pro_and_epi (__main__.TestPrologueFusion)

  #152560 opened Apr 30, 2025

• DISABLED test_comprehensive_signal_windows_hamming_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #152559 opened Apr 30, 2025

• DISABLED test_comprehensive_amin_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #152558 opened Apr 30, 2025

• PGO does not work on jobs for frameworks that copy code to different dirs at different attempts.

  #152555 opened Apr 30, 2025

• MPS varying seq len SDPA memory leak

  #152550 opened Apr 30, 2025

• FakeTensorUpdater does not trace nodes correctly

  #152548 opened Apr 30, 2025

• optree package status in PyTorch

  #152535 opened Apr 30, 2025

• AsyncCollectiveTensor doesn't trigger wait upon dtype cast

  #152534 opened Apr 30, 2025

• The 2.7.0 release tarball is missing `.ci/docker/ci_commit_pins/nccl-cu12.txt` required for building

  #152532 opened Apr 30, 2025

• [inductor][triton] Inductor is not compatible with the latest upstream Triton

  #152531 opened Apr 30, 2025

• flex attention does not leverage masking, memory error

  #152528 opened Apr 30, 2025

• can't reconstruct the communication group using PyTorch.

  #152527 opened Apr 30, 2025

• DISABLED test_comprehensive_lu_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #152520 opened Apr 30, 2025

• DISABLED test_comprehensive_repeat_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #152500 opened Apr 30, 2025

• UNSTABLE Lint / Lint URLs / linux-job

  #152489 opened Apr 29, 2025

• [dynamo] Try tracing into einops

  #152480 opened Apr 29, 2025

• [dynamo] Dynamo fails to run torch.cat() with FakeTensors because it can't confirm 's0 + s1*u0' is nonzero

  #152473 opened Apr 29, 2025

• DISABLED test_comprehensive_polygamma_polygamma_n_0_cuda_float16 (__main__.TestInductorOpInfoCUDA)

  #152469 opened Apr 29, 2025

• DISABLED test_comprehensive_polygamma_polygamma_n_1_cuda_float16 (__main__.TestInductorOpInfoCUDA)

  #152470 opened Apr 29, 2025

• torch.export with dynamic shapes on Static Cache HF LLama model fails

  #152465 opened Apr 29, 2025

• [dynamo] `torch.compile` prevents fsdp warning from getting generated

  #152451 opened Apr 29, 2025

• [dynamo] guard code generation triggers attribute error on DeviceMesh object

  #152447 opened Apr 29, 2025

• `torch.compile` causes assertion error in distributed checkpoint wrapper test

  #152442 opened Apr 29, 2025

• Inductor pattern matching on mutable ops

  #152441 opened Apr 29, 2025

• Newly added lint-urls jobs are very flaky

  #152439 opened Apr 29, 2025

• `nn.CrossEntropyLoss` accepts negative target probabilities

  #152437 opened Apr 29, 2025

• Fx Graph cache hit generates guards that does not exists in the original cached program causing recompilations only at cache hit.

  #152435 opened Apr 29, 2025

• [pt2] [AOTAutogradCache] Allow users to specify non torch functions as cacheable

  #152434 opened Apr 29, 2025

• [Manylinux 2.28] Migrate Docker container to use gcc 14

  #152426 opened Apr 29, 2025

• Silent incorrectness between static torch.compile vs eager

  #152425 opened Apr 29, 2025

• Invalid handling of nans in compiled torch.quantile / torch.nanquantile on cuda

  #152423 opened Apr 29, 2025

• torch.nn.functional.ctc_loss raises cuDNN error in PyTorch versions >=2.5.0

  #152421 opened Apr 29, 2025

• DISABLED test_comprehensive_index_select_cuda_int32 (__main__.TestInductorOpInfoCUDA)

  #152416 opened Apr 29, 2025

• DISABLED test_input_moved_to_cuda_device_script (__main__.TensorPipeCudaRemoteModuleTest)

  #152415 opened Apr 29, 2025

• [DTensor] Calling .item() on DTensor with Partial placement results in local value

  #152406 opened Apr 29, 2025

• [CPU][UT] 16 UT of test/inductor/test_cpu_select_algorithm.py failed with PyTorch 2025-04-028 nightly wheel

  #152398 opened Apr 29, 2025

• Illegal Instruction Caused by `grid_sample` Under Windows

  #152385 opened Apr 28, 2025

• Outdated contents in dynamo overview

  #152381 opened Apr 28, 2025

• TORCH_COMPILE_DEBUG=1 does not consistently generate debug logs

  #152374 opened Apr 28, 2025

• DISABLED test_remote_cache_load_function_device_cuda_float32_dynamic_False_bundle_triton_True_use_static_cuda_launcher_False (__main__.TestFxGraphCache)

  #152370 opened Apr 28, 2025

• DISABLED test_reduce_stress_cuda (__main__.ProcessGroupGlooTest)

  #152367 opened Apr 28, 2025

• [AOTI] Package lowered with package_constants_in_so=False still uses lots of memory when loaded

  #152356 opened Apr 28, 2025

• Pin setuptools runtime dependency

  #152355 opened Apr 28, 2025

• DISABLED test_e2e_compile_True_model_type1 (__main__.TestE2ESaveAndLoad)

  #152349 opened Apr 28, 2025

• torch.nonzero_static is not documented on the website

  #152347 opened Apr 28, 2025

• compile generates inefficient code for mutations on small slices of inputs

  #152346 opened Apr 28, 2025

• cudagraphs: `static_input_indices` incorrectly including SymInt graph args when using tensor subclasses + dynamic shapes

  #152343 opened Apr 28, 2025

• Unusually slow draft_export time

  #152337 opened Apr 28, 2025

• pin_memory crashes for big tensors and leaks page locked memory

  #152335 opened Apr 28, 2025

• compile generates inefficient code when mutating small slice of a graph input

  #152323 opened Apr 28, 2025

• DISABLED test_comprehensive_pca_lowrank_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #152318 opened Apr 28, 2025

• [dynamo] torch._dynamo crashes on `self.value.__module__` inside SkipFunctionVariable.call_function() (PyTorch 2.7, works 2.6)

  #152316 opened Apr 28, 2025

• [DCP] failure case of save method

  #152310 opened Apr 28, 2025

• Softmax Decomp Causes Incorrect Gradients when Using `torch.compile` with `F.multi_head_attention_forward`

  #152309 opened Apr 28, 2025

• bizarre behavior with torch module's Attribute Error

  #152308 opened Apr 28, 2025

• Recompile issue after fp8 conversion

  #152307 opened Apr 28, 2025

• NCCL out of memory error after updating to PyTorch 2.7

  #152302 opened Apr 28, 2025

• Unexpected result from `torch.xpu.is_bf16_supported()` when XPU is unavailable

  #152301 opened Apr 28, 2025

• Unexpected behavior when using dist.all_reduce(x, op=dist.ReduceOp.SUM)

  #152300 opened Apr 28, 2025

• `torch.compile()` produces incorrect results for `asinh_()` operation on large/small values

  #152299 opened Apr 28, 2025

• Flex attention: batch-index-dependent block mask causes error with changing batch size

  #152297 opened Apr 28, 2025

• `vmap` not working on `torch.arange`, `torch.scalar_tensor`, and `torch.ones`

  #152295 opened Apr 28, 2025

• Unexpected overflow behavior when using `torch.addcmul`

  #152294 opened Apr 28, 2025

• `torch.sparse.log_softmax` output mismatch between CPU and CUDA

  #152293 opened Apr 28, 2025

• `torch==2.6` broke `nn.Module.dtype` typing

  #152292 opened Apr 28, 2025

• [Intel GPU][PT2.8]scaled_dot_product_attention returns wrong output

  #152290 opened Apr 28, 2025

• Error after successful build: No module named 'torch._C._distributed_c10d'

  #152285 opened Apr 27, 2025

• Forward compatibility in torch.export

  #152283 opened Apr 27, 2025

• Update `torch/nn/modules/conv.py` to use Literal for support padding modes

  #152280 opened Apr 27, 2025

• Make scaler.step() return if step was skipped or not

  #152279 opened Apr 27, 2025

• MPS: Conv1d fails with NotImplementedError for output_channels > 65536

  #152278 opened Apr 27, 2025

• `setup.py develop` command is disappearing soon from `setuptools`

  #152276 opened Apr 27, 2025

• [cudagraphs][HF][torch 2.7] Excessive cudagraph re-recording for HF LLM models

  #152275 opened Apr 27, 2025

• Question about that support of torch.compile for a custom CUDA operator?

  #152270 opened Apr 27, 2025

• Arbitrary Code Execution Risk in `torch.distributed.utils.overload` When Misused in Type Annotations

  #152269 opened Apr 27, 2025

• `iter()` and `reversed()` do not raise `StopIteration` when exhausted in torch.compile

  #152262 opened Apr 26, 2025

• Context Parallel -- unsharded output doesn't match output without CP.

  #152261 opened Apr 26, 2025

• [FR] Support BSHM-layout scaled_dot_product_attention without transpose.

  #152257 opened Apr 26, 2025

• Windows inductor genarated code without function declaration, and compile failed on MSVC.

  #152251 opened Apr 26, 2025

• [DTensor] [distributed]: Operator aten.masked_fill_.Scalar does not have a sharding strategy registered

  #152249 opened Apr 26, 2025

• NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'SparseCUDA' backend.

