group_key,title,collapse_kind,primary_bucket,attention_score,confidence,source,issue_count,active_pr_count,context_issue_count,context_pr_count,open_issue_numbers,active_pr_numbers,context_issue_numbers,context_pr_numbers,component_labels,labels,work_thread_keys,maintainer_time_saved,maintainer_next_action,evidence_count,coverage_note,top_issue_titles,group_hash
issue-collapse:3e1e22c0568b25dd,Review thread: 1 issue + 1 PR - pr #188801: [MPS] Leak MetalShaderLibrary bundled singleton to avoid exit-time destructor crashes,linked work thread,waiting on contributor,151,high,work_thread_graph,1,1,0,0,188812,188801,,,"module: crash, module: flaky-tests, module: macos, module: mps","bot-triaged, module: crash, module: flaky-tests, module: macos, module: mps, needs reproduction, triage review",thread:issue:188812:pr:188801,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188801; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188812 [MPS] MetalShaderLibrary::getBundledLibrary singleton can crash at process exit,06454fcca47637d37bc2a70816fea07e9e4235df83855c1340723cba787f6b74
issue-collapse:c338000f54583443,Review thread: 2 issues + 1 PR - pr #189122: Fix sparse-dense mul dropping data when broadcasting a size-1 sparse dim,linked work thread,ready for maintainer decision,139,high,work_thread_graph,2,1,0,0,"158861, 188900",189122,,,"module: correctness (silent), module: sparse","bot-triaged, high priority, module: correctness (silent), module: sparse, triage review, triaged",thread:issue:158861:issue:188900:pr:189122,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#158861 Different behavior between sparse and dense tensors with broadcasting multiplication., #188900 `sparse.mul`: broadcasting a size-1 dimension of the sparse operand silently drops data",754c6b6bbf39227f140eea6702dbc586700f1c978f3dc57331c762939cd27805
issue-collapse:f5a7ee822ea80de7,Review thread: 2 issues + 1 PR - pr #185730: Fix dynamic shapes for variadic kwargs,linked work thread,PR blocked,121,high,work_thread_graph,2,1,0,0,"150022, 150371",185730,,,"module: regression, oncall: export, oncall: pt2","export-triaged, module: regression, oncall: export, oncall: pt2",thread:issue:150022:issue:150371:pr:185730,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#150022 Dynamic Shapes with **kwargs, #150371 Dynamic shapes doesn't work with kwargs",654b4e3f63b28a9a784473ddf072b81ee4e55f900cd6cd49409183ea1d04a583
issue-collapse:26c2a09b2790af05,Review thread: 1 issue + 1 PR - pr #188948: Fix triangular_solve for sparse CPU tensors on non-MKL platforms,linked work thread,waiting on contributor,107,high,work_thread_graph,1,1,0,0,153410,188948,,,"oncall: export, oncall: pt2","export-triaged, good first issue, oncall: export, oncall: pt2",thread:issue:153410:pr:188948,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188948; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#153410 [Feature request] `torch.export` .save/.load could support `safetensors` and/or `weights_only=True`,9dde14fa28f3c9e3e68ace3d18e724a5a9be264f74d138f8c4fac3282d7bd751
issue-collapse:ab8b0a963ea4d873,Review thread: 3 issues + 2 PRs - pr #181720: [MPS] Make pin_memory return CPU-aliased storage backed by a unified MTLBuffer,linked work thread,has linked issue,106,high,work_thread_graph,3,2,0,0,"180397, 181374, 188970","181720, 189256",,,"module: crash, module: inductor, module: macos, module: mps, module: performance","bot-triaged, feature, module: crash, module: inductor, module: macos, module: mps, module: performance, needs reproduction, triaged",thread:issue:180397:issue:181374:issue:188970:pr:181720,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#180397 [MPS] Add graph capture/replay API to eliminate per-op CPU dispatch overhead, #181374 `torch.empty((1,1), device=""cpu"", pin_memory=True).device` gives back MPS device on mac, #188970 [MPS] SIGSEGV in MPSStream::copy with triton.cudagraphs (reduce-overhead/max-autotune) on deep fp16 graphs",89fc35f84cfbfbe51890bbc1e5816814f943bd199a94e048232308bcd856260e
issue-collapse:e37d3d3231bd24d3,Review thread: 1 issue + 1 PR - pr #189043: Preload full bundled cuDNN set with RTLD_GLOBAL to prevent sublibrary version mismatch,linked work thread,waiting on contributor,106,high,work_thread_graph,1,1,0,0,188892,189043,,,"module: binaries, module: convolution, module: crash, module: cudnn","bot-triaged, has workaround, module: binaries, module: convolution, module: crash, module: cudnn, needs reproduction, triaged",thread:issue:188892:pr:189043,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #189043; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188892 nvidia-cudnn-cu13==9.20.0.48 bundled by torch is incomplete → CUDNN_STATUS_SUBLIBRARY_VERSION_MISMATCH,ac746ec42adb60385f5b0f28a7935ea6dedac4c14ebc3034cebbe95c4010a45e
issue-collapse:f180143ef37a413c,Review thread: 21 issues + 14 PRs - pr #181726: [xpu][1/4]Implement scaled_mm_v2 for MXFP8/MXFP4/NVFP4 on XPU,linked work thread,PR blocked,94,high,work_thread_graph,21,14,0,0,"178040, 183988, 186348, 187988, 188477, 188675, 188704, 188706, 188707, 188708, 188709, 188710, 188711, 188712, 188713, 188714, 188715, 188716, 188717, 188721, 188891","181726, 181727, 181728, 183511, 184276, 187315, 187318, 187989, 188268, 188931, 188996, 189149, 189186, 189297",,,"module: advanced indexing, module: arm, module: autograd, module: ci, module: crash, module: cuda, module: decompositions, module: floatx (formerly float8), module: functorch, module: inductor, module: linear algebra, module: macos, module: mps, module: nn, module: performance, module: tests, oncall: pt2","bot-triaged, enhancement, high priority, matrix multiplication, module: advanced indexing, module: arm, module: autograd, module: ci, module: crash, module: cuda, module: decompositions, module: floatx (formerly float8), module: functorch, module: inductor, module: linear algebra, module: macos, module: mps, module: nn, module: performance, module: tests, oncall: pt2, repro:agent:h100:repro_success:reproduced, skipped, topic: fuzzer, triaged, unstable",thread:issue:178040:issue:183988:issue:186348:issue:187988,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 21 open issues; raw issue rows remain available for audit.,"#178040 `torch.compile` raises RuntimeError on valid `torch.addmm` with shape mismatch where eager succeeds, #183988 [inductor] `torch.compile` crashes on `x[torch.randperm(n)]` — pattern matcher missing `slice_shape` input, #186348 torch.mm slower than pointwise mul+sum when K and N are small (under 8), #187988 [MPS] linear backward crashes (SIGABRT) with a 1-D weight: 'mps.matmul' op contracting dimensions differ, #188477 [CI][B200] smoke_b200 test_nv_universal_gemm fails: cutlass.cute.arch has no attribute 'ProxyKind' (unpinned cutlass_api vs pinned cutlass-dsl), #188675 DISABLED test_main_loop_scaling_shape2_use_fast_accum_True_scaling_block_sizes2_cuda (__main__.TestFP8LoweringCUDA)",57ad558f85f674703d75df9de913c43b82d17c11dd1d40ef812126c3979f1d12
issue-collapse:fa12328e61a1943f,Review thread: 3 issues + 3 PRs - pr #184481: Guard synthetic-base input alias layouts,linked work thread,waiting on contributor,94,high,work_thread_graph,3,3,0,0,"93617, 178680, 188133","184481, 184694, 185891",,,"module: aotdispatch, module: inductor, module: pt2-dispatcher, oncall: cpu inductor, oncall: pt2","bot-triaged, feature, high priority, internal ramp-up task, module: aotdispatch, module: inductor, module: pt2-dispatcher, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:178680:issue:188133:issue:93617:pr:184481,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #184481, #184694; keep related issue/PR context attached.",6,Represents 3 open issues; raw issue rows remain available for audit.,"#93617 [Inductor] Input Buffers Should Be Representable As Storage And Layout, #178680 torch.compile (inductor/aot_eager) produces nondeterministic / incorrect results on model with `add_`, `clamp_`, and `squeeze`, #188133 Pallas CPU mishandles synthetic-base mutated aliased inputs",93167f6519c6158351d4c28610b358ade4e8ae41b423a4a48a91f32ff139c137
issue-collapse:79173b600fb13f89,Review thread: 3 issues + 2 PRs - pr #185946: [CUDA] Warn instead of assert in `reportProcessMemoryInfo` to prevent crash on Tegra OOM,linked work thread,PR blocked,94,high,work_thread_graph,3,2,0,0,"16706, 185240, 186374","185946, 186375",,,"module: arm, module: crash, module: cuda, module: cudacachingallocator, module: error checking, module: jetson, module: random, module: tests","bot-triaged, module: CUDACachingAllocator, module: arm, module: crash, module: cuda, module: error checking, module: jetson, module: random, module: tests, triaged",thread:issue:16706:issue:185240:issue:186374:pr:185946,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#16706 Very poor Uniform() sampling near floating 0.0, #185240 CUDACachingAllocator.cpp:995 NVML internal assert fires instead of clean OOMError when cudaMalloc fails on Tegra (Jetson), #186374 [CUDA] `TestCudaAllocator.test_garbage_collect_expandable` failing on Orin",76e327e40cb98b5306cef725780d92f2c8a3f251afcadb9f1595535596fe14ff
issue-collapse:6377c52ccbcd44d0,Review thread: 2 issues + 3 PRs - pr #185846: [decomp] Match eager's device-dependent threshold comparison dtype,linked work thread,PR blocked,94,high,work_thread_graph,2,3,0,0,"185470, 185484","185846, 186358, 187609",,,"module: bfloat16, module: correctness (silent), module: decompositions, module: inductor, module: type promotion, oncall: pt2","bot-triaged, high priority, module: bfloat16, module: correctness (silent), module: decompositions, module: inductor, module: type promotion, oncall: pt2, triaged",thread:issue:185470:issue:185484:pr:185846:pr:186358,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#185470 [Inductor] `F.threshold` bf16 miscompilation: scalar threshold not cast to tensor dtype, #185484 [Inductor] Forward output mismatch between Eager and Inductor for F.softshrink with bfloat16 and float scalar",7e37f4c16f42fc81e564661489a662af0b18b2723a09861fb47ac31ebf42d34b
issue-collapse:c3688c4d5bd64629,Review thread: 3 issues + 1 PR - pr #189045: Fix NaN gradients in nn.MultiheadAttention for fully-masked rows (#41…,linked work thread,has linked issue,94,high,work_thread_graph,3,1,0,0,"40932, 41508, 55056",189045,,,"module: nans and infs, module: nn","enhancement, high priority, module: NaNs and Infs, module: nn, triaged",thread:issue:40932:issue:41508:issue:55056:pr:189045,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#40932 MultiheadAttention set(-inf) cause 'Nan' in loss computing, #41508 nn.MultiheadAttention causes gradients to become NaN under some use cases, #55056 [FR] Safe softmax",ebe21f8d5db35671f2f89c2821c13d3ca790853568d55468d6668f0d0616a596
issue-collapse:17d199035ead3b69,Review thread: 3 issues + 1 PR - pr #188998: [inductor] Fix bf16 matmul accumulation precision in Triton templates (#188492),linked work thread,PR blocked,94,high,work_thread_graph,3,1,0,0,"188492, 188841, 189287",188998,,,"module: bfloat16, module: ci, module: correctness (silent), module: cuda, module: dependency bug, module: floatx (formerly float8), module: inductor, module: numerical-stability, module: pt2 accuracy, module: regression, module: tests, module: vllm, oncall: pt2","bot-triaged, high priority, module: bfloat16, module: ci, module: correctness (silent), module: cuda, module: dependency bug, module: floatx (formerly float8), module: inductor, module: numerical-stability, module: pt2 accuracy, module: regression, module: tests, module: vllm, oncall: pt2, triaged, upstream triton",thread:issue:188492:issue:188841:issue:189287:pr:188998,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 3 open issues; raw issue rows remain available for audit.,"#188492 swin_base_patch4_window7_224 fails inductor accuracy (bf16 inference, H100) after Triton 3.8 pin bump, #188841 triton 3.8 pin update regressing compile + float8 training numerics, #189287 [vllm] vllm-x-pytorch-test multimodal pooling tests fail with EngineCore ""CUDA driver initialization failed"" since Triton 3.8 bump (#188251)",66a49382c2123953dace49ae850f2803ac533e5563dd6c42a01501799c0941d1
issue-collapse:8feecc6cd94aa812,Review thread: 3 issues + 1 PR - pr #184621: Fix Dynamo torch.Size from tensor shape inputs,linked work thread,waiting on contributor,94,high,work_thread_graph,3,1,0,0,"182649, 182651, 182652",184621,,,"module: crash, module: dynamo, module: faketensor, module: pt2-dispatcher, module: viewing and reshaping, oncall: pt2","bot-triaged, module: crash, module: dynamo, module: fakeTensor, module: pt2-dispatcher, module: viewing and reshaping, oncall: pt2, triaged",thread:issue:182649:issue:182651:issue:182652:pr:184621,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184621; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#182649 torch.compile Inductor crashes on onnx2torch-converted Add-Sub-Reshape model with FakeTensor shape input, #182651 torch.compile Inductor crashes on onnx2torch-converted ASPP-style Conv-Add-Reshape model with FakeTensor shape input, #182652 torch.compile Inductor crashes on onnx2torch-converted Conv-Reshape model with FakeTensor shape input",44347836999f44d69595bb06c23f67de7ea8d895ddf93292b22e271f2013b705
issue-collapse:f065e866ec34cec0,Review thread: 2 issues + 2 PRs - pr #189294: [CI] Realign scipy after nvmath install to fix H100 smoke NumPy ABI crash,linked work thread,PR blocked,94,high,work_thread_graph,2,2,0,0,"173059, 189034","189294, 189295",,,"module: ci, module: regression","bot-triaged, high priority, module: ci, module: regression, triaged",thread:issue:173059:issue:189034:pr:189294:pr:189295,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#173059 pandas/numpy binary incompatibility breaks test_dataloader and test_datapipe, #189034 Limited CI on H100 (smoke) red since 2026-07-03: test_foreach import crash from NumPy 2.x / SciPy ABI mismatch",05842e7c445cd980a793b6af7944ba6fb0a8269ac443f24f100118cacabdf945
issue-collapse:473412795e13bfad,Review thread: 2 issues + 2 PRs - pr #187354: Match Inductor i0/i1 infinities to eager CUDA,linked work thread,ready for maintainer decision,94,high,work_thread_graph,2,2,0,0,"187332, 188545","187354, 188556",,,"module: correctness (silent), module: cuda, module: decompositions, module: edge cases, module: nans and infs, module: special, oncall: pt2","bot-triaged, high priority, module: NaNs and Infs, module: correctness (silent), module: cuda, module: decompositions, module: edge cases, module: special, oncall: pt2, triaged",thread:issue:187332:issue:188545:pr:187354:pr:188556,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#187332 [PT2] Output mismatch between eager mode and torch.compile for torch.special.i0 with inf inputs on CUDA, #188545 [PT2] Output mismatch between eager and torch.compile for torch.special.bessel_y0 with +inf input on CUDA",3531c39d18d85f4a1e7786774e950c42b22c287d2e757b733d44dbe63ad797fc
issue-collapse:2af31c3345225c97,Review thread: 2 issues + 2 PRs - pr #185756: [clamp] Fix float16 scalar overflow check inconsistency between CPU and GPU,linked work thread,waiting on contributor,94,high,work_thread_graph,2,2,0,0,"171356, 187429","185756, 187908",,,"module: correctness (silent), module: edge cases, module: error checking, module: half, module: regression","bot-triaged, module: correctness (silent), module: edge cases, module: error checking, module: half, module: regression, topic: fuzzer, triaged",thread:issue:171356:issue:187429:pr:185756:pr:187908,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #187908; keep related issue/PR context attached.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#171356 `torch.clip` has checks for float16 scalar overflow on CPU but not on GPU, #187429 torch.where silently casts out-of-range Python float to +/-Inf in fp16 (inconsistent with full_like / full / fill_, which raise)",5655ff8713e9ed44f3ab7ddb77029d33d1b766d9d54327e1d7df827582a0415a
issue-collapse:1816febbaf15b36b,Review thread: 2 issues + 1 PR - pr #185496: Fix export autograd.grad saved tensor tracing,linked work thread,waiting on contributor,94,high,work_thread_graph,2,1,0,0,"146719, 155044",185496,,,"oncall: export, oncall: pt2","PT2-Bug-Bash, high priority, oncall: export, oncall: pt2, triaged",thread:issue:146719:issue:155044:pr:185496,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185496; keep related issue/PR context attached.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#146719 Operations that fail under `torch.export.export`(`torch.autograd.grad`) -> `torch.compile`, #155044 Execution of an `ExportProgram` of a model with `torch.autograd.grad(torch.sqrt(x), x, torch.ones_like(torch.sqrt(x)))` returns a `FakeTensor` instead of a real tensor",632e37bd27c9702533bc3daf8b9c8f2a3bdb4179e9824f51fe928e1bd40d801b
issue-collapse:34f2bc0ac321c982,Review thread: 2 issues + 1 PR - pr #184632: Fix CUDA fake strides for mixed dtype pointwise ops,linked work thread,waiting on contributor,94,high,work_thread_graph,2,1,0,0,"182200, 184101",184632,,,"module: aotdispatch, module: complex, module: crash, module: fft, module: inductor, module: meta tensors, module: pt2-dispatcher, oncall: pt2","bot-triaged, has workaround, module: aotdispatch, module: complex, module: crash, module: fft, module: inductor, module: meta tensors, module: pt2-dispatcher, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:182200:issue:184101:pr:184632,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184632; keep related issue/PR context attached.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#182200 `torch.compile` crashes on `stft`/`fftn`/`rfftn`/`ifftn` backward — meta kernel stride mismatch, #184101 `torch.compile` backward crashes for `abs()`/`angle()` on transposed complex tensors",ed311272de891a671e25668d1538bdf6f5024ded02e5b52531743be37908899e
issue-collapse:0645e2e2e4135752,Review thread: 2 issues + 1 PR - pr #184628: Fix SingletonInt static guard evaluation,linked work thread,PR blocked,94,high,work_thread_graph,2,1,0,0,"182217, 183369",184628,,,"module: crash, module: dynamic shapes, module: guards, module: nestedtensor, module: norms and normalization, oncall: pt2","bot-triaged, has workaround, module: crash, module: dynamic shapes, module: guards, module: nestedtensor, module: norms and normalization, oncall: pt2, triaged",thread:issue:182217:issue:183369:pr:184628,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#182217 `torch.compile` crashes on `F.layer_norm(NestedTensor)` — `InternalTorchDynamoError` in guard building, #183369 torch.compile crashes on nested_tensor_from_jagged + layer_norm with InternalTorchDynamoError",c8b7946edf82ed67370d211aa8ef115679f94eff06ae25870b3390430566c606
issue-collapse:05df422a9424e820,Review thread: 2 issues + 1 PR - pr #184392: Fix ModularIndexing printer semantics,linked work thread,PR blocked,94,high,work_thread_graph,2,1,0,0,"119883, 187027",184392,,,"module: correctness (silent), module: dynamic shapes, module: inductor, oncall: pt2","bot-triaged, high priority, module: correctness (silent), module: dynamic shapes, module: inductor, oncall: pt2, triaged",thread:issue:119883:issue:187027:pr:184392,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#119883 ModularIndexing does not have well-defined semantics, #187027 flip after repeat_interleave generates negative (out-of-bounds) index under dynamic shapes",ca449bc48679cd990b2766399dc97bff222121ca53aec1e586cb7e215221ff71
issue-collapse:408e6afa7c46b2af,Review thread: 2 issues + 1 PR - pr #183654: Fix Inductor CPU group norm affine precision,linked work thread,PR blocked,94,high,work_thread_graph,2,1,0,0,"181696, 183120",183654,,,"module: aotdispatch, module: correctness (silent), module: decompositions, module: inductor, module: norms and normalization, module: pt2-dispatcher, oncall: pt2","bot-triaged, high priority, module: aotdispatch, module: correctness (silent), module: decompositions, module: inductor, module: norms and normalization, module: pt2-dispatcher, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:181696:issue:183120:pr:183654,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#181696 Eval-mode `BatchNorm`/`GroupNorm` produces wrong output under `torch.compile(backend='inductor')`, #183120 torch.compile Inductor produces wrong output for BatchNorm2d + ELU + GroupNorm + log clamp chain",d25de1452596bbcc3009b2499b568752f2a3da1caafb1976bb7b3e5d29059c5d
issue-collapse:258c52f4333ccf3b,Review thread: 1 issue + 1 PR - pr #189273: Fix nn.functional.threshold OpInfo reference for strict NumPy uint8 casting,linked work thread,draft/noise,94,high,work_thread_graph,1,1,0,0,189267,189273,,,"module: nn, module: numpy, module: regression, module: tests, module: unsigned int","bot-triaged, module: nn, module: numpy, module: regression, module: tests, module: unsigned int, triaged",thread:issue:189267:pr:189273,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#189267 nn.functional.threshold OpInfo reference raises OverflowError on unsigned dtypes with NumPy >= 2.5,44e42ed015de77241bf54d9b325aaee37906857a7d54c5ed8b6fe989201c8fd3
issue-collapse:e162f6dbd466d9e4,Review thread: 1 issue + 1 PR - pr #189233: [MPS] Fix avg_pool2d/adaptive_avg_pool2d backward on channels_last inputs,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,175190,189233,,,"module: autograd, module: crash, module: memory format, module: mps, module: nn, module: pooling","bot-triaged, module: autograd, module: crash, module: memory format, module: mps, module: nn, module: pooling, triaged",thread:issue:175190:pr:189233,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#175190 [MPS] AvgPool2d/AdaptiveAvgPool2d backward crashes (SIGABRT) on channels_last inputs,f9534ac2ded6e2572c179067fc91c0bb7c0e29e70479d074fce9a471c1f3af93
issue-collapse:4c020c1b0e0e3a2f,Review thread: 1 issue + 1 PR - pr #189129: inductor: Fix non-deterministic gradients in slice_scatter backward fusion,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,188890,189129,,,"module: correctness (silent), module: determinism, module: regression, oncall: pt2","bot-triaged, module: correctness (silent), module: determinism, module: regression, oncall: pt2, triaged",thread:issue:188890:pr:189129,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188890 Non-deterministic gradients when using torch.compile and PyTorch >= 2.12,0731883f79fed1dba5f09e47c0261f9e6993d2efb9736e0e0716203fe402a503
issue-collapse:41d7088d8ea079f9,Review thread: 1 issue + 1 PR - pr #188791: Fix MKL VML first-call dispatch race producing silent low-precision results,linked work thread,has linked issue,94,high,work_thread_graph,1,1,0,0,188792,188791,,,"module: correctness (silent), module: cpu, module: intel, module: mkl, module: multithreading","bot-triaged, high priority, module: correctness (silent), module: cpu, module: intel, module: mkl, module: multithreading, triaged",thread:issue:188792:pr:188791,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188792 CPU float unary ops can silently return low-precision results: MKL VML first-call kernel dispatch race,d1285917d82bdffc3cb78a664d78e314fc777b0a80c8ac092f5a9ebc15be3e6d
issue-collapse:271132f2bbf18586,Review thread: 1 issue + 1 PR - pr #188605: Fix infinite recursion with pybind11 enum conversions,linked work thread,waiting on contributor,94,high,work_thread_graph,1,1,0,0,188547,188605,,,"module: crash, module: dynamo, module: pybind, oncall: pt2","bot-triaged, module: crash, module: dynamo, module: pybind, oncall: pt2, triaged",thread:issue:188547:pr:188605,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188605; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#188547 [dynamo] Infinite recursion when returning a Pybind11 enum from a compiled function,e7a21a2862a10e28c58ce545379bf049163f54b7d8515e1b338653001b1b59f4
issue-collapse:73d73dc317707e26,Review thread: 1 issue + 1 PR - pr #188546: [Inductor] Fix log_sigmoid_forward flushing float32 subnormals to +0.0 on CUDA,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,188541,188546,,,"module: correctness (silent), module: cuda, module: edge cases, module: inductor, module: nn, oncall: pt2","bot-triaged, high priority, module: correctness (silent), module: cuda, module: edge cases, module: inductor, module: nn, oncall: pt2, triaged",thread:issue:188541:pr:188546,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#188541 [Inductor][CUDA] F.logsigmoid flushes negative float32 subnormal outputs to +0.0, changing sign-sensitive behavior",6a015a1399d6db0710a8032af4a5e7a6982c46ce450bb7b352e2da0d67b69b3c
issue-collapse:15536c701a487e22,Review thread: 1 issue + 1 PR - pr #188114: Fix CK SDPA flash attention on gfx1200/gfx1201 (RDNA4),linked work thread,ready for maintainer decision,94,high,work_thread_graph,1,1,0,0,188113,188114,,,"module: crash, module: rocm, module: sdpa","bot-triaged, enhancement, module: crash, module: rocm, module: sdpa, triaged",thread:issue:188113:pr:188114,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#188113 CK SDPA flash attention kernel asserts on gfx1200/gfx1201 (RDNA4),7797b2bb88604f9000696504f1bf466f6fbac22a6d2817c641e6e9d77b9107fe
issue-collapse:75879e5f08b58785,Review thread: 1 issue + 1 PR - pr #188047: fix(dlpack): raise Python RuntimeError instead of aborting for negative DLPack strides,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,188023,188047,,,"module: crash, module: dlpack, module: error checking, module: numpy","bot-triaged, module: crash, module: dlpack, module: error checking, module: numpy, triaged",thread:issue:188023:pr:188047,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#188023 torch.from_dlpack aborts the Python process for NumPy arrays with negative strides,38a2de9ca26ff544e3a119645f6e0abd7c507ecf0afd46d98b982c250f088aca
issue-collapse:2d9a5763d10375ad,Review thread: 1 issue + 1 PR - pr #188039: autograd: fix PyObject* refcount leak in setup_context path,linked work thread,has linked issue,94,high,work_thread_graph,1,1,0,0,188034,188039,,,"module: autograd, module: memory usage, module: pybind","bot-triaged, high priority, module: autograd, module: memory usage, module: pybind, triaged",thread:issue:188034:pr:188039,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188034 [BUG] Reference leak in autograd Function.apply — setup_context return value never released,7558f5c22491e655992e9ebfda918ec10d1c5e7c249785c5e6099f52de84f658
issue-collapse:c0537dcb98a5963f,Review thread: 1 issue + 1 PR - pr #187955: Make expandable-segment Handle a move-only RAII owner and exception-safe,linked work thread,waiting on contributor,94,high,work_thread_graph,1,1,0,0,188008,187955,,,"module: crash, module: cuda, module: cudacachingallocator","bot-triaged, module: CUDACachingAllocator, module: crash, module: cuda, triaged",thread:issue:188008:pr:187955,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #187955; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188008 expandable_segments: partial mapAndSetAccess failure → bulk cuMemUnmap of unmapped range → terminate in destructor,af772d63c1d2d82cda79ff345b3cd52f16999a3ec76e4916300937de3ed01072
issue-collapse:56b4f028469da4cb,Review thread: 1 issue + 1 PR - pr #187493: Fix Inductor nextafter for low precision CUDA tensors,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,187340,187493,,,"module: correctness (silent), module: inductor, oncall: pt2","high priority, module: correctness (silent), module: inductor, oncall: pt2, triaged",thread:issue:187340:pr:187493,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187340 [torch.compile][Inductor][CUDA] torch.nextafter is incorrectly compiled as no-op for non-contiguous float16 tensors,a80ca489823538674750ccc838e68830351156dc7d88a74a7896f3a6caea21ff
issue-collapse:d9d327996457ad16,Review thread: 1 issue + 1 PR - pr #187486: Fix unbacked memo reuse during fake replay,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,187415,187486,,,"module: dynamic shapes, module: inductor, oncall: pt2","bot-triaged, high priority, module: dynamic shapes, module: inductor, oncall: pt2, triaged",thread:issue:187415:pr:187486,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187415 [Inductor] torch.compile fails on CUDA with bool mask indexing after torch.eye constructor movement: AssertionError in _rename_unbacked_to,baeb96ad4107e904a5ccc280d7128e0624911e6faf83c8e72b5b58446bc2eecf
issue-collapse:4cb82ac2237d516b,Review thread: 1 issue + 1 PR - pr #187311: Preserve forward AD tangents for Inductor dual inputs,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,187284,187311,,,"module: aotdispatch, module: correctness (silent), module: dynamo, module: forward ad, module: inductor, module: pt2-dispatcher, oncall: pt2","bot-triaged, high priority, module: aotdispatch, module: correctness (silent), module: dynamo, module: forward ad, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:187284:pr:187311,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187284 `torch.compile` with `inductor` backend silently drops forward-mode AD tangents (returns `None`),bc699f0618e667631b4c4cad14ecc25040bb679700cac272c0a4b084f69debc0
issue-collapse:658fc6aadaa0cb17,Review thread: 1 issue + 1 PR - pr #187104: Fix embedding_bag CPU segfault when offsets is empty and indices is not,linked work thread,has linked issue,94,high,work_thread_graph,1,1,0,0,175370,187104,,,"module: cpu, module: crash, module: embedding, module: empty tensor","bot-triaged, module: cpu, module: crash, module: embedding, module: empty tensor, topic: fuzzer, triaged",thread:issue:175370:pr:187104,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#175370 `F.embedding_bag` segfaults with float64 weight and empty offsets,3bd238d406b8e1252c37649749c55f95e84babd490900c75b7646986b913e537
issue-collapse:b41ebe7c293f990b,Review thread: 1 issue + 1 PR - pr #186622: Fix bf16 autocast precision barriers in Inductor,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,168126,186622,,,"module: aotdispatch, module: decompositions, module: inductor, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, high priority, module: aotdispatch, module: decompositions, module: inductor, module: pt2-dispatcher, oncall: pt2, pt2: numerics-discrepancy, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:168126:pr:186622,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#168126 Large numeric divergence for torch compile vs eager in bf16,10f904968e5ef477b48ef278212eb8b6e2634d00759e88c0dd22ee6221f139e2
issue-collapse:d9a09b43b8da85a2,Review thread: 1 issue + 1 PR - pr #186583: Propagate contextvars to Python autograd functions,linked work thread,ready for maintainer decision,94,high,work_thread_graph,1,1,0,0,186537,186583,,,"module: bc-breaking, module: dynamo, module: multithreading, module: regression, oncall: pt2","bot-triaged, module: bc-breaking, module: dynamo, module: multithreading, module: regression, oncall: pt2, triaged",thread:issue:186537:pr:186583,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#186537 torch 2.12 regression: torch.compile thread-local config - even backward thread has separate config,e0a85e43c4fcdce1b677d4fcaaacad4a92fcf61d4363e091dd8b11a4b7b52856
issue-collapse:a1419144e5fd8c4f,Review thread: 1 issue + 1 PR - pr #186454: Warn on unsupported custom op schema types,linked work thread,ready for maintainer decision,94,high,work_thread_graph,1,1,0,0,124863,186454,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, better-engineering, high priority, module: custom-operators, module: pt2-dispatcher, months, oncall: pt2, triaged",thread:issue:124863:pr:186454,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124863 Deprecate unsupported types in operator registration,3de2ead0951368696d24b38db787345c9c716a97e40fc2df6d83607bb198e139
issue-collapse:07fad9a68d7a7490,Review thread: 1 issue + 1 PR - pr #186395: Preserve overlap deps without blocking fusion,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,186084,186395,,,"module: inductor, oncall: pt2","high priority, module: inductor, oncall: pt2, triaged",thread:issue:186084:pr:186395,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#186084 [inductor] insert_overlap_deps wraps non-overlap-critical compute with control_deps, blocking fusion and causing silent bf16 accuracy regression",a21e4b4f908ea66cd25c8253275a9770cf8fca4dd28d0089fef10dd4f1c3b5a4
issue-collapse:e1042777b2d21163,Review thread: 1 issue + 1 PR - pr #186252: [inductor] Fix truncdiv off-by-one on CUDA with _div_rn,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,184408,186252,,,"module: inductor, oncall: pt2","high priority, module: inductor, oncall: pt2, triaged",thread:issue:184408:pr:186252,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#184408 `torch.div(rounding_mode='trunc')` produces off-by-one results under `torch.compile` on CUDA,6d8cb807131554420e1efc8915af676d2bd043295b15c839b363ca152bf61a0b
issue-collapse:02b4ca25889244ed,Review thread: 1 issue + 1 PR - pr #185349: Fix flex decoding BLOCK_M validation for short GQA,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,160018,185349,,,"module: crash, module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: crash, module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:160018:pr:185349,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160018 Triton kernel generated for Flex Attention throws InductorError: RuntimeError: No valid triton configs when seq length < 128,0e6afe3e2422bf710d81a9fe4d4e9a37f40a6810e1186f200f8c6b1cb683a501
issue-collapse:97fd8ff05e7a7e43,Review thread: 1 issue + 1 PR - pr #184970: Preserve pin_memory in generated view copy kernels,linked work thread,waiting on contributor,94,high,work_thread_graph,1,1,0,0,171894,184970,,,"module: cuda graphs, module: dynamo, oncall: pt2","PT2-Bug-Bash, high priority, module: cuda graphs, module: dynamo, oncall: pt2, triaged",thread:issue:171894:pr:184970,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184970; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#171894 `aot_eager` backend does not support cpu pinned memory and device copy in cudagraph,2cf9f172ad8438b0b9bcb6ec7e6dcce3971595c90a6df31302cbf1eb403fd5a9
issue-collapse:2b11dddfd86f4089,Review thread: 1 issue + 1 PR - pr #184848: Fix nearest upsample scale shortcuts in compile,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,175154,184848,,,"module: decompositions, module: dispatch, oncall: pt2","PT2-Bug-Bash, high priority, module: decompositions, module: dispatch, oncall: pt2, triaged",thread:issue:175154:pr:184848,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175154 [inductor/aot_eager] interpolate under torch.compile produces incorrect result when mode is `nearest`,8e8c60027b589076d7db6e5b310c65d25a9f50dda3a7387bc73d0fabb026c67c
issue-collapse:c649134baeff4437,Review thread: 1 issue + 1 PR - pr #184153: [Inductor] Clone cudagraph-owned grads before replay,linked work thread,waiting on contributor,94,high,work_thread_graph,1,1,0,0,146569,184153,,,"module: inductor, module: regression, oncall: pt2","module: inductor, module: regression, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:146569:pr:184153,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184153; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146569 torch.compile with fullgraph=True causes overwritten variable error in versions later than torch==2.5.1,115aeefb371ef279fe14e6742cd759d8c70a88cfd7162547f37fbb3d624d83c2
issue-collapse:91d07b3ae788b03f,Review thread: 1 issue + 1 PR - pr #183881: Fix online softmax epilogue fusion for T5,linked work thread,waiting on contributor,94,high,work_thread_graph,1,1,0,0,177118,183881,,,"module: cuda, module: inductor, module: performance, module: regression, oncall: pt2","bot-triaged, module: cuda, module: inductor, module: performance, module: regression, oncall: pt2, triaged",thread:issue:177118:pr:183881,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #183881; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#177118 [Inductor]T5 models performance dropped 10% on inductor mode after pattern match changes,707610a2065eb6a84370041ccc2a07d007a0d3b050709b496e2f215b9bd2ce29
issue-collapse:a5ba5aa0259177e8,Review thread: 1 issue + 1 PR - pr #183701: Fix CPU vectorized reduction tail stores,linked work thread,ready for maintainer decision,94,high,work_thread_graph,1,1,0,0,181690,183701,,,"module: crash, module: inductor, module: rnn, oncall: pt2","bot-triaged, module: crash, module: inductor, module: rnn, oncall: pt2, repro:agent:h100:repro_success:reproduced, repro:agent:repro_success:reproduced, triaged",thread:issue:181690:pr:183701,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#181690 Heap corruption when compiling models with `nn.LSTM` or `nn.GRU` using `torch.compile(backend='inductor')`,288800b5000b6e960b4b33aba8f522eb8fdf89e4786e665a32c46276f3fb32d4
issue-collapse:af3ec23c139b7535,Review thread: 1 issue + 1 PR - pr #181516: Fix heap-buffer-overflow in torch.linalg.eig on subnormal real inputs,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,162358,181516,,,"module: crash, module: error checking, module: linear algebra, module: sanitizers","module: crash, module: error checking, module: linear algebra, module: sanitizers, topic: fuzzer, triaged",thread:issue:162358:pr:181516,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162358 heap-buffer-overflow in torch.linalg.eig,48e68cb1ae3a8a204c48593e9ca6069e9b3c4943680bbb92a7daf7f8d1550ebf
issue-collapse:ddbaefadfbfb9222,Review thread: 1 issue + 1 PR - pr #181036: Clamp spdiags diagonal length to zero for out-of-range offsets,linked work thread,ready for maintainer decision,94,high,work_thread_graph,1,1,0,0,178089,181036,,,"module: crash, module: error checking, module: sparse","bot-mislabeled, bot-triaged, module: crash, module: error checking, module: sparse, triaged",thread:issue:178089:pr:181036,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#178089 bad-free in torch.sparse.spdiags due to missing bounds check on diagonal offsets,ae0dcca6fc5b4d9d5e272cb5493eff0e76bd29ee4fa4a354bcb47afe7a963b9f
issue-collapse:560f11f957271c7e,Review thread: 1 issue + 1 PR - pr #176234: [autograd] Explicitly error when layer_norm computes 3rd order derivatives,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,186256,176234,,,"module: autograd, module: correctness (silent), module: derivatives, module: double backwards, module: norms and normalization","actionable, bot-triaged, high priority, module: autograd, module: correctness (silent), module: derivatives, module: double backwards, module: norms and normalization, triaged",thread:issue:186256:pr:176234,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#186256 `native_batch_norm` silently produces incorrect 3rd+ order derivatives in training mode with `create_graph=True`,3d99fdcb53803c77e0794f326cfa20b1963ada06774f54ce9ed728a9b76345ea
issue-collapse:f0ea32cd359097e0,Review thread: 1 issue + 1 PR - pr #176070: [rnn] Prevent LSTMCell segfault on input/hidden shape mismatch,linked work thread,PR blocked,94,high,work_thread_graph,1,1,0,0,175978,176070,,,"module: crash, module: error checking, module: rnn","actionable, bot-mislabeled, bot-triaged, module: crash, module: error checking, module: rnn, triaged",thread:issue:175978:pr:176070,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175978 The following code snippet causes a segmentation fault when running torch.nn.lstm_cell:,8b39d8d58d3df526d8ab8eb8a8ed6bbba4ff1f7a3408fa77590fdf1e6c48ab93
issue-collapse:84c0fb59bae58aa1,Review thread: 3 issues + 1 PR - pr #184983: Make SDPA backend checks safe for unbacked head dims,linked work thread,ready for maintainer decision,89,high,work_thread_graph,3,1,0,0,"124502, 170127, 174608",184983,,,"module: dynamic shapes, oncall: export, oncall: pt2","PT2-Bug-Bash, module: dynamic shapes, oncall: export, oncall: pt2, onnx-needs-info",thread:issue:124502:issue:170127:issue:174608:pr:184983,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#124502 [export] Llama3 export with dynamic shapes fails with constraint violations, #170127 Could not gaurd on data-dependent expression for SDPA, #174608 AOTInductor compile crashes with CUDA illegal memory access on exported SAM3 model with math SDPA",99ebba87109a3e12d2254ea0ba798cc567aaf5140fbc15ec0f37d99e6ef3b7ac
issue-collapse:b5d73344677ecdb6,Review thread: 2 issues + 1 PR - pr #184508: Fix export with unused DTensor param attrs,linked work thread,PR blocked,89,high,work_thread_graph,2,1,0,0,"175467, 184229",184508,,,"module: dtensor, oncall: distributed, oncall: export, oncall: pt2","bot-triaged, module: dtensor, oncall: distributed, oncall: export, oncall: pt2, triaged",thread:issue:175467:issue:184229:pr:184508,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 2 open issues; raw issue rows remain available for audit.,"#175467 [DTensor] torch.export.export() fails with RuntimeError: wrapped_call_impl - DTensorSpec not registered as pytree constant, #184229 torch.export with prefer_deferred_runtime_asserts_over_guards=True crashes on DTensor-parallelized models (IndexError in set_missing_meta_vals)",237079a1d514c39e8a8b12013cc2cb2b156bb727544362d0d362978e8c395b86
issue-collapse:5309dd1937c162d5,Review thread: 1 issue + 1 PR - pr #189124: [inductor] Cache DSR-expanded single-config reductions,linked work thread,has linked issue,89,high,work_thread_graph,1,1,0,0,189121,189124,,,"module: compile-time, module: determinism, module: inductor, module: performance, module: reductions, oncall: pt2","bot-triaged, module: compile-time, module: determinism, module: inductor, module: performance, module: reductions, oncall: pt2, triage review",thread:issue:189121:pr:189124,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,"#189121 [inductor] Single-config reductions extended by _dynamic_scale_rblock are permanently uncacheable — re-benchmark {R0_BLOCK, R0_BLOCK/2} every process; near-tied winners flip perf and numerics",28993303dd970aa17169dc7e8079ed51bbce3a27f2e7c173ca1fa7b72c0aa4e0
issue-collapse:fe8e0c72ddac553b,Review thread: 1 issue + 1 PR - pr #188965: Fix non-deterministic node ordering in get_source_partitions(),linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,147170,188965,,,"oncall: export, oncall: pt2, oncall: quantization","good first issue, oncall: export, oncall: pt2, oncall: quantization",thread:issue:147170:pr:188965,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#147170 [BUG][PyTorch 2.0 Export][quant]:get_source_partitions() may return different matches with same input graph,0ddfbc915432e85040dd20deafbdde49d746dda0a48d1a9f14111791bd483c48
issue-collapse:a467eb9363cada9c,Review thread: 1 issue + 1 PR - pr #188698: Make Fsspec filesystem checkpointing public and exposing cache_staged_state_dict to enable async save,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,188772,188698,,,oncall: distributed,"bot-triaged, oncall: distributed",thread:issue:188772:pr:188698,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#188772 Make Fsspec filesystem checkpointing public,e3978c59311419181e5947606457ba2521bc15c9345d203d99481dd79649c161
issue-collapse:016de146b8f0dc1a,Review thread: 1 issue + 1 PR - pr #187481: Add overflow check to FBGEMM version of quantize_val_arm,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,187480,187481,,,oncall: quantization,"bot-triaged, oncall: quantization",thread:issue:187480:pr:187481,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#187480 `TestQuantizedTensor.test_qtensor_fill_per_channel_nhwc` fails on GB200,dc73e9dbb31cd1ae70dea18969e6000425f41b066d014e27fd08546936697017
issue-collapse:a15a6a99173a2d81,Review thread: 1 issue + 1 PR - pr #186043: Lower symbolic bool not to torch.sym_not,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,143157,186043,,,"module: dynamic shapes, oncall: export, oncall: pt2","module: dynamic shapes, oncall: export, oncall: pt2",thread:issue:143157:pr:186043,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143157 can we make not(x) --> sym_not(x) automatically during tracing.,1cc52bed8f06aa23ec79a3645957454b7cc05137793e710eb5fc949c76b5e113
issue-collapse:c642f3c5a7cf7d0e,Review thread: 1 issue + 1 PR - pr #185892: Improve export device mismatch errors,linked work thread,waiting on contributor,89,high,work_thread_graph,1,1,0,0,184841,185892,,,"oncall: export, oncall: pt2","bot-triaged, oncall: export, oncall: pt2",thread:issue:184841:pr:185892,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185892; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#184841 `torch.export.export` raises `RuntimeError: Unhandled FakeTensor Device Propagation` for CUDA/MPS example inputs instead of a clear user-facing error,0f99bca1d42619e75873051fc958a4d9ddd617fd7d61ed7e73f0d717ca488da5
issue-collapse:908a9d6650ddbbdf,Review thread: 1 issue + 1 PR - pr #185804: Fix non-strict export tensor len handling,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,147326,185804,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:147326:pr:185804,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#147326 torch.export.export fails when one input is a class inheriting from torch.nn.Module,1806eb4b5675b239c596411157e9bb41baacd07052cf17216b33c43e5fa75cc5
issue-collapse:9d621096a7a27a4b,Review thread: 1 issue + 1 PR - pr #185691: Avoid UB in float to signed integer casts,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,151510,185691,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",thread:issue:151510:pr:185691,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#151510 [eager] [cpu] [edge case] When processing `torch.nan_to_num-.long()`, inductor outputs the `reciprocal` of eager",722a994c46daba8ea45b9a339947e142bb0096b78456f8d052fe32b368c58980
issue-collapse:b74b00d0dab0844e,Review thread: 1 issue + 1 PR - pr #185324: Reject mixed-device fake convolution inputs,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,160399,185324,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",thread:issue:160399:pr:185324,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#160399 `torch.compile` causes segmentation fault with simple `ModuleList` model after multiple compile calls,570d41672e726e629fcf1c21b905aa23f670e76ead5d783433adb74842d4de02
issue-collapse:c61bc23fbbbc59d0,Review thread: 1 issue + 1 PR - pr #185108: Prevent non-strict export from leaking fake buffers,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,163189,185108,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:163189:pr:185108,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163189 [export] Silent fake tensor leakage when doing buffer mutation,6d06d1c4cbceeda6ec0abab19d74b63cde5e0ed57b12d49db71dcbe73b45eca9
issue-collapse:c671fb453f43df74,Review thread: 1 issue + 1 PR - pr #185069: Fix AOTI Windows cross-compile package loading,linked work thread,waiting on contributor,89,high,work_thread_graph,1,1,0,0,166093,185069,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:166093:pr:185069,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185069; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#166093 aoti cross compile for windows failed with undefined reference to WinMain,a26b08c6711fd1ecf27eb651735bf6741951651231469cbbb7a178f6ee668e53
issue-collapse:40564bfab4f5f317,Review thread: 1 issue + 1 PR - pr #184967: Fix export CIA decomps with backend kernels,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,171897,184967,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:171897:pr:184967,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171897 Torch decomposition has no effect on torch.ops.aten._fused_rms_norm,b379c927e53d74f84642109779730b077a0e14cd998eec750a9e6d767bceaffe
issue-collapse:c8c8e8705878b5be,Review thread: 1 issue + 1 PR - pr #184966: Preserve unflattened module forward signatures,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,171964,184966,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:171964:pr:184966,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171964 `torch.export.export` on unflattened module fails for dynamic shapes,00554ad9bba6508b1769e8ab16c89bafeab129af4311104350e53fe510921f4b
issue-collapse:728a8dc88dbe6336,Review thread: 1 issue + 1 PR - pr #184964: Handle tensor scalar slice bounds in export setitem,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,171974,184964,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:171974:pr:184964,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171974 `torch.export` fails on models with KV-cache-like stateful updates using data-dependent slice assignment,2c9193515106d80d9f797bc36136dccd88fd5cb820f74b8b775be4c674b59f0e
issue-collapse:82dcc58143b7f4db,Review thread: 1 issue + 1 PR - pr #184963: Fix FakeTensor shallow copy state,linked work thread,waiting on contributor,89,high,work_thread_graph,1,1,0,0,171977,184963,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:171977:pr:184963,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184963; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#171977 `torch.export` triggers internal state corruption (`AssertionError: cpu`) via `copy.copy` on FakeTensor,40d3e936f4ed78b99ab1886e9554a3311d6cc60b7b81e27547705d6ed6de5ac7
issue-collapse:62a8c66c15e8b70b,Review thread: 1 issue + 1 PR - pr #184961: Fix packed RNN export fake metadata,linked work thread,waiting on contributor,89,high,work_thread_graph,1,1,0,0,171978,184961,,,"oncall: export, oncall: pt2","PT2-Bug-Bash, oncall: export, oncall: pt2",thread:issue:171978:pr:184961,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184961; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171978 `torch.export` fails on models using `pack_padded_sequence` with LSTM due to FakeTensor in dimension,c98a337ec485c3a44b9b34b74024ee531a58208340d329e205d7dc6ae9ab8796
issue-collapse:b84e57353b6f3831,Review thread: 1 issue + 1 PR - pr #184950: Fix mkldnn transposed conv pointwise shape inference,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,172711,184950,,,"module: correctness (silent), oncall: cpu inductor, oncall: pt2","module: correctness (silent), oncall: cpu inductor, oncall: pt2, topic: fuzzer",thread:issue:172711:pr:184950,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#172711 torch.compile produces wrong output channels for mkldnn._convolution_transpose_pointwise (FakeTensor shape inference bug),fde46ab11d2b0634fc275a27a74803157e1d11b3224c35dd23b7b9dbece96d9e
issue-collapse:ad214527d854be44,Review thread: 1 issue + 1 PR - pr #184948: Fix dynamo export for tensor closures in output pytrees,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,172839,184948,,,"oncall: export, oncall: pt2","PT2-Bug-Bash, oncall: export, oncall: pt2",thread:issue:172839:pr:184948,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#172839 _dynamo_graph_capture_for_export doesn't work on pytree output with tensor closures,670f98e2786ef0c6c1443c774c50164ebdf699930f3e6ae3c4753f01e586ad5f
issue-collapse:74bc9406e03bcff9,Review thread: 1 issue + 1 PR - pr #184787: [export] Deserialize PT2 schema in C++ during load,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,176428,184787,,,"oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, oncall: export, oncall: pt2",thread:issue:176428:pr:184787,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#176428 torch.export.load() slow due to Python-side schema deserialization,d7f0bbd291fca588f3103c80143696e6e6c816b0783e890a4619b593fb5dc585
issue-collapse:c05dd3e306dce0cc,Review thread: 1 issue + 1 PR - pr #184667: Fix cond branch tensor constants in run_decompositions,linked work thread,ready for maintainer decision,89,high,work_thread_graph,1,1,0,0,180354,184667,,,"oncall: export, oncall: pt2","bot-triaged, oncall: export, oncall: pt2",thread:issue:180354:pr:184667,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#180354 `torch.cond` with tensors created in `[true/false]_fn` fails during `run_decompositions`,da578291aba6bf195beb4162b83627b3baf50bb75e30a11c3f91052224c13035
issue-collapse:d9ea3d1e883791aa,Review thread: 1 issue + 1 PR - pr #182224: Enable skipped static_runtime tests on AArch64,linked work thread,PR blocked,89,high,work_thread_graph,1,1,0,0,178522,182224,,,"module: arm, oncall: quantization","bot-triaged, module: arm, oncall: quantization",thread:issue:178522:pr:182224,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178522 AArch64 static runtimes failures,3940a455c4ca8d1900c2a9143cb6568a3fe5a0973f4224e68a2076a880b58606
issue-collapse:e0fb00e66db5a7a0,Review thread: 1 issue + 1 PR - pr #178844: [c10d] Support stacked output tensors in Gloo all_gather_into_tensor …,linked work thread,has linked issue,89,high,work_thread_graph,1,1,0,0,178798,178844,,,"oncall: distributed, oncall: distributed infra","bot-triaged, oncall: distributed, oncall: distributed infra, ptd-bot-triaged",thread:issue:178798:pr:178844,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#178798 all_gather_into_tensor in gloo backend does not support stacking,0e0075ca06e236f8e8560f96200f6de7c86d1bbe9e2fa6b83b5876eb3a39f817
issue-collapse:694158513a191b15,Review thread: 1 issue + 1 PR - pr #185300: Fix AttrProxy module reuse in Dynamo,linked work thread,PR blocked,83,high,work_thread_graph,1,1,0,0,161053,185300,,,"module: onnx, oncall: export, oncall: pt2","module: onnx, needs reproduction, oncall: export, oncall: pt2",thread:issue:161053:pr:185300,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161053 [export] Non-strict export of `torch.cond` fails with `UnspecializedNNModuleVariable is already tracked for mutation`,c97828d9c88084769c7713714a5d8c3b08b8685dc18f07c469d13aabdec9ea85
issue-collapse:7ac62d17248e76e2,Review thread: 1 issue + 1 PR - pr #185025: [export][serde] Initialize symbol ranges before constants,linked work thread,ready for maintainer decision,83,high,work_thread_graph,1,1,0,0,167872,185025,,,"oncall: export, oncall: pt2","needs reproduction, oncall: export, oncall: pt2",thread:issue:167872:pr:185025,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167872 exported program cannot be loaded with load_pt2: GraphModuleDeserializer' object has no attribute 'symbol_name_to_range',eaa9b273e6281bdac8dcef1ef5cbb8ee715972b4d283cd2d6d6b335b3756f3e0
issue-collapse:5ae304b30e3d8b4f,Review thread: 3 issues + 1 PR - pr #166876: [ci][cpu] Update AArch64 manylinux compiler to GCC14,linked work thread,PR blocked,76,high,work_thread_graph,3,1,0,0,"168288, 172445, 172632",166876,,,"module: arm, module: build, module: crash, module: third_party","module: arm, module: build, module: crash, module: third_party, triaged",thread:issue:168288:issue:172445:issue:172632:pr:166876,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#168288 ICE in gcc < 13.5 for AArch64 on Neoverse-V1, #172445 AArch64 GCC14 Issues Tracker, #172632 AArch64 GCC14 - inductor/test_mkldnn_pattern_matcher.py FAIL - internal compiler error: Segmentation fault",75c9cbfc873113c00cee73854caef27bc3f0fd86e6821b79c991cc42b56e237e
issue-collapse:271a6e96d360a0ea,Review thread: 1 issue + 1 PR - pr #186021: [Inductor][CPU] Guard compiled integer div by zero,linked work thread,PR blocked,76,high,work_thread_graph,1,1,0,0,143649,186021,,,"module: crash, oncall: cpu inductor, oncall: pt2","module: crash, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:143649:pr:186021,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143649 Floating Point Exception (core dumped) when running floordiv/remainder/fmod under torch.compile,191884e208fc4e7d481b938d04a6682db4cf5f3726ca4f2c6b3c5cedc30ec5a6
issue-collapse:6c8fd09a1175e20c,Review thread: 1 issue + 1 PR - pr #185984: Fix Dynamo out= resize for internal tensors,linked work thread,waiting on contributor,76,high,work_thread_graph,1,1,0,0,145220,185984,,,"module: dynamo, module: empty tensor, module: regression, oncall: pt2","module: dynamo, module: empty tensor, module: regression, oncall: pt2, triaged",thread:issue:145220:pr:185984,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185984; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#145220 Regression in the compilation of the torch.all operation in PyTorch version 2.6.0 compared to 2.5.1,0db71486dfa658de5a135a119d32f8689b9903cc5613140616dff72c9689f114
issue-collapse:9e20adbfcfc62e2b,Review thread: 1 issue + 1 PR - pr #185809: Fix as_strided_copy storage bounds in Inductor,linked work thread,PR blocked,76,high,work_thread_graph,1,1,0,0,147077,185809,,,"module: crash, module: error checking, oncall: pt2","module: crash, module: error checking, oncall: pt2, topic: fuzzer, triaged",thread:issue:147077:pr:185809,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#147077 [inductor] SIGSEGV when using `torch.compile` with `torch.as_strided_copy`,ddc14ba82928b9af4f7ab604ac0d058d750ea57733642476f435ad4bcafb1208
issue-collapse:9493b002d23110bf,Review thread: 1 issue + 1 PR - pr #181062: Validate grad_output shape in multi_margin_loss_backward,linked work thread,has linked issue,76,high,work_thread_graph,1,1,0,0,146790,181062,,,"module: crash, module: error checking","module: crash, module: error checking, topic: fuzzer, triaged",thread:issue:146790:pr:181062,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#146790 Segmentation Fault in `torch.ops.aten.multi_margin_loss_backward` with Empty `grad_output`,3a04dd3ef1addd0df73a5a6debb60c6b644f02f155cc73956bf4894d8ffcada1
issue-collapse:7ee5d1bb6794bcce,Review thread: 10 issues + 7 PRs - pr #183884: Fix compile worker quiesce draining,linked work thread,waiting on contributor,71,high,work_thread_graph,10,7,0,0,"130182, 136216, 137520, 140842, 144360, 145093, 145231, 158088, 186031, 186136","183884, 184161, 185433, 185976, 186011, 186094, 186361",,,"module: aotinductor, module: autograd, module: compile ux, module: correctness (silent), module: dynamic shapes, module: dynamo, module: faketensor, module: higher order operators, module: inductor, module: multiprocessing, module: pt2-dispatcher, module: testing, module: tests, module: torchcomms, oncall: distributed, oncall: distributed infra, oncall: export, oncall: pt2","actionable, bot-mislabeled, bot-triaged, export-triaged, module: aotinductor, module: autograd, module: compile ux, module: correctness (silent), module: dynamic shapes, module: dynamo, module: fakeTensor, module: higher order operators, module: inductor, module: multiprocessing, module: pt2-dispatcher, module: testing, module: tests, module: torchcomms, oncall: distributed, oncall: distributed infra, oncall: export, oncall: pt2, ptd-bot-triaged, triaged",thread:issue:130182:issue:136216:issue:137520:issue:140842,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #184161, #185433, #185976; keep related issue/PR context attached.",6,Represents 10 open issues; raw issue rows remain available for audit.,"#130182 autograd.Function HOP backward input names could be better, #136216 [torch.export] Detect internal constrains, #137520 Support torch.cond with differing (but same dimensionality) sizes on true/false branches, #140842 torch.cond with unbacked symint failed to compile in AOTI, #144360 Skip empty frames recursively when top-level is empty, #145093 Inductor aten.clone lowering ignores Conjugate and Negative dispatch keys",2be3d53afd4f142fae8de7dc54932d7c630b30bdcc4e25d4462dc18b19a12550
issue-collapse:33d57d8cfb0b2231,Review thread: 7 issues + 5 PRs - pr #184975: Support FakeTensor module materialization via to_empty,linked work thread,ready for maintainer decision,71,high,work_thread_graph,7,5,0,0,"145529, 153056, 170137, 170770, 170781, 186796, 187008","184975, 184976, 185953, 186801, 187090",,,"module: dynamo, module: faketensor, module: guards, module: nn, module: pt2-dispatcher, module: python frontend, oncall: export, oncall: profiler, oncall: pt2","bot-triaged, has workaround, module: dynamo, module: fakeTensor, module: guards, module: nn, module: pt2-dispatcher, module: python frontend, oncall: export, oncall: profiler, oncall: pt2, tensor subclass, triaged",thread:issue:145529:issue:153056:issue:170137:issue:170770,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#145529 Module.to() fail in dynamo when swap_module_params_on_conversion is true, #153056 Export doesn't move embedding to correct device, #170137 PyTorch memory profiler conflicts with torch.util.swap_tensors due to strict reference counter check, #170770 FakeTensorMode induced Cannot swap t1 because it has weakref associated with it, #170781 meta_model.to_empty(...) equivalent for FakeTensor, #186796 weakrefs don't when tensor swap is used",7539230a869466909a039f465e51785232a1d6a48bdda807b6e163b507713fab
issue-collapse:22280bab0193948e,Review thread: 3 issues + 2 PRs - pr #184978: Improve export error for untraceable lifted constants,linked work thread,ready for maintainer decision,71,high,work_thread_graph,3,2,0,0,"167007, 170672, 188153","184978, 185059",,,"oncall: export, oncall: pt2","bot-triaged, export-triaged, has workaround, oncall: export, oncall: pt2, triaged",thread:issue:167007:issue:170672:issue:188153:pr:184978,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 3 open issues; raw issue rows remain available for audit.,"#167007 torch.export failed with tensor subclass, worked in 2.9.0, #170672 torch.export.export Fails with RuntimeError and ""Please file an issue on github"" for In-Place Parameter Updates in forward, #188153 torch.export.export Fails with RuntimeError and ""Please file an issue on github""",fb90118d9b7172d66c03c13285dce7abf8422408d62f5dc0b1866ee729dc9b43
issue-collapse:639845f728297cc0,Review thread: 3 issues + 1 PR - pr #186016: Reject invalid bitwise dtypes in Inductor lowering,linked work thread,PR blocked,71,high,work_thread_graph,3,1,0,0,"139124, 144247, 144362",186016,,,"module: error checking, module: inductor, module: structured kernels, module: unsigned int, oncall: cpu inductor, oncall: pt2","module: error checking, module: inductor, module: structured kernels, module: unsigned int, oncall: cpu inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:139124:issue:144247:issue:144362:pr:186016,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#139124 shift operators not supporting integral types, #144247 [inductor][cpu] torch.bitwise_and/or/xor incorrectly accepts float32 tensors, #144362 Some operators miss dtype check when using `torch.compile`",66e9fe776bbf82ebac08950fd0a9c9bedd9fc38e077af7084150fb7750c0fff7
issue-collapse:6043a8c37f0dc289,Review thread: 3 issues + 1 PR - pr #185939: Fix int8 WoQ GEMM residual epilogue fusion,linked work thread,PR blocked,71,high,work_thread_graph,3,1,0,0,"145687, 145709, 145711",185939,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",thread:issue:145687:issue:145709:issue:145711:pr:185939,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 3 open issues; raw issue rows remain available for audit.,"#145687 [Inductor-CPU] `add` is not being fused with templated int8 WoQ GEMM while running LLaMA2, #145709 [Inductor-CPU] Templated codegened kernel names are ambiguous in PyTorch profiler results, #145711 [Inductor-CPU] Templated codegened kernel name may not correctly reflect epilogue fusions",e3d681530bf60c511ac962d382df1dd7917f789908a9807ef11bf7f0499728f7
issue-collapse:46c22a4de40268a3,Review thread: 3 issues + 1 PR - pr #185124: Fix prims.broadcast_in_dim functionalization,linked work thread,waiting on contributor,71,high,work_thread_graph,3,1,0,0,"157610, 162734, 163037",185124,,,"module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2","export-triaged, module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:157610:issue:162734:issue:163037:pr:185124,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185124; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#157610 `torch.compile` fails on `prims.broadcast_in_dim` with alias annotation error, #162734 Run decomposition fails for prims.broadcast_in_dim, #163037 [compile/inductor][functionalization] `prims.broadcast_in_dim` triggers alias-annotation error under `torch.compile` (eager OK)",147a00614176cae2910aa4fa158598432e01290bdda021d6284f7faf87fc8890
issue-collapse:90070d7e6ee618de,Review thread: 2 issues + 2 PRs - pr #185170: Fix ExportedProgram pickle after graph_module edits,linked work thread,ready for maintainer decision,71,high,work_thread_graph,2,2,0,0,"128554, 161671","185170, 186418",,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",thread:issue:128554:issue:161671:pr:185170:pr:186418,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 2 open issues; raw issue rows remain available for audit.,"#128554 aot_export_module produces unserializable GraphModules, #161671 Saving a modified ExportedProgram",d261d7f7ae51164cb7bc9120675481e0144ee664adcf3223cd91ebde1d7e85aa
issue-collapse:5aa0af760820480d,Review thread: 2 issues + 1 PR - pr #185803: Fix joint export for tied parameter aliases,linked work thread,ready for maintainer decision,71,high,work_thread_graph,2,1,0,0,"101192, 147380",185803,,,"module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2","export-triaged, module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:101192:issue:147380:pr:185803,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#101192 AOTAutograd export path does not support training graphs with parameters that do not receive gradients., #147380 [export] fail to export joint graph of a model with tied weights using experimental `_export_forward_backward` API",4933ccbaba86e636916de16d91d9c26510408c49bca2951f834cafbf420b27fc
issue-collapse:e72605c3eb4abae6,Review thread: 1 issue + 1 PR - pr #186900: Improve pytree missing registration error,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,140615,186900,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:140615:pr:186900,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140615 Non-actionable error message when output type is not pytree-able for aoti artifact,68fd53c9ad6317d074f8591cfb3eb747917ef1837a2204989c147ab2c6b03763
issue-collapse:f3c85b2628123c82,Review thread: 1 issue + 1 PR - pr #186898: Add always-on AOTI input rank checks,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,141115,186898,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:141115:pr:186898,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141115 [AOTInductor] AOT compiled models do not consistently check input shapes,5ecb21a4b6c494ce28b87d5a7aa915961a543589d8e41f5a44ca3d775cc6821b
issue-collapse:8e357dfd47f86713,Review thread: 1 issue + 1 PR - pr #186894: Fix non-strict export of vmap tensor indexing,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,158540,186894,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:158540:pr:186894,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158540 can't torch.export.export tinyllama model,3f97b2b97a1defddcc07cb671a121fecbbd52f1e7cc184a7964047945eb1883a
issue-collapse:2e2c25a1a4b583d4,Review thread: 1 issue + 1 PR - pr #186879: Fix export dynamic shapes for variadic kwargs,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,164313,186879,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:164313:pr:186879,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#164313 Torch export with dynamic shapes does not work with Hugging Face models that has TransformersKwargs in forward signature,4c2edd80fb4ed91b7ccb4653a896eadba81919cd5b7983240e85f346f754a1be
issue-collapse:23cf2d3fc6d0e6b2,Review thread: 1 issue + 1 PR - pr #186833: Remove stale export OpInfo xfails,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,170539,186833,,,"module: ci, oncall: export, oncall: pt2","module: ci, oncall: export, oncall: pt2",thread:issue:170539:pr:186833,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#170539 test_export_opinfo: 9 tests fail with ""Unexpected success"" due to outdated xfail markers",59d5a9e8cfc37bf1fd8e9282a8da7a023f696892b2e786d923406193be5f2e9d
issue-collapse:fe3c6581a05785f9,Review thread: 1 issue + 1 PR - pr #186119: Improve Dynamo observed exception messages,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,140607,186119,,,"module: compile ux, module: dynamo, oncall: export, oncall: pt2","module: compile ux, module: dynamo, oncall: export, oncall: pt2",thread:issue:140607:pr:186119,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140607 Bad error message for malformed args during export,ca8bec7129a0c189c20d31bf07a92ff0f97f897d3dd6954055838c4ec2d84738
issue-collapse:add4ac51278d6e7a,Review thread: 1 issue + 1 PR - pr #186118: Preserve guard eval defaulting without callback churn,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,140683,186118,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",thread:issue:140683:pr:186118,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140683 [inductor][cpu]maml fp32/amp_fp16 performance regression in 2024-11-11 nightly release,8bcd502a28cb6a613e2187215a8ab6284159b9f2effb7b4c5e3774a349caef76
issue-collapse:a932ff1762c7cecb,Review thread: 1 issue + 1 PR - pr #186058: Preserve export graph metadata during normalization,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,141640,186058,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:141640:pr:186058,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141640 Relaxing Export IR node.meta requirements,86f5d27472eda8038b124f92083469aae893378ba676a94374119f411965c250
issue-collapse:7b30199c5456e133,Review thread: 1 issue + 1 PR - pr #185952: Support compiled async all_reduce wait,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,145574,185952,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:145574:pr:185952,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145574 [compile / strict export] torch._dynamo.exc.Unsupported: CollectiveFunctionRewriteVariable can't support async_op=True for <function all_reduce at 0x7f40be5724d0>,3fd2236d0cd67c68ac24620d9af3e61ff5a5adf5da1a7cd2f00f9ca07dfcd7e5
issue-collapse:af9f9722ab84f239,Review thread: 1 issue + 1 PR - pr #185926: Preserve tuple arguments in export serde,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,145785,185926,,,"oncall: export, oncall: pt2","export-triage-review, oncall: export, oncall: pt2",thread:issue:145785:pr:185926,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145785 export serde turn hop's tuple arg into list,5ac0d6b8df3895298c7c429c0cd10ac945637390237e669edb6298a612fcc2a1
issue-collapse:edfffbfc131a060a,Review thread: 1 issue + 1 PR - pr #185568: Fix export of arange with data-dependent float step,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,153247,185568,,,"oncall: export, oncall: pt2","data dependent error, oncall: export, oncall: pt2",thread:issue:153247:pr:185568,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153247 Cannot use torch.arange in torch.ops.higher_order.scan,eee9ee191353d3b9b4c20089f2ca8e8fbb3e7e0b8d7a9161b78927aade17ab8f
issue-collapse:8d598e8f4484688b,Review thread: 1 issue + 1 PR - pr #185564: Handle pre-dispatch export of subclass tensor constants,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,153387,185564,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:153387:pr:185564,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153387 Pre-dispatch export doesn't work with non-param/buffer tensor subclasses,819782460ee462220470c20546a4cfcef036e2360c064e424e2c07c4222b4c4f
issue-collapse:d1435a135d770b10,Review thread: 1 issue + 1 PR - pr #185545: Handle symbolic bounds in slice decomposition,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,154137,185545,,,"module: dynamo, module: guards, oncall: export, oncall: pt2","export-triaged, module: dynamo, module: guards, oncall: export, oncall: pt2",thread:issue:154137:pr:185545,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154137 [Dynamo][Inductor] `detectron2_fcos_r_50_fpn` in export config failure on dashboard,945ebd4c3af2101150652a97525d69be34d5ed35adcaee6b796f7d21901f68a4
issue-collapse:8764022ded63e1c1,Review thread: 1 issue + 1 PR - pr #185504: Support unbacked SymInt closures in cond export,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,154559,185504,,,"oncall: export, oncall: pt2","data dependent error, oncall: export, oncall: pt2",thread:issue:154559:pr:185504,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154559 [cond] support for unbacked symbols in compiled region,87ae36979415b55da73345f3ef2c65544301f16cbb2dbe1f873d6c96c33f7fae
issue-collapse:2735b1db576a8ad3,Review thread: 1 issue + 1 PR - pr #185499: Fix unbacked memo reuse during decomposition retrace,linked work thread,waiting on contributor,71,high,work_thread_graph,1,1,0,0,154647,185499,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:154647:pr:185499,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185499; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#154647 Recreation of unbacked symint leads to ""possible memo disaster"" when running decompositions",39517be82fc07157b20441102d65507abb8adfa80fe75fadb596499af415e200
issue-collapse:572f0b5f2ad3581b,Review thread: 1 issue + 1 PR - pr #185460: Fix CPU perf smoketest baseline drift checks,linked work thread,draft/noise,71,high,work_thread_graph,1,1,0,0,156511,185460,,,"module: infra, oncall: cpu inductor, oncall: pt2","module: infra, oncall: cpu inductor, oncall: pt2",thread:issue:156511:pr:185460,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156511 functorch_maml_omniglot is a bad CPU performance smoketest model,3ed43208d2d65b87d67cb4fc4ced584f3db3422281f49732837a853f24d9da8d
issue-collapse:d60850eaf6ac99b2,Review thread: 1 issue + 1 PR - pr #185450: Fix reinplace alias checks for index_put,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,156786,185450,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",thread:issue:156786:pr:185450,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156786 [inductor] [cpu] `torch.nn.Embedding-torch.index_copy` outputs inconsistent results on cpu inductor,e83a46faf2701bb8afb49e27650d167f852c5dfb574fed263c2b241f34d51b18
issue-collapse:12a738b161f6788e,Review thread: 1 issue + 1 PR - pr #185404: [inductor] Fold small-row expanded bmm to mm in post-grad,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,159346,185404,,,"module: aotinductor, oncall: export, oncall: pt2","actionable, module: aotinductor, oncall: export, oncall: pt2",thread:issue:159346:pr:185404,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#159346 [AOTI] Severe Performance Regression with FP16 Autocast in AOTInductor for Small Batch Sizes,94a23e42158ec6d5fbd0bc8749ffc08f5bba6d625f493fb9cf3f4d5d46b4603c
issue-collapse:a0012627c036a493,Review thread: 1 issue + 1 PR - pr #185371: Fix symbolic antialiased resize export shapes,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,159468,185371,,,"module: onnx, oncall: export, oncall: pt2","module: onnx, oncall: export, oncall: pt2",thread:issue:159468:pr:185371,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#159468 Dynamo ONNX Export of resize operation doesn't work as expected,94840dc83ef4e562443db5145b75e9d88279d4c778b4e8cc449a5fbcd724928b
issue-collapse:b3f2cecb6cb11ecb,Review thread: 1 issue + 1 PR - pr #185353: Normalize export state names for subclass decomposition,linked work thread,waiting on contributor,71,high,work_thread_graph,1,1,0,0,159918,185353,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:159918:pr:185353,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185353; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#159918 [export] Tensor subclass decomposition breaks state dict contracts,290fcf96ac3fc54518541b449c7a0d1800a5067c391f2f9dd3dbce1e8ded5cc8
issue-collapse:5ac1d2264c66dfb8,Review thread: 1 issue + 1 PR - pr #185150: Preserve OrderedDict outputs in FX tracing,linked work thread,waiting on contributor,71,high,work_thread_graph,1,1,0,0,161807,185150,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:161807:pr:185150,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185150; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#161807 torch.fx.experimental.optimization.optimize_for_inference changes OrderedDict outputs to immutable_dict in torch.export CallSpec,0656ac35f78dff69986c5461fb32761ab8e5d5810234f54229ad273d48e43176
issue-collapse:dc51bf9e4803e56b,Review thread: 1 issue + 1 PR - pr #185140: Preserve reentrant checkpoint in non-strict export,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,162386,185140,,,"module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2","export-triaged, module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2",thread:issue:162386:pr:185140,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162386 Cannot trace out joint from non-strict export for model with activation checkpointing (AC),e7c3623ce69fafddfd067c07d3f9469fe1c042e6b074deaee5b6b49067f5512f
issue-collapse:e3c8ff21f6e4708e,Review thread: 1 issue + 1 PR - pr #185106: Fix non-strict export state aliases,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,163303,185106,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:163303:pr:185106,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163303 [export] non-strict export creates extra tensor constants when forward runs with a container of params,8ca3f903cc78c3bd6676eab23d358ee2c51e4893071450cc5022409c115b95ae
issue-collapse:900073ec9485f207,Review thread: 1 issue + 1 PR - pr #185105: Fix export move_to_device_pass device rewrites,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,163323,185105,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:163323:pr:185105,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163323 [export] Bug Report: module.to(device) on exported programs causes assertion failures (device-agnostic export broken since 2.8),9aa37f744be4c4b618015716ca0f055dc3eaf4ca0cb5725cce37583ecdb2235e
issue-collapse:6e278cc9d505e785,Review thread: 1 issue + 1 PR - pr #185100: Keep fake tensor cache synthesis out of proxy traces,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,163713,185100,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:163713:pr:185100,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163713 [export] 8*s72 is not tracked with proxy for torch.fx.experimental.proxy_tensor._ModuleStackTracer object,23e78874dbea61900991275d1432703c0695074cac99085261302eaefbd7acd0