  #152226 opened Apr 25, 2025

• DISABLED test_remote_cache_load_function_device_cuda_bfloat16_dynamic_False_bundle_triton_False_use_static_cuda_launcher_False (__main__.TestFxGraphCache)

  #152222 opened Apr 25, 2025

• DISABLED test_pending_fusions_multiple (__main__.TestPrologueFusion)

  #152221 opened Apr 25, 2025

• [C10D] Allow NCCL single P2P ops to use parent/collective communicator

  #152220 opened Apr 25, 2025

• Have compiled autograd config API support nested compilation

  #152219 opened Apr 25, 2025

• Outdated install commands

  #152213 opened Apr 25, 2025

• Have cherry-pick bot always add the current release to the PR

  #152212 opened Apr 25, 2025

• DISABLED test_reduce_stress_cuda (__main__.ProcessGroupGlooLazyInitTest)

  #152201 opened Apr 25, 2025

• HUD Dashboard sort by perf speedup doesn't do anything

  #152190 opened Apr 25, 2025

• The input for layers other than the first layer should be the hidden state from the previous layer.

  #152188 opened Apr 25, 2025

• GroupNorm compilation errors on UNet-based architecture on torch >= 2.6.0

  #152185 opened Apr 25, 2025

• write a custom ViewAndMutationmeta.__repr__

  #152183 opened Apr 25, 2025

• GH200/GB200 NCCL Build Pytorch

  #152182 opened Apr 25, 2025

• Raise an Error when File Not Found in `torch.jit.load()`

  #152178 opened Apr 25, 2025

• Add description of several params in the basic usage of `torch.min()`, `torch.max()`, `torch.all()` and `torch.any()`

  #152176 opened Apr 25, 2025

• /usr/local/lib/python3.11/dist-packages/torch/autograd/graph.py:825: UserWarning: grid_sampler_2d_backward_cuda does not have a deterministic implementation, but you set 'torch.use_deterministic_algorithms(True, warn_only=True)'.

  #152171 opened Apr 25, 2025

• DISABLED test_e2e_compile_True_model_type2 (__main__.TestE2ESaveAndLoad)

  #152169 opened Apr 25, 2025

• DISABLED test_e2e_compile_True_model_type0 (__main__.TestE2ESaveAndLoad)

  #152168 opened Apr 25, 2025

510 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Inductor] FX backend via Wrapper IR

  #146942 commented on May 2, 2025 • 18 new comments

• [Cutlass] Integrate EVT into CUDACPPScheduling

  #150906 commented on May 2, 2025 • 17 new comments

• Random Batch Sampler Speedup

  #147706 commented on May 1, 2025 • 16 new comments

• [cp] dispatch flex_attention to CP impl in TorchDispatchMode

  #151497 commented on Apr 29, 2025 • 14 new comments

• [aotd] Support saved tensors hooks in aot_autograd

  #150032 commented on Apr 30, 2025 • 13 new comments

• [Inductor] Add decomposeK as an autotuning choice for mm

  #150654 commented on May 2, 2025 • 13 new comments

• [1/n][Optimus][Auto-AC] Support activation quantization without scaling

  #148380 commented on May 1, 2025 • 13 new comments

• Implement util function compute_global_tensor_shape for 1D device mesh

  #151990 commented on May 1, 2025 • 10 new comments

• [reland][ROCm] remove caffe2 from hipify

  #151845 commented on Apr 28, 2025 • 9 new comments

• [dynamo] replace `unimplemented` with `unimplemented_v2` in `variables/functions.py`

  #151277 commented on May 1, 2025 • 9 new comments

• [device_mesh] improve device selection logic

  #150897 commented on May 1, 2025 • 9 new comments

• [Inductor UT] Generalize device-bias code in `test_flex_attention.py`

  #151937 commented on May 2, 2025 • 8 new comments

• [ROCm][CI] Enabled fp8 distributed tests in test_micro_pipeline_tp.py for MI300

  #151977 commented on May 2, 2025 • 8 new comments

• Add infra to run CPython tests under Dynamo

  #150787 commented on May 1, 2025 • 7 new comments

• Cache code generation during triton template expansion and enable it for mm_template.

  #151773 commented on May 2, 2025 • 7 new comments

• Fix take_along_dim negative index handling (#146211)

  #152161 commented on Apr 29, 2025 • 6 new comments

• [pytree] Add public pytree module `torch.utils.pytree`

  #137400 commented on May 1, 2025 • 6 new comments

• Unify how we create random inputs for auto-tuning

  #152147 commented on Apr 30, 2025 • 6 new comments

• Refactoring FSDP2 (_composable/fsdp) test cases to be device agnostic

  #149848 commented on Apr 30, 2025 • 6 new comments

• [CUDA] Replace deprecated usages of cub iterators and thread operators

  #147493 commented on Apr 29, 2025 • 6 new comments

• dynamically set tags

  #152089 commented on Apr 29, 2025 • 6 new comments

• [SymmMem] Add all-to-all

  #151498 commented on May 2, 2025 • 5 new comments

• [associative_scan] Refactoring of input checking and dynamo invocation

  #148657 commented on May 1, 2025 • 5 new comments

• NUMA Binding Integration with torchrun

  #149334 commented on Apr 29, 2025 • 5 new comments

• [dynamo] replace `unimplemented` with `unimplemented_v2` in `variables/torch_functions.py`

  #151278 commented on May 2, 2025 • 5 new comments

• Avoid differing results in `linalg.(tensor_)solve`

  #151896 commented on Apr 30, 2025 • 4 new comments

• Do not cover up `__dunder`__ method type-hints from `.pyi` file

  #150875 commented on May 1, 2025 • 4 new comments

• Optimize LRScheduler docs

  #146684 commented on Apr 28, 2025 • 4 new comments

• Add `load_state_dict` hint doc about invoke order work with lr_scheduler

  #149942 commented on Apr 29, 2025 • 4 new comments

• autograd: Add VJP and JVP rules for aten::aminmax

  #151186 commented on May 2, 2025 • 4 new comments

• removed zero dim cpu logic from fake_tensor.py

  #147501 commented on May 1, 2025 • 3 new comments

• [Intel GPU] Support f32 intermediate dtype, headdim size <=576 and f32 causal mask for SDPA

  #152091 commented on Apr 28, 2025 • 3 new comments

• Move prologue_supported_inputs computations to def_kernal

  #150869 commented on May 2, 2025 • 3 new comments

• Add is_pinned to host allocator

  #151439 commented on Apr 29, 2025 • 3 new comments

• flex attention: fix dispatch order for tensor subclasses, avoid hardcoding call to faketensor impl in dynamo

  #151719 commented on Apr 30, 2025 • 3 new comments

• use vectorized loads and stores for all datatypes in torch.cat

  #151818 commented on Apr 28, 2025 • 3 new comments

• Enable type promotions in slice_scatter (pytorch#147842)

  #151911 commented on Apr 29, 2025 • 3 new comments

• Make `Adam`, `AdamW` work with nonzero-dim Tensor betas

  #149939 commented on May 1, 2025 • 3 new comments

• [WIP]: track remaining runtime time asserts for backward coddgen instead of trying to regenerate all

  #151919 commented on Apr 28, 2025 • 3 new comments

• auto functionalize base_hop

  #151067 commented on May 2, 2025 • 2 new comments

• Remove conda usage in condaenv.bat

  #151035 commented on May 2, 2025 • 2 new comments

• update get_default_device to also respect torch.device ctx manager

  #148621 commented on Apr 30, 2025 • 2 new comments

• [Inductor] Adjust boundary checking of dimensions using YBLOCK

  #149504 commented on Apr 28, 2025 • 2 new comments

• Parallelize sort using libstdc++ parallel mode

  #150195 commented on May 1, 2025 • 2 new comments

• API change for new enum in cusparseltsplitkmode-t for cusparseLT 0.7.0+

  #150536 commented on Apr 30, 2025 • 2 new comments

• [dynamic shapes] guard_or_false for infer_size

  #152146 commented on May 2, 2025 • 2 new comments

• Exempt overriding methods from docstring_linter (fix #151692)

  #151906 commented on Apr 29, 2025 • 2 new comments

• [OpenReg] Add _lazy_init and rng_state support for OpenReg

  #151914 commented on Apr 30, 2025 • 2 new comments

• Add AC_TRACER Infra TorchDispatchMode key

  #152158 commented on Apr 26, 2025 • 1 new comment

• [Graph Partition] Pass all cudagraph tree tests

  #152048 commented on Apr 30, 2025 • 1 new comment

• Replace `fw_metadata` info with trace log hint in hint message

  #147365 commented on Apr 25, 2025 • 1 new comment

• Improve cache key graph printing performance

  #151928 commented on Apr 29, 2025 • 1 new comment

• [Intel GPU][Inductor] Fallback embedding_dense_backward on XPU

  #151637 commented on Apr 29, 2025 • 1 new comment

• [ROCm] Maxpool forward NHWC Perf Improvement targeting Resnet scenarios

  #151727 commented on Apr 30, 2025 • 1 new comment

• Fix #150472 torch.library.custom_op doesn't handle single element tuples returns