issue-collapse:f096b576151f3997,Review thread: 1 issue + 1 PR - pr #185084: Avoid contiguity guards on maybe-singleton symbolic dims,linked work thread,PR blocked,71,high,work_thread_graph,1,1,0,0,165032,185084,,,"module: dynamic shapes, oncall: export, oncall: pt2","export-triaged, module: dynamic shapes, oncall: export, oncall: pt2",thread:issue:165032:pr:185084,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#165032 torch.export fails when you slice an input at a dynamic dimension and Dim(“t”, min=, max=)",7e886358a6e2a24ffa2922bbb23b8df6e9c4468063643dbc78dc780b8aa6c107
issue-collapse:19531d85e63cd9c9,Review thread: 1 issue + 1 PR - pr #185082: Preserve lifted SymInts during strict export capture,linked work thread,draft/noise,71,high,work_thread_graph,1,1,0,0,165073,185082,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:165073:pr:185082,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#165073 torch.export lift symint input to a graph during dynamo,819b807778a5ec9153c4593a3d7a34317015e4e9332f4d6d74043c29e60f7bbb
issue-collapse:14681775b5af77c6,Review thread: 1 issue + 1 PR - pr #185073: [export] Treat free new-tracer state as constants,linked work thread,ready for maintainer decision,71,high,work_thread_graph,1,1,0,0,165641,185073,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",thread:issue:165641:pr:185073,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#165641 strict export doesn't work with global constants,a3e7526f1c308f04b420c92a7f5db189fca470be26d2db946828e7f855e9cc58
issue-collapse:728740da9e984e9e,Review thread: 1 issue + 1 PR - pr #184985: Fix nested compile handling in AOT joint export,linked work thread,waiting on contributor,71,high,work_thread_graph,1,1,0,0,169991,184985,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2",thread:issue:169991:pr:184985,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184985; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#169991 aot_export_joint_with_descriptors don't work with compile inside due to dynamo config,7eb3ea5288c98115c63095d9987ba361a146010888dd2a1c7ab63e5bbd5c9183
issue-collapse:e572d5cf035379a4,Review thread: 183 issues + 2 PRs - pr #186051: Fix disabled no-op module pre-hooks in compile,linked work thread,PR blocked,70,high,work_thread_graph,183,2,0,0,"1359, 1369, 2129, 3877, 8853, 9222, 9674, 10043, 12308, 12498, 13246, 14945, 17902, 18904, 19149, 20433, 22402, 23159, 24041, 24185, 27312, 28090, 28307, 28308, 28318, 28319, 28320, 28329, 28341, 29137, 29758, 29973, 31837, 31945, 32097, 32288, 32293, 32570, 32998, 33041, 34279, 34651, 35529, 35641, 40471, 40568, 40570, 41128, 41793, 43116, 43949, 44634, 45833, 45897, 45996, 46164, 46225, 46544, 46829, 46999, 48088, 48628, 50010, 50342, 50343, 50344, 50345, 50382, 51450, 51509, 51720, 51962, 53441, 53704, 53900, 54766, 55279, 55299, 55366, 56356, 56485, 56586, 56697, 56764, 57610, 57947, 58489, 58734, 58736, 58741, 58742, 58743, 58745, 58770, 58828, 59058, 59099, 59101, 59104, 59545, 59786, 59787, 59868, 60858, 61417, 61453, 61474, 61486, 61490, 61523, 61582, 61653, 62352, 62784, 62812, 63293, 63870, 64023, 64208, 64254, 64359, 64897, 68105, 68291, 69491, 70906, 70910, 70914, 70915, 70916, 70919, 70920, 70921, 70924, 70925, 71209, 71386, 71403, 72824, 76644, 80553, 83818, 86055, 87085, 87358, 87366, 87448, 89255, 92141, 94779, 101699, 103352, 103581, 105465, 106584, 106614, 106667, 106991, 108565, 108977, 109017, 109108, 114951, 115819, 123082, 124505, 127622, 132559, 133676, 135704, 135724, 135826, 136264, 139792, 142358, 143588, 145624, 148077, 150612, 151912, 156308, 165612, 187389","186051, 188012",,,"module: 64-bit, module: advanced indexing, module: arm, module: autograd, module: batching, module: bc-breaking, module: boolean tensor, module: build, module: codegen, module: compile ux, module: complex, module: cpp, module: cpu, module: crash, module: cuda, module: custom-operators, module: data, module: dataloader, module: dependency bug, module: deprecation, module: derivatives, module: distributions, module: docs, module: dynamo, module: error checking, module: floatx (formerly float8), module: functionalization, module: half, module: inductor, module: initialization, module: internals, module: interpolation, module: lazy, module: linear algebra, module: loss, module: m1, module: macos, module: masked operators, module: memory usage, module: mkl, module: mkldnn, module: molly-guard, module: multiprocessing, module: nans and infs, module: nestedtensor, module: nn, module: numerical-stability, module: numpy, module: optimizer, module: performance, module: printing, module: python array api, module: python frontend, module: reductions, module: scatter & gather ops, module: scipy compatibility, module: sdpa, module: sorting and selection, module: sparse, module: special, module: tensor creation, module: tensoriterator, module: tests, module: type promotion, module: ux, module: viewing and reshaping, oncall: export, oncall: jit, oncall: pt2","OSS contribution wanted, actionable, better-engineering, bot-triaged, complex_autograd, enhancement, export-triage-review, ezyang's list, feature, function request, good first issue, has workaround, high priority, llm-amenable, low priority, module: 64-bit, module: NaNs and Infs, module: TensorIterator, module: advanced indexing, module: arm, module: autograd, module: batching, module: bc-breaking, module: boolean tensor, module: build, module: codegen, module: compile ux, module: complex, module: cpp, module: cpu, module: crash, module: cuda, module: custom-operators, module: data, module: dataloader, module: dependency bug, module: deprecation, module: derivatives, module: distributions, module: docs, module: dynamo, module: error checking, module: floatx (formerly float8), module: functionalization, module: half, module: inductor, module: initialization, module: internals, module: interpolation, module: lazy, module: linear algebra, module: loss, module: m1, module: macos, module: masked operators, module: memory usage, module: mkl, module: mkldnn, module: molly-guard, module: multiprocessing, module: nestedtensor, module: nn, module: numerical-stability, module: numpy, module: optimizer, module: performance, module: printing, module: python array api, module: python frontend, module: reductions, module: scatter & gather ops, module: scipy compatibility, module: sdpa, module: sorting and selection, module: sparse, module: special, module: tensor creation, module: tests, module: type promotion, module: ux, module: viewing and reshaping, needs design, needs reproduction, needs research, no-scrub, oncall: export, oncall: jit, oncall: pt2, open source, tensor subclass, todo, topic: bc breaking, tracker, triaged",thread:issue:10043:issue:101699:issue:103352:issue:103581,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 183 open issues; raw issue rows remain available for audit.,"#1359 Conjugate Gradient mentioned in docs, but not implemented, #1369 ""Sparsified"" mathematical operations, #2129 [Feature request] truncated normal initializer(sampler), #3877 Wrap Cephes library for mathematical special functions, #8853 Todo functions and autograd supports for Sparse Tensor, #9222 [feature request] Implementing Block Sparse Operations",9fb86107505c82159ec8febb1a73a7b8116c8ab1a36239e50d3499de6718628a
issue-collapse:fcf182999f3ddf61,Review thread: 81 issues + 1 PR - pr #186539: [Dynamo] Limit FX node growth from loop unrolling,linked work thread,PR blocked,70,high,work_thread_graph,81,1,0,0,"2628, 23756, 26288, 29697, 30291, 31557, 32021, 35666, 37410, 41081, 41243, 41353, 42959, 46168, 46184, 50122, 50688, 54389, 55135, 58997, 61686, 62448, 62530, 64730, 66073, 66707, 67586, 67760, 67761, 67958, 68332, 68648, 68978, 68979, 69972, 70173, 71683, 72146, 78917, 79197, 82464, 83159, 86590, 89116, 92884, 94652, 95408, 96136, 101082, 102839, 103444, 103588, 104849, 105255, 106624, 110761, 111003, 111441, 112303, 112901, 113287, 113914, 114206, 115626, 118758, 118966, 125531, 128046, 129889, 148819, 158371, 158968, 159765, 161302, 162591, 166208, 166209, 168044, 171302, 173559, 175891",186539,,,"module: activation checkpointing, module: aotdispatch, module: autograd, module: bootcamp, module: compile-time, module: cpu, module: cuda graphs, module: distributions, module: docs, module: dynamic shapes, module: dynamo, module: functional ux, module: functorch, module: graph breaks, module: half, module: higher order operators, module: inductor, module: internals, module: interpolation, module: lrscheduler, module: memory usage, module: mta, module: nn, module: norms and normalization, module: numerical-stability, module: numpy, module: optimizer, module: performance, module: pt2-dispatcher, module: python frontend, module: tests, module: type promotion, module: ux, module: xla, oncall: distributed, oncall: distributed parallelisms, oncall: jit, oncall: pt2, oncall: quantization","Stale, actionable, better-engineering, bot-triaged, enhancement, feature, function request, good first issue, has workaround, high priority, internal ramp-up task, low priority, module: LrScheduler, module: activation checkpointing, module: aotdispatch, module: autograd, module: bootcamp, module: compile-time, module: cpu, module: cuda graphs, module: distributions, module: docs, module: dynamic shapes, module: dynamo, module: functional UX, module: functorch, module: graph breaks, module: half, module: higher order operators, module: inductor, module: internals, module: interpolation, module: memory usage, module: mta, module: nn, module: norms and normalization, module: numerical-stability, module: numpy, module: optimizer, module: performance, module: pt2-dispatcher, module: python frontend, module: tests, module: type promotion, module: ux, module: xla, months, needs design, needs reproduction, needs research, oncall: distributed, oncall: distributed parallelisms, oncall: jit, oncall: pt2, oncall: quantization, ptd-bot-triaged, topic: new features, topic: performance, triage review, triaged, weeks",thread:issue:101082:issue:102839:issue:103444:issue:103588,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 81 open issues; raw issue rows remain available for audit.,"#2628 BatchNorm{1-2-3}d are redundant, #23756 [feature request] Core API for invertible/inplace and flow-like ops + memory-saving (hookless?) reversible sequential container for RevNets to allow for much larger batch-sizes in academic setting, #26288 Inplace and out arguments for BatchNorm (and other norm layers: InstanceNorm / LayerNorm / GroupNorm ...), #29697 MultiStepLR does not return good lr after load_state_dict, #30291 [feature request] A way to restore/assign tensor _version, #31557 [docs] F.ctc_loss docs to warn clearly about invalid inf-causing inputs; zero_infinity to become enabled by default",f5000bca8b5c3f4bcbea7d4a1b437a7d13cccc0af203f4d9c8126482b076a5ce
issue-collapse:3e824594340bbdf8,Review thread: 11 issues + 6 PRs - pr #180247: Migrate build system from setuptools to scikit-build-core,linked work thread,waiting on contributor,65,high,work_thread_graph,11,6,0,0,"108406, 119526, 157807, 160875, 163311, 164469, 169479, 179374, 180624, 183913, 188898","180247, 180248, 180250, 185208, 187633, 189263",,,"module: build, module: ci, module: cpp-extensions, module: devx, module: docs, module: flaky-tests, module: inductor, module: infra, oncall: profiler, oncall: pt2, oncall: releng","bot-triaged, easy, enhancement, large, module: build, module: ci, module: cpp-extensions, module: devx, module: docs, module: flaky-tests, module: inductor, module: infra, needs reproduction, needs research, oncall: profiler, oncall: pt2, oncall: releng, security, skipped, topic: docs, triaged",thread:issue:108406:issue:119526:issue:157807:issue:160875,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #180247, #180248, #187633; keep related issue/PR context attached.",6,Represents 11 open issues; raw issue rows remain available for audit.,"#108406 enhance documentation around the developer build, #119526 DISABLED test_source_multithreaded_multiple_preexisting_work_in_main_thread_False (__main__.TestProfiler), #157807 [RFC] Replace setuptools build backend with scikit-build-core, #160875 Add `uv` installation commands to pytorch.org, #163311 [build] `CMake (or ninja)` configuration can't index `uv` environment for incremental compilation, #164469 [RFC] Adopt Spin as developer cli",f3cf5eed03aa34402c7edb07da2f07d164a3e071e5e1f7e3d37fa51ce2e4d09e
issue-collapse:2b04004691e8917a,Review thread: 2 issues + 1 PR - pr #185085: [dynamo] Support datetime.now attrs as runtime inputs,linked work thread,PR blocked,65,high,work_thread_graph,2,1,0,0,"125171, 164922",185085,,,"module: dynamo, oncall: export, oncall: pt2","function request, module: dynamo, oncall: export, oncall: pt2, triaged",thread:issue:125171:issue:164922:pr:185085,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#125171 datettime.now() is not supported by Dynamo, #164922 `torch.compile` fails to trace `datetime.now()` with Dynamo guard check failure",5fa6b5a43c7ed166dd29eeabebd8d4519ccfccb411c997f08fdba25d4bcaacf0
issue-collapse:ba9c37825c81b730,Review thread: 1 issue + 2 PRs - pr #182377: [DTensor] SymmetricMemory-backed DTensors,linked work thread,ready for maintainer decision,65,high,work_thread_graph,1,2,0,0,182382,"182377, 182379",,,"module: dtensor, module: symm_mem, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: dtensor, module: symm_mem, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged",thread:issue:182382:pr:182377:pr:182379,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#182382 [RFC] [DTensor] Support for One-Sided Communication in DTensor,44ebc5b777a0919bd1d840baf36089b672551e5e22006721e30000ffa40694c5
issue-collapse:4771b44b72c1d78a,Review thread: 1 issue + 1 PR - pr #189097: [RFC] Expose Kineto custom logger registration,linked work thread,draft/noise,65,high,work_thread_graph,1,1,0,0,189098,189097,,,oncall: profiler,"bot-triaged, oncall: profiler",thread:issue:189098:pr:189097,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189098 [RFC] Expose Kineto custom logger registration and protocol-based export to PyTorch users,a714511dd727e48f6d171ed42fabc81d808fe87822e90468643136755549a91b
issue-collapse:e292a3c8cdb5f443,Review thread: 1 issue + 3 PRs - pr #176830: feat: add polyfills for setitem/delitem op,linked work thread,waiting on contributor,62,high,work_thread_graph,1,3,0,0,116396,"176830, 181357, 188966",,,"module: dynamo, oncall: pt2","good first issue, module: dynamo, oncall: pt2, triaged",thread:issue:116396:pr:176830:pr:181357:pr:188966,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188966; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#116396 [dynamo] Missing support for many trivial builtin functions in operator,d06daf65c24b00ab82e80fed9f61d237dc0e5ad68084e74cab320f894f5f2d3c
issue-collapse:ea6bf9df07a59151,Review thread: 2 issues + 1 PR - pr #187952: Fix out= variant recompilations for torch.max/min/topk,linked work thread,ready for maintainer decision,62,high,work_thread_graph,2,1,0,0,"135859, 187953",187952,,,"module: compile-time, module: dynamic shapes, module: faketensor, module: guards, module: linear algebra, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, bot-triaged, good first issue, module: compile-time, module: dynamic shapes, module: fakeTensor, module: guards, module: linear algebra, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:135859:issue:187953:pr:187952,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#135859 bmm, topk, cholesky, linalg.norm, max with out variants set causing recompilations in torch.compile, #187953 torch.compile: out= variants of CompositeImplicitAutograd linalg ops (linalg.norm, cholesky, ...) recompile on every input shape",714b55628e5c4cd970a2a90ae87bab3db98ee47ae5ef86fbd08db5c25a58b451
issue-collapse:a73b88b933b8dc5d,Review thread: 1 issue + 1 PR - pr #188967: Update logging docs to match current API,linked work thread,ready for maintainer decision,62,high,work_thread_graph,1,1,0,0,137285,188967,,,"module: logging, oncall: pt2","better-engineering, good first issue, module: logging, oncall: pt2, triaged",thread:issue:137285:pr:188967,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#137285 Docs are little bit outdated for torch logs,19f81c7268b679495f37af4766e0968b8e64d69064a70c49ebaab433ade456c7
issue-collapse:2fe91410ceafbccb,Review thread: 1 issue + 1 PR - pr #186488: Improve schema errors for symbolic int list mismatches,linked work thread,ready for maintainer decision,62,high,work_thread_graph,1,1,0,0,122129,186488,,,"module: dynamic shapes, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, good first issue, llm-amenable, module: dynamic shapes, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:122129:pr:186488,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122129 Obscure error: Expected a value of type 'List[int]' for argument 'sizes' but instead found type 'immutable_list',d59f536c8ca7c682ed4ae7c0218ff0902aab2d848a3bd30895c29a220f67731b
issue-collapse:1dc7f339dcab5404,Review thread: 1 issue + 1 PR - pr #185782: fix: [PT2] [Hardening] Track recompiles alongside graph breaks in our actu...,linked work thread,has linked issue,62,high,work_thread_graph,1,1,0,0,113040,185782,,,"module: dynamo, oncall: pt2","feature, good first issue, module: dynamo, oncall: pt2, triaged",thread:issue:113040:pr:185782,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#113040 [PT2] [Hardening] Track recompiles alongside graph breaks in our actual/expected comparison CI runs,3eb2936bb0e65e6e2f08b7dd81896a4837cf00ec0034f1b8478a053853b31086
issue-collapse:7cdf35634cbcf6a5,Review thread: 1 issue + 1 PR - pr #184460: Remove obsolete Python wrapper symbol extraction,linked work thread,PR blocked,62,high,work_thread_graph,1,1,0,0,102870,184460,,,module: inductor,"better-engineering, good first issue, module: inductor, triaged",thread:issue:102870:pr:184460,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#102870 Inductor: delete code that extracts out sizevars by inspecting tensor inputs to find a size that handled it,588d0c5f8d4bbcf6a70b8d3b0403ad1ee5db26ff20279321e28bca25c0074fe8
issue-collapse:314a3c74050162bb,Review thread: 1 issue + 1 PR - pr #179273: [optim] support of dynamic coefficients and normalization in Muon,linked work thread,has linked issue,62,high,work_thread_graph,1,1,0,0,179250,179273,,,module: optimizer,"bot-triaged, function request, module: optimizer, needs design, triaged",thread:issue:179250:pr:179273,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#179250 Muon: support of dynamic coefficients and normalization,d7552b1329a5c2e817e8066261cde4088daa8922a98b8411e8269e2fc5b12330
issue-collapse:0a45212bf08b96e1,Review thread: 2 issues + 1 PR - pr #185786: Handle non-natural SymPy powers in value range interp,linked work thread,ready for maintainer decision,56,high,work_thread_graph,2,1,0,0,"136628, 148003",185786,,,"oncall: export, oncall: pt2","export-triaged, needs reproduction, oncall: export, oncall: pt2, triaged",thread:issue:136628:issue:148003:pr:185786,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#136628 [torch.export.load] failed while executing `pow_by_natural`, #148003 Error : torch/utils/_sympy/interp.py:176] [0/2] failed while executing pow_by_natural([VR1, int_oo], VR[-1, -1]])",5aef38301280a7ae83b760960f5e6e75e25a8ad577a813db0460852f7bd28c1b
issue-collapse:028fbc20ddf85f60,Review thread: 1 issue + 1 PR - pr #188641: Mark linalg.eigh as cudagraph-unsafe in Inductor,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,188355,188641,,,"module: cuda graphs, module: inductor, module: linear algebra, oncall: pt2","bot-triaged, has workaround, module: cuda graphs, module: inductor, module: linear algebra, needs reproduction, oncall: pt2, triaged",thread:issue:188355:pr:188641,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188355 [Inductor][CUDA Graphs] max-autotune attempts to capture torch.linalg.eigh and crashes with cudaErrorStreamCaptureUnsupported,820942ec33b3fa6d690b966dd44407efb7acd03269b39c4be6225f7485e05c2b
issue-collapse:be48834ea9b38ffd,Review thread: 1 issue + 1 PR - pr #186849: Fix CuTeDSL grouped_mm descriptor init shape,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,176178,186849,,,"module: inductor, oncall: pt2","module: inductor, needs reproduction, oncall: pt2, repro:agent:h100:repro_success:different_error, triaged",thread:issue:176178:pr:186849,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#176178 [CuTeDSL] grouped_mm illegal address from TMA descriptor init/update shape mismatch,ba4e6627c56e8edb0734edfdd53a26bece8c9f8a47c7c6e80539d88fd5245c51
issue-collapse:63666ed02936a6c2,Review thread: 1 issue + 1 PR - pr #186261: Fix pre-dispatch tracing for saved tensor detach,linked work thread,ready for maintainer decision,56,high,work_thread_graph,1,1,0,0,137009,186261,,,"module: proxytensor, module: pt2-dispatcher, oncall: pt2","module: ProxyTensor, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",thread:issue:137009:pr:186261,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137009 pre-dispatch tracing can't interpose on detach() calls from the autograd engine.,568734ef32d1e5a53581724b5b109498459481f8c9819ced11ef495efa637029
issue-collapse:66ee2dd8b8277351,Review thread: 1 issue + 1 PR - pr #185784: Fix flex attention captured vector gradient scatters,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,148112,185784,,,"module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: flex attention, module: higher order operators, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",thread:issue:148112:pr:185784,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148112 MLA with Learnable RoPE Tensors is Broken with Flex Attention,83d0dd1eebf2e9fe144f8998540bcf76e5ffafddd444300ebfc90c5087bac703
issue-collapse:c2c17008370c183c,Review thread: 1 issue + 1 PR - pr #185742: Support None leaves in nonstrict_trace,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,149963,185742,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",thread:issue:149963:pr:185742,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149963 [ued][wan] nonstrict_trace does not support None inputs,ed4c60c3806add991f7e4d5f96166e4dc6fcb52778c39ed199a9d040135f4053
issue-collapse:be2b26ba6f1d915f,Review thread: 1 issue + 1 PR - pr #185732: Fix nested compile during subclass fake tracing,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,150262,185732,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","module: fakeTensor, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",thread:issue:150262:pr:185732,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150262 Fake mode mismatch when doing nested compile + tensor subclass,652db050215b90c84b3d52e0256c14a47fbf25ef5b29c1372292e47d7f756704
issue-collapse:4fe9e9587e468332,Review thread: 1 issue + 1 PR - pr #184977: Fix Dynamo wrapper method sources for nested compile,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,170696,184977,,,"module: dynamo, oncall: pt2","module: dynamo, needs reproduction, oncall: pt2, triaged",thread:issue:170696:pr:184977,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#170696 [torch.compile] `CatchErrorsWrapper' object has no attribute '__closure__` when wrapping a fullgraph'd function in fullgraph'd module forward,04b632158f1ea02cde2840c2affae7e1d1ec29738c2229f333495bc4675dbc42
issue-collapse:2a9fab3b5e064b9d,Review thread: 1 issue + 1 PR - pr #184934: Validate packed RNN batch_sizes before fallback,linked work thread,PR blocked,56,high,work_thread_graph,1,1,0,0,173046,184934,,,oncall: pt2,"needs reproduction, oncall: pt2, triaged",thread:issue:173046:pr:184934,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173046 [Segfault] compilion with eager for the rnn_tanh operator leads to a segfault,22f6e5a33d9831a0614b363a0f3c5f1586f81b9797e6d2efd9df8834b857512a
issue-collapse:b403c155b264f71e,"Review thread: 19 issues + 4 PRs - pr #184779: [MPS] Rewrite _bincount_mps with native Metal atomic-add (~18 s → 2.1 ms, 8,500× speedup)",linked work thread,waiting on contributor,52,high,work_thread_graph,19,4,0,0,"77764, 83264, 83968, 84489, 97310, 98169, 102852, 111173, 113062, 130430, 136692, 140570, 140718, 141287, 149261, 149325, 154052, 157648, 178492","184779, 184780, 186540, 188816",,,"module: 64-bit, module: correctness (silent), module: crash, module: hub, module: linear algebra, module: macos, module: mps, module: norms and normalization, module: performance, module: sdpa","bot-triaged, enhancement, feature, has workaround, low priority, module: 64-bit, module: correctness (silent), module: crash, module: hub, module: linear algebra, module: macos, module: mps, module: norms and normalization, module: performance, module: sdpa, needs research, tracker, triaged",thread:issue:102852:issue:111173:issue:113062:issue:130430,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #184779, #186540; keep related issue/PR context attached.",6,Represents 19 open issues; raw issue rows remain available for audit.,"#77764 General MPS op coverage tracking issue, #83264 Emulating FP64 and increased precisions on Apple silicon, #83968 Python3 Depletes 2021 M1 Mac Memory Running Training Ops For Model's M, L and X, #84489 For PyTorch Nightly, failure when changing MPS device to CPU after PYTORCH_ENABLE_MPS_FALLBACK occurs., #97310 MPS: `unique` and `unique_consecutive` extremely slow when `return_counts=True`, #98169 The operator 'aten::_weight_norm_interface' is not currently implemented for the MPS device.",e289c3dab0c8299da7cdc41d1a62c29ffced9cf312c7d8b507708afa1cb93a9c
issue-collapse:e4f42358d48aa7cc,Review thread: 13 issues + 22 PRs - pr #160585: Add dim=None support to torch.logsumexp,linked work thread,PR blocked,44,high,work_thread_graph,13,22,0,0,"122512, 156075, 162151, 167729, 167732, 168393, 170487, 170496, 170834, 171158, 174929, 182842, 187806","160585, 163503, 171826, 181680, 184974, 185033, 185034, 185149, 185524, 186100, 186476, 187468, 187469, 187531, 187532, 187707, 188173, 189091, 189092, 189234, 189291, 189324",,,"module: aotdispatch, module: autograd, module: ci, module: correctness (silent), module: custom-operators, module: dynamo, module: forward ad, module: functorch, module: nn, module: numerical-stability, module: pt2-dispatcher, module: python frontend, module: vmap, oncall: pt2","actionable, bot-mislabeled, bot-triaged, enhancement, has workaround, module: aotdispatch, module: autograd, module: ci, module: correctness (silent), module: custom-operators, module: dynamo, module: forward ad, module: functorch, module: nn, module: numerical-stability, module: pt2-dispatcher, module: python frontend, module: vmap, oncall: pt2, topic: fuzzer, triaged, unstable",thread:issue:122512:issue:156075:issue:162151:issue:167729,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#122512 Higher peak memory with torch.compile, #156075 `torch.logsumexp`: support `dim=None`, #162151 `permute` + `add` + `transpose` plays wrong on tensor slice, #167729 Dynamo should support torch.autograd.grad, #167732 Dynamo should support backward() capture, #168393 `register_autograd` broken for `torch.func`, fails to register `setup_context`",85e3b49bd3c695b033268edd5a79a902fb9ff2af3689c5c8abcedb00fb7f9364
issue-collapse:a19912a3aa1bc369,Review thread: 7 issues + 1 PR - pr #186517: Detect aliased subclass inputs after desugaring,linked work thread,ready for maintainer decision,44,high,work_thread_graph,7,1,0,0,"114397, 114403, 114411, 114412, 114413, 114414, 114415",186517,,,"module: __torch_dispatch__, module: aotdispatch, module: dynamic shapes, module: functionalization, module: higher order operators, module: pt2-dispatcher, oncall: pt2","better-engineering, feature, good first issue, module: __torch_dispatch__, module: aotdispatch, module: dynamic shapes, module: functionalization, module: higher order operators, module: pt2-dispatcher, oncall: pt2, tracker, triaged",thread:issue:114397:issue:114403:issue:114411:issue:114412,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#114397 __torch_dispatch__ + compile: run functionalization before the subclass, #114403 refactor TracingContext to take a more limited subset of ViewAndMutationMeta, #114411 higher order ops + torch.compile + __torch_dispatch__ subclasses, #114412 torch.compile <> __torch_dispatch__ support tracker issue, #114413 Support sending __torch_dispatch__ subclasses in torch.compile that do not desugar, #114414 [BE] remove SmallVector optimization in PyInterpreter.cpp when storing custom sizes",6e9dd0ef2d82e095744cd5a595e28e3df13d048197a21a640b4b5627c9684aff
issue-collapse:4193cce58f385915,Review thread: 3 issues + 3 PRs - pr #184997: Fix TVM backend scalar tensor inputs,linked work thread,waiting on contributor,44,high,work_thread_graph,3,3,0,0,"169111, 169188, 188866","184997, 189010, 189125",,,"module: backend, module: dependency bug, module: dynamo, module: third_party, oncall: pt2","bot-triaged, module: backend, module: dependency bug, module: dynamo, module: third_party, oncall: pt2, triaged",thread:issue:169111:issue:169188:issue:188866:pr:184997,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184997; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#169111 `torch.compile(backend=""tvm"", fullgraph=True, dynamic=True)` fails on PyTorch 2.9.0 with `RuntimeError: Tracer cannot infer type ... Only tensors ... supported, but got SymInt`, #169188 `torch.compile(backend=""tvm"")` produces wrong result on simple `nn.Parameter` addition (eager = 2.0, compiled = 0.0), #188866 [dynamo] TVM backend fails to import with all pip-installable TVM versions (relay removed in TVM 0.20)",1323c9f218de2f289c3a40c690c2c7ec57c9c25df9d77e761d46b26b6e1d7424
issue-collapse:b4e590fffac6870f,Review thread: 3 issues + 3 PRs - pr #184051: Honor custom op layout constraints in fallbacks,linked work thread,waiting on contributor,44,high,work_thread_graph,3,3,0,0,"161111, 161113, 161119","184051, 185280, 185288",,,"module: dynamic shapes, module: dynamo, module: inductor, module: vllm, oncall: pt2","empathy-day, module: dynamic shapes, module: dynamo, module: inductor, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:161111:issue:161113:issue:161119:pr:184051,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184051; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#161111 [UED][Falcon-7B x vLLM] errors tracing causal_conv_1d triton kernel grid fn, #161113 [UED][Falcon-7B x vLLM] selective_scan_fwd custom op runtime error, #161119 [UED][Falcon-7B x vLLM] excessive shape recompilations (static specialization)",9ca59cf35923bd77b5c9cc27f82767072d66310038eedbdb6df490f87ac2f1cb
issue-collapse:5154c2a0455158a1,Review thread: 3 issues + 3 PRs - pr #170092: [dynamo] Support constant ops on unrealized LazyConstantVariables,linked work thread,waiting on contributor,44,high,work_thread_graph,3,3,0,0,"174541, 187590, 187591","170092, 170644, 184858",,,"module: compile-time, module: dynamo, module: guards, oncall: pt2","bot-triaged, enhancement, feature, module: compile-time, module: dynamo, module: guards, oncall: pt2, triaged",thread:issue:174541:issue:187590:issue:187591:pr:170092,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184858; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#174541 torch.compile: ConstDictVariable should support runtime dict lookup for dynamic keys to avoid recompilation, #187590 [Dynamo] LazyConstants -- Avoid over-guarding on constant values, #187591 [Dynamo] Symbolic Strings",bc1d98c2e9d08e8ec0e3d9fae840f52b380bb6335d9392a76f2ebd2e5141d115
issue-collapse:7f5b90e4ad565a3f,Review thread: 1 issue + 4 PRs - pr #188100: Support callable msg lambda in TestCase.assert* methods beyond assertEqual(),linked work thread,PR blocked,44,high,work_thread_graph,1,4,0,0,189133,"188100, 188299, 188573, 189286",,,"module: flaky-tests, module: inductor, oncall: pt2","module: flaky-tests, module: inductor, oncall: pt2, skipped, triaged",thread:issue:189133:pr:188100:pr:188299:pr:188573,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#189133 DISABLED test_vmap (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd),e1d5fbd8386b0f5960effae662f37b73184d34cc1c2e8a26a214a5e95ee77308
issue-collapse:bf8f67da2eb0ee18,Review thread: 3 issues + 1 PR - pr #187243: Reject p < 1 in triplet_margin_loss to prevent NaN from p-norm overflow,linked work thread,has linked issue,44,high,work_thread_graph,3,1,0,0,"184036, 187178, 187179",187243,,,"module: correctness (silent), module: distance functions, module: edge cases, module: linear algebra, module: nn, module: norms and normalization, module: numerical-stability","bot-triaged, low priority, module: correctness (silent), module: distance functions, module: edge cases, module: linear algebra, module: nn, module: norms and normalization, module: numerical-stability, topic: fuzzer, triaged",thread:issue:184036:issue:187178:issue:187179:pr:187243,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 3 open issues; raw issue rows remain available for audit.,"#184036 torch.nn.PairwiseDistance returns +Inf for moderately large p (overflow in fp32, no log-space rewrite), #187178 torch.nn.functional.triplet_margin_loss returns NaN for p < 1 (p-norm overflow), #187179 torch.cdist returns +Inf for small p (p-norm overflow, docstring admits p in [0, inf])",0ee24f49bbe44f14d263ff2ba39528dab3b62ca14f8a7eea7fbc17526b7a87fc
issue-collapse:115400a0f171893f,Review thread: 3 issues + 1 PR - pr #186614: Delay RNN graph break until module call,linked work thread,PR blocked,44,high,work_thread_graph,3,1,0,0,"141184, 158007, 185135",186614,,,"module: compile-time, module: dynamo, module: inductor, module: performance, module: rnn, oncall: pt2","bot-triaged, dynamo-triage-jan2025, enhancement, low priority, module: compile-time, module: dynamo, module: inductor, module: performance, module: rnn, oncall: pt2, triaged",thread:issue:141184:issue:158007:issue:185135:pr:186614,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#141184 Dynamo graph break on RNN/LSTM doesn't resume the function correctly., #158007 `torch.compile` cannot compile a model with a basic `LSTM`, even on latest main, #185135 [inductor] Scheduler phases scale poorly on lowered aten.lstm/aten.gru graphs",921f8f62241e07e8dcd8f7f446ab467b6184d14df973b5b16f04bee6176523ff
issue-collapse:bdcd422c70c9999a,Review thread: 3 issues + 1 PR - pr #185347: Support torch.compile under user FakeTensorMode,linked work thread,waiting on contributor,44,high,work_thread_graph,3,1,0,0,"136446, 136586, 160057",185347,,,"module: dynamo, module: faketensor, module: inductor, module: pt2-dispatcher, oncall: pt2","feature, module: dynamo, module: fakeTensor, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:136446:issue:136586:issue:160057:pr:185347,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185347; keep related issue/PR context attached.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#136446 Allow Inductor to Compose with FakeTensorMode to Estimate Memory Usage, #136586 support FakeTensor input for torch.compile, #160057 Error raised when running torch.compile under FakeTensorMode context",7697530b07f6d33b907e4811526641ab82422dfd64278153265103d04a318cce
issue-collapse:846a6bd359d8f6ae,Review thread: 3 issues + 1 PR - pr #184990: [dynamo] Route fake RuntimeError through user exception handlers,linked work thread,PR blocked,44,high,work_thread_graph,3,1,0,0,"157217, 169538, 183887",184990,,,"module: dynamo, module: inductor, oncall: pt2","PT2-Bug-Bash, dynamo-triage-dec2025, module: dynamo, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, topic: fuzzer, triaged",thread:issue:157217:issue:169538:issue:183887:pr:184990,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 3 open issues; raw issue rows remain available for audit.,"#157217 [dynamo] dynamo is unable to enter `except RuntimeError` while eager can, #169538 `torch.compile` fails on `addmm_` with int64 tensors due to fake-tensor meta error (`expand: attempting to expand a dimension of length 2 -> 1`), #183887 torch.compile fails on int64 addmm/addmm_ comparison with fake tensor RuntimeError",97e56b7252d3e2caff03b81fe9118cd2258af1db885d7a016ecb7b3e7ed97624
issue-collapse:2e84c45e4e7513cf,Review thread: 2 issues + 2 PRs - pr #173948: Fixes LazyConvTranspose modules when used with torch.compile(dynamic=True),linked work thread,waiting on contributor,44,high,work_thread_graph,2,2,0,0,"167191, 173252","173948, 185050",,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:167191:issue:173252:pr:173948:pr:185050,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #173948; keep related issue/PR context attached.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#167191 [dynamo] [dynamic] `nn.LazyLinear` reject SymInt when direct using `torch.compile()`, #173252 [dynamo] LazyConvTranspose1d fails under torch.compile(dynamic=True) with “SymIntArrayRef expected to contain only concrete integers”",26df713b3f57669c22c86f12c107c4f40ec4bff5c22923f4cff4ff2ae011d47a
issue-collapse:1554a3d3e41ed255,Review thread: 2 issues + 1 PR - pr #189224: [xpu][skip] Skip test_combo_kernel_dynamic_scale_rblock on XPU,linked work thread,PR blocked,44,high,work_thread_graph,2,1,0,0,"187810, 187811",189224,,,module: xpu,"module: xpu, skipped, triaged",thread:issue:187810:issue:187811:pr:189224,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#187810 DISABLED test_combo_kernel_dynamic_scale_rblock (__main__.ComboKernelTestsPerSubkernelBlocks), #187811 DISABLED test_combo_kernel_dynamic_scale_rblock (__main__.ComboKernelTests)",5817e1e77a60e90b919bc008551852981f7d43cf54906ed69b8da1b7794463fd
issue-collapse:abd3af066852bbd0,Review thread: 2 issues + 1 PR - pr #189215: [MPS] Add `largeTensorTest` to `test_group_norm_backward_large_input`,linked work thread,waiting on contributor,44,high,work_thread_graph,2,1,0,0,"186605, 186867",189215,,,"module: flaky-tests, module: macos, module: mps","module: flaky-tests, module: macos, module: mps, skipped, triaged",thread:issue:186605:issue:186867:pr:189215,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #189215; keep related issue/PR context attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#186605 DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_False_mps_float16 (__main__.TestConsistencyMPS), #186867 DISABLED test_group_norm_backward_large_input_float16_trigger_32bit_overflow_True_mps_float16 (__main__.TestConsistencyMPS)",29d7cb1bf6c198c7e7e23c159996721b05fd2deadd71212285a476ffa6803b4d
issue-collapse:10aa38964a67467c,Review thread: 2 issues + 1 PR - pr #189145: [Dynamo] Preserve Stream/Event subclass types in graph reconstruction,linked work thread,has linked issue,44,high,work_thread_graph,2,1,0,0,"188325, 188405",189145,,,"module: cuda, module: dynamo, oncall: pt2","bot-triaged, module: cuda, module: dynamo, oncall: pt2, triaged",thread:issue:188325:issue:188405:pr:189145,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#188325 torch._dynamo: EventVariable.python_type() returns torch.Event instead of type(self.value), #188405 torch.compile reconstructs a graph-created torch.Stream/torch.Event subclass as the base class, losing the subclass type across a graph break",4bc376cf67067bad22afba9d7a130391d94754238bb0aae0b6ae0e8862e30bce
issue-collapse:f32d97f3f5541749,Review thread: 2 issues + 1 PR - pr #188636: Handle reciprocal SymPy powers in shape range eval,linked work thread,PR blocked,44,high,work_thread_graph,2,1,0,0,"188422, 188423",188636,,,"module: dynamic shapes, module: inductor, module: sdpa, oncall: pt2","bot-triaged, module: dynamic shapes, module: inductor, module: sdpa, oncall: pt2, triaged",thread:issue:188422:issue:188423:pr:188636,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#188422 Some test_sdpa_can_compile_dynamic in transformers fail (with torch.compile(dynamic=True)) with torch 2.13 RC but pass with torch 2.12, #188423 test_sdpa_can_compile_dynamic in transformers fails (with torch.compile(dynamic=True)) with torch 2.13 RC but pass with torch 2.12",b62ac1a52444c94b896ff1a3c3ee6aaf3ecd34d1efb128d302fc592ebd9effcb
issue-collapse:cf383a7a206ace33,Review thread: 2 issues + 1 PR - pr #187353: Preserve signed zero in Triton neg lowering,linked work thread,PR blocked,44,high,work_thread_graph,2,1,0,0,"187336, 187721",187353,,,"module: correctness (silent), module: edge cases, module: half, module: inductor, module: loss, module: reductions, module: special, oncall: pt2","bot-triaged, module: correctness (silent), module: edge cases, module: half, module: inductor, module: loss, module: reductions, module: special, oncall: pt2, triaged",thread:issue:187336:issue:187721:pr:187353,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#187336 [Inductor] torch.compile loses signbit in torch.special.log_ndtr, causing discrepancies in sign-sensitive ops (e.g., copysign, atan2), #187721 [Inductor] Silent correctness : F.kl_div reduction over -0.0 yields different signbits in Eager vs Compiled",802989af87d93a4560019ecf6f3d8816d036b999a973908443fc4e49c540125b
issue-collapse:da1c5b884f1ecc74,Review thread: 2 issues + 1 PR - pr #186403: Avoid CUDA init during CPU torch.compile,linked work thread,PR blocked,44,high,work_thread_graph,2,1,0,0,"97894, 129131",186403,,,"module: aotdispatch, module: dynamo, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: dynamo, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:129131:issue:97894:pr:186403,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#97894 Dropout traces poorly with AotAutograd/make_fx, #129131 Torch compile initialises CUDA context, even compiling CPU functions",a9519944f08ab3aaa19083378e70150e15af0a6ea520c39edd5a8ddeacf33846
issue-collapse:ae44c40a13da5f21,Review thread: 2 issues + 1 PR - pr #185718: Fix fake functorch-wrapped tensor handling in jacfwd vmap,linked work thread,ready for maintainer decision,44,high,work_thread_graph,2,1,0,0,"151196, 172428",185718,,,"module: custom-operators, module: functorch, module: pt2-dispatcher, oncall: pt2","dynamo-functorch, module: custom-operators, module: functorch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:151196:issue:172428:pr:185718,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#151196 compile of vmap of jacfwd fails, #172428 `torch.compile` on a `custom_op`'s fake when using `torch.func.vjp` crashes",a2e2c2bd54fa5e5b8b8bb33d408e238e13321440a6eae86c5fa684bf26b9c519
issue-collapse:2e6c5b22a85f1008,Review thread: 2 issues + 1 PR - pr #185369: Fix dynamic output sizes for adaptive avg pool lowering,linked work thread,ready for maintainer decision,44,high,work_thread_graph,2,1,0,0,"159550, 185575",185369,,,"module: dynamic shapes, module: inductor, module: interpolation, module: pooling, oncall: pt2","bot-triaged, internal ramp-up task, module: dynamic shapes, module: inductor, module: interpolation, module: pooling, oncall: pt2, triaged",thread:issue:159550:issue:185575:pr:185369,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#159550 _adaptive_avg_pool2d does not support dynamic shapes？, #185575 [inductor] Upsample(mode='area') fails under torch.compile(dynamic=True) — ""cannot determine truth value of Relational"" in _adaptive_avg_pool2d lowering",d52cee4f495890bfd23f7cd8d2c1f8886a2492f091dc774511330a5c4c83206d
issue-collapse:8710dc4b3652b3c6,Review thread: 2 issues + 1 PR - pr #185043: Functionalize mutations in deferred local_map,linked work thread,waiting on contributor,44,high,work_thread_graph,2,1,0,0,"167568, 182403",185043,,,"module: aotdispatch, module: functionalization, module: higher order operators, module: performance, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: aotdispatch, module: functionalization, module: higher order operators, module: performance, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:167568:issue:182403:pr:185043,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185043; keep related issue/PR context attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#167568 Autoparallel's local_map doesn't functionalize triton kernels properly, #182403 local_map deferred inlining can expose index_put_ and force default_partition fallback",769549066a8410b1bc942ba1b853664ef135806f03280891d0e4a7374ffdd562
issue-collapse:24789c942f9787d1,Review thread: 2 issues + 1 PR - pr #184958: Prevent escaped vjp TensorWrappers from reaching AOTAutograd,linked work thread,waiting on contributor,44,high,work_thread_graph,2,1,0,0,"172026, 179510",184958,,,"module: autograd, module: functorch, oncall: pt2","module: autograd, module: functorch, oncall: pt2, triaged",thread:issue:172026:issue:179510:pr:184958,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184958; keep related issue/PR context attached.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#172026 `torch.compile(torch.func.vjp)` crashes when creating a new `Tensor`, #179510 RuntimeError: Argument 'primals_out' of Node 'results' was used before it has been defined! Please check that Nodes in the graph are topologically ordered",d0af4d1b7dff49c9104e088f3a6bee3a7940afd8a5fde4968ca1b68dfc3326ec
issue-collapse:be5c6a5e19023307,Review thread: 2 issues + 1 PR - pr #184639: Support None returns from torch.cond,linked work thread,waiting on contributor,44,high,work_thread_graph,2,1,0,0,"181891, 182036",184639,,,"module: cuda graphs, module: higher order operators, module: pt2-dispatcher, oncall: pt2","bot-triaged, enhancement, module: cuda graphs, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:181891:issue:182036:pr:184639,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184639; keep related issue/PR context attached.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#181891 torch.cond()'s true_fn and false_fn are not allowed to return None, #182036 torch.while_loop()'s body_fn is not allowed to return nothing",ea6da55dc8687bdea87842e1592342ea00782d20a227f99e748a1d5f713fae9f
issue-collapse:9b9fd67a4588e713,Review thread: 2 issues + 1 PR - pr #183908: Fix deferred runtime assert symbol availability,linked work thread,ready for maintainer decision,44,high,work_thread_graph,2,1,0,0,"154146, 176770",183908,,,"module: dynamic shapes, module: inductor, oncall: pt2","module: dynamic shapes, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:154146:issue:176770:pr:183908,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#154146 inductor unbacked codegen results in undefined inputs, #176770 torch.compile error in unit tests, but test passes when ran individually",a4ea7685071c62fdb753d0b6c202bccb9ed6de4651362ae0365fafb3274de41e
issue-collapse:5c414e0658d482da,Review thread: 2 issues + 1 PR - pr #181895: [CUDA] Fix linspace integer dtype precision mismatch with CPU,linked work thread,waiting on contributor,44,high,work_thread_graph,2,1,0,0,"180156, 181807",181895,,,"module: cpu, module: cuda, module: edge cases, module: numerical-stability, module: tensor creation","bot-mislabeled, bot-triaged, module: cpu, module: cuda, module: edge cases, module: numerical-stability, module: tensor creation, triaged",thread:issue:180156:issue:181807:pr:181895,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #181895; keep related issue/PR context attached.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#180156 GPU two-pass variance: intermediate squared values overflow float32. CPU Welford online algorithm avoids overflow., #181807 torch.linspace with integer dtype produces inconsistent results on CPU and CUDA",56da3f369a9a75c961382bd11adff6ae2a396d1d937d42fa81200a1213b669d9
issue-collapse:b6d774cfac10ff79,Review thread: 1 issue + 2 PRs - pr #186205: [dynamo] Add as_strided_ handler to fix metadata propagation (#185888),linked work thread,waiting on contributor,44,high,work_thread_graph,1,2,0,0,185888,"186205, 187890",,,"module: dynamo, module: meta tensors, oncall: pt2","bot-triaged, module: dynamo, module: meta tensors, oncall: pt2, triaged",thread:issue:185888:pr:186205:pr:187890,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186205; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#185888 [Inductor] as_strided_ metadata mutation is not reflected on original tensor,2f9fa9f63b67bd911c3478ba407acde06b85a1771970cc2d0da6f55800c04367
issue-collapse:f442af3bec73e825,Review thread: 1 issue + 2 PRs - pr #185989: Bump onnx to 1.22.0rc1 to integrate ONNX 1.22.0 release branch,linked work thread,draft/noise,44,high,work_thread_graph,1,2,0,0,185955,"185989, 186969",,,module: onnx,"bot-triaged, enhancement, module: onnx, triaged",thread:issue:185955:pr:185989:pr:186969,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#185955 Integrate with ONNX 1.22.0 release branch,5716615116cea484b858c28a8fc6584bd30e41c82b537ff179f68d248899c900
issue-collapse:5eccccbc43fcbef4,Review thread: 1 issue + 2 PRs - pr #184308: Route privateuse1-overrideable ops through AOTI dispatcher (,linked work thread,draft/noise,44,high,work_thread_graph,1,2,0,0,184195,"184308, 186920",,,"module: aotinductor, module: privateuse1, module: sdpa, oncall: export, oncall: pt2","bot-triaged, module: PrivateUse1, module: aotinductor, module: sdpa, oncall: export, oncall: pt2, triaged",thread:issue:184195:pr:184308:pr:186920,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#184195 aoti_torch_cuda__scaled_dot_product_fused_attention_overrideable not implemented,06b12f65f00980d95cc41982ad8e3da51dd0fbce2d3f8627ec0ad58d0628c48a
issue-collapse:df8fd4cb1cfb93a5,Review thread: 1 issue + 2 PRs - pr #181096: inductor/flex_attention: fix fake block-mask generator for autotuner,linked work thread,waiting on contributor,44,high,work_thread_graph,1,2,0,0,180651,"181096, 186857",,,"module: flex attention, module: inductor, module: performance, oncall: pt2","bot-triaged, module: flex attention, module: inductor, module: performance, oncall: pt2, repro:agent:h100:repro_fail:skipped, triaged",thread:issue:180651:pr:181096:pr:186857,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #181096; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#180651 [Inductor][flex attention] flex_attention autotuner uses misleading block-mask, may causing overhead and slower configs",86e7b6ea20c985b66ee211b38633ac1d0f8c1f4d750c8631e75218f7b4e5df06
issue-collapse:4e8adca274680370,Review thread: 1 issue + 2 PRs - pr #179372: [Inductor] Fix view_as_complex stride requirement in Inductor backward by forcing contiguous layout,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,2,0,0,179368,"179372, 183862",,,"module: complex, module: inductor, module: memory format, oncall: pt2","bot-triaged, module: complex, module: inductor, module: memory format, oncall: pt2, triaged",thread:issue:179368:pr:179372:pr:183862,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#179368 torch.compile + view_as_complex: RuntimeError: Tensor must have a last dimension with stride 1,659d963908fa1adb980ba470c7b7d8b05cd7e7b7622d62bbd11e96c5db97b222
issue-collapse:48051b42faa031a1,Review thread: 1 issue + 2 PRs - pr #167224: [FlopCounter] Add skip_unsupported parameter for graceful handling of unsupported HOPs,linked work thread,waiting on contributor,44,high,work_thread_graph,1,2,0,0,134385,"167224, 187592",,,"module: flop counter, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: flop counter, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:134385:pr:167224:pr:187592,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #167224; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#134385 FlopCounterMode doesn't support HOP,489d698e0f51efb39173504406a1716d7412ccd837f9c8c1447bc5796f200f58
issue-collapse:b114cfca320bd6d0,Review thread: 1 issue + 1 PR - pr #189317: [doc] Add accelerator test reuse guide (#181474),linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,181474,189317,,,"module: docs, module: openreg, module: tests","bot-triaged, feature, module: docs, module: openreg, module: tests, triaged",thread:issue:181474:pr:189317,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#181474 Add a document for reusing the testcases in PyTorch.,ae102f0da3dbae80be30d59e1342710d263444cdd1b4b1618f7fa637301afdf1
issue-collapse:a87ecceb2aeb5aca,Review thread: 1 issue + 1 PR - pr #189313: Support aten::cumprod ONNX export via CumProd (opset 26),linked work thread,draft/noise,44,high,work_thread_graph,1,1,0,0,102948,189313,,,module: onnx,"OSS contribution wanted, module: onnx, triaged",thread:issue:102948:pr:189313,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#102948 [onnx] aten::cumprod cannot be exported to ONNX,1700ca7a06ec454607fd512e36a7a085dd504c886e9dff0124d2ef12777a34d8
issue-collapse:39a30ecb8560b071,Review thread: 1 issue + 1 PR - pr #189288: [inductor] Fix compute_ancestors KeyError from self-referential mutation copy,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,189271,189288,,,"module: custom-operators, module: dynamic shapes, module: inductor, module: pt2-dispatcher, module: viewing and reshaping, oncall: pt2","bot-triaged, has workaround, module: custom-operators, module: dynamic shapes, module: inductor, module: pt2-dispatcher, module: viewing and reshaping, oncall: pt2, triaged",thread:issue:189271:pr:189288,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189271 [inductor] KeyError in Scheduler.compute_ancestors: self-edge from realizing a reshaped in-place-mutated buffer,8c24987cd8dc0550c284b7ef719684f029a67447507e67f1c94269252dccbf36
issue-collapse:13618e9c15b4f225,Review thread: 1 issue + 1 PR - pr #189158: Skip GHA stats artifacts from other workflow attempts,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,189157,189158,,,module: ci,"bot-triaged, module: ci, triaged",thread:issue:189157:pr:189158,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#189157 GHA stats artifact downloader does not skip mismatched workflow run attempts,52ef5c54e267013054e148b947574a4a20b66db0667f8a41dc3bbfbd7948454c
issue-collapse:90ea2e497e92afb6,Review thread: 1 issue + 1 PR - pr #189142: fx: Fix incorrect return_annotation for tuple types in operator schemas,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,189106,189142,,,module: fx,"bot-triaged, module: fx, triaged",thread:issue:189106:pr:189142,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#189106 `get_signature_for_torch_op` reports incorrect `return_annotation` for `tuple`s,233fe615bd95aae2b2f12c11b2fee7147b8dd1bd9153130e93d1dd2e60cc6c88
issue-collapse:ffdcce76b81b7a8e,"Review thread: 1 issue + 1 PR - pr #189141: [test] Add opt-in sign-sensitive special-value parity check (copysign, atan2, reciprocal)",linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188680,189141,,,"module: tests, oncall: pt2","bot-triaged, enhancement, module: tests, oncall: pt2, triaged",thread:issue:188680:pr:189141,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188680 [test] OpInfo gap: allclose-based compile-vs-eager checks are blind to signed-zero and signed-infinity divergences,89aa3fd45d8577cb6f0d78372d45051fd1e0d645b4b45ef35fa6a95e77e963e1
issue-collapse:a49365085d65a97c,Review thread: 1 issue + 1 PR - pr #189127: [inductor] Decompose broadcast-bias baddbmm before pointwise,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,187093,189127,,,"module: inductor, module: performance, oncall: pt2","bot-triaged, enhancement, module: inductor, module: performance, oncall: pt2, triaged",thread:issue:187093:pr:189127,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#187093 [inductor] baddbmm with broadcast bias: autotune never considers the bmm + pointwise-bias decomposition (1.3-1.45x left on the table at MoE expert-GEMM shapes),fe9b1845833792812646c6963ce5f501faa6b6f84295077bdc38b8fb32e2bba3
issue-collapse:800d99ec59b34be9,Review thread: 1 issue + 1 PR - pr #189119: Fix test-jsons artifact upload path,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,189118,189119,,,"module: ci, module: tests","bot-triaged, module: ci, module: tests, triaged",thread:issue:189118:pr:189119,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#189118 tools/testing: upload test-jsons artifact from the generated JSON zip,40f207659478689a5ab9659d13deac56cefcbd9119a0bd4b5103d1e4f0722bfe
issue-collapse:ae6fc12f814bfbe0,Review thread: 1 issue + 1 PR - pr #189038: [MPS] Fix uint8 casting in MPS `copy_` kernels,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,160744,189038,,,"module: correctness (silent), module: edge cases, module: mps","module: correctness (silent), module: edge cases, module: mps, triaged",thread:issue:160744:pr:189038,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #189038; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#160744 MPS copy_ produces incorrect output when input has more than one element,0927d2734e3978cc09d84b614b4c9208238cb33b3255dc73d1c115983b1f919f
issue-collapse:b5a1a1c79a0910c5,Review thread: 1 issue + 1 PR - pr #189009: [xpu][fix] Enable test_fallback_kernel_with_symexpr_output on XPU with correct head_dim,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188131,189009,,,"module: inductor, module: xpu, oncall: pt2","bot-triaged, module: inductor, module: xpu, oncall: pt2, skipped, triaged",thread:issue:188131:pr:189009,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188131 DISABLED test_fallback_kernel_with_symexpr_output_xpu (__main__.AOTInductorTestABICompatibleGpu),e57f40c85b9777e33ec6f59c94ff053feda4c775433c854e3b008f922c3e762e
issue-collapse:418692834441a3b0,Review thread: 1 issue + 1 PR - pr #188994: [dynamic shapes] Preserve algebraic relationships in foreign unbacked SymInt transfer (#188723),linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,188723,188994,,,"module: dynamic shapes, module: faketensor, module: fx, module: pt2-dispatcher, oncall: pt2","bot-triaged, enhancement, function request, module: dynamic shapes, module: fakeTensor, module: fx, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:188723:pr:188994,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188994; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#188723 ShapeEnv lacks a metadata-tree transfer API that preserves foreign unbacked SymInt relationships,cf639701a216581f99c50184c7431882f76aef112f2f281a3a9e7d26a3316295
issue-collapse:b9966e2b5620bf6d,Review thread: 1 issue + 1 PR - pr #188990: [FSDP] Fix to_cpu when flat_param is already on CPU,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,188989,188990,,,"module: fsdp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: fsdp, oncall: distributed, oncall: distributed parallelisms, triaged",thread:issue:188989:pr:188990,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188989 [FSDP] summon_full_params(offload_to_cpu=True) fails with CPU-offloaded FSDP model,6130a975f35d0feff2f8cce0b43dd57f18098d37eba7f847ddab3929031589fe
issue-collapse:d7a48223422fcd8e,Review thread: 1 issue + 1 PR - pr #188980: Fix inconsistent accelerator guard in checkpoint staging test,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188790,188980,,,"oncall: distributed, oncall: distributed checkpointing","bot-triaged, oncall: distributed, oncall: distributed checkpointing, triaged",thread:issue:188790:pr:188980,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188790 Inconsistent device check in test_staging.py: guard uses torch.cuda.is_available() while block body uses torch.accelerator.is_available(),62adc13d6801e421ebd43a283d7afe9e2663b76d5329b52e5b0e32976c26cca7
issue-collapse:0594590f7806c505,Review thread: 1 issue + 1 PR - pr #188913: [inductor] Add triton.cudagraph_max_recorded_sizes to cap dynamic shape re-recording,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188150,188913,,,"module: compile ux, module: cuda graphs, module: dynamic shapes, module: performance, oncall: pt2","bot-triaged, enhancement, has workaround, module: compile ux, module: cuda graphs, module: dynamic shapes, module: performance, oncall: pt2, triaged",thread:issue:188150:pr:188913,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188150 torch.compile reduce-overhead: CUDAGraphs recompiles on every batch with dynamic padding (HF training loop),49da78a516ce4418efa93512649e8a7c9209b0e7eabe7ce97130a2c4f83a6ee5
issue-collapse:296248f8806415f5,Review thread: 1 issue + 1 PR - pr #188862: [inductor] Fix bool minimum/maximum crash in index propagation (#188862) (#188862),linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188230,188862,,,"module: inductor, oncall: pt2","bot-triaged, module: inductor, oncall: pt2, triaged",thread:issue:188230:pr:188862,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188230 [inductor] `index_propagation` `ValueError: The argument 'False' is not comparable`,bd5be2d3375c78258c535e248ff3745577f69a6c043ae19d4d716fb9ab40133d
issue-collapse:9de023a775aa2ab5,Review thread: 1 issue + 1 PR - pr #188828: Skip conv_transpose3d redispatch test on Windows CI - flaky test,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188757,188828,,,"module: __torch_function__, module: ci, module: cuda, module: tests, module: windows","bot-triaged, module: __torch_function__, module: ci, module: cuda, module: tests, module: windows, triaged",thread:issue:188757:pr:188828,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188757 Windows CI: test_overrides.py class TestTorchFunctionRedispatchOpsCUDA fails consistently (same as #182869 on Linux),2269c4cf2fd7f4daaeb53db4feb20b93b98b809c181d2c1ba2ab208699d99767
issue-collapse:61ba39700aeda090,Review thread: 1 issue + 1 PR - pr #188795: Fix `torch.asarray` to preserve an input's device when a default device is set,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,150199,188795,,,"module: python array api, module: python frontend","module: python array api, module: python frontend, triaged",thread:issue:150199:pr:188795,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#150199 `asarray`: device does not propagate from input to output after `set_default_device`,508c5e46e480b995c26fde759ad9841d47d6166a07f697ee0311829dd86ce1b5
issue-collapse:053c97850916c330,Review thread: 1 issue + 1 PR - pr #188784: Fix `torch.asarray` for buffer-protocol objects with no dtype,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188128,188784,,,"module: correctness (silent), module: python array api, module: python frontend, module: tensor creation","bot-triaged, module: correctness (silent), module: python array api, module: python frontend, module: tensor creation, triaged",thread:issue:188128:pr:188784,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#188128 `asarray` does not work for boolean dtype when using buffer protocol,53b43cd5bbe7adb21344a002c87df3733ee27c5c1a10a56f41a0306be88e02ad
issue-collapse:4b6959406141427f,Review thread: 1 issue + 1 PR - pr #188783: [xpu][fix]Fix bmm_outer_product override to support XPU tensors,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188773,188783,,,module: xpu,"module: xpu, skipped, triaged",thread:issue:188773:pr:188783,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188773 DISABLED test_cow_inputs_accepted_by_override (__main__.TestBmmOuterProduct),5f1fdaa12c54bfb5605864df78fb27117c800b9dd1c51572cd545df5aa75628e
issue-collapse:148a45484d0584bd,Review thread: 1 issue + 1 PR - pr #188719: Fix Inductor KeyError on view_as_complex of non-contiguous CUDA tensors,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,188718,188719,,,"module: complex, module: cuda, module: inductor, oncall: pt2","bot-triaged, has workaround, module: complex, module: cuda, module: inductor, oncall: pt2, triaged",thread:issue:188718:pr:188719,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188718 torch.compile (Inductor) fails with KeyError: 'complex64' on view_as_complex of a valid non-contiguous CUDA tensor,5e38732d63de21b0d582c30ad727497921676c0a11551133657d772ea3187830
issue-collapse:c33b5ba0f429d112,Review thread: 1 issue + 1 PR - pr #188632: Fix Dynamo opaque object staticmethod guards,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,188544,188632,,,"module: dynamo, module: guards, oncall: pt2","bot-triaged, module: dynamo, module: guards, oncall: pt2, triaged",thread:issue:188544:pr:188632,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#188544 [opaque object] Value Opaque Object errors with instance static method.,1a2038f950b1f3dea055bfb110acafe8d01d6e107ced05b6e758bfff7b5aa01c
issue-collapse:5aaace5de5c50283,Review thread: 1 issue + 1 PR - pr #188327: [test] Add error_inputs for nn.Conv2d module,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,174183,188327,,,"module: nn, module: tests","actionable, bot-triaged, enhancement, module: nn, module: tests, triaged",thread:issue:174183:pr:188327,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#174183 [test] Add error_inputs for nn.Conv2d module,b9c1165974ac79fb39ef01ee298ea9bc3789238bf17adbc1182933c0426d3f17
issue-collapse:a4898f8c900d731f,Review thread: 1 issue + 1 PR - pr #188066: test: add F.normalize zero-input correctness tests (fixes #184575),linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,184575,188066,,,"module: autograd, module: edge cases, module: norms and normalization","bot-triaged, module: autograd, module: edge cases, module: norms and normalization, triaged",thread:issue:184575:pr:188066,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#184575 `F.normalize` returns ~1e12 gradient at zero-vector input instead of NaN, while the forward pass already returns NaN",7a1a6671110d8449495427c871df8a5ba53134768a126865088cb1665773099a
issue-collapse:d74020d3fb82c825,Review thread: 1 issue + 1 PR - pr #188053: test: add svdvals NaN propagation test (fixes #187759),linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,187759,188053,,,"module: correctness (silent), module: edge cases, module: linear algebra, module: nans and infs, module: numpy","bot-triaged, module: NaNs and Infs, module: correctness (silent), module: edge cases, module: linear algebra, module: numpy, triaged",thread:issue:187759:pr:188053,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#187759 torch.linalg.svdvals swallows NaN (returns finite values), disagreeing with torch.linalg.svd; matrix_rank then reports full rank for a NaN matrix",6b29c0334022a2adb733c13b83c9b11b07a26ddff2fc29395be1687675a79424
issue-collapse:8e4059f5baced801,Review thread: 1 issue + 1 PR - pr #188049: Fix Inductor div floor nonfinite CUDA results,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,188048,188049,,,"module: correctness (silent), module: cuda, module: edge cases, module: inductor, oncall: pt2","bot-triaged, module: correctness (silent), module: cuda, module: edge cases, module: inductor, oncall: pt2, triaged",thread:issue:188048:pr:188049,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#188048 [Inductor][CUDA] torch.div(rounding_mode=""floor"") mismatches eager for infinite inputs",49718b87bbcfacba0831560a2e754ba4f20285971a6bc44817a2128c8bc20ef7
issue-collapse:f2d722d61099a9b6,Review thread: 1 issue + 1 PR - pr #187956: Propagate FX subgraph replacement meta values,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,187945,187956,,,"module: fx, oncall: pt2","bot-triaged, has workaround, module: fx, oncall: pt2, triaged",thread:issue:187945:pr:187956,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#187945 `replace_pattern` and `replace_pattern_with_filters` do not propagate `meta[""val""]` to replacement nodes, causing downstream filters to silently reject matches",3141d8667d183c78b069e8f512f71ab2e60d17207a130adaee11da871795e49b
issue-collapse:db53c8fd2e59341e,Review thread: 1 issue + 1 PR - pr #187614: Use sym_numel in linalg_cond string overload,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,186028,187614,,,"module: dynamic shapes, module: faketensor, module: linear algebra, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: dynamic shapes, module: fakeTensor, module: linear algebra, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:186028:pr:187614,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#186028 [dynamic] torch.linalg.cond(p=""fro"") fails with symbolic sizes while dynamic=False succeeds",e9dcde827e25c22e40e49f36d71ebda116873f74f896e0e3b34498f1c7befafe
issue-collapse:3a42098eb1083c2a,Review thread: 1 issue + 1 PR - pr #187604: Fix fake slice pending unbacked replacements,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,185185,187604,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:185185:pr:187604,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185185 torch.compile(fullgraph=True) fails with PendingUnbackedSymbolNotFound when tensor-derived scalar values are used as slice bounds,e08e6f89dae68c9a6784d0a72c8d361e4982092773b2c8bc1f5d4c214b7a740c
issue-collapse:7c140abb41332feb,Review thread: 1 issue + 1 PR - pr #187601: [functorch] Fix dynamic layers under inference_mode,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,177750,187601,,,"module: correctness (silent), module: functionalization, module: functorch, module: pt2-dispatcher, module: vmap, oncall: pt2","bot-triaged, inference mode, module: correctness (silent), module: functionalization, module: functorch, module: pt2-dispatcher, module: vmap, oncall: pt2, triaged",thread:issue:177750:pr:187601,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#177750 [functorch] vmap and functionalize under inference_mode,700948919cc880313ba376a3f7ba3a4b7705e9f528a9102a130ed616a9d26186
issue-collapse:acab68c4ce0e1c96,Review thread: 1 issue + 1 PR - pr #187452: Polyhedral Optimization Pass for PyTorch Inductor (Phase 1),linked work thread,draft/noise,44,high,work_thread_graph,1,1,0,0,187451,187452,,,"module: inductor, module: performance, oncall: pt2","bot-triaged, feature, large, module: inductor, module: performance, oncall: pt2, triaged",thread:issue:187451:pr:187452,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#187451 RFC: Polyhedral Optimization Pass for PyTorch Inductor,eb475c6a442d6dd157a22f78ff79b26f1b517bd76075b6157518afa176d29cbb
issue-collapse:4664251506b3182b,Review thread: 1 issue + 1 PR - pr #187242: Validate LocalResponseNorm parameters to prevent silent inf output,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,187177,187242,,,"module: error checking, module: nn, module: norms and normalization","bot-triaged, enhancement, module: error checking, module: nn, module: norms and normalization, triaged",thread:issue:187177:pr:187242,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#187177 torch.nn.LocalResponseNorm returns +/-Inf when k=0 and alpha=0 (zero denominator, no validation)",6523960cf428619c04f4b03b4c740c2e7f4d0f4b8bc706d2b74c25b0477e01ba
issue-collapse:20c62b890462fbc7,Review thread: 1 issue + 1 PR - pr #187171: [inductor] Fix test_comprehensive_addbmm_cpu_float16 precision mismatch (#187170),linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,187170,187171,,,"module: half, module: inductor, module: tests, oncall: pt2","bot-triaged, module: half, module: inductor, module: tests, oncall: pt2, triaged",thread:issue:187170:pr:187171,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#187170 [inductor] test_comprehensive_addbmm_cpu_float16 fails due to overly strict float32 reference comparison,bc0cfcfbe4a699464bbd14e9d2a21026760a599594bc6cbbee615c335942a5f4
issue-collapse:2b54067f5716b1d4,Review thread: 1 issue + 1 PR - pr #187161: [FSDP] Cache post-forward DeviceMesh to avoid duplicate NCCL communicators per layer,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,187155,187161,,,"module: fsdp, module: nccl, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: fsdp, module: nccl, oncall: distributed, oncall: distributed parallelisms, triaged",thread:issue:187155:pr:187161,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #187161; keep related issue/PR context attached.,4,Represents 1 open issue; raw issue rows remain available for audit.,#187155 [FSDP] Cache post-forward DeviceMesh to deduplicate NCCL communicators,1355503705a1ce45387b83a564e312856031e7788c0ab3e052318e353f3e2c77
issue-collapse:cc6bc2f2277df1f1,Review thread: 1 issue + 1 PR - pr #187057: Fix value-opaque graph output replay,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,187041,187057,,,"module: dynamo, module: library, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: dynamo, module: library, module: pt2-dispatcher, oncall: pt2, tensor subclass, triaged",thread:issue:187041:pr:187057,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187041 [opaque object] Bug with storing in global dict and consuming in TensorSubclass,fe994d186cbfea2a7df72e95530d19e6a0266f2ec6067cbc74d35430869959cb
issue-collapse:dca79e78d8648a75,Review thread: 1 issue + 1 PR - pr #187012: Support optional Tensor list schemas,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,160094,187012,,,"module: custom-operators, module: pt2-dispatcher, module: vllm, oncall: pt2","actionable, internal ramp-up task, module: custom-operators, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:160094:pr:187012,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160094 torch.library.infer_schema doesn't have full support for list types,ec6c8ca0b581fcdc122ebc322a0e5da7daa004f3acd0fe130c7c4bb0666436f6
issue-collapse:2f49742484a712e3,Review thread: 1 issue + 1 PR - pr #186953: Fix CPU amin/amax backward rematerialization mismatch,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,186799,186953,,,"module: autograd, oncall: cpu inductor, oncall: pt2","module: autograd, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:186799:pr:186953,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#186799 `torch.compile` silently produces wrong gradient for `atan2(x, y).amin(dim)` on float32",c73a62bf7597c0145e522d1a955d7122befd11e8417f8d839b1def0b7fef3161
issue-collapse:b5038e8fd037b56e,Review thread: 1 issue + 1 PR - pr #186942: [inductor] Round fmod/remainder scalars for bf16/fp16 CPU,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,186875,186942,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2, pt2: numerics-discrepancy, triaged",thread:issue:186875:pr:186942,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#186875 [Inductor] fmod/remainder with Python float scalar diverge from eager for bf16/fp16 on CPU (same class as #185517),a7b37a1a32cc4cec89bd9b16fb66df9b256db550638b3bbfcd0047fcda9860f8
issue-collapse:bba72173e07407d1,Review thread: 1 issue + 1 PR - pr #186923: Graph break on custom op fake tensor failures,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185141,186923,,,"module: dynamo, oncall: pt2","bot-triaged, module: dynamo, oncall: pt2, triaged",thread:issue:185141:pr:186923,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#185141 [Dynamo] torchvision.io functions raise ""RuntimeError when making fake tensor call"" instead of graph breaking",24b91a87411873d06f60682f45853ae897af8b6cde339cb021b773e15c3c0416
issue-collapse:3fd26ecfaa6c246b,Review thread: 1 issue + 1 PR - pr #186913: Fix int64 value_expr arithmetic in Triton codegen,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,183901,186913,,,"module: correctness (silent), module: cuda, module: inductor, module: int overflow, oncall: pt2","bot-triaged, module: correctness (silent), module: cuda, module: inductor, module: int overflow, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:183901:pr:186913,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#183901 torch.compile produces int32-overflowed results for int64 arange multiplication on CUDA,55ac8620240c00f5e1134ac46127a51d95090a42fbb128c936637b0fee77f3e2
issue-collapse:c6f7c11b4ceb527f,Review thread: 1 issue + 1 PR - pr #186859: Recover fake/proxy views from functional tensor bases,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,181304,186859,,,"module: aotdispatch, module: faketensor, module: functorch, module: pt2-dispatcher, oncall: pt2","bot-triaged, dynamo-functorch, module: aotdispatch, module: fakeTensor, module: functorch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:181304:pr:186859,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#181304 torch.compile fails with FakeTensor Assertion Error in torch/_subclasses/fake_tensor.py,d52a3f7d3e99fcbb29093878e26b295d61f139e2e8c245c8c699492b3865af49
issue-collapse:d059f69cbf704d68,Review thread: 1 issue + 1 PR - pr #186845: [make_fx] Decompose detach by default for higher-order grads,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,175477,186845,,,"module: autograd, module: decompositions, module: fx, oncall: pt2","bot-mislabeled, bot-triaged, module: autograd, module: decompositions, module: fx, oncall: pt2, triaged",thread:issue:175477:pr:186845,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175477 make_fx symbolic tracing produces wrong 2nd-order gradients through torch.sqrt,c006507f53dc16f783853b81cade0f4cea56b89224c5bfaf3c22c09442175c34
issue-collapse:90a2d752753eda91,Review thread: 1 issue + 1 PR - pr #186764: fix: Use typing_extensions.TypeAliasType for better reexport of `__module__`,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,171905,186764,,,"module: lint, module: typing","actionable, module: lint, module: typing, triaged",thread:issue:171905:pr:186764,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186764; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#171905 Use typing_extensions.TypeAliasType for better reexport of `__module__`,ec435c72b6c6325618c76443ef153516e80350c445b41b8041caf2304c63a52d
issue-collapse:fc0886609c1246da,Review thread: 1 issue + 1 PR - pr #186656: Autotune XBLOCK for mix-order reductions,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185814,186656,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:185814:pr:186656,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,"#185814 [Inductor] Mix-Order Reduction: XBLOCK Derivation Ignores N in RMSNorm Backward, Causing Bandwidth Cliff",f4838bf43541b9bf3dcf1efaa1f3869242f6ca776b9b672153df10985ab0aa62
issue-collapse:cf3b6a8e95b41f82,Review thread: 1 issue + 1 PR - pr #186561: Preserve autograd metadata for functionalized inplace views,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,90507,186561,,,"module: bootcamp, module: functionalization, module: pt2-dispatcher, oncall: pt2","internal ramp-up task, module: bootcamp, module: functionalization, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:90507:pr:186561,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#90507 Functionalization on inplace_views should properly reflect autograd metadata,6d9646d83c2af286d516d1de99ae0388cc8e9b4a6131f26bfaf143c4f4b20605