  #151408 commented on Apr 29, 2025 • 1 new comment

• [AOTI][reland] Remove typedef for half and bfloat16

  #151109 commented on Apr 30, 2025 • 1 new comment

• [training] Adding NUMA support for pytorch

  #150597 commented on Apr 29, 2025 • 1 new comment

• [ROCm] Add support for SymmetricMemory

  #150580 commented on May 1, 2025 • 1 new comment

• [Inductor] Restrict block analysis to only match integer dims and strides

  #149615 commented on Apr 28, 2025 • 1 new comment

• Inductor logging + analysis of torch.profile

  #149697 commented on May 2, 2025 • 1 new comment

• [dynamic shapes] guard_or_false for computeStorageNbytes

  #150483 commented on Apr 25, 2025 • 1 new comment

• Add MPS support for getHostAllocator API

  #151913 commented on Apr 29, 2025 • 1 new comment

• [map] always turn on dynamo for map

  #152041 commented on Apr 29, 2025 • 1 new comment

• Enable AArch64 CI scripts to be used for local dev

  #143190 commented on Apr 30, 2025 • 1 new comment

• [pytree] simplify public API exposition with `__module__`

  #148328 commented on May 1, 2025 • 0 new comments

• [ATen][CUDA] Optimize 128 bit vectorization

  #148320 commented on May 2, 2025 • 0 new comments

• handle jk for emulation runs

  #148240 commented on May 2, 2025 • 0 new comments

• [Intel CPU] Fix issue #143483.

  #144854 commented on May 2, 2025 • 0 new comments

• Enable `_lazy_clone` between CPU and MPS

  #148408 commented on May 1, 2025 • 0 new comments

• set non_blocking to true in torch._foreach_copy_ to improve performance

  #148431 commented on Apr 28, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `torch/[p-z]*/` to `ruff format`

  #144552 commented on May 1, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `torch/ao/` to `ruff format`

  #148185 commented on May 1, 2025 • 0 new comments

• [pytree] add another simplified pytree module `torch.pytree`

  #148180 commented on May 1, 2025 • 0 new comments

• Improvement with comprehensive docstrings and implementation of class method for the code.

  #148170 commented on Apr 29, 2025 • 0 new comments

• [Don't merge]Upgrade submodule oneDNN to v3.7 (#147498)(Z7)

  #148163 commented on Apr 30, 2025 • 0 new comments

• Replacing explicit backend search with api call

  #144944 commented on May 1, 2025 • 0 new comments

• Enable fp16 linear layers in PyTorch via ACL

  #144992 commented on Apr 26, 2025 • 0 new comments

• draft

  #148160 commented on Apr 29, 2025 • 0 new comments

• Checks kv pair indexing in OrderedPreservingDictTest.test_range_insert

  #148136 commented on May 1, 2025 • 0 new comments

• Use myst_nb in docs

  #148105 commented on Apr 28, 2025 • 0 new comments

• Enable XPU distributed test for PT2.8

  #149916 commented on Apr 29, 2025 • 0 new comments

• [WIP] no normalizations abstractions

  #149899 commented on Apr 28, 2025 • 0 new comments

• [Dynamo] Add easydict support

  #149851 commented on Apr 26, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `torch/[e-n]*/` to `ruff format`

  #144553 commented on May 1, 2025 • 0 new comments

• [cuDNN][SDPA] cuDNN SDPA refactor/cleanup, nested tensor backward, test priority bump for `sm90`, `sm100`

  #149282 commented on May 1, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `torch/[a-c]*/` to `ruff format`

  #144554 commented on May 1, 2025 • 0 new comments

• [BE][Ez]: Update CU126 to CUDNN 12.8 too

  #149254 commented on Apr 30, 2025 • 0 new comments

• [Easy] update pip sources for CUDA in nightly pull tool

  #149143 commented on May 1, 2025 • 0 new comments

• Update the heuristic for AArch64 bmm/baddbmm

  #149122 commented on Apr 29, 2025 • 0 new comments

• Unrestrict some onlyCPU tests

  #149095 commented on Apr 29, 2025 • 0 new comments

• [test] bigger runnner

  #149003 commented on Apr 30, 2025 • 0 new comments

• [Inductor] Record Triton’s Base32 Cache Key in .best_config for Debugging

  #148981 commented on Apr 29, 2025 • 0 new comments

• Move token linter code into tools/linter/adaptors/_linter/

  #148959 commented on May 1, 2025 • 0 new comments

• cpp_wrapper: build non-performance-sensitive code at O1

  #148773 commented on May 1, 2025 • 0 new comments

• Trunk workflow for Windows Arm64

  #148753 commented on May 2, 2025 • 0 new comments

• [BE][PYFMT] remove `black`: finish `black -> ruff format` migration

  #144557 commented on May 1, 2025 • 0 new comments

• [inductor] lowering for fractional_max_pool3d

  #148630 commented on Apr 28, 2025 • 0 new comments

• Adjust CMake code for Eigen

  #148628 commented on May 1, 2025 • 0 new comments

• [BE][pytree] cleanup parameterized pytree tests

  #148569 commented on May 1, 2025 • 0 new comments

• [triton hash update] update the pinned triton hash

  #148492 commented on May 2, 2025 • 0 new comments

• [Intel CPU] Fix issue #143482.

  #144760 commented on May 2, 2025 • 0 new comments

• Fix issue #146018: Improve CachingAutotuner handling

  #147580 commented on Apr 25, 2025 • 0 new comments

• Update pybind11 submodule to 3.0.0-dev test

  #147524 commented on Apr 30, 2025 • 0 new comments

• removed check for ConvTranspose3D on MPS

  #145366 commented on Apr 28, 2025 • 0 new comments

• Fix pxtas warnings on sm_120

  #147491 commented on Apr 25, 2025 • 0 new comments

• Documentation: fix RNN example for multiple layers

  #147490 commented on Apr 25, 2025 • 0 new comments

• [test] sccache log

  #147470 commented on May 1, 2025 • 0 new comments

• handle default in _NamedOptimizer

  #147357 commented on Apr 25, 2025 • 0 new comments

• [NOT_FOR_COMMIT] Try Triton-cpu-arm

  #147341 commented on Apr 26, 2025 • 0 new comments

• Fix torch.compile Fallback for Meta Device Tensors

  #147339 commented on Apr 27, 2025 • 0 new comments

• Optimize `Sequential` methods description

  #147304 commented on May 1, 2025 • 0 new comments

• [Easy] update pip sources for ROCm in nightly pull tool

  #145685 commented on May 1, 2025 • 0 new comments

• Do not use username for inductor default_cache_dir

  #147291 commented on Apr 26, 2025 • 0 new comments

• Fix clang-tidy warnings in torch/jit

  #147253 commented on May 2, 2025 • 0 new comments

• Add quantized BatchNorm1d module

  #147113 commented on Apr 25, 2025 • 0 new comments

• Porting Pytorch to AIX Operating System.

  #146983 commented on Apr 30, 2025 • 0 new comments

• cmake: fix detection logic when using system XNNPACK

  #145853 commented on Apr 26, 2025 • 0 new comments

• [ARM] Fix TestDataLoader.test_segfault unexpected success on Aarch6[4

  #146090 commented on Apr 29, 2025 • 0 new comments

• [CI] Get rid of UCC builds

  #146173 commented on May 2, 2025 • 0 new comments

• Fix non-bitwise type annotations for Tensor operators (see #145838)

  #146845 commented on May 1, 2025 • 0 new comments

• [Inductor-CPU] FP16 X int8 WoQ GEMM for M <= 4 with FP16 accum & compute

  #146781 commented on Apr 27, 2025 • 0 new comments

• Enable pt2e quantization path for arm

  #146690 commented on Apr 28, 2025 • 0 new comments

• [2/N] Fix cppcoreguidelines-init-variables suppression

  #146237 commented on Apr 29, 2025 • 0 new comments

• Update code_template.py re.compile() is directly applied to the regex…

  #146489 commented on Apr 29, 2025 • 0 new comments

• [not for land] temp changes to enable 'simple_fsdp'

  #146558 commented on Apr 28, 2025 • 0 new comments

• [HOP] Mutation and alias rework

  #146658 commented on Apr 30, 2025 • 0 new comments

• Refactor layout constraint selection logic

  #148104 commented on May 2, 2025 • 0 new comments

• Fix test_tensorboard when started w/o tensorboard package

  #148079 commented on Apr 29, 2025 • 0 new comments

• use identity op for alpha=inf in torch.celu and quantized_celu

  #148066 commented on Apr 29, 2025 • 0 new comments

• Support `contextlib.suppress`

  #147990 commented on Apr 29, 2025 • 0 new comments

• xpu: test py_limited_api with SyclExtension

  #147984 commented on Apr 27, 2025 • 0 new comments

• [aot] reset aot counter on torch._dynamo.reset

  #147915 commented on Apr 27, 2025 • 0 new comments

• [DONOTLAND] Fix partial + scalar issue

  #147910 commented on Apr 27, 2025 • 0 new comments

• [PT2][Optimus][Opportunity Finder][1/n] Add opportunity finder in the inductor for GEMM horizonal fusion search

  #147908 commented on Apr 28, 2025 • 0 new comments

• Remerge of #144974

  #147903 commented on Apr 27, 2025 • 0 new comments

• Change persistent reduction threshold to 32

  #147899 commented on Apr 28, 2025 • 0 new comments

• Back out "use copy2d in h2d/d2h copy when possible (#146256)"

  #147808 commented on Apr 27, 2025 • 0 new comments

• test

  #147800 commented on Apr 28, 2025 • 0 new comments

• Upgrade to DLPack 1.0.