issue-collapse:3245105c37334b31,Review thread: 1 issue + 1 PR - pr #186557: Resolve scalar overloads before AOT compilation,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,90923,186557,,,"module: aotdispatch, module: pt2-dispatcher, oncall: export, oncall: pt2","feature, has workaround, module: aotdispatch, module: pt2-dispatcher, months, oncall: export, oncall: pt2, triaged",thread:issue:90923:pr:186557,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#90923 dynamo.optimizations.training.aot_autograd does not trace correct overload,abdff2d50d93384863e29a86e96d9240b63b999d5194ad0f066965044afa55cb
issue-collapse:e64f925b82ac077a,Review thread: 1 issue + 1 PR - pr #186554: Preserve hooks for dependent AOTAutograd outputs,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,91468,186554,,,"module: aotdispatch, module: autograd, module: correctness (silent), module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: autograd, module: correctness (silent), module: pt2-dispatcher, months, no-scrub, oncall: pt2, triaged",thread:issue:91468:pr:186554,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#91468 torch.compile incorrect when imperative autograd APIs are used,39c1d44e7ea4f71054aa4c8ad0ca7be6812d58f587889f51ca2742204c9a7c25
issue-collapse:abfc2dbb49efd759,Review thread: 1 issue + 1 PR - pr #186529: Preserve grad state in Dynamo minifier repros,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,111414,186529,,,"module: minifier, oncall: pt2","module: minifier, needs research, oncall: pt2, triaged",thread:issue:111414:pr:186529,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111414 Minifier doesn't transfer execution states like @torch.no_grad to repro,35d6689d56203df650aeeff285bb91389bafd31e08d79b6c2de3665fe363fa31
issue-collapse:75a6fa884b9b2499,Review thread: 1 issue + 1 PR - pr #186522: Trace Tensor.requires_grad_ in pre-dispatch make_fx,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,113899,186522,,,"module: proxytensor, module: pt2-dispatcher, oncall: pt2","module: ProxyTensor, module: pt2-dispatcher, oncall: pt2, pre_dispatch tracing, triaged",thread:issue:113899:pr:186522,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186522; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#113899 Tensor.requires_grad_() does not trace with make_fx,02937fbe6a176fedaed82457642276f45608bd54d32d70141a5797dd75ca53f9
issue-collapse:f98f7eeca3183667,Review thread: 1 issue + 1 PR - pr #186510: Fix Dynamo low-precision addc opmath,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,116202,186510,,,"module: dynamo, module: optimizer, module: tests, oncall: pt2","module: dynamo, module: optimizer, module: tests, oncall: pt2, triaged",thread:issue:116202:pr:186510,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186510; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#116202 Dynamo'd `test_mixed_device_dtype` needs higher tolerances for Adagrad, RMSProp",ec3cefe04fdce605f43345a22a03d1155ffae402f5d0c484df4193b668804ee2
issue-collapse:7894026d14099387,Review thread: 1 issue + 1 PR - pr #186506: Register missing CIA decomposition py_impls,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,117555,186506,,,"module: aotdispatch, module: decompositions, module: pt2-dispatcher, module: python dispatcher, oncall: pt2","better-engineering, module: aotdispatch, module: decompositions, module: pt2-dispatcher, module: python dispatcher, oncall: pt2, triaged",thread:issue:117555:pr:186506,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#117555 Update decompositions registered to `Autograd` dispatch key to also register `CompositeImplicitAutograd`,2786e044a75148a24d6f74087245fa246024b88a20453d89f549958bfcaa187e
issue-collapse:c5e76370eda5a124,Review thread: 1 issue + 1 PR - pr #186501: Collapse foreach input mutation copies in AOTAutograd,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,119191,186501,,,"module: aotdispatch, module: functionalization, module: mta, module: pt2-dispatcher, oncall: pt2","good first issue, internal ramp-up task, module: aotdispatch, module: functionalization, module: mta, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:119191:pr:186501,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#119191 AOTAutograd: functionalization should emit foreach_copy_() instead of copy_() when possible,e7b1d0e9700e638d22f1e50a5357532e309d8364312115169651f795eb5df5db
issue-collapse:2435e78d74ea620d,Review thread: 1 issue + 1 PR - pr #186493: Optimize AOTAutograd compiled forward wrapper,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,122029,186493,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","better-engineering, module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:122029:pr:186493,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122029 AOTAutograd has high fixed overheads,c6db18301939b3b798cdca63a21f52f60f3eafad21384f9277598de26df0966b
issue-collapse:df2c4abd48d32ae9,Review thread: 1 issue + 1 PR - pr #186478: Fix AOTAutograd resized out tensor mutation replay,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,122386,186478,,,"module: functionalization, module: pt2-dispatcher, oncall: pt2","module: functionalization, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:122386:pr:186478,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122386 [histogram] expand: attempting to expand a dimension of length 6!,640d53a4053819ac2b463cff0d0c23982c5a7fb69a6a373b182898a53d48625e
issue-collapse:50edf2f11f1605f1,Review thread: 1 issue + 1 PR - pr #186470: Raise user error for joint parameter mutation export,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,122860,186470,,,"module: functorch, module: pt2-dispatcher, oncall: export, oncall: pt2","better-engineering, module: functorch, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:122860:pr:186470,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122860 Do not use assert for user triggered issues,68741e5ebdeb5e29d34702596b7c903a8f9e0809e1a207bdba591fe8e3c7f406
issue-collapse:6db540561f311db6,Review thread: 1 issue + 1 PR - pr #186459: Fix fake view strides for AOT group norm tracing,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,124181,186459,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:124181:pr:186459,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124181 AOTAutograd breaks diffusers VAE decoder,1f819d103b469deb4c9588b3da070478d69cf957dd39231167b6345b0ba59325
issue-collapse:7b37b2a16ffa72b0,Review thread: 1 issue + 1 PR - pr #186457: Fix torch.ops float parsing for scalar tensor args,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,124344,186457,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:124344:pr:186457,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124344 Inconsistent argument parsing behavior for torch.ops bindings between Tensor and FakeTensor,f0a33ae16342b9e33626e7f3934afab4c5da8373c87fe1a48294e12bab021411
issue-collapse:0f45880325d5f35a,Review thread: 1 issue + 1 PR - pr #186431: Fix ModuleTracker with Dynamo runtime modules,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,127112,186431,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:127112:pr:186431,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#127112 [dynamo] Module tracker test errors in dynamo at runtime,b1bf7edf48ca64031b48b6f2a08fe1a4b72e7cfd840ada9b632e83987deb36d2
issue-collapse:5e1860101ebb5630,Review thread: 1 issue + 1 PR - pr #186425: Skip Dynamo inside non-traceable subclass dispatch hooks,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,128160,186425,,,"module: dynamo, module: pt2-dispatcher, oncall: pt2","dynamo-tensor-subclasses, module: dynamo, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:128160:pr:186425,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#128160 [dynamo] Dynamo traces through __torch_dispatch__ on custom tensor subclasses,9bce8247a4605aba5e74e8b9ad7c9dc9de953162424a18dcd0426e93c79d1a1c
issue-collapse:858afd86247cbd16,Review thread: 1 issue + 1 PR - pr #186388: Avoid native eager router graph breaks under Dynamo,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,186354,186388,,,"module: dynamo, module: graph breaks, oncall: pt2","bot-triaged, has workaround, matrix multiplication, module: dynamo, module: graph breaks, oncall: pt2, triaged",thread:issue:186354:pr:186388,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#186354 Dynamo graph break in native bmm_outer_product eager router due to _is_cow_tensor / call_boxed,b85d405fbf274249494ed28543eb5db6c73faa19ae7ee7338b9169d389fbc660
issue-collapse:a19bcc958141154d,Review thread: 1 issue + 1 PR - pr #186248: Simplify Min/Max using ShapeEnv value ranges,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,137096,186248,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:137096:pr:186248,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137096 Expressions involving Min/Max don't simplify with value ranges information,552439246d5fa95965a70bb641f1fba16032a1286dd1de4c2137e85398b62878
issue-collapse:704a2e5c965a2b0b,Review thread: 1 issue + 1 PR - pr #186246: Fix stale scalar value in addcdiv graph-break decomposition,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185382,186246,,,"module: correctness (silent), module: inductor, oncall: pt2","has workaround, module: correctness (silent), module: inductor, oncall: pt2, triaged",thread:issue:185382:pr:186246,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185382 `torch.compile`: Inductor silently produces wrong results for `addcdiv_`/`addcmul_` when `.item()` graph break occurs in a loop,50f6530dbf4b93d4bf434d31e8d58e57ed3fa66247f5c0f8886f566144e6f3a5
issue-collapse:20778d604f57553f,Review thread: 1 issue + 1 PR - pr #186210: Fix cond branch-local shape assertions,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,138141,186210,,,"module: dynamic shapes, module: higher order operators, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, module: dynamic shapes, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:138141:pr:186210,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138141 torch.cond with shape condition in predicate doesn't allow for assertions,c1366eef8f8685149755ddc07e0f6b3e89815a34c98ff1606e0771b833658e55
issue-collapse:ffa9a7ec64ff3497,Review thread: 1 issue + 1 PR - pr #186201: Deduplicate symbolic cond inputs after export passes,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,138229,186201,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:138229:pr:186201,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138229 [cond] Cond lifts duplicate symint inputs,80d0e565c000ed2c937eee122a800dde16014ef4e0ece468c1e679fbf47366f9
issue-collapse:c33b77ffa267f992,Review thread: 1 issue + 1 PR - pr #186199: Fix functionalization return mapping for mixed mutable out ops,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,138264,186199,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:138264:pr:186199,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138264 `_native_batch_norm_legit`: `aot_eager` backend incorrectly raising an error.,2c02db72023041efd17785b3fc3409545c1eb7d68389b41da214fda62694f268
issue-collapse:5c31195e6df469b0,Review thread: 1 issue + 1 PR - pr #186166: Respect fake/functional modes for backend autograd kernels,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,139707,186166,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:139707:pr:186166,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#139707 Adding backend-specific autograd kernel for aten builtin operators breaks `torch.compile`,1d063b6a0e8ab7f2c032af37ccf64e07b8d0c4e0c82825ff3ec30b017b49afe7
issue-collapse:33c246c15cde58f7,Review thread: 1 issue + 1 PR - pr #186111: Fix stale mutated input history in AOTAutograd,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,140707,186111,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:140707:pr:186111,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140707 mutating a buffer to change its requires_grad-ness can cause RuntimeError,e2eb221d609635369488ec5bb765dc88df6afab91ec23e25dd62585cdfb9dfb4
issue-collapse:2baae740e453cc5e,Review thread: 1 issue + 1 PR - pr #186007: Preserve eager post accumulate hook ordering in AOTAutograd,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,144376,186007,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:144376:pr:186007,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#144376 torch.compile post_accumulate_grad_hook ordering is wrong for tiebreakers,bf512d891e30169f818940be4e506dd5799a6699ac046ec5901c7f92fefa70d0
issue-collapse:50b06af45f4382ef,Review thread: 1 issue + 1 PR - pr #185972: Handle malformed MSVC help output,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,145383,185972,,,"module: windows, oncall: pt2","module: windows, oncall: pt2, triaged",thread:issue:145383:pr:185972,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145383 Windows Pytorch compiler crash some version of cl.exe. Fix provided,f5a28406d34acc2196b6244405666df4fa6de5324893f81004c239c67dbbf629
issue-collapse:828b121cf680f85d,Review thread: 1 issue + 1 PR - pr #185929: Fix Dynamo benchmark model state isolation,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,145773,185929,,,"module: benchmark, module: cuda graphs, module: dynamo, oncall: pt2","PT2-Bug-Bash, module: benchmark, module: cuda graphs, module: dynamo, oncall: pt2, triaged",thread:issue:145773:pr:185929,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145773 Dynamo performance test benchmarks reuse model state between eager and compiled,2c48aa25213e17ca549c67fcdc551a1f77d5fba030a4a42ee634da9d8de2f6a1
issue-collapse:612ad553ea4aa21f,Review thread: 1 issue + 1 PR - pr #185916: Fix FlexAttention nested index backward,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,146896,185916,,,"module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:146896:pr:185916,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146896 Assertion error in Flex Attention backward pass when indexing a parameter,2e65ce2a66bb8d9bd217abc0483365026a6de8013ea56027432fc2f9bbd24f43
issue-collapse:3cc92d9c80e2f0ce,Review thread: 1 issue + 1 PR - pr #185914: Log aot_export fw_metadata instead of printing it,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,147135,185914,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","better-engineering, module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:147135:pr:185914,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#147135 Don't print fw_metadata in ""Found a graph input that requires gradients""",32b9d3f920dd8872f4c013da30ca29ca0cd222f7e52914ffd47fb6529fd79f77
issue-collapse:c28d6efc9ca4ed22,Review thread: 1 issue + 1 PR - pr #185897: Support concrete enum types in torch.library.define schemas,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,183957,185897,,,"module: custom-operators, module: library, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: library, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:183957:pr:185897,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183957 enum custom op input is a little weird with torch.library.define,50e4869818b5b5b58d5cfa5c3881ab7ea9928bbd101d755aef08f3eeaed711c3
issue-collapse:8cddd113f21addb2,Review thread: 1 issue + 1 PR - pr #185878: Fix Inductor expm1 on float subnormal inputs,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185480,185878,,,"module: bfloat16, module: correctness (silent), module: cuda, module: edge cases, module: inductor, oncall: pt2","bot-triaged, module: bfloat16, module: correctness (silent), module: cuda, module: edge cases, module: inductor, oncall: pt2, triaged",thread:issue:185480:pr:185878,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185480 [Inductor][CUDA][bf16] `F.celu` returns `+0.0` where eager returns a negative bf16 subnormal,53260006595fe0011791cea2724f1fd716d629730d7bc324530606e3e0fad5bf
issue-collapse:6b2d96281c36a1c5,Review thread: 1 issue + 1 PR - pr #185876: Prevent saving mutated grad inputs in AOT min-cut,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185497,185876,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:185497:pr:185876,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185497 [pt2] RuntimeError: variable modified by inplace operation during backward in compiled mode (succeeds in eager),dc0ff116c95843af406dce725dd4336971452c8c03090e72a052580063d5506a
issue-collapse:7eef0db486fd9ec9,Review thread: 1 issue + 1 PR - pr #185873: Preserve mutation renames after scheduler recompute,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185509,185873,,,"module: correctness (silent), module: inductor, oncall: pt2","bot-triaged, module: correctness (silent), module: inductor, oncall: pt2, triaged",thread:issue:185509:pr:185873,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#185509 [inductor] Silent incorrect gradients with interpolate, conv2d, max_pool2d and torch.where",5750739faa549d94bded32d89581a6a7b45af910b9cbef1420c732091577a5c7
issue-collapse:9da32de01294b7fd,Review thread: 1 issue + 1 PR - pr #185871: Fix symbolic stride-order sorting without explicit ShapeEnv,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185510,185871,,,"module: dynamic shapes, module: inductor, oncall: pt2","bot-triaged, module: dynamic shapes, module: inductor, oncall: pt2, triaged",thread:issue:185510:pr:185871,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#185510 [Dynamo] TypeError: cannot determine truth value of Relational during stride sorting with dynamic=True,658165d70c2d59ecf8de4b31ff03a12490afdb4211312a69987223be2c39c219
issue-collapse:44235b1bc268bb9f,Review thread: 1 issue + 1 PR - pr #185866: Fix CPU Inductor numerics for BatchNorm erf normalize,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185589,185866,,,"module: correctness (silent), module: dynamic shapes, module: inductor, module: numerical-stability, oncall: pt2","bot-triaged, module: correctness (silent), module: dynamic shapes, module: inductor, module: numerical-stability, oncall: pt2, triaged",thread:issue:185589:pr:185866,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#185589 [inductor] Severe numerical inconsistency: diag_embed + erf + normalize(p=-1) under torch.compile(dynamic=True) — max diff up to 565,248",82d05692cbd18bf2fcd8fae78c183552dfb943e7a7c6ff3d06126c09fb6a8534
issue-collapse:16f5821611cdab5e,Review thread: 1 issue + 1 PR - pr #185861: Fix ViewMeta serialization for special views,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,185630,185861,,,"module: aotdispatch, module: dynamic shapes, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: aotdispatch, module: dynamic shapes, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:185630:pr:185861,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185630 [PT2] RuntimeError with garbage shape in apply_view_meta_sequence when using torch.compile(dynamic=True),2a96092915c54cc3a0fbfef376a1901fdbd23762f76be95a70303a217cbb1d64
issue-collapse:76e349798b4722bb,Review thread: 1 issue + 1 PR - pr #185829: Fix autograd.grad graph breaks under aot_eager,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,145899,185829,,,"module: aotdispatch, module: dynamo, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: dynamo, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:145899:pr:185829,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#145899 torch.compile + torch.autograd.grad: grad can be implicitly created only for scalar outputs,eb00e9e52a2a9a79e0736756650980f1ffd3ca58b1b077bf61e4a8be55f0bb4e
issue-collapse:29011a4b680cf46d,Review thread: 1 issue + 1 PR - pr #185775: Avoid storage access for no-storage tensors in MetaConverter,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,148695,185775,,,"module: autograd, module: faketensor, module: pt2-dispatcher, oncall: pt2","module: autograd, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:148695:pr:185775,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#148695 dynamo fakification errors with opaquetensorimpl,42e22b06f89e9e4f3b10587b3a580503d47cb7d55d966d8e4b85d002d4e1020b
issue-collapse:e4d013703d045a38,Review thread: 1 issue + 1 PR - pr #185762: [Dynamo] Support _AutoDispatchBelowAutograd context manager,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,149586,185762,,,"module: compiled autograd, module: dynamo, module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: compiled autograd, module: dynamo, module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:149586:pr:185762,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149586 UserWarning: Dynamo does not know how to trace the builtin `None.pybind11_object.__new__.`,8c36c15bedb5c72f96fc0f4ca43de2d6c23c6fe7797dc61115f442a6fc3f08ce
issue-collapse:0fb4180ca90d1414,Review thread: 1 issue + 1 PR - pr #185759: Improve float32 flex attention backward tiling,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,149767,185759,,,"module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:149767:pr:185759,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185759; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#149767 `flex_attention` slower than manual attention implementation,2d6cc4eede78d9abf3a98641e8d545c935160f1b5fc4d9484e8f5693c5a11f0f
issue-collapse:a02863dc37d8463c,"Review thread: 1 issue + 1 PR - pr #185751: [nn] Raise ValueError early for invalid (ndim, pad_size) in non-constant F.pad modes",linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,184845,185751,,,"module: docs, module: error checking, module: nn, module: padding","bot-triaged, enhancement, module: docs, module: error checking, module: nn, module: padding, topic: fuzzer, triaged",thread:issue:184845:pr:185751,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#184845 `torch.nn.functional.pad` raises `NotImplementedError` for valid `(ndim, pad_size)` combinations in non-constant modes — constraint not documented in the API",408384f1a67a78a5777e487b050d7fa6d191c132f0639e0ed4df1bf09f271489
issue-collapse:9a766bbb56f7a08d,Review thread: 1 issue + 1 PR - pr #185734: Preserve torch._check messages in runtime asserts,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,150063,185734,,,"module: dynamic shapes, module: dynamo, oncall: pt2","PT2-Bug-Bash, module: dynamic shapes, module: dynamo, oncall: pt2, triaged",thread:issue:150063:pr:185734,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150063 Torch.check does not preserve original error message in the deferred runtime assert,27df9839ae8bc55e3993dcc6af4c063c676e7a7f77669a8b67a703b8ffd54032
issue-collapse:747a3a01aaf38dc6,Review thread: 1 issue + 1 PR - pr #185604: Improve infer_size unbacked broadcast error,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,152656,185604,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:152656:pr:185604,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#152656 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.",c572fd75e226debaddafee09981c9cdaab00abe73920442c18bb6ac36e7cd058
issue-collapse:1cc8a05ea5ae12c8,Review thread: 1 issue + 1 PR - pr #185501: Fix Issue #185396 inductor complex64 triton signature,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185396,185501,,,"module: compile ux, module: complex, module: cuda, module: inductor, oncall: pt2","bot-triaged, module: compile ux, module: complex, module: cuda, module: inductor, oncall: pt2, triaged",thread:issue:185396:pr:185501,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#185396 [Inductor][CUDA] Internal KeyError: 'complex64' during Triton signature generation for graph with complex intermediates,516a5dc5b7c10d639f227bee25c57d04a466c16caee260b2ca60fa7f7e1720c3
issue-collapse:d3529067dfa431c6,Review thread: 1 issue + 1 PR - pr #185304: Fix while_loop stride metadata preservation,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,160939,185304,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:160939:pr:185304,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160939 Strange failure modes of while_loop under torch.compile -- possibly due to internal array stride,f65df2754a757098d4a4eb3d109b6c6ac67cb6c07cb030c401ed4aa7a62b776e
issue-collapse:f7c47e8542731268,Review thread: 1 issue + 1 PR - pr #185231: [C++] Ensure wrapped tensors can be printed,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,185200,185231,,,"module: cpp, module: printing","actionable, module: cpp, module: printing, triaged",thread:issue:185200:pr:185231,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185200 [C++] Print tensors without storage,5ef3457f587815e2b8eb5e65c750aef74e8f58c4aa1cd2b39efaf0668316260c
issue-collapse:a3912bef88eb27ee,Review thread: 1 issue + 1 PR - pr #185054: [autograd] Validate forward-mode jacobian/hessian require vectorize=T…,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,184842,185054,,,"module: autograd, module: error checking, module: forward ad","bot-triaged, enhancement, module: autograd, module: error checking, module: forward ad, small, triaged",thread:issue:184842:pr:185054,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#184842 `torch.autograd.functional.jacobian` raises `NotImplementedError` at execution time for `strategy='forward-mode'` with `vectorize=False` instead of raising `ValueError` at argument validation,bbaff5bff09699681a1716c3d3b5f0d51f055a71818bfa9cfb6f3c0fb929a3a2
issue-collapse:3b8bc8f6db9da656,Review thread: 1 issue + 1 PR - pr #185052: Fix Inductor reuse for disjoint strided slices,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,167098,185052,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:167098:pr:185052,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167098 request for faster torch.compile kernels for casting to nvfp4,7eb3ffb0808d452ecb0930c4a9a2eb444bb77a640059153e7cc94ee4e6952866
issue-collapse:efaccccf09683fbf,Review thread: 1 issue + 1 PR - pr #184995: Fix AOT getitem replay under inference_mode,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,169477,184995,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:169477:pr:184995,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#169477 torch.compile on @torch.inferece_mode() cause RuntimeError: <weakref ... to 'torch.storage.UntypedStorage' ...>,959c87319c3ac3b0004bd0b73fb75f3c448182be3da049d2ec67b647f565bf4f
issue-collapse:cc8bb773221e80c7,Review thread: 1 issue + 1 PR - pr #184981: Fix torch._numpy bool numeric subtract,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,170509,184981,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:170509:pr:184981,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#170509 Dynamo fails on unit tests: test/test_torch.py.test_geometric_kstest,9dcf045f93cae1b8edcdb2f6c96eb09f26f9b61d3720de180d11b7279c9c9d1f
issue-collapse:996a478b4a905aa4,Review thread: 1 issue + 1 PR - pr #184944: Make OpaqueBase usable as a pybind base,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,172869,184944,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:172869:pr:184944,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184944; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#172869 [opaque obj] Figure out subclassing OpaqueBase in pybinded classes,5c76f06658ae14a6497a8ad59af7454e9773f47cf44374218409ea36c769e8ea
issue-collapse:2d1456276152c78b,Review thread: 1 issue + 1 PR - pr #184938: Fix AOTAutograd forward output overwritten by backward mutation,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,172956,184938,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:172956:pr:184938,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184938; keep related issue/PR context attached.,4,Represents 1 open issue; raw issue rows remain available for audit.,#172956 Mutation during backward on tensor saved for backward can create invalid graph,949ee542bab377ce014810813e4c3dd3bd6260fdb2810713291ee73fa9264129
issue-collapse:d90822a71a481b81,Review thread: 1 issue + 1 PR - pr #184928: Fix linalg_pinv tensor tolerance broadcasting,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,173052,184928,,,"module: decompositions, module: faketensor, module: pt2-dispatcher, oncall: pt2","module: decompositions, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:173052:pr:184928,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173052 [Inconsistency] aten.linalg_pinv returns inconsistent output shape between eager and inductor,588fba66b9d85ac5bc5bb45ae8c5fcd958568657ed63887b279f5e113b69b587
issue-collapse:229acb9594ce3d6f,Review thread: 1 issue + 1 PR - pr #184920: [dynamo] Store FunctionIdSet entries by identity weakrefs,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,173578,184920,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, topic: improvements, triaged",thread:issue:173578:pr:184920,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173578 FunctionIdSet should store weakrefs instead of object ids.,01f46d727033b7741feb4022434dd47a0f070405166d0abf8a99f36df059702e
issue-collapse:412c19d257a91e5a,Review thread: 1 issue + 1 PR - pr #184917: Make FX traceback preserve state thread-local,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,173816,184917,,,"module: dynamo, module: fx, oncall: pt2","bot-triaged, module: dynamo, module: fx, oncall: pt2, triaged",thread:issue:173816:pr:184917,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173816 make should_preserve_node_meta and relevant variable TLS,8faead50f47bf1ce7592b55f7916a654de89655b76eeed27aa623a5e837980ac
issue-collapse:18ba7596986edbfe,Review thread: 1 issue + 1 PR - pr #184910: Reject direct aten.set in Dynamo,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,174176,184910,,,"module: functionalization, module: pt2-dispatcher, oncall: pt2","module: functionalization, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:174176:pr:184910,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184910; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#174176 Inductor crashes with INTERNAL ASSERT in FunctionalizeFallbackKernel,a30b8736218a9d20a1b022cca2c407dcd9f00679ae317e62cb01cd84c5b7ce04
issue-collapse:febeef48837da657,Review thread: 1 issue + 1 PR - pr #184864: Error on overlapping as_strided mutations in compile,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,174371,184864,,,"module: functionalization, module: pt2-dispatcher, oncall: pt2","module: functionalization, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:174371:pr:184864,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#174371 [inductor][aot_eager] Incorrect results with torch.compile when using clone, as_strided followed by add_",09ea7a6433e01473378313ce56d1cec9395d3e560cc6b991d2893f6df050e9f5
issue-collapse:2199ed7cf1187ff3,Review thread: 1 issue + 1 PR - pr #184838: Fix Dynamo raw descriptor building,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,175264,184838,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:175264:pr:184838,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175264 [dynamo] Investigate why we create VariableTrackers for only sourceful objects in UserDefinedClassVariable,beeefb528d9d3498d514252a1cbeb3d582aa37c52b29219dd20b41703d758d16
issue-collapse:8636ea28dd2ab103,Review thread: 1 issue + 1 PR - pr #184830: [Dynamo] Guard source-backed identity comparisons,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,175267,184830,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:175267:pr:184830,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175267 [dynamo] Audit `x is y` code path,ce0ba97506e1438f1c6882979d1da830b68fbc359435c2ef0ec15b5f310c1266
issue-collapse:9f134144401b69c0,Review thread: 1 issue + 1 PR - pr #184820: [dynamo] respect custom metaclass __call__ during class construction,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,175292,184820,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:175292:pr:184820,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#175292 [dynamo] Metaclass not supported correctly,19a837de1955db3ae52362669aa089152863d94fe4fb61889ffbe791579f84f9
issue-collapse:c459831c3d17c731,Review thread: 1 issue + 1 PR - pr #184809: Improve custom op Tensor subclass fake error,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,175408,184809,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: pt2-dispatcher, oncall: pt2, tensor subclass, triaged",thread:issue:175408:pr:184809,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175408 custom_op + TensorSubclass + torch.compile interaction : Better error message or doc,14312db4043299040c759094d35158c91be9770550d168278597606518f2a45f
issue-collapse:4859543c778ba6de,Review thread: 1 issue + 1 PR - pr #184803: [dynamo] Avoid key-set guard for dict setitem,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,175858,184803,,,"module: compile ux, module: compile-time, module: dynamo, oncall: pt2","bot-triaged, feature, function request, module: compile ux, module: compile-time, module: dynamo, oncall: pt2, triaged",thread:issue:175858:pr:184803,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175858 RFC: `torch._dynamo.mark_dynamic_dict` — skip key-set guards on dict mutations,b27a15c0ccf41eba0ca23421c934000338725a97717edfd81b118b32aa7390b7
issue-collapse:05783334bdef01de,Review thread: 1 issue + 1 PR - pr #184793: Fix no-tensor-aliasing recompile reason for stale sources,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,176211,184793,,,"module: dynamo, module: guards, oncall: pt2","module: dynamo, module: guards, oncall: pt2, triaged",thread:issue:176211:pr:184793,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#176211 TypeError in guard when working with list of tensors,350abe32ddd225c9eb448800e405a76ad8acf498732e95dbe3c04e8630186586
issue-collapse:3b2ba3735ca950dc,Review thread: 1 issue + 1 PR - pr #184747: Remove no-op SDPA masks before backend dispatch,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,177712,184747,,,"module: dynamo, module: performance, module: sdpa, oncall: pt2","bot-triaged, module: dynamo, module: performance, module: sdpa, oncall: pt2, triaged",thread:issue:177712:pr:184747,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#177712 [dynamo] [torch.compile] sdpa become slower after torch._dynamo.run in dynamo benchmark when attention_mask is None,521e844f1e7892548e7f1474f13448ff23cf9bc9b58d6fef2fc4f3d3c8497595
issue-collapse:e40a697828b6492f,Review thread: 1 issue + 1 PR - pr #184716: Fix conv_transpose2d meta output padding validation,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,178127,184716,,,"module: convolution, module: error checking, module: meta tensors, oncall: pt2","bot-triaged, module: convolution, module: error checking, module: meta tensors, oncall: pt2, triaged",thread:issue:178127:pr:184716,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178127 torch.conv_transpose2d meta execution fails to validate invalid output_padding/stride/dilation combination,f56d450757389f928f0e4acb7db9086beb1e9b7079659bbdc85b484f6d6b1033
issue-collapse:00c684ee9c6142c6,Review thread: 1 issue + 1 PR - pr #184713: Support Dynamo class setattr for hook handles,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,178248,184713,,,"module: dynamo, oncall: pt2","bot-triaged, module: dynamo, oncall: pt2, triaged",thread:issue:178248:pr:184713,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178248 [TorchDynamo] Failed to trace builtin operator 'setattr' when registering forward hook inside forward method,c838f428c2982e37dfc96b7c3a6ffe9207a3eb9a321286d9311ded8f7e3551f9
issue-collapse:c8f7a63f07021639,Review thread: 1 issue + 1 PR - pr #184709: Fix Dynamo tensor subclass constructor side effects,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,178388,184709,,,"module: dynamo, oncall: pt2","bot-triaged, module: dynamo, oncall: pt2, tensor subclass, triaged",thread:issue:178388:pr:184709,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178388 [torch.compile] Dynamo fails to trace assignment of custom tensor subclass to module attribute:self.custom_weights = CustomTensor(self.layer1.weight.data)``,47918c042a18fcc402ae0d08d4b9d3df16765667b48c2bf1ecd8a66bb0ea8e17
issue-collapse:b8dc4826a6bfb1f7,Review thread: 1 issue + 1 PR - pr #184695: Fix all-masked TransformerEncoder compile consistency,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,178677,184695,,,"module: correctness (silent), module: dynamo, module: edge cases, module: nestedtensor, module: nn, oncall: pt2","bot-triaged, module: correctness (silent), module: dynamo, module: edge cases, module: nestedtensor, module: nn, oncall: pt2, topic: fuzzer, triaged",thread:issue:178677:pr:184695,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178677 `torch.compile` silently succeeds on `TransformerEncoder` with all-masked `src_key_padding_mask` where eager raises RuntimeError,d2b821265817d3ca53dbbfeb7fa80c93defb84565501eb26b7757369640b7470
issue-collapse:65669e4537794564,Review thread: 1 issue + 1 PR - pr #184656: Fix CPU SDPA higher-order grad compile,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,181177,184656,,,"module: aotdispatch, module: cpu, module: functorch, module: pt2-dispatcher, module: sdpa, oncall: pt2","bot-triaged, dynamo-functorch, module: aotdispatch, module: cpu, module: functorch, module: pt2-dispatcher, module: sdpa, oncall: pt2, triaged",thread:issue:181177:pr:184656,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#181177 [aot_autograd] RuntimeError: derivative for aten::_scaled_dot_product_flash_attention_for_cpu_backward is not implemented,d461acdd0ac772ebec239d902e56bd7a810a0fc1e63f6c8c806d97de0ab567fa
issue-collapse:382d0c98d23b6bb7,Review thread: 1 issue + 1 PR - pr #184602: Preserve frozen ShapeEnv during fake tensor guards,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,182940,184602,,,"module: dynamic shapes, module: faketensor, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: dynamic shapes, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:182940:pr:184602,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184602; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#182940 Eager fake-tensor execution on symbolic torch.export metadata can specialize shapes and add guards outside tracing,f7ff16b76a4d83117b8d5fcf079355e926a4faed358b613f5e8b63b74003395d
issue-collapse:3d3fdc554f1d3e50,Review thread: 1 issue + 1 PR - pr #184595: Allow functorch grad fake propagation in Dynamo,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,183015,184595,,,"module: functorch, oncall: pt2","bot-triaged, dynamo-functorch, has workaround, module: functorch, oncall: pt2, triaged",thread:issue:183015:pr:184595,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183015 `torch.compile(torch.vmap(torch.func.hessian(f)))` crashes with TorchRuntimeError,3cfc3fa98512a11e2468d37a278ef5738167d150c242b1bce8111250368381e1
issue-collapse:9d3549361a396de6,Review thread: 1 issue + 1 PR - pr #184590: Fix uncached NestedTensor seqlens under compile,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,183266,184590,,,"module: dynamic shapes, module: dynamo, module: nestedtensor, oncall: pt2","bot-triaged, module: dynamic shapes, module: dynamo, module: nestedtensor, oncall: pt2, triaged",thread:issue:183266:pr:184590,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183266 [NestedTensor] torch.compile fails when accessing _max_seqlen with GuardOnDataDependentSymNode,242fcbdc04f37e95a3148e2b1a4e5de84dd61f21b88fd4a4cd9944105be9ccee
issue-collapse:c50326c829236ed2,Review thread: 1 issue + 1 PR - pr #184573: Fix compile support for CTC loss tensor lengths,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,183396,184573,,,"module: cudnn, module: faketensor, module: loss, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: cudnn, module: fakeTensor, module: loss, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:183396:pr:184573,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184573; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#183396 `torch.compile` fails on `nn.CTCLoss` with missing fake impl for `aten._use_cudnn_ctc_loss.Tensor`,67fc074ba696d6afa0e13ace18aaa4ec8ee12f6ccd73d73ee9a6b89c2b98ff4a
issue-collapse:1e6dfe3e8b21d9b3,Review thread: 1 issue + 1 PR - pr #184367: Fix AotAutograd aliased-output replay for traceable wrapper subclasses,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,184002,184367,,,"module: aotdispatch, module: dtensor, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, has workaround, module: aotdispatch, module: dtensor, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triaged",thread:issue:184002:pr:184367,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#184002 [DTensor][Compile] Compile fails when DTensor function returns an aliased output,4e95eb0622efe177b8ab4e179fb8cc2e85f0e5740c001430215fb0506a7bbabb
issue-collapse:c169b8c847c8a4a5,Review thread: 1 issue + 1 PR - pr #184336: fix RPATH in libtorch extraction from wheel,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,183971,184336,,,"module: cpp, module: cuda","bot-mislabeled, bot-triaged, module: cpp, module: cuda, ptd-bot-triaged, triaged",thread:issue:183971:pr:184336,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184336; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#183971 `ldd libtorch_nvshmem.so` in 2.12 libtorch results in `libnvshmem_host.so.3 => not found`,5e03c5f184ffe57a6aa6faaa5a21b4206f03e5798a874159548cecf91ce2a5c8
issue-collapse:09085b5b61ccff8b,Review thread: 1 issue + 1 PR - pr #184256: [inductor] Cap reduction XBLOCK to max block,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,138301,184256,,,module: inductor,"module: inductor, repro:agent:h100:repro_fail:not_reproduced, triaged",thread:issue:138301:pr:184256,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138301 max_block['X'] not synced with max XBLOCK size (error shows up in `create_block_mask`),8ca58f68bd4c9c7df76715b097bed54c7be5afb60c3e5120414a144eb3e7029c
issue-collapse:be016b0b9677285a,Review thread: 1 issue + 1 PR - pr #184130: Avoid torch imports in Inductor compile sidecar,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,148651,184130,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:148651:pr:184130,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184130; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148651 Avoid fork for TORCHINDUCTOR_COMPILE_THREADS > 1,8fdc7c34ad95022e4a19b1c4989733866d0bff9b9baf666b2d815c073c751a04
issue-collapse:5c52b637df6b827a,Review thread: 1 issue + 1 PR - pr #184124: Fix Inductor multi-output replacement insertion order,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,148751,184124,,,"module: inductor, oncall: pt2","inductor_pattern_match, module: inductor, oncall: pt2, triaged",thread:issue:148751:pr:184124,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#148751 [Inductor] Torch inductor pattern match breaks topological sort after replace the pattern,b4153c8371fd02b41855b71bab1c10f3796ddee84220deaed794da6ac1cb8a30
issue-collapse:4e0684536930de14,Review thread: 1 issue + 1 PR - pr #184109: Fix Inductor replacement patterns across view-to-reshape,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,151649,184109,,,"module: inductor, module: vllm, oncall: pt2","module: inductor, module: vllm, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged, vllm-compile",thread:issue:151649:pr:184109,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#151649 Inductor pattern matcher replaces aten.reshape with aten.view in pattern,012ee11d9980bcb0cf19f713fc2dac81970fc3e4afaac38a7c8070bf804771ea
issue-collapse:dcf20c933eb730e3,Review thread: 1 issue + 1 PR - pr #184094: Clone aliased mm/addmm inputs on CPU to avoid in-place gemm,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,85852,184094,,,module: docs,"module: docs, triaged",thread:issue:85852:pr:184094,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#85852 `torch.mm` produces wrong result on cpu when using in-place computation,237418e117ee5ca15b714291160c2419543572ed42e87ae89b25f4289e482a0d
issue-collapse:5bcfbeb2534cc7c2,Review thread: 1 issue + 1 PR - pr #184092: Skip tiny cudagraph partitions in max-autotune,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,154301,184092,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:154301:pr:184092,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184092; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#154301 cudagraph copies in 'max-autotune' can add overhead on small model,17d23a2046040ca85d73619890369bc9cf25b7f1cd7b1937bc2338ef156c33bd
issue-collapse:61b4a92c1718ec0a,Review thread: 1 issue + 1 PR - pr #184091: Handle cudagraph liveness changes during replay,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,154306,184091,,,"module: cuda graphs, module: inductor, oncall: pt2","module: cuda graphs, module: inductor, oncall: pt2, triaged",thread:issue:154306:pr:184091,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184091; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#154306 graph recording observed an input tensor deallocate during graph recording that did not occur during replay,68cbfab040a0f7131891a462e59ed983ff5a6a1465de3f284f93b23471a29aaa
issue-collapse:f332c26a4cf44ea5,Review thread: 1 issue + 1 PR - pr #184059: Fix split reduction kernel metadata,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,160124,184059,,,"module: fx, module: inductor, oncall: pt2","module: fx, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:160124:pr:184059,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160124 FX graph segments for multiple triton kernels from reduction node are broken,279db12d51ba9d2f5e082fc122ffa4612f7ee7bc9a2408ed1931464f877eddcb
issue-collapse:3411846c7f36ebf9,Review thread: 1 issue + 1 PR - pr #184048: Enable Inductor OpInfo CPU coverage,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,161132,184048,,,"module: inductor, module: tests, oncall: cpu inductor, oncall: pt2","PT2-Bug-Bash, module: inductor, module: tests, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:161132:pr:184048,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161132 The Inductor test suite test_torchinductor_opinfo was incorrectly disabled on all devices except CUDA.,a57677ea6afd97935b1eff07b585f0d051b23991cb6ad899c982d1dd2564c0c1
issue-collapse:5476f3882bc2702b,Review thread: 1 issue + 1 PR - pr #184015: [inductor] Handle scalar workaround args in pattern checks,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,166604,184015,,,"module: inductor, oncall: pt2","inductor_pattern_match, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:166604:pr:184015,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#166604 [inductor][pattern_matcher] RuntimeError in joint graph passes when compiling simple `randperm` + advanced indexing pipeline,2049ea81399d0973284710ddd30418611e3787e52f119b510aa140db3167ecb8
issue-collapse:a71c88b63d7b40d3,Review thread: 1 issue + 1 PR - pr #184009: Handle SymPy boolean Inductor graph inputs,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,167070,184009,,,"module: dynamic shapes, module: inductor, module: pt2-dispatcher, oncall: pt2","module: dynamic shapes, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:167070:pr:184009,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184009; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#167070 Inductor doesn't accept Sympy.Eq as input,f0274cec10d9da64b11f1043fff9d345b2687f914084e8c3a89961cd2f2f31dd
issue-collapse:c3f05406512db811,Review thread: 1 issue + 1 PR - pr #184007: [inductor] Clear stale subprocess cache env vars,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,167725,184007,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, repro:agent:h100:incorrect, repro:agent:h100:repro_fail:not_reproduced, triaged",thread:issue:167725:pr:184007,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167725 `torch._inductor.utils.fresh_cache` fails due to persistant process pool,1a3512ec56be196142b8d73a687bb9ec25860ee0b6a0de1867ebfc76874a8a57
issue-collapse:a5996db516ccc9c6,Review thread: 1 issue + 1 PR - pr #183994: Specialize numpy scalar module attrs,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,172942,183994,,,"module: cuda graphs, module: inductor, oncall: pt2","PT2-Bug-Bash, module: cuda graphs, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:172942:pr:183994,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#172942 cuda graphs dont play nice with numpy numbers,6a37096729bb94fe4bcaf1b3fcff5dfcd7691352369a1f7ddbbfde9c7e2b6e2b
issue-collapse:6e5aef3a1683417b,Review thread: 1 issue + 1 PR - pr #183969: Fix CSE across unrelated mutation regions,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,174472,183969,,,"module: inductor, module: vllm, oncall: pt2","bot-triaged, has workaround, module: inductor, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:174472:pr:183969,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #183969; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#174472 [Inductor][CSE] `split_with_sizes` not deduplicated on B200 with FP8 quantization,097b5415c3aef007d87b58ee440c4addef6857ede70fe8bf2571def0a3bc3c6e
issue-collapse:5d219186ad73d743,Review thread: 1 issue + 1 PR - pr #183958: Enable SDPA constraint cache by default,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,175833,183958,,,"module: inductor, module: performance, module: sdpa, oncall: pt2","bot-triaged, module: inductor, module: performance, module: sdpa, oncall: pt2, repro:agent:h100:repro_fail:not_reproduced, triaged",thread:issue:175833:pr:183958,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#175833 Duplicate SDPA Input Constraints,5307cd68d0d100b50f8a13fddc017b41a8765047ea09e8ba72694330c0a18a3f
issue-collapse:408e700dfeb906ef,Review thread: 1 issue + 1 PR - pr #183954: Disallow mid-trace reference opaque creation,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,175968,183954,,,"module: custom-operators, module: inductor, module: pt2-dispatcher, oncall: pt2","bot-triaged, module: custom-operators, module: inductor, module: pt2-dispatcher, oncall: pt2, tensor subclass, triaged",thread:issue:175968:pr:183954,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #183954; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#175968 [opaque object] Running the compiled function with opaque object for the second time fails.,4e53f58232cba1ded086ee823c93ab57e45ced9cff883b670030e065ce83dca4
issue-collapse:492bc6fe43f50f43,Review thread: 1 issue + 1 PR - pr #183904: Fix CPU addcmul tensor-value decomposition numerics,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,176929,183904,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, repro:agent:h100:repro_success:mitigation, triaged",thread:issue:176929:pr:183904,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #183904; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#176929 Addcmul CPU bitwise numerics not matching eager w/ torch.compile,94b914215ef6abee506faab0aa03594356ee855a6c52e3cb82024424cd9d4fc0
issue-collapse:da33a51529d2a088,Review thread: 1 issue + 1 PR - pr #183879: [inductor][cpu] Enable int8 WOQ GEMM template on Arm,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,177327,183879,,,"module: arm, module: cpu, module: inductor, module: linear algebra, oncall: cpu inductor, oncall: pt2","bot-triaged, module: arm, module: cpu, module: inductor, module: linear algebra, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:177327:pr:183879,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#177327 AArch64 Unit Test Failure - test_int8_woq_mm_*,37ec4f49045b63f0e7e90690665360a5ed117649e7e46787762163298b745386
issue-collapse:9a079e5b0e000571,Review thread: 1 issue + 1 PR - pr #183871: [inductor] Match eval BatchNorm CUDA numerics,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,178096,183871,,,"module: inductor, module: pt2 accuracy, oncall: pt2","bot-triaged, module: inductor, module: pt2 accuracy, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:178096:pr:183871,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #183871; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#178096 torch.compile(dynamic=True) on CUDA: large eager vs compiled mismatch for BatchNorm2d + Conv2d,4892594abd86f290477847702cb6ced803727778dc5258a406b6e6927f5e12d9
issue-collapse:a5e2b37ea38bac82,Review thread: 1 issue + 1 PR - pr #183866: Preserve native matmul lowp output precision,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,178317,183866,,,"module: inductor, module: tests, oncall: pt2","bot-triaged, module: inductor, module: tests, oncall: pt2, triaged",thread:issue:178317:pr:183866,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#178317 [inductor] Tolerances in test_remove_no_ops have to be lowered with native matmul,cac9b81df5c8cf80d3ca7a664668ce1689a65ebd4140c127110d558d918e3d0c
issue-collapse:54f54d81075dbeca,Review thread: 1 issue + 1 PR - pr #183744: Update ACL dependencies for BF16 inner product,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,180447,183744,,,"module: arm, module: bfloat16, module: cpu, module: inductor, module: mkldnn, module: performance, oncall: pt2","bot-triaged, module: arm, module: bfloat16, module: cpu, module: inductor, module: mkldnn, module: performance, oncall: pt2, triaged",thread:issue:180447:pr:183744,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#180447 Poor BF16 torch.compile + Freezing perf on AArch64 CPUs,897187e817b8b954e17cb7fbdc36af13a0477cf6289c838bf41e2a01904acac3
issue-collapse:1473c79316f09bbe,Review thread: 1 issue + 1 PR - pr #183731: [inductor] Validate fallback output dtype metadata,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,180807,183731,,,"module: correctness (silent), module: inductor, oncall: pt2","bot-triaged, module: correctness (silent), module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:180807:pr:183731,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#180807 [inductor][static launcher] Missing runtime dtype validation for pointer-typed Triton args can silently produce incorrect results,05efd38c85d31372b96fa6869b14d22ec5ac92150c37b5d7a51c0dd27a731c9b
issue-collapse:334ea35b57b30693,Review thread: 1 issue + 1 PR - pr #183652: Fix dynamic slice lowering for generic views,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,183259,183652,,,"module: dynamic shapes, module: inductor, oncall: pt2","bot-triaged, module: dynamic shapes, module: inductor, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:183259:pr:183652,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#183259 InductorError on in-place multidimensional dynamic slicing with Tensor-derived slice bounds,1323f95d226c4bfc40e40ed3f0a1e41f74f2c6cd3c6326d7ef2e2ef961056163
issue-collapse:9343644585c6a196,Review thread: 1 issue + 1 PR - pr #183628: Fix fractional max pool RNG in Inductor,linked work thread,ready for maintainer decision,44,high,work_thread_graph,1,1,0,0,183249,183628,,,"module: correctness (silent), module: inductor, module: pooling, module: random, oncall: pt2","bot-triaged, has workaround, module: correctness (silent), module: inductor, module: pooling, module: random, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:183249:pr:183628,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183249 `torch.compile` produces wrong results for `fractional_max_pool2d` when `_random_samples` is not provided,72503033bd53f0571f270cbab64c2889a4345453a57260e750b0031467e26ba6
issue-collapse:77dba34e3fbc38bf,Review thread: 1 issue + 1 PR - pr #182972: added graph break on exception based on if user code would catch it,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,175608,182972,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:175608:pr:182972,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #182972; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#175608 torch.compile fails to trace try-except semantic and throw `RuntimeError`,b0f5e4a71016a5ef88870ced98b15d409dcc3c8d985b3de04412492805bb9d2e
issue-collapse:6cd3b127180d58e0,Review thread: 1 issue + 1 PR - pr #182273: [c10d] Fix device_id not propagated through ProcessGroupWrapper,linked work thread,has linked issue,44,high,work_thread_graph,1,1,0,0,178977,182273,,,"module: c10d, oncall: distributed","bot-triaged, module: c10d, oncall: distributed, triaged",thread:issue:178977:pr:182273,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#178977 `device_id` not honored when creating `process_group` with `$TORCH_DISTRIBUTED_DEBUG=DETAIL`,1c9631fc82395bdadbcd4ab93b60c29852feb93ada7350fcad119de80b64c0b3
issue-collapse:0a49c433efdbc90a,Review thread: 1 issue + 1 PR - pr #181860: [c10] Add InlineEventBase to unify event APIs in DeviceGuardImplInter…,linked work thread,draft/noise,44,high,work_thread_graph,1,1,0,0,163836,181860,,,"module: openreg, module: privateuse1","module: PrivateUse1, module: openreg, triaged",thread:issue:163836:pr:181860,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#163836 [OpenReg] Using InlineEvent (or its base) for Carrying More Event Metadata in DeviceGuardImplInterface APIs,8a42fda4bb52453ed773bf91247afc849e2f1586b6a6650a5861fa2201ba8201
issue-collapse:b55ff6ff1f2481a0,Review thread: 1 issue + 1 PR - pr #181823: Docs: update torch.multinomial num_samples requirement note for bette…,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,181347,181823,,,module: docs,"module: docs, triaged",thread:issue:181347:pr:181823,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#181347 [Feedback about torch.multinomial][Docs] Inconsistency in torch.multinomial: Note says ""lower than"" but example allows ""equal to"" for num_samples without replacement",a3d096ee2f7e65fd8762cf511f596c9a2e9f0d489cc3b031cacc20f87818187f
issue-collapse:16471da23b671bcc,Review thread: 1 issue + 1 PR - pr #180353: [MPS] Replace isin MPSGraph implementation with native Metal kernels,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,180833,180353,,,"module: mps, module: performance","bot-triaged, enhancement, module: mps, module: performance, triaged",thread:issue:180833:pr:180353,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #180353; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#180833 [MPS] Replace isin MPSGraph implementation with native Metal kernel and binary-search path for large inputs,1ade642253511f9c08c316ee4b2d1b3daf8ca627c5cc8ec914ca59da68c0900a
issue-collapse:c32ebcdd338b31fb,Review thread: 1 issue + 1 PR - pr #178298: torch.linalg.solve_triangular: support for eligible non-dense args,linked work thread,PR blocked,44,high,work_thread_graph,1,1,0,0,178379,178298,,,"module: correctness (silent), module: cuda, module: linear algebra","actionable, bot-triaged, module: correctness (silent), module: cuda, module: linear algebra, triaged",thread:issue:178379:pr:178298,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,"#178379 [bug]`torch.linalg.solve_triangular(A, B, *, out=B)`: wrong result with `B.is_neg() == True`",42c05f3b32c1dea39533b5d0eed16f7516bff8d628c218dbf62629e06c1c4e3c
issue-collapse:98dac75c80e2db9b,Review thread: 1 issue + 1 PR - pr #172074: Fixes grad_dtype property being lost in fsdp,linked work thread,waiting on contributor,44,high,work_thread_graph,1,1,0,0,170648,172074,,,"module: fsdp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: fsdp, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:170648:pr:172074,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #172074; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#170648 fsdp2 support grad_dtype,85d78a70ad89c34ddf5950f2599d87da63ae687c1943bc8b85ce07fb75fcd3db
issue-collapse:0d6f96fca746bd7d,Review thread: 1 issue + 1 PR - pr #189319: Enable flake8-bugbear B007 (unused loop control variable),linked work thread,PR blocked,38,high,work_thread_graph,1,1,0,0,106571,189319,,,module: lint,"better-engineering, enhancement, good first issue, module: lint, triaged",thread:issue:106571:pr:189319,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#106571 Enable more flake8-bugbear lints,93c4c1e02e9af13aee32527ca8a0c18f47ef1c71cd86bc944b450724e12803a8
issue-collapse:fabea4a4fbdd557a,Review thread: 1 issue + 1 PR - pr #186549: Fix fake deepcopy for custom module mappings,linked work thread,PR blocked,38,high,work_thread_graph,1,1,0,0,97078,186549,,,oncall: pt2,"ezyang's list, needs reproduction, oncall: pt2, triaged",thread:issue:97078:pr:186549,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#97078 [compile] TypeError: __init__() missing 1 required positional argument: 'parent_module',7c8aa25f56fcb8d43e46f22cf4cc78534eb5afe1d5f8ee81e10bd11086ff78c1
issue-collapse:8a197f50d7bb5a2d,Review thread: 1 issue + 1 PR - pr #186512: [dynamo] Graph break on shape-changing Parameter.data assignment,linked work thread,PR blocked,38,high,work_thread_graph,1,1,0,0,115484,186512,,,"module: dynamo, oncall: pt2","dynamo-triage-june2024, module: dynamo, needs reproduction, oncall: pt2, triaged",thread:issue:115484:pr:186512,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#115484 torch.compile() breaks when using DeepSpeed ZeRO Level 3 sharding,aaffe637ef8142ad9f7a51fe26ec826a241ac48ee70d1ddda40c19a13e0ef6dc
issue-collapse:47f7d552be4f1af1,Review thread: 1 issue + 1 PR - pr #185813: [ROCm][Windows] Audit license-files: explicit enumeration + audit test (#183434),linked work thread,waiting on contributor,38,high,work_thread_graph,1,1,0,0,183434,185813,,,"module: binaries, module: third_party","bot-triaged, enhancement, module: binaries, module: third_party, needs design, triaged",thread:issue:183434:pr:185813,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185813; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183434 Audit license-files glob in pyproject.toml for over-collection,c32365b3a0d3414d915bddd29ad28d5daae9e011af2428f5af59eeec95313404
issue-collapse:998fa468d6c59c67,Review thread: 1 issue + 1 PR - pr #184178: Fix masked Triton conv load pointers,linked work thread,ready for maintainer decision,38,high,work_thread_graph,1,1,0,0,143412,184178,,,"module: inductor, oncall: pt2","module: inductor, needs reproduction, oncall: pt2, triaged",thread:issue:143412:pr:184178,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143412 [Inductor test failure] torch:inductor/test_select_algorithm TestSelectAlgorithm.test_convolution1 with cuda 12.6.3,448d957bc8508ccce8cd09c9315049f85bc5f9a42ff5cdb9c6f4658fd8ff61a4
issue-collapse:034aec58eaed142a,Review thread: 1 issue + 18 PRs - pr #182109: [MPS] Replace MPSGraph LayerNorm backward with native Metal kernels,linked work thread,waiting on contributor,28,high,work_thread_graph,1,18,0,0,187455,"182109, 182731, 182736, 182749, 182791, 183688, 187456, 187457, 187458, 187553, 187556, 187557, 187559, 187787, 187976, 188156, 188333, 189146",,,"module: memory usage, module: mps, module: performance","bot-triaged, enhancement, module: memory usage, module: mps, module: performance, triaged",thread:issue:187455:pr:182109:pr:182731:pr:182736,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #187553, #187787, #187976; keep related issue/PR context attached.",6,Represents 1 open issue; raw issue rows remain available for audit.,#187455 [MPS] Migrate remaining MPSGraph-backed ops to native Metal to bound graph-cache memory,8cb576c381b6f4094f5f92ad79fa36d17b9b118e26db35756b0a2847ecf0ccb3
issue-collapse:4bf8b379afe5e034,Review thread: 12 issues + 1 PR - pr #119977: [NJT] Factory function support,linked work thread,stale/low urgency,28,high,work_thread_graph,12,1,0,0,"88137, 89124, 89125, 99142, 108567, 112398, 112509, 114090, 115978, 118107, 130538, 132212",119977,,,"module: loss, module: nestedtensor, module: serialization","enhancement, module: loss, module: nestedtensor, module: serialization, topic: new features, triaged",thread:issue:108567:issue:112398:issue:112509:issue:114090,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#88137 Add eq, to, masked_select, index_select, narrow to nested tensors, #89124 Zero-copy way to make flat tensor into a nested_tensor given a shape, #89125 binary_cross_entropy/bce_with_logits (+ other loss functions) for nested_tensor, #99142 More Nested Tensor Functionality (layer_norm, cross_entropy / log_softmax&nll_loss), #108567 Allow slicing of Nested Tensors along constant dimensions, #112398 [Tracker] Move nested tensors to beta",b9a47275f1c66c4b90d1cb11dbb62236a97452aa60408fde502ed1a8cb36cfeb
issue-collapse:98b262d8f95117dd,Review thread: 7 issues + 1 PR - pr #184193: Preserve FX graph cache guard provenance,linked work thread,PR blocked,28,high,work_thread_graph,7,1,0,0,"13716, 18631, 70954, 73764, 75747, 95604, 187473",184193,,,"module: convolution, module: cpu, module: cuda, module: cudnn, module: dependency bug, module: docs, module: performance, module: vllm, oncall: pt2, oncall: releng","module: convolution, module: cpu, module: cuda, module: cudnn, module: dependency bug, module: docs, module: performance, module: vllm, oncall: pt2, oncall: releng, triaged",thread:issue:13716:issue:18631:issue:187473:issue:70954,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#13716 depthwise convolution are slow on cpu, #18631 FP32 depthwise convolution is slow in GPU, #70954 Training grouped Conv2D is slow, #73764 Group convolution slower than manually running separate convolutions in CUDA streams, #75747 Depthwise Conv1d performance (a naive CUDA kernel is 10x faster), #95604 Performance bugs exists in multiple convolution operations(e.g., `Convtranspose2d`) when useing the `groups` argument",86f9d7a9a422c8dfc2bbe35715a96236259aecd90c32be814826b144b7612ff4
issue-collapse:67850527c62d9bb9,Review thread: 2 issues + 6 PRs - pr #185057: Support torch.cuda.use_mem_pool in compile,linked work thread,PR blocked,28,high,work_thread_graph,2,6,0,0,"162859, 167026","185057, 188755, 189123, 189173, 189175, 189230",,,"module: inductor, module: symm_mem, module: vllm, oncall: distributed, oncall: pt2","mempool-bug, module: inductor, module: symm_mem, module: vllm, oncall: distributed, oncall: pt2, triaged, vllm-compile",thread:issue:162859:issue:167026:pr:185057:pr:188755,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#162859 [RFC] support symmetric memory in torch.compile, #167026 Support torch.cuda.use_mem_pool in torch.compile",384514a5c0e8024d1b950f6e61caf9c6d7acccf2ace13771638b706d7cc942fa
issue-collapse:c6473b0ab5452b3c,Review thread: 5 issues + 2 PRs - pr #181109: tools/stats: add script to generate test-times.json from local JUnit XML reports,linked work thread,ready for maintainer decision,28,high,work_thread_graph,5,2,0,0,"129553, 141550, 147513, 171659, 180975","181109, 188999",,,"module: build, module: ci, module: cpu, module: performance, module: risc-v, module: vectorization","bot-triaged, enhancement, feature, large, module: build, module: ci, module: cpu, module: performance, module: risc-v, module: vectorization, proposal accepted, triaged",thread:issue:129553:issue:141550:issue:147513:issue:171659,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 5 open issues; raw issue rows remain available for audit.,"#129553 RVV support in PyTorch, #141550 RISC-V CI support, #147513 [RFC] Request for Feedback and Review on PRs Adding RISC-V and RVV Support, #171659 [RFC] RISC-V Architecture Support Roadmap for PyTorch, #180975 [Tracking] RISC-V PyTorch enablement",0f88210ab7ee5a8631b99b4554e10e8913621f1020cf51336ed2f0dc1e875b35
issue-collapse:36b00bbaeb2272a2,Review thread: 1 issue + 5 PRs - pr #176688: Enable eager tests for inductor-unittest - enable test_dtypes and align XPU op dyptes with OpInfo,linked work thread,waiting on contributor,28,high,work_thread_graph,1,5,0,0,188987,"176688, 176689, 178849, 179549, 179550",,,"module: logging, module: tests, module: xpu, oncall: pt2","bot-triaged, module: logging, module: tests, module: xpu, oncall: pt2, skipped, triaged",thread:issue:188987:pr:176688:pr:176689:pr:178849,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #178849; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#188987 DISABLED test_optimizer_non_static_param (__main__.LoggingTests),debf463f83d2a85e5a36b37c95310367248eaf95795b46fa0ba420efd72f9a5f
issue-collapse:d2266684ac43bbc0,Review thread: 4 issues + 1 PR - pr #184185: Use memory stats for persistent reductions,linked work thread,PR blocked,28,high,work_thread_graph,4,1,0,0,"139908, 141265, 141916, 142250",184185,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:139908:issue:141265:issue:141916:issue:142250,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 4 open issues; raw issue rows remain available for audit.,"#139908 Inductor vs. Liger Performance Track, #141265 rope perf improvement, #141916 rms_norm forward performance analysis, #142250 Embedding forward performance analysis",c15241fabb12897dae0bdc148a2ea6f3a881198e0c88dd168747916bdad817c7
issue-collapse:e4dde8729b88ee8b,Review thread: 4 issues + 1 PR - pr #183328: [special] Add arbitrary-order modified Bessel functions,linked work thread,PR blocked,28,high,work_thread_graph,4,1,0,0,"76324, 78065, 80152, 80157",183328,,,module: special,"feature, module: special, triaged",thread:issue:76324:issue:78065:issue:80152:issue:80157,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 4 open issues; raw issue rows remain available for audit.,"#76324 Bessel and Related Functions, #78065 Gamma and Related Functions, #80152 Orthogonal Polynomials, #80157 Elliptic Functions and Integrals",c1f961a1d68c32271d7d9bad692d2f8dfe215b008cd38f08115c0ad95164eaae
issue-collapse:a6c7884d55ca35ba,Review thread: 1 issue + 4 PRs - pr #169694: Increase precision for golden solution in transformer tests.,linked work thread,waiting on contributor,28,high,work_thread_graph,1,4,0,0,188670,"169694, 183691, 184531, 187050",,,"module: aotinductor, module: inductor, module: tests, oncall: export, oncall: pt2","bot-triaged, enhancement, module: aotinductor, module: inductor, module: tests, oncall: export, oncall: pt2, triaged",thread:issue:188670:pr:169694:pr:183691:pr:184531,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.","Wait for contributor update on PR #169694, #183691, #184531; keep related issue/PR context attached.",4,Represents 1 open issue; raw issue rows remain available for audit.,#188670 [RFC] Inductor / AOTI tests: decouple “correct codegen path” from tight eager-fp16 equality,9c2180745550f18f0ac5a51808955b152c78ca74f6b8aa94152bf567d2841c5c
issue-collapse:761896f258b8a367,Review thread: 3 issues + 1 PR - pr #188720: [ROCm] Enable cholesky_ex via hipSOLVER xgeev (ROCm >= 7.14),linked work thread,draft/noise,28,high,work_thread_graph,3,1,0,0,"181647, 181648, 181649",188720,,,"module: ci, module: inductor, module: rocm, oncall: pt2","bot-triaged, module: ci, module: inductor, module: rocm, oncall: pt2, triaged, unstable",thread:issue:181647:issue:181648:issue:181649:pr:188720,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#181647 UNSTABLE rocm-nightly / linux-noble-rocm-nightly-py3.12-gfx942 / test (default), #181648 UNSTABLE rocm-nightly / linux-noble-rocm-nightly-py3.12-gfx942 / test (inductor), #181649 UNSTABLE rocm-nightly / linux-noble-rocm-nightly-py3.12-gfx942 / test (distributed)",1c41764ee358b685c80e8a1730207e39d22ed8ce6dd5fbeccb0d23f234bcfce8
issue-collapse:f59c181068292abe,Review thread: 2 issues + 2 PRs - pr #168999: [Pt. 1] Make some aspects of Dynamo/Inductor global state thread safe,linked work thread,PR blocked,28,high,work_thread_graph,2,2,0,0,"126024, 168042","168999, 187589",,,"module: dynamic shapes, module: inductor, module: multithreading, oncall: pt2","module: dynamic shapes, module: inductor, module: multithreading, months, no-scrub, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",thread:issue:126024:issue:168042:pr:168999:pr:187589,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 2 open issues; raw issue rows remain available for audit.,"#126024 Can not use torch.compile with dynamic=Ture when using multi-threads, #168042 TorchInductor compile breaks when called in a multithreaded context with GIL off",380e79c33d1f70bb6d10410ad97411bb52133dbe02d974ab2471ebbd0903625e
issue-collapse:7d54f7a8a5f2abe2,Review thread: 1 issue + 3 PRs - pr #184355: [3/N][Test] Migrate TestBwdGradients device-agnostic skips to `@skipOps`,linked work thread,PR blocked,28,high,work_thread_graph,1,3,0,0,177259,"184355, 184356, 184678",,,module: tests,"bot-triaged, enhancement, module: tests, triaged",thread:issue:177259:pr:184355:pr:184356:pr:184678,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#177259 Migrate device-agnostic skips from `op_db` to `@skipOps`,188b36f159f6dcc7f6953206b28b0624fce6b4a2515e59cff9a9ccb2bf52621d
issue-collapse:70136fc3a860a0cc,Review thread: 1 issue + 3 PRs - pr #183276: [Inductor][Refactor] 2/N Introduce shared heuristic registry with codegen_heuristic support,linked work thread,PR blocked,28,high,work_thread_graph,1,3,0,0,178153,"183276, 183277, 183278",,,"module: inductor, module: performance, oncall: pt2","bot-triaged, enhancement, module: inductor, module: performance, oncall: pt2, triaged",thread:issue:178153:pr:183276:pr:183277:pr:183278,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#178153 [RFC] Refactor Triton Heuristics via Unified Device-Scoped Class Hierarchy,0f95d3084743c4995071f9c5521377e0bb010b013216a665981f211c624388f0
issue-collapse:e0447ec644d71f2f,Review thread: 1 issue + 3 PRs - pr #179286: Use C++20 concepts where it improves readability,linked work thread,PR blocked,28,high,work_thread_graph,1,3,0,0,176662,"179286, 186854, 188440",,,module: cpp,"module: cpp, topic: improvements, triaged",thread:issue:176662:pr:179286:pr:186854:pr:188440,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#176662 C++20 Rollout Tracking,4f4c302c57db4d66c69b5e0ea839219996d42f6337d9a3d655ce65452b92c4c9
issue-collapse:87e2098b4e035729,Review thread: 1 issue + 3 PRs - pr #168223: Add failing test for negative epoch crash in LR schedulers (ZeroDivisionError),linked work thread,waiting on contributor,28,high,work_thread_graph,1,3,0,0,177833,"168223, 169107, 174025",,,"module: error checking, module: lrscheduler, module: optimizer","bot-triaged, enhancement, module: LrScheduler, module: error checking, module: optimizer, triaged",thread:issue:177833:pr:168223:pr:169107:pr:174025,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #169107; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#177833 StepLR accepts step_size=0 and crashes with ZeroDivisionError on first step(),ca82381c38487cb2fbb057b331524b6127989a2d54f85daf197a396aec1f20ee
issue-collapse:03df0990faad2888,Review thread: 2 issues + 1 PR - pr #186245: [1/N] Optimize PyTorch CUDA build with LLVM BOLT,linked work thread,ready for maintainer decision,28,high,work_thread_graph,2,1,0,0,"188602, 189078",186245,,,"module: arm, module: build, module: ci, module: cuda, module: performance","bot-triaged, feature, module: arm, module: build, module: ci, module: cuda, module: performance, triaged",thread:issue:188602:issue:189078:pr:186245,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#188602 [RFC] Optimizing CUDA+AArch64 builds using LLVM-BOLT, #189078 [RFC] Build profiling for PyTorch",e187ce2321d07b8b3d67d018153f763e649d31e0789f7b5aa279a595f3dc3a70
issue-collapse:41540d7a3916a65d,Review thread: 2 issues + 1 PR - pr #185790: Fix torch.nn.functional.gelu returning NaN for infinite inputs,linked work thread,ready for maintainer decision,28,high,work_thread_graph,2,1,0,0,"185770, 187293",185790,,,"module: edge cases, module: nans and infs, module: nn","bot-triaged, module: NaNs and Infs, module: edge cases, module: nn, triaged",thread:issue:185770:issue:187293:pr:185790,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#185770 torch.nn.functional.gelu returns NaN for +inf input with approximate=""none"", #187293 torch.nn.functional.gelu(approximate=""none"") returns NaN for +Inf input",bbef55cdc902765bba6fd364296b3dbf6582cc6030da6f96b489cd38f13da309
issue-collapse:b4dab88afc66d90a,Review thread: 2 issues + 1 PR - pr #184987: Fix index_select refs bounds semantics,linked work thread,ready for maintainer decision,28,high,work_thread_graph,2,1,0,0,"169779, 187463",184987,,,"module: advanced indexing, module: decompositions, module: python frontend, module: scatter & gather ops, oncall: pt2","bot-triaged, module: advanced indexing, module: decompositions, module: python frontend, module: scatter & gather ops, oncall: pt2, triaged",thread:issue:169779:issue:187463:pr:184987,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 2 open issues; raw issue rows remain available for audit.,"#169779 [Inductor] Inconsistent handling of negative indices in index_select: CPU/CUDA/MPS, #187463 PyTorch has Inconsistent Negative Indexing Semantics",a60ba7b4bd39b0bfd9cce7a76bff378c784f38d5b21234925d266eff559f4f4d
issue-collapse:828f21df7b77fbb9,Review thread: 2 issues + 1 PR - pr #182712: [MPS] Enhance Linear tensor ops to avoid unnecessary .contiguous() calls on weight tensor,linked work thread,waiting on contributor,28,high,work_thread_graph,2,1,0,0,"181946, 183758",182712,,,"module: memory format, module: mps, module: performance, module: testing","bot-triaged, enhancement, large, module: memory format, module: mps, module: performance, module: testing, triaged",thread:issue:181946:issue:183758:pr:182712,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #182712; keep related issue/PR context attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#181946 [MPS] Remove Unnecessary Tensor Memory Gathers and Contiguous Calls for Stride-Aware MPS Backend Inputs, #183758 [Testing] Expand Testing for Non-Contiguous Tensor Ops Across All Platforms",9db2d9ca2bfacb3114ef45e416981125f0b5a39f37b6aec50c9926cea70ec304
issue-collapse:7d1cd4d054801a1e,Review thread: 2 issues + 1 PR - pr #165837: [OpenReg][Random] Enrich OpenReg random number generator module implementation add focused documentation,linked work thread,PR blocked,28,high,work_thread_graph,2,1,0,0,"158917, 172318",165837,,,"module: openreg, module: privateuse1","module: PrivateUse1, module: openreg, triaged",thread:issue:158917:issue:172318:pr:165837,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#158917 [Feat] Tracking for OpenReg Improvements, #172318 [OpenReg] Add DeviceTypeTestBase example with instantiate_device_type_tests",bbf882575023ab52083eb751b14e65eee77d301026e68a136f55239b7ae6596c