  #145000 commented on Apr 26, 2025 • 0 new comments

• Made partitioning more(?) deterministic

  #145024 commented on Apr 29, 2025 • 0 new comments

• [Inductor-CPU] Avoid memory allocator lock contention in the GEMM template

  #147797 commented on Apr 27, 2025 • 0 new comments

• [cuda] Add new gamma beta backwards kernel

  #147773 commented on Apr 26, 2025 • 0 new comments

• add pt2 testing for torch.float8_e8m0fnu

  #147770 commented on Apr 27, 2025 • 0 new comments

• [DCP][OSS] Rank local checkpointing in DCP without collectives

  #147758 commented on Apr 30, 2025 • 0 new comments

• Adding MVP of P1 INT16 Full

  #147747 commented on Apr 25, 2025 • 0 new comments

• Turn Stream into protocol and improve typing in torch/_C/__init__.pyi.in

  #145239 commented on Apr 30, 2025 • 0 new comments

• [Intel GPU] OneDNN primitive cache support for Int4 WOQ gemm on XPU

  #147693 commented on Apr 28, 2025 • 0 new comments

• [cuBLAS] restrict input range for `addmm` tests

  #147658 commented on Apr 28, 2025 • 0 new comments

• [sm100][sm120][fp8][CUDA] skip rowwise scaling tests on SM100+ for now

  #147645 commented on Apr 26, 2025 • 0 new comments

• Remove backend_type_map from Backend

  #147635 commented on Apr 27, 2025 • 0 new comments

• Refactor typing: Replace Any with ParamSpec for better type safety

  #147582 commented on Apr 26, 2025 • 0 new comments

• Add dynamo config to HOP-ify context managers

  #152159 commented on Apr 30, 2025 • 0 new comments

• [SymmMem] Add all_to_all_vdev

  #151819 commented on May 2, 2025 • 0 new comments

• Normalize dynamic size symbols in template codegen cache key.

  #151778 commented on Apr 28, 2025 • 0 new comments

• [Inductor] Modify TritonTemplate store_output function to support TMA stores

  #151775 commented on Apr 30, 2025 • 0 new comments

• [Inductor] Modify persistent+TMA template for Triton mm and admm to use new TMA API

  #151774 commented on Apr 30, 2025 • 0 new comments

• [2/n][Optimus][Auto-AC] Support activation quantization with scaling

  #151770 commented on May 1, 2025 • 0 new comments

• Add adaptive_avg_pool2d input and output_size check

  #151769 commented on Apr 25, 2025 • 0 new comments

• [Don't merge] Upgrade oneDNN to v3.8 for XPU build

  #151767 commented on Apr 25, 2025 • 0 new comments

• torch.testing._internal.optests - MPS Support

  #151758 commented on Apr 28, 2025 • 0 new comments

• Use gather in index_select

  #151715 commented on May 2, 2025 • 0 new comments

• [dtensor] op_schema recursive check for symints

  #151679 commented on Apr 28, 2025 • 0 new comments

• [Intel GPU] Use user-friendly err msg in mm

  #151655 commented on Apr 28, 2025 • 0 new comments

• Add OIDC perms to windows-[build|test] workflows

  #151596 commented on May 1, 2025 • 0 new comments

• Add OIDC permissions to linux-test workflow

  #151585 commented on May 1, 2025 • 0 new comments

• Add OIDC permissions to linux-build workflow

  #151581 commented on May 1, 2025 • 0 new comments

• Update OpenBLAS commit

  #151547 commented on Apr 28, 2025 • 0 new comments

• Fix `InstanceNorm` wrong suggestion in warning message

  #151534 commented on May 1, 2025 • 0 new comments

• [WIP] Deprecate getPinnedMemoryAllocator use getHostAllocator instead

  #151531 commented on Apr 29, 2025 • 0 new comments

• Add OIDC permissions to bazel workflow

  #151456 commented on Apr 25, 2025 • 0 new comments

• Allow to byteswap data when reading saved torch jit data

  #151447 commented on May 1, 2025 • 0 new comments

• Implement fast exp for AVX2 and AVX512 for the flash attention

  #151441 commented on Apr 29, 2025 • 0 new comments

• [Cutlass] Add epilogue inputs/outputs to def_kernel

  #151406 commented on May 2, 2025 • 0 new comments

• [ROCm] Upgrade ROCm CI to ROCm6.4

  #151368 commented on May 2, 2025 • 0 new comments

• Fix skipIfXpu and skipIfHpu disables tests when used on class

  #151315 commented on Apr 30, 2025 • 0 new comments

• Add inductor backend to device interface; make minifier_tests more device agnostic

  #151314 commented on Apr 29, 2025 • 0 new comments

• [Export] Remove to() from module generated form exported program

  #151307 commented on Apr 28, 2025 • 0 new comments

• Use Allocator API raw_allocate & raw_dealloc in CUDAAllocator

  #151305 commented on Apr 25, 2025 • 0 new comments

• WIP: divup op

  #152144 commented on Apr 25, 2025 • 0 new comments

• [inductor] pass reduction idx to scan inner_fns

  #152142 commented on Apr 28, 2025 • 0 new comments

• Remove some instances of uninitialized memory use

  #152132 commented on Apr 29, 2025 • 0 new comments

• Update _torch_docs.py to Fix torch.bernoulli()

  #152104 commented on Apr 28, 2025 • 0 new comments

• Migrate to new Windows Arm64 runners

  #152099 commented on Apr 28, 2025 • 0 new comments

• Switch to standard pep517 sdist generation

  #152098 commented on May 1, 2025 • 0 new comments

• Work around MPSGraph issue in backward pass of nn.ReplicationPad1d/2d

  #152094 commented on Apr 28, 2025 • 0 new comments

• Add optional device index to AOTIModelPackageLoader

  #152093 commented on May 2, 2025 • 0 new comments

• [cuDNN][SDPA] Fix head-dim 256 condition for SM 10.0

  #152076 commented on May 1, 2025 • 0 new comments

• Test

  #152055 commented on Apr 28, 2025 • 0 new comments

• unbreak fb:operator_benchmark_test

  #152049 commented on May 1, 2025 • 0 new comments

• Ignore unused structured arguments in member functions

  #152019 commented on Apr 29, 2025 • 0 new comments

• Add CPython complex tests

  #152015 commented on May 1, 2025 • 0 new comments

• [Kineto] Upgrade the kineto commit to fb36cce

  #152007 commented on Apr 29, 2025 • 0 new comments

• [UniformValueConstantFolder] deduce value on CPU rather than on device

  #151998 commented on Apr 28, 2025 • 0 new comments

• [SymmMem] Use cub's BlockScan instead of in-house impl for offset calculation

  #151993 commented on May 2, 2025 • 0 new comments

• [inductor] Remove usage of autotune_fallback_to_aten outside inductor code

  #151988 commented on Apr 29, 2025 • 0 new comments

• Add torchcheck for replication_pad3d_backward

  #151986 commented on Apr 25, 2025 • 0 new comments

• [Intel GPU] undo broadcast on zero stride tensor for SDPA

  #151976 commented on Apr 29, 2025 • 0 new comments

• Inductor Tiling Rewrite

  #151958 commented on Apr 29, 2025 • 0 new comments

• [inductor] fix lowering for cummin, cummax for one element tensors

  #151931 commented on May 1, 2025 • 0 new comments

• [ROCm][CI] Update dockerfile to use centos9

  #151929 commented on May 1, 2025 • 0 new comments

• Skip fuse attention on fp32 if not tf32

  #151924 commented on Apr 28, 2025 • 0 new comments

• [WIP] Deprecate AcceleratorHooksInterface isPinnedPtr, use at::getHostAllocator()->is_pinned instead

  #151916 commented on Apr 29, 2025 • 0 new comments

• Deprecated pkg_resources and use distributions instead

  #151915 commented on Apr 29, 2025 • 0 new comments

• Add `LinearLR` compute lr formula in doc

  #151894 commented on Apr 25, 2025 • 0 new comments

• [inductor] Clean typing in codegen/common.py and codecache.py

  #150767 commented on Apr 28, 2025 • 0 new comments

• [Dynamo][Typing] Enable `@override` for VTs [1/N]

  #150763 commented on Apr 25, 2025 • 0 new comments

• [BE][CI][Easy] Run `lintrunner` on generated `.pyi` stub files

  #150732 commented on May 2, 2025 • 0 new comments

• [BE] Resolve lint errors in `.pyi` stub files

  #150731 commented on May 2, 2025 • 0 new comments

• [BE] Ensure generated stub files by `gen_pyi` are properly formatted

  #150730 commented on May 2, 2025 • 0 new comments

• [BE] Add `__all__` to `torch/nn/functional.pyi` and `torch/return_types.pyi`

  #150729 commented on May 2, 2025 • 0 new comments

• [BE] Update `.pyi` stub template to use Generic TypeAlias (PEP 585) and Union Type (PEP 604)

  #150728 commented on May 2, 2025 • 0 new comments

• [torchgen] Refactor and simplify `gen_pyi.py` to use Generic TypeAlias (PEP 585) and Union Type (PEP 604)

  #150727 commented on May 2, 2025 • 0 new comments

• [torchgen] Refactor `torchgen.utils.FileManager` to accept `pathlib.Path`

  #150726 commented on May 2, 2025 • 0 new comments

• Avoid overwriting COW data in MPS code

  #150721 commented on May 2, 2025 • 0 new comments

• [export] add runtime assert messages to python torch checks

  #150719 commented on Apr 25, 2025 • 0 new comments

• [WIP] Support XPU in memory tracker

  #150703 commented on Apr 30, 2025 • 0 new comments

• Raise `BufferError` for DLPack buffer-related errors.