issue-collapse:e649c1c87614c45d,Review thread: 1 issue + 1 PR - pr #189162: [inductor] Use non-NaN-propagating fmax in softmax persistent reduction,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,189161,189162,,,"module: inductor, module: performance, oncall: pt2","bot-triaged, enhancement, module: inductor, module: performance, oncall: pt2, triaged",thread:issue:189161:pr:189162,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189161 [Inductor] Remove Redundant NaN Propagation in `max` op in Softmax,28f9f231846cc87e5e3a5af5a18815c40598e194219dea1068416c1177b456c3
issue-collapse:c29455d264c8bfdf,Review thread: 1 issue + 1 PR - pr #188951: Add ROCm Docker release images,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,188319,188951,,,"module: binaries, module: ci, module: docker, module: rocm","bot-triaged, feature, module: binaries, module: ci, module: docker, module: rocm, triaged",thread:issue:188319:pr:188951,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#188319 [RFC] Publish official ROCm Docker images on pytorch/pytorch,608269386b7e7cf28788a367e22bdee99688ab884890b3d5fd99127631e4d0ed
issue-collapse:6d162d0394c06795,"Review thread: 1 issue + 1 PR - pr #188681: [Inductor] Use CUDA toolkit's libdevice by default, fallback to Triton's bundled if not found",linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,188672,188681,,,"module: inductor, module: performance, oncall: pt2","bot-triaged, enhancement, module: inductor, module: performance, oncall: pt2, triaged",thread:issue:188672:pr:188681,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188672 [Inductor] Prefer CUDA Toolkit libdevice for Triton-backed Inductor kernels,05793a6fbac5d26b78ec1fc71a36c8df06e891806fe41a57378e78da4f447cd4
issue-collapse:a916962ebed1e1fd,Review thread: 1 issue + 1 PR - pr #188500: Allow pre-split pipeline microbatches,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,188727,188500,,,"module: pipelining, oncall: distributed parallelisms","bot-triaged, module: pipelining, oncall: distributed parallelisms, triaged",thread:issue:188727:pr:188500,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188727 torch.distributed.pipelining _split_block_mask hides batch offset from non-strict tracing,34204ab99a49a3a29c7d126ae570e037257f8b3c70709f52471624ac5aa3fefb
issue-collapse:40821ee9b10207d3,Review thread: 1 issue + 1 PR - pr #188263: (fix) NNPACK PeachPy air gapped build,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,188652,188263,,,"module: build, module: nnpack","bot-triaged, enhancement, module: build, module: nnpack, triaged",thread:issue:188652:pr:188263,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188652 Pytorch can't be built in an air-gapped environment,6c9d44df1cc0df8829fe81ab317222123f7acff52234aca966e7b461a61bdb2a
issue-collapse:ae980ca03a281e92,Review thread: 1 issue + 1 PR - pr #187696: Add DataLoader worker signal message affixes,linked work thread,waiting on contributor,28,high,work_thread_graph,1,1,0,0,188020,187696,,,module: dataloader,"bot-triaged, feature, module: dataloader, triaged",thread:issue:188020:pr:187696,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #187696; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188020 Allow platform context in DataLoader worker fatal signal messages,1f818f1c42268a315389f23af0cedef2909b469166eb976e1ec21db699670723
issue-collapse:f40d6bcc0bd813a0,Review thread: 1 issue + 1 PR - pr #187328: Add fusion provenance to profiler timelines,linked work thread,PR blocked,28,high,work_thread_graph,1,1,0,0,189270,187328,,,"module: inductor, module: trace, oncall: pt2","bot-triaged, enhancement, module: inductor, module: trace, oncall: pt2, triaged",thread:issue:189270:pr:187328,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189270 [inductor] Emit per-kernel physical op structure (with rematerialization) in provenance,1faaab11bcf8b1a455ee7c7901197d0bb21f5761576d2fea90641b4c7917fae3
issue-collapse:5b0aee7d65276f79,Review thread: 1 issue + 1 PR - pr #186965: [RFC][dynamo] Add ContextVar.get() tracing support (Phase 1),linked work thread,waiting on contributor,28,high,work_thread_graph,1,1,0,0,186449,186965,,,"module: compile ux, module: dynamo, module: graph breaks, oncall: pt2","bot-triaged, feature, large, module: compile ux, module: dynamo, module: graph breaks, oncall: pt2, triaged",thread:issue:186449:pr:186965,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186965; keep related issue/PR context attached.,2,Represents 1 open issue; raw issue rows remain available for audit.,#186449 [RFC]: Contextvars Support in TorchDynamo,58510a4f7d2f4bd83cfa520550a8ac0047b9c69865374523af3d54b88c74d5f3
issue-collapse:9d0ec5a2305aa9e8,Review thread: 1 issue + 1 PR - pr #186835: [RFC]: Auto-select and run Operator Performance Benchmarks from PR Diffs,linked work thread,draft/noise,28,high,work_thread_graph,1,1,0,0,186837,186835,,,"module: ci, module: performance","bot-triaged, feature, module: ci, module: performance, op-bench, triaged",thread:issue:186837:pr:186835,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#186837 [RFC]: Auto-select and run Operator Performance Benchmarks from PR Diffs,d6e267db548bc94e074e3eb427e33dc519af7e5cbdee7862d958aed96931d302
issue-collapse:d1c617582bd0360b,Review thread: 1 issue + 1 PR - pr #186209: Unify triton-xpu CD nightly build into 'triton' package with version suffix,linked work thread,draft/noise,28,high,work_thread_graph,1,1,0,0,186350,186209,,,"module: binaries, module: rocm, module: xpu","bot-triaged, feature, module: binaries, module: rocm, module: xpu, triaged",thread:issue:186350:pr:186209,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#186350 [Feature Request] Distribute triton with device-local version identifiers instead of separate package names,817d46382e15bd51abc25c99260e5f63203d45d603fca73fd8e724b1924ff74d
issue-collapse:390adbb49c2db483,Review thread: 1 issue + 1 PR - pr #185486: [OpenReg] Add minimal implementation for inductor integration in openreg,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,181093,185486,,,"module: backend, module: dynamo, module: openreg, module: privateuse1, oncall: pt2","bot-triaged, feature, module: PrivateUse1, module: backend, module: dynamo, module: openreg, oncall: pt2, triaged",thread:issue:181093:pr:185486,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#181093 [RFC] Add Minimal torch.compile Backend to Validate PrivateUse1 Integration,650f258d405b33ef07def80e21a4a8ad6892208f30fd3ea94fc6bbbd42724e0e
issue-collapse:9ad399bc847e1267,Review thread: 1 issue + 1 PR - pr #183487: Update XNNPACK submodule to 0b66c9f,linked work thread,PR blocked,28,high,work_thread_graph,1,1,0,0,183795,183487,,,module: xnnpack,"bot-triaged, module: xnnpack, triaged",thread:issue:183795:pr:183487,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183795 Clean up XNNPACK_NO_CODE_CACHE,fe3141dd5d52a754de4e5738fc7e3d5c10e354d0271327563db8fe1df60c8693
issue-collapse:f17a38c2f933ce4d,Review thread: 1 issue + 1 PR - pr #182376: [CUDA][CPU] Unify CPU/GPU relu/max behavior,linked work thread,PR blocked,28,high,work_thread_graph,1,1,0,0,181801,182376,,,"module: correctness (silent), module: cuda, module: edge cases, module: nn","bot-triaged, module: correctness (silent), module: cuda, module: edge cases, module: nn, triaged",thread:issue:181801:pr:182376,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#181801 Inconsistent Signed-Zero Handling Between CPU and CUDA Backends,d6d620ecec698f3a316a91e166b849c0805bff3ef8cdee288977718f4ccad3cb
issue-collapse:e441ae044b063039,Review thread: 1 issue + 1 PR - pr #180208: Update slow tests,linked work thread,PR blocked,28,high,work_thread_graph,1,1,0,0,187183,180208,,,"module: ci, module: tests","bot-triaged, enhancement, module: ci, module: tests, triaged",thread:issue:187183:pr:180208,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187183 Update slow tests job removes tests that may still be slow,449867c66f2fde471e896a7522482eec817e6c2bb9fa7da8ba0ea38eada7a820
issue-collapse:5f558581c226565f,Review thread: 1 issue + 1 PR - pr #180020: Add a debug tool to warn on NULL stream use,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,167885,180020,,,"module: cuda, module: performance","module: cuda, module: performance, triaged",thread:issue:167885:pr:180020,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167885 Ideas for Simple CUDA Dynamic Analysis Tools for Common Bugs,91bda4573c74ac8009138b96d4e31a162faa060b08b7c15304a623ab797bf007
issue-collapse:9c731f756d660f08,Review thread: 1 issue + 1 PR - pr #176678: [CI][Windows] Add NVIDIA RTX workflow,linked work thread,waiting on contributor,28,high,work_thread_graph,1,1,0,0,170753,176678,,,"module: ci, module: windows","module: ci, module: windows, triaged",thread:issue:170753:pr:176678,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #176678; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#170753 [CI][Windows][Infrastructure] Support for RTX GPU workflows on Windows,a4b7a3a70c5cbd90ab176c10c027da9dd8e6d6c93d28d5176b46486b4eb23fe7
issue-collapse:e69baa705af6f1a9,Review thread: 1 issue + 1 PR - pr #176317: [MPS] Leveraging unified memory memcpy instead of blit encoding for tensor device transfers,linked work thread,ready for maintainer decision,28,high,work_thread_graph,1,1,0,0,172987,176317,,,"module: memory usage, module: mps","enhancement, module: memory usage, module: mps, triaged",thread:issue:172987:pr:176317,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#172987 Leveraging Unified Memory for MPS Tensors on Apple Silicone,87197de160e2d1f3e1d16e9210c148f8174784d5669e6fb2a4120c17b4b4ddba
issue-collapse:b08445aa8d62a976,Review thread: 1 issue + 1 PR - pr #176265: Add backward batching rules for SDPA and fix attn_bias under vmap,linked work thread,has linked issue,28,high,work_thread_graph,1,1,0,0,187266,176265,,,"module: functorch, module: performance, module: sdpa, module: vmap","bot-triaged, feature, has workaround, module: functorch, module: performance, module: sdpa, module: vmap, triaged",thread:issue:187266:pr:176265,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187266 scaled_dot_product_attention falls back to math for unbatched (3D) inputs under vmap,159964cf378e1f1489a7639cbca65301f77ee1bb0e3b1c199745c166cd5c70f1
issue-collapse:e8b350ce58dfac88,Review thread: 1 issue + 1 PR - pr #173519: Add on-demand ppc64le wheel build support,linked work thread,waiting on contributor,28,high,work_thread_graph,1,1,0,0,180362,173519,,,"module: build, module: ci, module: power","bot-triaged, feature, large, module: POWER, module: build, module: ci, triaged",thread:issue:180362:pr:173519,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #173519; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#180362 RFC: Power (ppc64le) CI and Build Infrastructure Plan,9185fd3413a0645056fbe03e4fe3bc52e3f52bd47ecabe4ce3f38598055c1a2c
issue-collapse:df7ec3ff78712040,Review thread: 1 issue + 1 PR - pr #162706: Fix accelerator-related skip-decorators for tests,linked work thread,PR blocked,28,high,work_thread_graph,1,1,0,0,162179,162706,,,"module: fsdp, module: tests, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: fsdp, module: tests, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:162179:pr:162706,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#162179 test_fully_shard_logging.py is not being run,4a285937d25de19ff02bd11bfa3e44887fd29d000b4c143419ac4e5b4175cbdd
issue-collapse:f384b46703f6558b,"Review thread: 1 issue + 1 PR - pr #153557: [PP] wip, allow grad to be None",linked work thread,draft/noise,28,high,work_thread_graph,1,1,0,0,153484,153557,,,"module: pipelining, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: pipelining, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:153484:pr:153557,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153484 Pytorch PP requires all parameters to have grad in backward,64e36c5276f53d092282e7100525162df92f8643737446949a5af6bd2f496736
issue-collapse:9ccb01161f568980,Review thread: 3 issues + 2 PRs - pr #185902: Enable torch.compile for RNN modules by default,linked work thread,PR blocked,26,high,work_thread_graph,3,2,0,0,"91439, 140845, 149909","185902, 186091",,,"module: decompositions, module: dynamo, oncall: pt2","feature, low priority, module: decompositions, module: dynamo, oncall: pt2, triaged",thread:issue:140845:issue:149909:issue:91439:pr:185902,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 3 open issues; raw issue rows remain available for audit.,"#91439 Pytorch2.0 doesn't support compiling GRU and RNN model, #140845 compiled LSTM performance is worse than not compiled, #149909 [ued][kokoro] RNN/LSTMS do not work with torch.compile",97629fc785b948f2902e2bd9b2a3c0fc6e88f74d32c996e06101c2109a3e1c31
issue-collapse:38e9687e3303ee71,Review thread: 3 issues + 1 PR - pr #188006: [ROCm] Fix torch.pdist correctness on ROCm; re-enable test_pdist_norm_large,linked work thread,waiting on contributor,26,high,work_thread_graph,3,1,0,0,"168861, 168863, 168868",188006,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168861:issue:168863:issue:168868:pr:188006,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #188006; keep related issue/PR context attached.,5,Represents 3 open issues; raw issue rows remain available for audit.,"#168861 TestModule: test/test_torch.py::TestTorchDeviceType, #168863 Test: TestTorchDeviceType.test_cov, #168868 Test: TestTorchDeviceType.test_pdist_norm_large",46a9313e17541ca527c423199bf565905792c161eacaa7a82af8402e623f1800
issue-collapse:d90effe53697e1f6,Review thread: 2 issues + 2 PRs - pr #185815: Fix Dynamo NumPy scalar identity handling,linked work thread,PR blocked,26,high,work_thread_graph,2,2,0,0,"141258, 146951","185815, 186069",,,"module: dynamo, module: numpy, oncall: pt2","dynamo-user-empathy-day, module: dynamo, module: numpy, oncall: pt2, triaged",thread:issue:141258:issue:146951:pr:185815:pr:186069,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 2 open issues; raw issue rows remain available for audit.,"#141258 `torch.compile` fails with runtime error on Numpy scalar operations, #146951 [dynamo] `x is x` gets incorrectly interpreted to `False`",e2627cda7ffbf55effe50de8f33e1e33270ca73aeac52930d211b72adf7a7b39
issue-collapse:e085e1cbbc380dad,Review thread: 2 issues + 2 PRs - pr #185132: Fix custom op autograd defaults with tensorlist outputs,linked work thread,ready for maintainer decision,26,high,work_thread_graph,2,2,0,0,"123470, 162687","185132, 186466",,,"module: autograd, module: custom-operators, module: dynamo, module: pt2-dispatcher, oncall: pt2","module: autograd, module: custom-operators, module: dynamo, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:123470:issue:162687:pr:185132:pr:186466,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#123470 torch.compile x custom ops: op that accepts float also accepts Tensor in eager-mode, #162687 custom op with torch.compile backward error when op has default arguments",3d5a7ab3d480dffc757989bca3647fddb00a45ff72c3558f0f5ac30b3583bfec
issue-collapse:9e34e27b2606196b,Review thread: 2 issues + 1 PR - pr #186538: Fix CUDA driver linking for Inductor CUTLASS builds,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"103417, 103891",186538,,,"module: docker, oncall: pt2","module: docker, oncall: pt2, triaged, upstream triton",thread:issue:103417:issue:103891:pr:186538,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#103417 Torch model compile error ""/usr/bin/ld: cannot find -lcuda"" though cuda is installed via run file, #103891 Docker images: faster linker for `torch.compile`",945ea57872b4d7ae658b1f68cc4e50ad16524672e98ee58ededa6cfa462e5661
issue-collapse:cb26a3799b708e83,Review thread: 2 issues + 1 PR - pr #186406: Fix invalid source in AOT repro generation,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"128830, 134885",186406,,,"module: dynamo, module: minifier, oncall: pt2","module: dynamo, module: minifier, oncall: pt2, triaged",thread:issue:128830:issue:134885:pr:186406,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#128830 TORCHDYNAMO_REPRO_AFTER=aot produces invalid repro code, #134885 SyntaxError when Running repro.py",3566e90a643273decbbf22c3ef556c22c2c768582847ba64cbbbac8c7ceb1911
issue-collapse:4069d26745d0dbf5,Review thread: 2 issues + 1 PR - pr #186366: Support vmapped compiled autograd.Functions,linked work thread,waiting on contributor,26,high,work_thread_graph,2,1,0,0,"129845, 133493",186366,,,"module: dynamo, module: functorch, module: pt2-dispatcher, oncall: pt2","dynamo-autograd-function, dynamo-ctx-manager, dynamo-functorch, module: dynamo, module: functorch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:129845:issue:133493:pr:186366,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186366; keep related issue/PR context attached.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#129845 vmap + autograd.Function [generate_vmap_rule=True] + torch.compile don't work together, #133493 vmap + autograd.Function [generate_vmap_rule=False] + torch.compile don't work together",11f592ef2717955bfd700aede96bd03b09e2d2604051beecbb7b7c2203af6ba7
issue-collapse:5eaf78a8bb1e51e3,Review thread: 2 issues + 1 PR - pr #186180: Fix Dynamo cProfile under active Python profilers,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"137869, 139232",186180,,,"module: dynamo, module: vllm, oncall: pt2","actionable, module: dynamo, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:137869:issue:139232:pr:186180,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#137869 TORCH_COMPILE_CPROFILE does not work for python 3.12, #139232 TORCH_COMPILE_CPROFILE=1 doesn't work on python 3.12",9b346376fb79b6f24d86b38c623ec6a6ad40b8a35151e0cba73176e5226a9083
issue-collapse:8f20367bbc153247,Review thread: 2 issues + 1 PR - pr #185528: Fix Dynamo tracing of compile support checks,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"120115, 154259",185528,,,"module: dynamo, oncall: pt2","better-engineering, module: dynamo, oncall: pt2, triaged",thread:issue:120115:issue:154259:pr:185528,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#120115 `detectron2_maskrcnn` OOMs on eager with A100 40G., #154259 torch.compile regression blocks torchvision/torchbench pin upgrade",e064572d6dea8047adf2e3f46db31cfca212182404c607d6ca6712c87fe89baf
issue-collapse:2301fd7a0b15dfa2,Review thread: 2 issues + 1 PR - pr #185462: Fix NumPy ndarray operators under default device contexts,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"140884, 156162",185462,,,"module: dynamo, module: numpy, oncall: pt2","module: dynamo, module: numpy, oncall: pt2, triaged",thread:issue:140884:issue:156162:pr:185462,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#140884 `torch.compile` tries to invoke a torch op on numpy.ndarray in nightly release, #156162 torch2.7.1 issue for torch.compile numpy",bfcd7c2cc1e2f7698200d755650309151380fba72be31c45c97b069224170541
issue-collapse:567f131da95223e9,Review thread: 2 issues + 1 PR - pr #185356: Fix DTensor embedding mask state under compile,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"159843, 160697",185356,,,"module: dtensor, module: dynamo, oncall: distributed, oncall: pt2","module: dtensor, module: dynamo, oncall: distributed, oncall: pt2, triaged",thread:issue:159843:issue:160697:pr:185356,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#159843 [DTensor] nn.Embedding Compile Failure when Creating FX Graph, #160697 [DTensor] nn.Embedding Compile - Empty Masked Buffer from AOT Autograd",6da627741f0124f51a1364b5961d178914453bcad2353cd8e1fa973aa8d35c45
issue-collapse:bc0266f0f1fa1124,Review thread: 2 issues + 1 PR - pr #185116: Thread effect tokens through cond branches,linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"163143, 165981",185116,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:163143:issue:165981:pr:185116,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#163143 [torch.compile] Side-Effectful Tokens doesn't work well with torch.cond, #165981 Fails to use side_effect op inside torch.cond branch",b5b207fbd1e1298bca7c29af2e27f9442b3c2f5967ce50285c1cc11fc8223c6f
issue-collapse:21fa4efdf0d2d10e,Review thread: 2 issues + 1 PR - pr #184223: Run Inductor fallbacks below autograd,linked work thread,waiting on contributor,26,high,work_thread_graph,2,1,0,0,"139521, 139629",184223,,,"module: dispatch, module: inductor, oncall: pt2","module: dispatch, module: inductor, oncall: pt2, triaged",thread:issue:139521:issue:139629:pr:184223,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184223; keep related issue/PR context attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#139521 boxing-unboxing overhead seems significant, #139629 Inductor should run fallbacks without Autograd-related dispatch keys",3853104790e67e8839efc6fe889c0c252e1b229b90d2518557a5f27cbbd95e8b
issue-collapse:f43f2500dae4e1f5,Review thread: 2 issues + 1 PR - pr #184121: Improve persistent tiled reduction defaults,linked work thread,waiting on contributor,26,high,work_thread_graph,2,1,0,0,"148682, 149982",184121,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:148682:issue:149982:pr:184121,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184121; keep related issue/PR context attached.,6,Represents 2 open issues; raw issue rows remain available for audit.,"#148682 can we make inductor create faster fusions for tiled reductions across dim0 and dim1?, #149982 request for faster inductor kernels for blockwise reduction across dim1 -> write",ad1e8d4127636136c22f3693b8a9d844719974a981842de4d05e94da3b83643a
issue-collapse:172a259f81ab8f82,Review thread: 2 issues + 1 PR - pr #184016: Fix Inductor DCE for fallback alias mutations,linked work thread,waiting on contributor,26,high,work_thread_graph,2,1,0,0,"166009, 166014",184016,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:166009:issue:166014:pr:184016,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184016; keep related issue/PR context attached.,5,Represents 2 open issues; raw issue rows remain available for audit.,"#166009 [Inductor] DCE breaks with fallback operators that produce aliases + mutation, #166014 Make Inductor Fallback Nodes Less Reliant on Invariants from Functionalization / AOT Autograd",e6ef619825b69ab5bfd30ed6ff655f7fb9e300f72d2ff69ff9e4e86cb7c1f7c2
issue-collapse:80c401c18f46adf6,"Review thread: 2 issues + 1 PR - pr #132135: [torch.special] Adding betainc, betaincc, betaincinv, betainccinv, betaln and beta with backward operation",linked work thread,PR blocked,26,high,work_thread_graph,2,1,0,0,"132133, 145014",132135,,,"module: dynamo, module: special","module: dynamo, module: special, triaged",thread:issue:132133:issue:145014:pr:132135,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#132133 Adding betainc with backward operation, #145014 Add ATen functions in native_functions.yaml to torch_in_graph_functions list automatically",e094c241f55e417d52c21bde1f1f167311e2bb3c9dcb02d3c769fff13f3c851f
issue-collapse:289509dc89b16c0f,Review thread: 1 issue + 1 PR - pr #189174: Fix dcp privateuse1 stream support,linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,165097,189174,,,"module: privateuse1, oncall: distributed, oncall: distributed checkpointing","module: PrivateUse1, oncall: distributed, oncall: distributed checkpointing, triaged",thread:issue:165097:pr:189174,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#165097 Distributed checkpoint should not expect private use 1 to be cuda compatible,9ab78c5a061863f1c71210ed1643d233abcd240a5f485c990de7036098eda13c
issue-collapse:aa2f1bb3baa931ee,Review thread: 1 issue + 1 PR - pr #188906: Add torch.utils.checkpoint.register_determinism_check,linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,162980,188906,,,module: activation checkpointing,"actionable, module: activation checkpointing, triaged",thread:issue:162980:pr:188906,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#162980 Adding custom function to _allowed_determinism_checks_to_fns from model code,474656843cb4c50290fb699526c53fb15fb9bcdad8d66cd90242c7a67718cc03
issue-collapse:71039ca0e345a966,Review thread: 1 issue + 1 PR - pr #186897: Warn when custom op int schema specializes SymInts,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,154851,186897,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:154851:pr:186897,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186897; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#154851 Improve warning when specializations happen due to operator,059ad45369f9be549edfde9615daab377a6bbea7a35ad6ce6f6fbbdbe297f2a7
issue-collapse:1847c0275f55899d,Review thread: 1 issue + 1 PR - pr #186893: Fix dynamic size-one tensor recompiles,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,160271,186893,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:160271:pr:186893,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#160271 Error when using torch.compile with dynamic=True: tensor 'guidance' stride mismatch at index 0,79f16e21444ddb3bbd262d9913a6b159ba285ecad6c8f11a9d13278e45aef030
issue-collapse:cc284cda4a69c203,Review thread: 1 issue + 1 PR - pr #186886: Fix keyed user Triton autotune in Inductor,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,164124,186886,,,"module: dynamic shapes, module: user triton, oncall: pt2","module: dynamic shapes, module: user triton, oncall: pt2, triaged",thread:issue:164124:pr:186886,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#164124 torch.compile compiles multiple Triton autotune kernels, but uses the wrong ones",ec9080e0fe2520c2d991997212066406db54d16cb1db0ad05b9e89d9ed244e7e
issue-collapse:15b38c240f9dbca6,Review thread: 1 issue + 1 PR - pr #186564: Skip uninitialized outputs during accuracy minification,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,93437,186564,,,"module: minifier, oncall: pt2","module: minifier, oncall: pt2, small, triaged",thread:issue:93437:pr:186564,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#93437 Accuracy minifier can find spurious accuracy failures involving uninitialized memory,805593367b56612847b1d718cbde3e268d4d600035a3dc0f8fb34c60a18ee530
issue-collapse:0a121f0e530eeaea,Review thread: 1 issue + 1 PR - pr #186553: [dynamo] Graph break on packed sequence lower ops,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,93501,186553,,,"module: dynamo, module: performance, oncall: pt2","mlperf, module: dynamo, module: performance, oncall: pt2, triaged",thread:issue:93501:pr:186553,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#93501 _pack_padded_sequence fails in dynamo due to requiring a non-fake 2nd argument,af176ba174abcc583da627a064f85579c4acab3faa5f09b275833e13f87e2b56
issue-collapse:d3d00eb7fa395d7a,Review thread: 1 issue + 1 PR - pr #186551: Preserve symbolic tensor metadata in after-AOT repros,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,93367,186551,,,"module: minifier, oncall: pt2","module: minifier, oncall: pt2, triaged",thread:issue:93367:pr:186551,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#93367 Aot accuracy minifier with dynamic shapes doesn't work,f09da2e9fdfceea962d3bfaae0e35ca768be13fba65ee570fdb110d5e3d75a9e
issue-collapse:f08d139375ed865d,Review thread: 1 issue + 1 PR - pr #186546: Preserve original Dynamo minifier backend failures,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,97750,186546,,,"module: minifier, oncall: pt2","module: minifier, oncall: pt2, triaged",thread:issue:97750:pr:186546,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#97750 irrelevant error output for Minified repro,83d3f610da1e353d9f859efb63986283097c990cf8d506ec01e5666ea39974e0
issue-collapse:b32277f74ad2f0f8,Review thread: 1 issue + 1 PR - pr #186534: Account for fanout in Inductor reuse realization,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,105768,186534,,,"module: convolution, module: memory usage, oncall: pt2","module: convolution, module: memory usage, oncall: pt2, triaged",thread:issue:105768:pr:186534,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#105768 torch.compile uses more memory when using less parameters,d6eca10ee8643d1b32e6ae032fae9da5de515fcee53953fec557b7700c526622
issue-collapse:b9a29af7767e75ae,Review thread: 1 issue + 1 PR - pr #186533: Fix SymInt vector dispatcher signature detection,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,106135,186533,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:106135:pr:186533,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#106135 Registering function that takes const std::vector<c10::SymInt>& to SymInt[] schema gives confusing error message,bd449bb82fdeb77147c1687c9c7f54c814e3f45b04c6b3ad5a67922920b21996
issue-collapse:afbbe8dba48040f5,Review thread: 1 issue + 1 PR - pr #186531: Remove stale Dynamo skip for max_pool1d pooling test,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,111320,186531,,,"module: memory usage, oncall: pt2","module: memory usage, oncall: pt2, triaged",thread:issue:111320:pr:186531,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111320 test_max_pool1d reliably OOMs after https://github.com/pytorch/pytorch/pull/111216/,e3f0ea6cd73471af91facd77f6a455f4957626f199c4e84e878e9df90b71028d
issue-collapse:45cbe1048125b59c,Review thread: 1 issue + 1 PR - pr #186530: [dynamo] Restore autocast dtype after trace-time updates,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,111385,186530,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:111385:pr:186530,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111385 Torch Compile Dynamic fails on sample on diffusers VAE,de7e43a2c6a36373010259ac312a0648175eca59a3963368f05cd232eabca993
issue-collapse:887c67917ad6b7ab,Review thread: 1 issue + 1 PR - pr #186520: Preserve SymInt bindings in dynamic-shape minifier repros,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,114296,186520,,,"module: dynamic shapes, module: minifier, oncall: pt2","module: dynamic shapes, module: minifier, no-scrub, oncall: pt2, triaged",thread:issue:114296:pr:186520,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#114296 Minifier doesn't work with dynamic shapes,13111e0f07a05fee57e63c9e87616dc660f6f4d8b5316413d5e5c24851657818
issue-collapse:f12d75c8f878a7e6,Review thread: 1 issue + 1 PR - pr #186507: Invalidate Dynamo caches when module hooks change,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,117265,186507,,,"module: dynamo, oncall: pt2","dynamo-must-fix, internal ramp-up task, module: dynamo, oncall: pt2, triaged",thread:issue:117265:pr:186507,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#117265 [dynamo][inline-inbuilt-nn-modules]torch.compile silently incorrect with full_backward_pre_hook,aa66703da8292c5abffeb80e3c17cc29f1306e51ed4b6ca6269591169a16aa54
issue-collapse:149e42c40d41d707,Review thread: 1 issue + 1 PR - pr #186505: Simplify mod-one symbolic shape guards,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,118332,186505,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:118332:pr:186505,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#118332 Further simplifying guards,75d47e4a4e5a2953ce8e581150dcd087660c694870b2922610eaf8c8b8716f60
issue-collapse:37bc899ddff267c3,Review thread: 1 issue + 1 PR - pr #186504: Allow HOP speculation through torch autograd.Function skipfiles,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,118334,186504,,,"module: dynamo, oncall: pt2","dynamo-triage-june2024, module: dynamo, oncall: pt2, triaged",thread:issue:118334:pr:186504,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#118334 Using autograd.Functions defined in torch/ cause graph breaks,a9fb8274b02e297dc52b195ad1937c9551f1ed6b9d5e91b3bf68324364e89219
issue-collapse:0b6f899c5089441d,Review thread: 1 issue + 1 PR - pr #186503: Preserve input aliases for multi-output view returns,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,118739,186503,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:118739:pr:186503,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#118739 Getting ""RuntimeError: Trying to backward through the graph a second time"" when calling backward on compiled unbind / split ops",1ec90bc18df22613fa5a3d49a54edba1f2afef38ffcbc44442796b084e931077
issue-collapse:4480e7938db2e03f,Review thread: 1 issue + 1 PR - pr #186502: Make context decorator frames identify the context manager,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,118747,186502,,,"module: logging, oncall: pt2","module: logging, oncall: pt2, triaged",thread:issue:118747:pr:186502,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#118747 decorate_context results in uninformative backtraces,66355ff1bb68f94a36808bd89150221198c0c5c426062d77fb8cf480e2d1daec
issue-collapse:c6c50d5b985cddc3,Review thread: 1 issue + 1 PR - pr #186500: Fix Dynamo resume after f-string logging graph breaks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,120375,186500,,,"module: dynamo, oncall: pt2","dynamo-must-fix, dynamo-triage-june2024, module: dynamo, oncall: pt2, triaged",thread:issue:120375:pr:186500,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120375 Dynamo doesn't generate resume calls after graph breaking on log calls,58591e0fdf2c9b105e1258f6fd6ccd979584340b702a27a70e229cef317608ce
issue-collapse:e00a85443d7a0f45,Review thread: 1 issue + 1 PR - pr #186498: Add dynamic shape test assertions for input dims,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,120911,186498,,,"module: dynamic shapes, module: tests, oncall: pt2","module: dynamic shapes, module: tests, oncall: pt2, triaged",thread:issue:120911:pr:186498,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120911 Symbolic shape tests don't actually test that arguments aren't specialized,3d8b5714374d6737385f7cd0ba178d59807424746d54a5fffbdfaaae9434e023
issue-collapse:7767134448e73804,Review thread: 1 issue + 1 PR - pr #186495: Initialize compile context for non-strict export tracing,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,121318,186495,,,"module: logging, oncall: export, oncall: pt2","module: logging, oncall: export, oncall: pt2, triaged",thread:issue:121318:pr:186495,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#121318 torch/export/_trace.py and related non-strict entrypoints do not initialize TORCH_LOGS/TORCH_TRACE,8fe54b45465de6337b8ca8daa7427a8049a81f1dc8146f264d1b524fe43c08d8
issue-collapse:cbaba83d0988a5b2,Review thread: 1 issue + 1 PR - pr #186494: [Dynamo] Constant-fold CUDA Tensor.is_pinned(),linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,121526,186494,,,"module: user triton, oncall: pt2","module: user triton, oncall: pt2, triaged",thread:issue:121526:pr:186494,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#121526 Compiling RMSNorm Triton Kernal gives error,4bfcd2033dc9953d3cc7d4c6983660fa887d660e0792c8c89a73bb492b38c693
issue-collapse:336acd685a0923c5,Review thread: 1 issue + 1 PR - pr #186487: Run Dynamo compilation with empty functorch stack,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,122200,186487,,,"module: dynamo, module: functorch, module: pt2-dispatcher, oncall: pt2","module: dynamo, module: functorch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:122200:pr:186487,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#122200 Dynamo compilation should always run with empty functorch interpreter stack, even if ambient context has functorch",aa4f65e5800336da10baaa20550c4c327f73da7e8f46896ab426b56cace5832e
issue-collapse:883c731d962b4d11,Review thread: 1 issue + 1 PR - pr #186485: Fix export dynamic dims with size-one examples,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,122294,186485,,,"module: dynamic shapes, oncall: export, oncall: pt2","module: dynamic shapes, oncall: export, oncall: pt2, triaged",thread:issue:122294:pr:186485,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#122294 AOIInductor: Dynamic Shapes Specificiaton fails for SAM,65bcd384360469cb1224c999a1d719c725ad886a55dea8c46bc1220f55e46e60
issue-collapse:4df37735bbb48c44,Review thread: 1 issue + 1 PR - pr #186472: Graph break on dynamic unspecialized ModuleList indexing,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,122578,186472,,,"module: dynamo, oncall: pt2","dynamo-triage-june2024, module: dynamo, oncall: pt2, triaged",thread:issue:122578:pr:186472,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122578 Recompiles and cache_size_limit from detectron2 CycleBatchNormList,39174a092ba32cafee2d4a93b8df821d7df5011f9c191e9b0fe561aac3ec0f2d
issue-collapse:ea119d7e2930810e,Review thread: 1 issue + 1 PR - pr #186471: Handle symbolic tensors in overlap status,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,122773,186471,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:122773:pr:186471,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122773 torch.gather - Cannot call numel() on tensor with symbolic sizes/strides,02f20ebea91b5a2d5478c75d68dbf4cb67cdf1ac6fe69903a29388be9ae41506
issue-collapse:eccbdc96818fd075,Review thread: 1 issue + 1 PR - pr #186469: Wrap Dynamo-generated bytecode runtime errors,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,123374,186469,,,"module: dynamo, oncall: pt2","low priority, module: dynamo, oncall: pt2, triaged",thread:issue:123374:pr:186469,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123374 [dynamo] incorrect error traceback for runtime errors when executing dynamo codegen,9ed78fec87d64ab0abc1db80872cb646c7153bfdf585f0c81fe2c77f6d90de3f
issue-collapse:bb06f778d14933dd,Review thread: 1 issue + 1 PR - pr #186467: Preserve non-persistent buffers in compiled modules,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,123411,186467,,,"module: dynamo, oncall: pt2","dynamo-nn-modules, dynamo-triage-june2024, module: dynamo, oncall: pt2, triaged",thread:issue:123411:pr:186467,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123411 dynamo/fx doesn't honor 'non-persistent' buffers,e4aa8ac0ae2287a249cc11358f64f3403660149963b66e9a66b3917f05167a14
issue-collapse:7ab2844a35dc698d,Review thread: 1 issue + 1 PR - pr #186464: Fix singleton quotient divisibility replacement,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,123651,186464,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:123651:pr:186464,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123651 Divisibility substitution does the wrong thing with non-trivial max that would be < 2 after division,cefb10fbb6f964068d5c1bd1197933504a460aa1594eb8d72ed7f1f41c652883
issue-collapse:49aeda0540cb3d9c,Review thread: 1 issue + 1 PR - pr #186463: Split foreach bool inplace error samples,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,123663,186463,,,"module: mta, module: testing, oncall: pt2","module: mta, module: testing, oncall: pt2, triaged",thread:issue:123663:pr:186463,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123663 Improve Foreach OpInfo test infra to enable meta coverage,c99256e72aa461d10cc5f374974ad1cdf0aff32e3b0f4bfbac1c0d317326feaf
issue-collapse:2f811c83f20d506a,Review thread: 1 issue + 1 PR - pr #186462: Allow cond branch mutation during export,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,123972,186462,,,"module: higher order operators, module: onnx, oncall: pt2","module: higher order operators, module: onnx, oncall: pt2, triaged",thread:issue:123972:pr:186462,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123972 Dynamo export: limited support in Torch.cond,1aa08ec706f4bcdf6608d4ed58f7f4e0b5aef54d8962dce6b8f91a0d6e7f5e8b