  #150691 commented on Apr 26, 2025 • 0 new comments

• AOTI: add all fallback ops that are missing from C-shim

  #150673 commented on May 1, 2025 • 0 new comments

• [Inductor] Fix CUDA memory usage for CPU only compile

  #150669 commented on Apr 28, 2025 • 0 new comments

• Refactor `torch/utils/data/datapipes/gen_pyi.py` with `torchgen`

  #150626 commented on May 2, 2025 • 0 new comments

• Fix nn.LazyModuleMixin examples

  #150596 commented on May 1, 2025 • 0 new comments

• fix dynamic shapes for kwargs

  #150583 commented on Apr 30, 2025 • 0 new comments

• Enable lazy cloning in `Tensor.to` between CPU and MPS

  #150569 commented on May 2, 2025 • 0 new comments

• Inductor respects exact strides on custom ops by default

  #150511 commented on May 2, 2025 • 0 new comments

• [DLPack] Add support for missing keyword-arguments.

  #150218 commented on Apr 26, 2025 • 0 new comments

• Fix DLPack stream logic.

  #150217 commented on Apr 26, 2025 • 0 new comments

• [DLPack] add NumPy exchange tests.

  #150216 commented on Apr 26, 2025 • 0 new comments

• AOTI freezing: fix test issues and enable by default

  #149961 commented on May 1, 2025 • 0 new comments

• [inductor] Add more typing to _inductor/ir.py

  #149959 commented on Apr 28, 2025 • 0 new comments

• [inductor] Add typing to _inductor/ir.py

  #149958 commented on Apr 29, 2025 • 0 new comments

• [3/N] Use internal linkage in C++ files

  #151297 commented on Apr 30, 2025 • 0 new comments

• [WIP][SymmMem] Add sendrecv op

  #151262 commented on Apr 29, 2025 • 0 new comments

• [dynamo] keep C++ symbolic shape guards disabled for benchmarks

  #151225 commented on May 1, 2025 • 0 new comments

• Implement MKLGenerator

  #151218 commented on May 1, 2025 • 0 new comments

• Update slow tests

  #151207 commented on Apr 28, 2025 • 0 new comments

• Fix DWConv in QNNPACK for aarch32

  #151191 commented on Apr 26, 2025 • 0 new comments

• [dynamo] Prevent lazy variable realization on STORE_FAST

  #151184 commented on May 2, 2025 • 0 new comments

• Guard additional use of DriverAPI

  #151125 commented on Apr 28, 2025 • 0 new comments

• TESTING: IGNORE

  #151116 commented on May 1, 2025 • 0 new comments

• Update auto-tuning support for _scaled_grouped_mm

  #150944 commented on May 1, 2025 • 0 new comments

• fix shard tensor gather when a local tensor on certain ranks has zero elements

  #150914 commented on Apr 27, 2025 • 0 new comments

• [DO NOT MERGE] Throwaway changes

  #150910 commented on May 2, 2025 • 0 new comments

• [Inductor] Fix cuda_template.py typing

  #150909 commented on May 2, 2025 • 0 new comments

• [Inductor] Fix cuda_kernel typing

  #150908 commented on Apr 29, 2025 • 0 new comments

• [Cutlass] Changes to gemm template for EVT

  #150907 commented on May 2, 2025 • 0 new comments

• [device_mesh] replace dim_group_info with group_name

  #150898 commented on Apr 30, 2025 • 0 new comments

• Add CPython tests for iter/sort

  #150797 commented on May 1, 2025 • 0 new comments

• Add CPython generator/contextlib tests

  #150796 commented on May 2, 2025 • 0 new comments

• Add CPython int/float tests

  #150795 commented on May 1, 2025 • 0 new comments

• Add CPython math/cmath tests

  #150794 commented on May 1, 2025 • 0 new comments

• Add CPython string tests

  #150793 commented on May 1, 2025 • 0 new comments

• Add CPython set tests

  #150792 commented on May 1, 2025 • 0 new comments

• Add CPython dict tests

  #150791 commented on May 1, 2025 • 0 new comments

• Add CPython list/tuple tests

  #150790 commented on May 1, 2025 • 0 new comments

• Add CPython exception tests

  #150789 commented on May 1, 2025 • 0 new comments

• Add CPython tests for unittest

  #150788 commented on May 1, 2025 • 0 new comments

• The operator 'aten::_linalg_eigh.eigenvalues' is not currently implemented for the MPS device

  #151874 commented on Apr 28, 2025 • 0 new comments

• CUDA deps cannot be preloaded under Bazel

  #117350 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_int8 (__main__.TestForeachCUDA)

  #150837 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_inductor_all_to_all_single (__main__.CompileTest)

  #147795 commented on Apr 28, 2025 • 0 new comments

• Bug with "make latexpdf"

  #135420 commented on Apr 28, 2025 • 0 new comments

• DeepSeek: mixed precision optimizers

  #146542 commented on Apr 28, 2025 • 0 new comments

• `2` and `-2` for `ord` argument of `linalg.norm()` should be explained more clearly

  #136453 commented on Apr 28, 2025 • 0 new comments

• [inductor] [cuda] [fake tensor] `torch.ones(x.size(0))` becomes a fake tensor for `torch.diagonal_scatter`

  #151670 commented on Apr 28, 2025 • 0 new comments

• Graph break on .t() when Tensor._make_subclass

  #151771 commented on Apr 28, 2025 • 0 new comments

• Inductor generates wrong code for `torch.embedding`

  #151918 commented on Apr 28, 2025 • 0 new comments

• Tracking issue: Incorrect Meta Strides / Turn On PyDispatcher in FakeTensor Mode

  #145094 commented on Apr 28, 2025 • 0 new comments

• Label tracking meta-issue (edit me to get automatically CC'ed on issues! cc bot)

  #24422 commented on Apr 28, 2025 • 0 new comments

• Add maxcount Parameter to torch.unique and torch.unique_consecutive

  #151722 commented on Apr 28, 2025 • 0 new comments

• ExpandableMemorySegments not working on H100s/A100s

  #122057 commented on Apr 28, 2025 • 0 new comments

• [ROCm] QR decomposition is much slower on MI300x than A100

  #151066 commented on Apr 28, 2025 • 0 new comments

• Unaccaptable OOMs all the time.

  #152135 commented on Apr 28, 2025 • 0 new comments

• distrubuted: false positive Grad strides vs Bucket strides warning

  #152042 commented on Apr 28, 2025 • 0 new comments

• [inductor] nan_asserts doesn't work for FP8, "RuntimeError: "isinf" not implemented for 'Float8_e4m3fn'"

  #149002 commented on Apr 28, 2025 • 0 new comments

• Support loading and executing a ExportedProgram from torch.export in C++ environment

  #144663 commented on Apr 28, 2025 • 0 new comments

• [C10D] Autograd Support for Collectives

  #152131 commented on Apr 28, 2025 • 0 new comments

• [torch.compile][export] `PendingUnbackedSymbolNotFound` for `torch.full`

  #151980 commented on Apr 28, 2025 • 0 new comments

• FlopCounterMode doesn't support HOP

  #134385 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_inductor_all_reduce_non_contig_input (__main__.CompileTest)

  #147733 commented on Apr 28, 2025 • 0 new comments

• torch.compile doesnot support index with tensor

  #151997 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_comprehensive_native_layer_norm_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #152056 commented on Apr 28, 2025 • 0 new comments

• [Dynamo] Exception raised inside torch.autocast causes crash AttributeError: 'NoneType' object has no attribute 'is_python_constant

  #152012 commented on Apr 28, 2025 • 0 new comments

• linear + relu don't fuse

  #152101 commented on Apr 28, 2025 • 0 new comments

• FlexAttention + Export / AOTI

  #152128 commented on Apr 28, 2025 • 0 new comments

• Compiling attention (SDPA) with nested tensors fails when using DDP

  #152068 commented on Apr 28, 2025 • 0 new comments

• RFC: Torch Native Runtime

  #152034 commented on Apr 28, 2025 • 0 new comments

• standalone_compile with training errors with no cache artifacts

  #152022 commented on Apr 28, 2025 • 0 new comments

• Adam optimizer ValueError: beta1 as a Tensor

  #149508 commented on Apr 26, 2025 • 0 new comments

• [BE] Consolidate PR labeling logic

  #151579 commented on Apr 29, 2025 • 0 new comments

• [Infra] Jobs got frequently cancelled, sometimes mid-checkout

  #151669 commented on Apr 29, 2025 • 0 new comments

• The docstring linter should not force overridden methods to be documented

  #151692 commented on Apr 29, 2025 • 0 new comments

• fbgemm packages are compiled in torchinductor torchbench tests

  #152024 commented on Apr 29, 2025 • 0 new comments

• [inductor] [silent incorrectness] `torch.nn.PairwiseDistance(p=2)` outputs incorrect results with eager

  #151198 commented on Apr 29, 2025 • 0 new comments

• Major perf regression with `BatchNorm2d` + `torch.compile` with `reduce-overhead` + DDP

  #139207 commented on Apr 29, 2025 • 0 new comments

• associative scan is incorrect for certain shapes/kwargs

  #137943 commented on Apr 29, 2025 • 0 new comments

• Support AC with graph break

  #139989 commented on Apr 29, 2025 • 0 new comments

• torch.compile + Huggingface GenerationMixin

  #141196 commented on Apr 29, 2025 • 0 new comments

• nn.MultiheadAttention causes gradients to become NaN under some use cases

  #41508 commented on Apr 29, 2025 • 0 new comments

• Expanding subset of tensor reads wrong memory

  #151799 commented on Apr 29, 2025 • 0 new comments

• aten.grid_sampler_3d.default is missing a c-shim implementation, using proxy executor as fallback

  #147625 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_inductor_inplace_op_on_view (__main__.CompileTest)

  #147852 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_uint8 (__main__.TestForeachCUDA)

  #150878 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_int8 (__main__.TestForeachCUDA)

  #149859 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_inductor_reduce_scatter_tensor_single (__main__.CompileTest)

  #147911 commented on Apr 29, 2025 • 0 new comments

• Expand Tag Set: views & reductions

  #129020 commented on Apr 29, 2025 • 0 new comments

• [inductor] [cpu] `torch.nn.RReLU()` doesn't respect `fallback_random` flag

  #147255 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_serialized_patterns_up_to_date (__main__.TestPatternMatcher)

  #135476 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_comprehensive_scatter_reduce_prod_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #140294 commented on Apr 29, 2025 • 0 new comments

• torch.compile fails in FSDP due to .data assignment with different floating type

  #152162 commented on Apr 29, 2025 • 0 new comments

• Update quantization to make source files complient with /Zc:lambda

  #92600 commented on Apr 29, 2025 • 0 new comments

• [Inductor] define custom pass as list

  #151876 commented on Apr 29, 2025 • 0 new comments

• Loss parallel's override of log_softmax doesn't support negative dims

  #152016 commented on Apr 29, 2025 • 0 new comments

• `RuntimeError: UR error` with XPU

  #149953 commented on Apr 29, 2025 • 0 new comments

• "RuntimeError: makeDeviceForHostname(): unsupported gloo device" with nightly torch 2.8

  #150381 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_distributed_checkpoint_state_dict_type0_cuda (__main__.TestDistributedCheckpointCUDA)

  #145807 commented on Apr 29, 2025 • 0 new comments

• TorchInductor CPU Performance Dashboard

  #93531 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_comprehensive_nn_functional_glu_cuda_float16 (__main__.TestInductorOpInfoCUDA)

  #140383 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_comprehensive_nanquantile_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #139593 commented on Apr 29, 2025 • 0 new comments

• [dynamo] Some inefficiencies around handling __torch_function__

  #151776 commented on Apr 28, 2025 • 0 new comments

• `torch.bmm` is slow on non-contiguous BF16 CPU tensors

  #151934 commented on Apr 28, 2025 • 0 new comments

• Negative index support for `take_along_dim`

  #146211 commented on Apr 26, 2025 • 0 new comments

• Add a `TORCH_LOGS_RANK=0` env var that integrates with `TORCH_LOGS`

  #146913 commented on Apr 26, 2025 • 0 new comments

• [inductor] [assertion error] `torch.select_scatter` crashes on inductor but passes on eager

  #151296 commented on Apr 26, 2025 • 0 new comments

• [mem profiler] mem fragmentation and pynvml view

  #150574 commented on Apr 25, 2025 • 0 new comments

• [AotInductor][Export][Triton] how to export custom triton kernels when use torch.export.export

  #151746 commented on Apr 25, 2025 • 0 new comments

• AOTI cannot move tensors between cuda devices

  #152130 commented on Apr 25, 2025 • 0 new comments

• AOTInductor package can only be loaded on the first GPU (cuda:0) in C++ via AOTIModelPackageLoader

  #152087 commented on Apr 25, 2025 • 0 new comments

• [feature request][AOTI] Expand check input assertions to cover input guards created during compilation?

  #151925 commented on Apr 25, 2025 • 0 new comments

• Exported Module cannot call train() or eval()

  #151726 commented on Apr 25, 2025 • 0 new comments

• [export] deserialization for unbacked ranges is wrong

  #151809 commented on Apr 25, 2025 • 0 new comments

• DISABLED test_item_to_inputs_kernel_nobreak_cuda (__main__.TestInductorDynamicCUDA)

  #119538 commented on Apr 25, 2025 • 0 new comments

• Stack trace from pytest is very far away and far too find on some tests

  #141204 commented on Apr 25, 2025 • 0 new comments

• Some Performance Bug in `tol` of `torch.lobpcg()`

  #152154 commented on Apr 25, 2025 • 0 new comments

• Note some limit in docstring of `padding` in Poolnd

  #152156 commented on Apr 25, 2025 • 0 new comments

• Some Doc Issue about `torch.lobpcg()`

  #152107 commented on Apr 25, 2025 • 0 new comments

• [Scaled MM] Update to support on B200 TN, NT, NN, TT Layouts are supported

  #152150 commented on Apr 25, 2025 • 0 new comments

• padding_mode `reflect` works different from others in Conv

  #152152 commented on Apr 25, 2025 • 0 new comments

• Documentation Clarification Needed for Clamping of Scale Coefficient in clip_grads_with_norm_

  #151554 commented on Apr 25, 2025 • 0 new comments

• [Inductor] track block shape of intermediary variables

  #149905 commented on Apr 25, 2025 • 0 new comments

• torch.nested.narrow() or torch.nested.to_padded_tensor() breaks backwards pass - invalid gradient

  #145837 commented on Apr 25, 2025 • 0 new comments

• Enable TorchInductor to Generate Matmuls Natively via `tl.dot`

  #151705 commented on Apr 25, 2025 • 0 new comments

• DISABLED test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_bfloat16 (__main__.TestForeachCUDA)

  #148965 commented on Apr 25, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_int16 (__main__.TestForeachCUDA)

  #150772 commented on Apr 25, 2025 • 0 new comments

• Exporting the operator 'aten::lift_fresh' to ONNX - not supported

  #151932 commented on Apr 25, 2025 • 0 new comments

• [ONNX] Flip `dynamo` default to True in torch.onnx.export

  #151693 commented on Apr 25, 2025 • 0 new comments

• Pinned memory doubles memory usage for tensors slightly over 128MB

  #150517 commented on Apr 25, 2025 • 0 new comments

• Inductor pattern matcher replaces aten.reshape with aten.view in pattern

  #151649 commented on Apr 25, 2025 • 0 new comments

• DISABLED test_ranks_and_tag (__main__.CompileTest)

  #147974 commented on Apr 25, 2025 • 0 new comments

• Xcode 16+: duplicate LC_RPATH '@loader_path'

  #151592 commented on Apr 25, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_float64 (__main__.TestForeachCUDA)

  #150752 commented on Apr 25, 2025 • 0 new comments

• [Inductor] atomic_add does not support bf16

  #97016 commented on Apr 25, 2025 • 0 new comments

• standalone compile FakeTensor from_graph detection with tensor subclass outputs

  #151945 commented on Apr 28, 2025 • 0 new comments

• Add operator name to the size/strides/alignment assertion

  #151930 commented on Apr 28, 2025 • 0 new comments

• Runtime assertion not generated in inductor for input unbacked symints

  #151879 commented on Apr 28, 2025 • 0 new comments

• Optimize printing sympy expressions during logging and cache key computation

  #151823 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_int64 (__main__.TestForeachCUDA)

  #150822 commented on Apr 28, 2025 • 0 new comments

• Support Delay Loading of c10.dll in when using libtorch as a thirdparty library.

  #105058 commented on Apr 28, 2025 • 0 new comments

• Torch 2.1 compile + FSDP (mixed precision) + LlamaForCausalLM: `RuntimeError: attempting to assign a gradient with dtype 'c10::BFloat16' to a tensor with dtype 'float'.`

  #111317 commented on Apr 28, 2025 • 0 new comments

• [inductor] [cpu] [edge case] When processing `torch.nan_to_num-.long()`, inductor outputs the `reciprocal` of eager

  #151510 commented on Apr 28, 2025 • 0 new comments

• [dynamo] Replace `unimplemented` with `unimplemented_v2`

  #147913 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_while_loop_schema_gen (__main__.TestHopSchema)

  #141202 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_rng (__main__.TestCompilerBisector)

  #139590 commented on Apr 28, 2025 • 0 new comments

• `torch.jit.script` does not respect `torch.set_default_dtype`

  #150607 commented on Apr 28, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_outplace_cuda_int32 (__main__.TestForeachCUDA)

  #150800 commented on Apr 28, 2025 • 0 new comments

• Python 3.10 + intel-openmp failed to use numactl after import torch._C

  #136307 commented on Apr 28, 2025 • 0 new comments

• Torch compile issue, AttributeError: 'NoneType' object has no attribute 'store_cubin'

  #150980 commented on Apr 27, 2025 • 0 new comments

• Elastic training crashes on killed agent

  #150916 commented on Apr 27, 2025 • 0 new comments

• [Torch Profiler] Only two streams captured in CUDA graph but multiple streams shown in Torch Profiler

  #152114 commented on Apr 27, 2025 • 0 new comments

• DeepSeek: a2a communication with metadata on the GPU

  #146329 commented on Apr 27, 2025 • 0 new comments

• ROCm+gcc 15 asserts

  #145608 commented on Apr 27, 2025 • 0 new comments

• [FSDP2][DTensor] numeric bug for DTensor + python float in gradient clipping

  #149768 commented on Apr 26, 2025 • 0 new comments

• caching keys+values in TransformerDecoderLayer for faster inference

  #107573 commented on Apr 26, 2025 • 0 new comments

• [release] Make pytorch source distribution package respect pep-0517

  #150461 commented on Apr 26, 2025 • 0 new comments

• Change the type hint for nn.Module.__call__ to be friendly to overrides.

  #74746 commented on Apr 26, 2025 • 0 new comments

• [inductor] [cuda] [silent incorrectness] `F.softmax-torch.argsort` output silent incorrectness when tensor input is very large

  #151745 commented on Apr 26, 2025 • 0 new comments

• [feature request] [ux] Frontend methods for fused elementwise affine transform: mul+add+dtype convert + support `integer_tensor.mul_(float_constant)` and `float_tensor.mul(some_constant, out = integer_tensor)` maybe via new args `rounding_mode=...` and `dtype=...` + maybe support OpenCV-style saturated dtype conversions (e.g. `clamp_` before conversion)

  #106624 commented on Apr 26, 2025 • 0 new comments

• [feature request] torch.mix function to generalize/symmetrize addcmul

  #104849 commented on Apr 26, 2025 • 0 new comments

• `torch.lerp` to support argument type promotion / broadcasting - including of `input` / `end` arguments

  #57947 commented on Apr 26, 2025 • 0 new comments

• Wrong formula for CosineAnnealingLR

  #152081 commented on Apr 26, 2025 • 0 new comments

• Build pytorch for rocm failed

  #148167 commented on Apr 26, 2025 • 0 new comments

• [RFC] Add new CPP builder for inductor on pytorch Windows

  #124245 commented on Apr 26, 2025 • 0 new comments

• Torch Inductor Windows Path Escape Characters

  #135954 commented on Apr 26, 2025 • 0 new comments

• Compilation of the post-training quantized model using Nvidia ModelOpt is failing with the error: Unsupported — 'inline in skipfiles: QuantLinearConvBase.quantize_weight

  #151450 commented on Apr 26, 2025 • 0 new comments

• refine fp32 precision api

  #125888 commented on Apr 30, 2025 • 0 new comments

• [vision hash update] update the pinned vision hash

  #125806 commented on May 2, 2025 • 0 new comments

• Automated submodule update: FBGEMM

  #115316 commented on May 2, 2025 • 0 new comments

• [pytree] support PyStructSequence types for Python pytree

  #113258 commented on May 1, 2025 • 0 new comments

• Automated submodule update: kineto

  #106149 commented on May 2, 2025 • 0 new comments

• [ATen][Sparse] Use Third-Party Eigen for sparse addmm

  #101814 commented on Apr 30, 2025 • 0 new comments

• Performance regression on modded-nanogpt torch-2.7.0.dev20250208→torch-2.7.0.dev20250209

  #147463 commented on May 2, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_int16 (__main__.TestForeachCUDA)

  #150309 commented on May 2, 2025 • 0 new comments

• DISABLED test_remove_noop_view_dtype_cuda (__main__.GPUTests)

  #151541 commented on May 2, 2025 • 0 new comments

• DISABLED test_remove_noop_view_dtype_cpu (__main__.CpuTests)

  #151540 commented on May 2, 2025 • 0 new comments

• DISABLED test_comprehensive_floor_cuda_float16 (__main__.TestInductorOpInfoCUDA)

  #152058 commented on May 2, 2025 • 0 new comments

• DISABLED test_comprehensive_nansum_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #140693 commented on May 2, 2025 • 0 new comments

• DISABLED AotInductorTest.FreeInactiveConstantBufferRuntimeConstantFoldingCuda (build.bin.test_aoti_inference)

  #150299 commented on May 2, 2025 • 0 new comments

• DISABLED AotInductorTest.FreeInactiveConstantBufferCuda (build.bin.test_aoti_inference)

  #149495 commented on May 2, 2025 • 0 new comments

• DISABLED test_comprehensive_bitwise_right_shift_cuda_int32 (__main__.TestInductorOpInfoCUDA)

  #152057 commented on May 2, 2025 • 0 new comments

• [ued] Slow start up time for `torch.compile` on GGUF Auraflow

  #150706 commented on May 2, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float64 (__main__.TestForeachCUDA)

  #150298 commented on May 2, 2025 • 0 new comments

• `context_parallel` fails for training with `RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation`

  #149306 commented on May 2, 2025 • 0 new comments

• The state of sparse Tensors

  #9674 commented on May 1, 2025 • 0 new comments

• [Feature request] Exclusive prefix sum, `torch.cumsum(input, dim=0, exclusive=True)`

  #76191 commented on May 1, 2025 • 0 new comments

• DISABLED test_remove_noop_view_default_cuda (__main__.GPUTests)

  #151511 commented on May 1, 2025 • 0 new comments

• `view()` + modify-in-place fails silently with DTensor

  #147570 commented on May 1, 2025 • 0 new comments

• LoadHIP.cmake should find_package(composable_kernel)

  #149809 commented on May 1, 2025 • 0 new comments

• at::BlasBackend::Ck does not handle all ROCm BLAS gpus

  #150187 commented on May 1, 2025 • 0 new comments

• [ROCm] PyTorch slow on TTS

  #150168 commented on May 1, 2025 • 0 new comments

• Support SDPA flash attention/ memory efficant attn on ROCm gfx908

  #141958 commented on May 1, 2025 • 0 new comments

• [RFC] : Dynamically Quantized 8-bit Matrix Multiplication support

  #149500 commented on May 1, 2025 • 0 new comments

• dynamo cannot trace global op_set .__contains__

  #145761 commented on May 1, 2025 • 0 new comments

• Investigate FlexAttention performance degradation on low precision inputs

  #147336 commented on May 1, 2025 • 0 new comments

• When using torch to convert to oxxn model, testing the inference results with actual images shows tensor mismatch

  #152097 commented on May 1, 2025 • 0 new comments

• Dynamo unsupported: dynamic padding

  #123855 commented on May 1, 2025 • 0 new comments

• [ROCm] MI300X FP8 scaled_mm is extremely slow on nightly

  #143465 commented on Apr 30, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `torch/_[a-h]*/` to `ruff format`

  #144551 commented on May 1, 2025 • 0 new comments

• [BE][PYFMT] migrate PYFMT for `{torch,test}/{nn,optim}/**` to `ruff format`

  #144548 commented on May 1, 2025 • 0 new comments

• codecache: Remove cpp_prefix.h duplication per build, then precompile it

  #144293 commented on May 1, 2025 • 0 new comments

• [ci] Add riscv opt-int build

  #143979 commented on Apr 27, 2025 • 0 new comments

• Make init_method deprecated to fix TCP connection refused error

  #143858 commented on Apr 25, 2025 • 0 new comments

• Make functionalization `ViewMeta` serializable with pickle.

  #143712 commented on Apr 27, 2025 • 0 new comments

• [Draft][WIP] Enable XPU path for FlexAttention

  #143553 commented on Apr 29, 2025 • 0 new comments

• [while_loop][jit inductor] auto-unspecialize int input and output to unbacked symints

  #143457 commented on Apr 30, 2025 • 0 new comments

• Fix type annotation of `Linear.bias`

  #142326 commented on Apr 26, 2025 • 0 new comments

• remove redundant assign

  #140399 commented on Apr 27, 2025 • 0 new comments

• [pt2e][quant] Add simple cond support for annotate, prepare and convert

  #140323 commented on Apr 27, 2025 • 0 new comments

• cpu: enable gemm-bf16f32 for SDPA BF16

  #140159 commented on Apr 29, 2025 • 0 new comments

• Support LOAD_BUILD_CLASS opcode in dynamo

  #139561 commented on Apr 29, 2025 • 0 new comments

• Use the device interface for detecting Triton availability

  #139171 commented on May 1, 2025 • 0 new comments

• Fix bug of torch.nn.functional.kl_div when broadcast happened

  #138810 commented on Apr 27, 2025 • 0 new comments

• Add overflow check for negtive integer div_floor and div_trunc on CPU

  #138684 commented on Apr 30, 2025 • 0 new comments

• Extending SVE VEC Backend Support in PyTorch to SVE128 and SVE512.

  #138388 commented on Apr 29, 2025 • 0 new comments

• [pytree] add `treespec_{leaf,tuple,dict}` functions for args_spec modification

  #138214 commented on May 1, 2025 • 0 new comments

• Add TORCH_CHECK_INDEX in convert_indices_from_coo_to_csr_cpu

  #138068 commented on Apr 29, 2025 • 0 new comments

• [BE]: Try turning on LTO in CMake in CI

  #137866 commented on Apr 25, 2025 • 0 new comments

• Load cuda deps more aggressively

  #137059 commented on Apr 28, 2025 • 0 new comments

• Add back DistributedDataParallel types that were lost when pyi was removed

  #136835 commented on Apr 29, 2025 • 0 new comments

• [Don't Merge] Try to build custom ops with MKL XPU

  #133658 commented on Apr 29, 2025 • 0 new comments

• add ranking for grouped benchmarks

  #133287 commented on May 2, 2025 • 0 new comments

• [torch.special] Adding betainc, betaincc, betaincinv, betainccinv, betaln and beta with backward operation

  #132135 commented on Apr 29, 2025 • 0 new comments

• [1/N] Use 3.25.3 as the minimum CMake version

  #130522 commented on May 1, 2025 • 0 new comments

• [pytree] implement key path APIs for CXX pytree

  #130141 commented on Apr 29, 2025 • 0 new comments

• Remove direct dependency of protobuf from CMake

  #127919 commented on Apr 27, 2025 • 0 new comments

• [inductor] enable bf32 for mkldnn linear pointwise/binary in inductor

  #127294 commented on Apr 30, 2025 • 0 new comments

• [inductor] enable bf32 test for mkldnn conv

  #127293 commented on Apr 30, 2025 • 0 new comments

• allow to use bf16 as fp32 internal precision for mkldnn conv backward

  #126054 commented on Apr 30, 2025 • 0 new comments

• allow to use bf16 as fp32 internal precision for mkldnn conv

  #126050 commented on Apr 30, 2025 • 0 new comments

• Profiler doesn't seem to work on AMD CPUs

  #150052 commented on Apr 30, 2025 • 0 new comments

• [ROCm] sdpa group query attention bf16 numeric error

  #139352 commented on Apr 30, 2025 • 0 new comments

• [NJT] `.bmm`'s BmmBackward0 fails compilation when second arg requires grad

  #152122 commented on Apr 30, 2025 • 0 new comments

• [RFC] zentorch Integration

  #150296 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_remove_noop_slice1_cuda (__main__.GPUTests)

  #151381 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_remove_noop_slice1_cpu (__main__.CpuTests)

  #151379 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_remove_noop_slice_scatter_cuda (__main__.GPUTests)

  #151378 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_remove_noop_slice_scatter_cpu (__main__.CpuTests)

  #151382 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_bool (__main__.TestForeachCUDA)

  #150120 commented on Apr 30, 2025 • 0 new comments

• Implementation of a numerically stable log(1 - softmax) function in PyTorch

  #129657 commented on Apr 30, 2025 • 0 new comments

• Release artifacts for rc releases

  #124759 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_fake_registration (__main__.TestOpProfiles)

  #151301 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_remove_noop_slice_cuda (__main__.GPUTests)

  #151383 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_bfloat16 (__main__.TestForeachCUDA)

  #150119 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_matrix_rank_basic_cuda_float32 (__main__.TestLinalgCUDA)

  #150406 commented on Apr 30, 2025 • 0 new comments

• Grad strides do not match bucket view strides.

  #47163 commented on Apr 30, 2025 • 0 new comments

• [torch/elastic] unexpected behavior of torch elastic

  #147064 commented on Apr 30, 2025 • 0 new comments

• [ONNX] Create a message to suggest users setting dynamo=True when exporting

  #152025 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_parity__foreach_acos_fastpath_inplace_cuda_bfloat16 (__main__.TestForeachCUDA)

  #150902 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_aoti (__main__.TestMemoryPlanning)

  #145211 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_comprehensive_nansum_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #139710 commented on Apr 30, 2025 • 0 new comments

• Error when using torch.fx on bert

  #67970 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_parity__foreach_abs_fastpath_inplace_cuda_bfloat16 (__main__.TestForeachCUDA)

  #148966 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_repeated_calling_cuda (__main__.AOTInductorTestABICompatibleGpu)

  #146185 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_inductor_reuse_buffer_after_inplace_collective (__main__.CompileTest)

  #147950 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_vdd_clamp_dynamic_shapes_cpu (__main__.DynamicShapesCpuTests)

  #134445 commented on Apr 30, 2025 • 0 new comments

• [C10D] Make collectives backwards throw an error

  #152127 commented on Apr 30, 2025 • 0 new comments

• Numerical inaccuracies in "ddp_apply_optim_in_backward" unit tests for gloo backend

  #111834 commented on Apr 29, 2025 • 0 new comments

• DISABLED test_duplicate_registration_impl (__main__.TestOpProfiles)

  #151281 commented on Apr 29, 2025 • 0 new comments

• Is it possible to remove NCCL submodule and use only nccl binaries from pypi instead ?

  #144768 commented on Apr 29, 2025 • 0 new comments

• `version.txt` mismatch with tags in release branch

  #151425 commented on Apr 29, 2025 • 0 new comments

• Status of pip wheels with _GLIBCXX_USE_CXX11_ABI=1

  #51039 commented on May 1, 2025 • 0 new comments

• DTensor slicing on sharded dimension leads to replication

  #149447 commented on May 1, 2025 • 0 new comments

• [graph pickler] [inductor compile async] imprecise filter for non standard op?

  #151904 commented on May 1, 2025 • 0 new comments

• DISABLED test_parity__foreach_acos_fastpath_inplace_cuda_complex128 (__main__.TestForeachCUDA)

  #150933 commented on May 1, 2025 • 0 new comments

• DISABLED test_binary_op_with_scalar_self_support__foreach_pow_is_fastpath_True_cuda_complex128 (__main__.TestForeachCUDA)

  #149323 commented on May 1, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float32 (__main__.TestForeachCUDA)

  #150208 commented on May 1, 2025 • 0 new comments

• DISABLED test_remove_noop_view_default_cpu (__main__.CpuTests)

  #151512 commented on May 1, 2025 • 0 new comments

• RFC: The State of Custom CUDA extensions in PyTorch

  #152032 commented on May 1, 2025 • 0 new comments

• RuntimeError: "_amp_foreach_non_finite_check_and_unscale_cuda" not implemented for 'BFloat16'

  #127176 commented on May 1, 2025 • 0 new comments

• AOTI packaged model fails with generic error when run in for loop but succeeds on individual sample

  #146524 commented on May 1, 2025 • 0 new comments

• MPS operator coverage tracking issue (2.6+ version)

  #141287 commented on May 1, 2025 • 0 new comments

• DISABLED test_int64_upsample3d_cuda_bfloat16 (__main__.TestTorchDeviceTypeCUDA)

  #146007 commented on May 1, 2025 • 0 new comments

• DISABLED test_comprehensive_special_xlog1py_cuda_float32 (__main__.TestInductorOpInfoCUDA)

  #140648 commented on May 1, 2025 • 0 new comments

• Ability to do aot/inductor compilation from a jit model (or torch.exported model)

  #127928 commented on May 1, 2025 • 0 new comments

• General MPS op coverage tracking issue

  #77764 commented on May 1, 2025 • 0 new comments

• Memory Leak in MPS Backend During LSTM Iterations (Out of Memory Error)

  #145374 commented on May 1, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_float16 (__main__.TestForeachCUDA)

  #150173 commented on May 1, 2025 • 0 new comments

• [Async TP] all-gather-matuls not fusing properly when rowwise scales are used

  #149990 commented on May 1, 2025 • 0 new comments

• [Tracker] Nested tensor op coverage requests

  #118107 commented on May 1, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_complex64 (__main__.TestForeachCUDA)

  #150161 commented on May 1, 2025 • 0 new comments

• DISABLED test_comprehensive_nanmean_cuda_float64 (__main__.TestInductorOpInfoCUDA)

  #140339 commented on May 1, 2025 • 0 new comments

• DISABLED test_inductor_all_gather_into_tensor_coalesced (__main__.CompileTest)

  #146806 commented on May 1, 2025 • 0 new comments

• DISABLED test_remove_noop_slice_cpu (__main__.CpuTests)

  #151384 commented on May 1, 2025 • 0 new comments

• DISABLED test_is_isnot (__main__.TestScript)

  #120694 commented on May 1, 2025 • 0 new comments

• [CI] No workflows scheduled on PRs

  #151322 commented on Apr 30, 2025 • 0 new comments

• [inductor] Improve codegen for argmax+max

  #146643 commented on Apr 30, 2025 • 0 new comments

• DISABLED test_foreach_copy_with_multi_dtypes__foreach_copy_cuda_complex128 (__main__.TestForeachCUDA)

  #150141 commented on Apr 30, 2025 • 0 new comments

• missing docs for torch.Tag

  #126518 commented on Apr 30, 2025 • 0 new comments

• Quantile is limited to 16 million elements and have poor performance.

  #64947 commented on Apr 30, 2025 • 0 new comments

• enhance documentation around the developer build

  #108406 commented on Apr 30, 2025 • 0 new comments

• Training/Fine-tuning fails with PyTorch 2.8 + 4x 5090 GPUs using DDP/FSDP/DeepSpeed

  #150734 commented on Apr 30, 2025 • 0 new comments

• torch.compile on MPS progress tracker

  #150121 commented on Apr 30, 2025 • 0 new comments