issue-collapse:393f4acd24caa755,Review thread: 1 issue + 1 PR - pr #186461: Fix SymBool equality handling in symbolic shapes,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,124110,186461,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:124110:pr:186461,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124110 Equality between SymBool is extremely busted,326e8e4e766d763a8473e750cb90c708b0444d8c1cb4ceff1ae332148aef4022
issue-collapse:3c10233a6a119705,Review thread: 1 issue + 1 PR - pr #186460: Support dynamic mode in Dynamo explain,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,124163,186460,,,"module: dynamo, oncall: pt2","low priority, module: dynamo, oncall: pt2, triaged",thread:issue:124163:pr:186460,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124163 Support `dynamic=True` in torch._dynamo.explain,66d43c598ec2585faadc52377685f2d877677c0eff967922c956d94d4c1a8b9b
issue-collapse:aa285788cf43de10,Review thread: 1 issue + 1 PR - pr #186455: Fix AOTAutograd Dynamo memory leak repro,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,124747,186455,,,"module: ci, module: dynamo, oncall: pt2","module: ci, module: dynamo, oncall: pt2, triaged",thread:issue:124747:pr:186455,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124747 TestAOTAutograd.test_mem_leak_from_save_for_bw fails locally but not on CI when run with dynamo,a320dd35fb37212a24b27405cfd9c68d01536cbb4e253662f378868aad1225c3
issue-collapse:ad65f9d9fc081b6c,Review thread: 1 issue + 1 PR - pr #186446: Fix fake overlap handling for Dynamo try/except,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,126474,186446,,,"module: correctness (silent), module: dynamo, oncall: pt2","dynamo-side-effects, dynamo-triage-june2024, module: correctness (silent), module: dynamo, oncall: pt2, triaged",thread:issue:126474:pr:186446,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#126474 [dynamo] Handle inplace op aliasing errors,ced8014558e270a5b5ca327bf879d104ef44616d453e8ea29972b8e3f172267f
issue-collapse:1dc6e77bf69b8d36,Review thread: 1 issue + 1 PR - pr #186444: Fix Adam capturable zero-lr NaNs,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,126514,186444,,,"module: optimizer, oncall: pt2","module: optimizer, oncall: pt2, triaged",thread:issue:126514:pr:186444,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186444; keep related issue/PR context attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#126514 Weights become NaN with torch.compile optimizer capturable=True, lr=0.0, nn.Embedding",f2691d552403d47291c9d578fc0024fa4c4661a9331fcdd25448095a6fc62cac
issue-collapse:1db568e0995a50ff,Review thread: 1 issue + 1 PR - pr #186443: Validate random_ and uniform_ meta arguments,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,126834,186443,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:126834:pr:186443,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#126834 [dynamo] Check for invalid random_ arguments while tracing,7ade5c0dfcc483c923af4ded638c7590e2d7d63fab08692aae2993dced3e7a49
issue-collapse:96c37159c130f594,Review thread: 1 issue + 1 PR - pr #186435: Fix mkldnn FakeTensor error handling,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,127111,186435,,,"module: dynamo, module: faketensor, oncall: pt2","low priority, module: dynamo, module: fakeTensor, oncall: pt2, triaged",thread:issue:127111:pr:186435,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#127111 [dynamo] Detect mkldnn_max_pool2d/mkldnn view errors at tracing time,f2e7ccb9e0c014f9829abc4c0a53c530a53b3cbed3b2971fb49ad725bd93255f
issue-collapse:a0228e0032952664,Review thread: 1 issue + 1 PR - pr #186429: Fix ONNX export of pad_sequence with symbolic split lengths,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,127153,186429,,,"module: onnx, oncall: export, oncall: pt2","module: onnx, oncall: export, oncall: pt2, triaged",thread:issue:127153:pr:186429,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#127153 [ONNX] pad_sequence() is not exportable, with neither legacy onnx.export nor with dynamo_export",e7f9c271404042cea9b98e527209e0719c7eaaaaa04b59a3b98e656cdffa8ba9
issue-collapse:c51a9bc20854038a,Review thread: 1 issue + 1 PR - pr #186420: Run torchrec tests in Dynamo CI,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,128531,186420,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:128531:pr:186420,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186420; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#128531 test/dynamo/test_torchrec.py doesn't run in OSS CI (and won't run locally),a19cc3a18036ecc8d8c37f615c7f5053de9dacc67dbf5c103cb74e5071a5af7b
issue-collapse:6c77c8cc053aee45,Review thread: 1 issue + 1 PR - pr #186415: Fix Dynamo Parameter construction in new modules,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,128556,186415,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:128556:pr:186415,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#128556 Dynamo graph breaks when passing a constant value to a nn module initialized within a function,2c6d93824b025c11f37b801f7b92ea3e81517295a1f84bd7d2ab7a473ee965ea
issue-collapse:1eb4fc2c0c66bd30,Review thread: 1 issue + 1 PR - pr #186409: Fix subclass view replay with dynamic metadata,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,128649,186409,,,"module: dynamic shapes, module: nestedtensor, oncall: pt2","module: dynamic shapes, module: nestedtensor, oncall: pt2, tensor subclass, triaged",thread:issue:128649:pr:186409,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#128649 [Nested Tensor / subclasses] view(-1) and splitting dimensions with view() failures,e96eeb670c45bd9e53b034decc55db555d1f41ff90ad19ae0489b7f878d05ded
issue-collapse:045bd1b53a4d510e,Review thread: 1 issue + 1 PR - pr #186397: Keep constant-false torch._check in compiled graph,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,129327,186397,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:129327:pr:186397,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186397; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#129327 [export/dynamo] torch._check fails at compile time when the condition evaluates to False,c10401692df50305c55b3ce815226d25fd924627b9e62db5b09450cb809b0948
issue-collapse:138dda73b24eebf8,Review thread: 1 issue + 1 PR - pr #186380: Run python decompositions under functionalization,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,129418,186380,,,"module: functionalization, module: pt2-dispatcher, oncall: export, oncall: pt2","module: functionalization, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:129418:pr:186380,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#129418 functionalization should run underneath python decomps,0509d4eb003c80c870fa5de0be56c439f6a3e1036b366cc541aaab814573e1e6
issue-collapse:a1bb40afc7da4d03,Review thread: 1 issue + 1 PR - pr #186373: Disable duck sizing by default,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,129456,186373,,,"module: dynamic shapes, module: onnx, oncall: pt2","module: dynamic shapes, module: onnx, oncall: pt2, triaged",thread:issue:129456:pr:186373,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#129456 [dynamic][inline-inbuilt-nn-modules] Extra compilation - pytest test/onnx/dynamo/test_dynamo_with_onnxruntime_backend.py::TestDynamoWithONNXRuntime::test_llama_attention_with_local_backend_0,d90159ed83df1b0a8021db0c05414b570a7c97d010df5fd6fcd613b4cc8556c9
issue-collapse:9f993888ef344d7d,Review thread: 1 issue + 1 PR - pr #186372: Fix torch.Tensor construction from traced NumPy arrays,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,129648,186372,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:129648:pr:186372,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#129648 `torch.compile` mode and eager mode are different in generating tensor API behavior,f349614fcc3f84c5f551fa4f7993accf838f3d2458bf54e1e184a5a41bceb898
issue-collapse:2a9e243969ee6c03,Review thread: 1 issue + 1 PR - pr #186364: Prevent guard validation from mutating ShapeEnv,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,129936,186364,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:129936:pr:186364,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#129936 [symbolic-shapes] Guard validation in Dynamo changes the shape_env,2745a0dda2ff0e5e8320efb22c931bc7a48cec78b64a523f84cff7e3739067f7
issue-collapse:9fb8ebd19f18127c,Review thread: 1 issue + 1 PR - pr #186355: Fix AOTAutograd unused output backward pruning,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,131794,186355,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:131794:pr:186355,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#131794 [torch.compile] Unused differentiable outputs lead to unnecessary gradient computation when compared to eager,e6fc1025eb1f4ca8568bd5bb454030eda4d6cbbd7618fb9e00f185174cb1fda1
issue-collapse:18d0554dafec6a03,Review thread: 1 issue + 1 PR - pr #186349: Improve Dynamo logger method graph-break guidance,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,132635,186349,,,"module: dynamo, oncall: pt2","dynamo-logging, module: dynamo, oncall: pt2, triaged",thread:issue:132635:pr:186349,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#132635 [Dynamo] support logging via reorderable_logging_functions,700e1ea3bcc2978ea33f7a968c6e0a450afadbf8dd7b650e67f0d5fa30e2ea17
issue-collapse:ff50e3bc2e7006cb,Review thread: 1 issue + 1 PR - pr #186340: Ensure explicit compile attempts skipped code,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,132923,186340,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:132923:pr:186340,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#132923 If you compile a function/module, we should definitely attempt to compile it (don't try to skip it)",0f5211f89e0748a52031c8c5a4a3fe42f182c1fde7390b6d47e277543e9465f8
issue-collapse:f2698f61670925ce,Review thread: 1 issue + 1 PR - pr #186291: Fix export distribution validation assertions,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,135061,186291,,,"module: dynamic shapes, oncall: export, oncall: pt2","module: dynamic shapes, oncall: export, oncall: pt2, triaged",thread:issue:135061:pr:186291,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#135061 [export] non-strict export error on torch.distributions.Normal,03625cf84f0a8bb3420c45324c9e4f4fcffece488e1144eec95b770971c032d8
issue-collapse:ac32e0fbbca53a5d,Review thread: 1 issue + 1 PR - pr #186287: Reject non-functional export decomposition entries,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,135759,186287,,,"module: pt2-dispatcher, oncall: export, oncall: pt2","export-triaged, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:135759:pr:186287,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#135759 core_aten_decompositions table have non-functional or CIA entries.,ce6f3499b23c7196d697443e9c38cf8299e1f15a2fa4d7ca405d23a2bfe97fad
issue-collapse:a22dd70952c2b448,Review thread: 1 issue + 1 PR - pr #186283: Trace data_ptr through tensor constructors,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,136271,186283,,,"module: dynamo, module: user triton, oncall: pt2","module: dynamo, module: user triton, oncall: pt2, triaged",thread:issue:136271:pr:186283,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#136271 [triton x pt2] Dynamo should trace data_ptr accesses,eb885c882489c637eb62247d4b59a12bd437b594ac9e3ab9ca31bbfe2aec5d4b
issue-collapse:3d324e57442764ab,Review thread: 1 issue + 1 PR - pr #186280: Disable profiler Python tracing during compile,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,136404,186280,,,"module: vllm, oncall: profiler, oncall: pt2","actionable, module: vllm, oncall: profiler, oncall: pt2, triaged, vllm-compile",thread:issue:136404:pr:186280,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#136404 Profiling Torch.compile (with_stack=True) leads to frozen program with massive memory.,de68bb0355bbbeb1d65519106340946ffe11a6a039b49c1aacab4b620457a961
issue-collapse:f954e02299e0c45a,Review thread: 1 issue + 1 PR - pr #186263: Increase small fake tensor constant limit,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,136642,186263,,,"module: dynamic shapes, module: faketensor, module: pt2-dispatcher, oncall: export, oncall: pt2","module: dynamic shapes, module: fakeTensor, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:136642:pr:186263,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#136642 Linespace does not support SymInt step. (but int),9a1c683e09df4041fa14d8e2694975091c21002185ee1ec67d122791257f3241
issue-collapse:ec443013dd3e4a1c,Review thread: 1 issue + 1 PR - pr #186249: [dynamo] Support time.time as a runtime scalar input,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,137068,186249,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:137068:pr:186249,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137068 UserWarning: Graph break due to unsupported builtin time.time,cee3bc5785b058b656a6acc4efd4c2ec9c6821fa39d9b2f1431976a239b630a8
issue-collapse:108869d1e09207ef,Review thread: 1 issue + 1 PR - pr #186225: Preserve NJT nested ints in AOT subclass outputs,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,137275,186225,,,"module: nestedtensor, oncall: pt2","module: nestedtensor, oncall: pt2, triaged",thread:issue:137275:pr:186225,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137275 Device transfer for NJT within torch.compile allocates a new nested int,c2a336c43706446f48ab287e547b9736a27a9a4351cc72e4c2c31c70fbc1e32a
issue-collapse:3c067ccbe4a8e1db,Review thread: 1 issue + 1 PR - pr #186222: Include epilogue guards in Dynamo code_parts,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,137388,186222,,,"module: dynamo, module: guards, oncall: pt2","module: dynamo, module: guards, oncall: pt2, triaged",thread:issue:137388:pr:186222,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137388 some guards are not shown in the code_parts,0e88797cb892edd478dbe9f8b5afba6744556bc3f56502c44397aa15ec2852a6
issue-collapse:ffda45201599cde8,Review thread: 1 issue + 1 PR - pr #186217: Move FX graph cache hash details out of debug logs,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,137875,186217,,,oncall: pt2,"compile-cache, oncall: pt2, triaged",thread:issue:137875:pr:186217,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137875 TORCH_LOGS=+inductor spams too much information about cache,611f744afdcc8ba567c6c9074c03b4cfe29da72853e9caf0d835bf96018fb48a
issue-collapse:59202d3a2549fb63,Review thread: 1 issue + 1 PR - pr #186214: Profile lazy backward compilation with cProfile,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,137996,186214,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:137996:pr:186214,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#137996 cprofile on compiler should trigger on backwards,747944253c30d3df260edc67a89928345c31f6631afefb6735de9a487e4a03bc
issue-collapse:4de09831874539c7,Review thread: 1 issue + 1 PR - pr #186194: Canonicalize trunc of symbolic integer division by one,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,138926,186194,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",thread:issue:138926:pr:186194,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138926 torch.cond complains two tensors have different metadata,ed9a129dfbdb2aa91fe2d11592414be7712f99848e213a9c5231c74948189e81
issue-collapse:da2026c4799eca24,Review thread: 1 issue + 1 PR - pr #186192: Fix CompiledFxGraph callable state across threads,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,138961,186192,,,"module: multithreading, oncall: pt2","module: multithreading, oncall: pt2, triaged",thread:issue:138961:pr:186192,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138961 CompiledFxGraph.current_callable is not thread-safe,500aa9345e73695a366a8b94bfe82b6a545c38e967472017dd3e25bc2f20cb16
issue-collapse:0f421b5cdc2c6f7a,Review thread: 1 issue + 1 PR - pr #186189: Report dynamo expected-failure XPASSes,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,139080,186189,,,"module: dynamo, module: tests, oncall: pt2","module: dynamo, module: tests, oncall: pt2, triaged",thread:issue:139080:pr:186189,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#139080 [dynamo] `dynamo_expected_failures` is silent on certain tests that ended up passing,be126c940ac23c599a6755e55dec591ac932a9c77371b439797ddadc30df8b6a
issue-collapse:7a3f7c1eb1b87829,Review thread: 1 issue + 1 PR - pr #186188: Fix FakeTensor wrapping for meta tensor constructor,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,139092,186188,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","actionable, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:139092:pr:186188,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#139092 `tensor` not a `FakeTensor` under `FakeTensorMode` and `device('meta')`,761df1d0a32b47d0d1b0b5ac2c430960b5c187eb0b473a4acc481c0bc3345a73
issue-collapse:813fdf4a4589ce65,Review thread: 1 issue + 1 PR - pr #186168: Fix Dynamo tracing of torch.random.fork_rng,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,139603,186168,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:139603:pr:186168,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#139603 `torch.compile` doesn't support `set` arguments,83c04e261f225704a62ab40902e76f819a07068b21c1db378f97eedfdae3a859
issue-collapse:b4829f463bf87d06,Review thread: 1 issue + 1 PR - pr #186160: Fix FakeTensor forward AD batch_norm meta rewrapping,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,139756,186160,,,"module: forward ad, oncall: pt2","module: forward ad, oncall: pt2, triaged",thread:issue:139756:pr:186160,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #186160; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#139756 Bug with torch.compile + fwAD,031f1ca7465d04ae0275feb82377740b9187e1d2ff7e710fc9066a438e84370e
issue-collapse:084ef428410c57c8,Review thread: 1 issue + 1 PR - pr #186150: Fix compile storage semantics for as_strided scatter,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,140171,186150,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","low priority, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:140171:pr:186150,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 1 open issue; raw issue rows remain available for audit.,#140171 Compiling `as_strided` call on a view input errors.,95191f20c049f152d3219347e71b5bc4ac7afb277eb8cf8fdc7a3ff8b6436bd8
issue-collapse:f3819264dd6fb8e2,Review thread: 1 issue + 1 PR - pr #186082: Improve SymIntArrayRef concrete-int diagnostics,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,140960,186082,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:140960:pr:186082,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140960 Better Error Message for SymIntArrayRef expected to contain only concrete integers,6bc6ab73a5200e45451f13945b15f97ccbf31dc3d311fd265b42d7c66a3ca412
issue-collapse:b4192b78a89928b9,Review thread: 1 issue + 1 PR - pr #186078: Fix OptimizedModule _parameters access,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,141005,186078,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:141005:pr:186078,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141005 OptimizedModule _parameters attr getting and setting are not working as expected,e1b3fc5ed6558459a8a31e2b18df3c79ad931d1b029dbaeec0caf485ee28ec17
issue-collapse:1cac6719d740b7c6,Review thread: 1 issue + 1 PR - pr #186074: Fix multilabel_margin_loss decomposition sentinel handling,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,141017,186074,,,"module: pt2 accuracy, oncall: pt2","module: pt2 accuracy, oncall: pt2, triaged",thread:issue:141017:pr:186074,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141017 multilabel_margin_loss gives incorrect results in torch.compile,a1d3e987d8e9fc491af6fd451183ea57a907154661c1d3542fe857cbe2c8f81d
issue-collapse:28de99403d301268,Review thread: 1 issue + 1 PR - pr #186072: Make constant warnings reorderable in Dynamo by default,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,141162,186072,,,"module: dynamo, oncall: pt2","empathy-day, module: dynamo, oncall: pt2, triaged",thread:issue:141162:pr:186072,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141162 Dynamo should reorder warnings (and other logging) by default if they don't involve tensor values,2ea22d3976095f9e491bc20c6c5ca476c87b2a3190fbb98615b7c1923e7e682e
issue-collapse:860debbbb89d48ad,Review thread: 1 issue + 1 PR - pr #186063: [dynamo] Route OptimizedModule delegated methods through compiled call,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,141473,186063,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:141473:pr:186063,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141473 How to use torch.compile + HF model?,7fdfbe28196d1b46c1cf637fc3b0b8355c40eb8f6d398227b3317438a5ae84e2
issue-collapse:8fa8c2b5a872e1b2,Review thread: 1 issue + 1 PR - pr #186053: Fix AOTAutograd cache save for call-free lazy backwards,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,141790,186053,,,"module: aotdispatch, oncall: pt2","module: aotdispatch, oncall: pt2, triaged",thread:issue:141790:pr:186053,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141790 AOTAutograd cache always misses even on simple program,cda9af8b7fa2158ca8710648ccfcc8a3d7e10c12096311bc18032477c1d792ef
issue-collapse:4ea38d35e6052a92,Review thread: 1 issue + 1 PR - pr #186052: Fix Inductor C++ pybinding source literal escaping,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,142321,186052,,,"module: windows, oncall: pt2","module: windows, oncall: pt2, triaged",thread:issue:142321:pr:186052,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#142321 torch.compile 路径处理错误导致 C++ 编译失败,2afbd875756978257f4f465b62c130a684b32453a8b6c75148ea39bb5d21a1df
issue-collapse:b1a999bc4713d16b,Review thread: 1 issue + 1 PR - pr #186044: Stop Dynamo from importing optional libraries at import time,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,142489,186044,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:142489:pr:186044,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#142489 Stop importing arbitrary third-party libraries in Dynamo,d4656db7234a1abc00e263e991c2aef7f6205aff32f6cb103855ff48c126a9c7
issue-collapse:4639a98d2a00d520,Review thread: 1 issue + 1 PR - pr #186039: Fix Dynamo verbose guard checks for frame locals,linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,143308,186039,,,"module: dynamo, module: guards, oncall: pt2","module: dynamo, module: guards, oncall: pt2, triaged",thread:issue:143308:pr:186039,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#143308 [dynamo, guards] Implement FrameLocalsMapping version of check_verbose_nopybind",30241bf8d8f0971844be451c629b6b426df93526a5284016d91752e3b52eb95c
issue-collapse:f6e4000242195a4a,Review thread: 1 issue + 1 PR - pr #186032: Avoid reshape metadata guards on symbolic size-one checks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,143463,186032,,,"module: dynamic shapes, module: dynamo, oncall: export, oncall: pt2","module: dynamic shapes, module: dynamo, oncall: export, oncall: pt2, triaged",thread:issue:143463:pr:186032,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143463 unreasonable ConstraintViolationError when using torch dynamo to compile torch model,52e3a7f3be3e5a09d667af89908b2fc82ce388c6f198e601092b7efc1491ecea
issue-collapse:22bea6e90973420d,Review thread: 1 issue + 1 PR - pr #186027: Fix symbolic broadcast inference for unbacked sizes,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,143495,186027,,,"module: faketensor, oncall: pt2","module: fakeTensor, oncall: pt2, triaged",thread:issue:143495:pr:186027,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#143495 infer_size(a, b) fails when it could return a value",76331de70cfc247ee4bda5554556f3c7f288d823fbfdbaa503548d88a188be45
issue-collapse:5ffdc1770c91e3d1,Review thread: 1 issue + 1 PR - pr #186003: Fix Dynamo condition-dependent frame skip caching,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,144820,186003,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:144820:pr:186003,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#144820 [dynamo] Do not always skip code objects unconditionally,39b05068941373492ef29b444c5a801c762ecdcb2ec51defd2de8b60aa4343a4
issue-collapse:220ad4a91f096aa2,Review thread: 1 issue + 1 PR - pr #185971: Fix Dynamo random.Random replay on cache hits,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,145445,185971,,,"module: dynamo, oncall: pt2","dynamo-triage-jan2025, module: dynamo, oncall: pt2, triaged",thread:issue:145445:pr:185971,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145445 [dynamo] `random.Random` gives wrong result on second call,0a26c4b7c88fa6a3aee8202c38508902017e7a84946a6beec98aae34752bcc24
issue-collapse:6822aa897dcf051d,Review thread: 1 issue + 1 PR - pr #185947: Fix Dynamo logit with tensor eps,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,145596,185947,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:145596:pr:185947,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145596 [Dynamo] compile torch.logit with different data types,b1291b274726c0f5be5a79cbe5bfdabbc65a962b5ca78ec18b95a188efc45f0b
issue-collapse:9e78e40bebad9b8b,Review thread: 1 issue + 1 PR - pr #185922: Fix fake one_hot invalid class count checks,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,146274,185922,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","actionable, module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:146274:pr:185922,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146274 torch.nn.functional.one_hot has inconsistent behavior between eager and torch.compile when num_classes=0,995b38a26bc791fa45b40e8bfccc7cb23f0baa7d1486b3af41fdf27fcb59b99f
issue-collapse:f002cdfa5cf92ccd,Review thread: 1 issue + 1 PR - pr #185920: [Dynamo] Detect tensor mutations when reconstructing generators,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,146628,185920,,,"module: dynamo, oncall: pt2","dynamo-side-effects, module: dynamo, oncall: pt2, triaged",thread:issue:146628:pr:185920,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146628 Dynamo should consider tensor mutation when reconstructing generator,ae9c34e88c1ff9afbb5a85314573a11bac4f0b64ad94e0ef94e1a31d08ef884c
issue-collapse:1014b4617ef83de0,Review thread: 1 issue + 1 PR - pr #185908: [Dynamo] Preserve live dict view semantics around mappingproxy mutation,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149583,185908,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149583:pr:185908,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149583 Easydict support,284133aa12d342e0909f786a7435a1fb9a409eb817cbd1dc2a5b9be3f15ec436
issue-collapse:4da59ffa7fbb8c95,Review thread: 1 issue + 1 PR - pr #185899: Fix Dynamo data mutation on functorch tensors,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,149911,185899,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149911:pr:185899,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185899; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#149911 [dynamo] `torch.compile` doesn't respect `GradTrackingTensor`'s data attribute mutation check,e503e59bcc4efeeb527afc8b8fe2f561dd0a0c50c20e1e08ebe4ab1742519148
issue-collapse:f83c7b9d7be6b5a8,Review thread: 1 issue + 1 PR - pr #185823: [Dynamo] Handle class method_descriptor calls generically,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,146673,185823,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:146673:pr:185823,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#146673 Dynamo should handle `class.method_descriptor(instance, *args, **kwargs)` generically",452766b41a892eca94914c9d03923ad0d5e4f501d37fe4c643d7fcb7a8641712
issue-collapse:556f2c3d23b39df8,Review thread: 1 issue + 1 PR - pr #185800: Fix TORCH_LOGS stderr handling at process exit,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,147486,185800,,,"module: logging, module: testing, oncall: pt2","module: logging, module: testing, oncall: pt2, triaged",thread:issue:147486:pr:185800,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#147486 remote_cache.py loggings cause a logging error after pytest exit,cc36e8060452f3d7c453574915180228aafb289398a7389ac17c866ce05395e3
issue-collapse:07344516f5712462,Review thread: 1 issue + 1 PR - pr #185799: Avoid dataclass constructors in Dynamo and make_fx graphs,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,147530,185799,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:147530:pr:185799,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#147530 [fx][dynamo][be] Don't allow arbitrary dataclass in the graph,b3791814a676cd705518438ddc840d103408a2ac73409847139ff21fa773eecc
issue-collapse:7fa0c9af4cbcfc91,Review thread: 1 issue + 1 PR - pr #185795: Preserve overlapping input layouts when cloning Dynamo inputs,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,147843,185795,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:147843:pr:185795,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#147843 The opt_output in `fx_graph_runnable.py` is inconsistent with the actual output when testing run_repro(acc=True).,01a8d851447cc3527b3c17f16594d6ed870e6de674bc2b68b82a93fc242e6ce7
issue-collapse:cd807ddba4a7c225,Review thread: 1 issue + 1 PR - pr #185781: Avoid unbound symbols in insignificant output strides,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,148199,185781,,,"module: dynamic shapes, module: dynamo, oncall: cpu inductor, oncall: pt2","module: dynamic shapes, module: dynamo, oncall: cpu inductor, oncall: pt2, triaged",thread:issue:148199:pr:185781,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148199 [Inductor-CPU] qlinear_binary output may have undefined strides with dynamic shape support,2c4c95d99275b94f1e5f3d25d84810b45f5961b66116529002f1a0f9c7c535b4
issue-collapse:cb1b8bd03892d288,Review thread: 1 issue + 1 PR - pr #185776: Propagate fake RuntimeErrors to Dynamo handlers,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,148475,185776,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:148475:pr:185776,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148475 Dynamo replaces exception by hard error in `run_node`,43bb33e417af2dfd740845389bbc7b739836b3e27b993ec656a82fa50f30a52e
issue-collapse:62c3b62e70bec13b,Review thread: 1 issue + 1 PR - pr #185774: Fix non-strict export for Tensor.numpy,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,148779,185774,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2, triaged",thread:issue:148779:pr:185774,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148779 [Export Benchmark] non-strict export doesn't work with .numpy on Kokoro model,8ee8fa772b06b7c2ad19b032b8d40e8c462c176481557c8abc602463ce828793
issue-collapse:27157549e8ad5c2a,Review thread: 1 issue + 1 PR - pr #185773: Use max dynamic bounds for Inductor autotuning,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,148842,185773,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:148842:pr:185773,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148842 We should use max size instead of hint size when autotuning,0127552c95c362974027c39a803de93f76584f39d153df03f2e0d593a07e4104
issue-collapse:42fac1f08e3717de,Review thread: 1 issue + 1 PR - pr #185772: Fix CPU small-output bilinear upsample precision in Inductor,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,148859,185772,,,"module: pt2 accuracy, oncall: pt2","module: pt2 accuracy, oncall: pt2, triaged",thread:issue:148859:pr:185772,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148859 [Inductor] Output mismatch shape after compilation,b959f2bacd070a04d3380990b23a340e997befcdfc0ffc24ba003952e3387423
issue-collapse:35231f592ecc9e74,Review thread: 1 issue + 1 PR - pr #185771: Support Dynamo tracing through textwrap.indent string formatting,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149010,185771,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149010:pr:185771,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149010 [XFORMERS] torch._dynamo.exc.Unsupported,f351d63a7468eb0d5321231812c1242bad4608691b729334f0fc4dbbcf37dbb9
issue-collapse:34da1f90f12a498e,Review thread: 1 issue + 1 PR - pr #185769: Preserve generator return values in Dynamo,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149037,185769,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:149037:pr:185769,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149037 (Will PR if ok) Support generator returning values,c43e42dee36c47321cf0713f65b5aa9a9c40a1758c96882e1cc3ce6b4d52f587
issue-collapse:c030b967f6d22624,Review thread: 1 issue + 1 PR - pr #185768: Improve Dynamo error for dynamic module hook registration,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149047,185768,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149047:pr:185768,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149047 Unsupported: call_method NNModuleVariable() register_forward_hook [NestedUserFunctionVariable()] {},a9ed206d3222361add3707a857a93754c6ca2c4388309c2fab62fcbc31223c4e
issue-collapse:452a44cdfb7e87fb,Review thread: 1 issue + 1 PR - pr #185766: Dynamo: support torch.jit.isinstance without graph breaks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149556,185766,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149556:pr:185766,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149556 [ued][f5-tts][dynamo] dont graph break on `torch.jit.isinstance`,b4cb8ee689e35e98fe5280655f59240466a8ee268d256f2f76b4f01606a8e242
issue-collapse:530d485c7ee9b472,Review thread: 1 issue + 1 PR - pr #185764: Recommend nonstrict_trace for skipped C extensions,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,149565,185764,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:149565:pr:185764,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185764; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#149565 Should (eventually) be recommending nonstrict_trace instead of allow_in_graph in error messages:,56cbd658e53f96c0c9e92d7a2b0eb8b65c475aad939b43d3608a74a38a3d9111
issue-collapse:8429f105a27a734d,Review thread: 1 issue + 1 PR - pr #185763: Improve disable API graph break messages,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149566,185763,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149566:pr:185763,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149566 User-torch._dynamo.disable annotations may be misleading,5a910b4402ad7baa293b1dec09b6d1a603c08ca7f05c29e038bc547d2cacb609
issue-collapse:48ce7a8ef6386532,Review thread: 1 issue + 1 PR - pr #185755: Use gmem efficient attention kernel for long fp32 KV,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,149857,185755,,,"module: sdpa, oncall: pt2","module: sdpa, oncall: pt2, triaged",thread:issue:149857:pr:185755,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149857 SDPA (`EFFICIENT_ATTENTION`) slower than torch.compile decomposition on `tf32`,28808baae588370c405e06bd9214bcef958daa637dbb2a52ce3ebe42837ec19f
issue-collapse:6e90aff5a04509f5,Review thread: 1 issue + 1 PR - pr #185752: Make OptimizedModule state_dict use public keys,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149894,185752,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149894:pr:185752,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149894 [ued][f5-tts][dynamo] `torch.compile` changes state dict,47f85c1a6903786131b0db8cd747a42ba81ac68a70e9d1672e8605934f289041
issue-collapse:5ad4137bce2c8573,Review thread: 1 issue + 1 PR - pr #185749: Fix Dynamo construction for classes with __getattribute__,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,149895,185749,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149895:pr:185749,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149895 [Dynamo] Cannot instantiate class if `__getattribute__` is defined,aa8dce6470a2c1a6574e25e840a345409c34f30cd62fce2c6e70460a17a70c4b
issue-collapse:dd4403eb6e604d9e,Review thread: 1 issue + 1 PR - pr #185748: [dynamo] Model RemovableHandle hook registration side effects,linked work thread,draft/noise,26,high,work_thread_graph,1,1,0,0,149921,185748,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149921:pr:185748,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149921 [ued][whisper][dynamo] Graph break - setattr on class object,f2d1a9c3e68aa91bb2206d82dd69938277fbdb7ebaf456d50e39d1979089f0b3
issue-collapse:a509719fae889d46,Review thread: 1 issue + 1 PR - pr #185740: Support inspect.signature on nested functions in Dynamo,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,149968,185740,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:149968:pr:185740,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149968 [ued][qwen][dynamo] inspect.signature with non-constant function,8c02474167a1a7a07e1e2b4026ac4e391cde408d6d1d0c52eacf11eab80a9033
issue-collapse:2663768b7779463a,Review thread: 1 issue + 1 PR - pr #185731: Clear stale prefix exception table entries in Dynamo,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,150319,185731,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:150319:pr:185731,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150319 [dynamo] Bad error message when trying to compile async functions due to bad bytecode jump targets,4ff972a439302a8ddffec3c8fc8f0133bb6c4aa06d3138156c062cfc9584af5b
issue-collapse:e0983944e7e1de13,Review thread: 1 issue + 1 PR - pr #185726: Improve Dynamo stack for nonstrict_trace input errors,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,150465,185726,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:150465:pr:185726,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150465 `torch._dynamo.nonstrict_trace` has confusing user code stacktrace,e017e783e0c766fb8cafa4c1dfb9bd508c9fe223a9b1dc3e807e626fcc7e650d
issue-collapse:0fb9de7b8f7f9b02,Review thread: 1 issue + 1 PR - pr #185724: Fix scalar searchsorted decomposition for unbacked values,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,150613,185724,,,"module: decompositions, module: dynamic shapes, oncall: pt2","module: decompositions, module: dynamic shapes, oncall: pt2, triaged",thread:issue:150613:pr:185724,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#150613 Torch compile for `torch.searchsorted` failed when capturing scalar outputs, with scalar `values` taken from `sorted_sequence`",b67de44ec4d2914ab3e523d7516177222792eb3eb5c27d3d169eed968ecf0454
issue-collapse:e54ced4c228e62b9,Review thread: 1 issue + 1 PR - pr #185722: [dynamo] Wrap module.compile leaf module call frames,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,150915,185722,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:150915:pr:185722,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150915 `module.compile()` behaves differently from `torch.compile(module)`,df9824f691eb6b7729ab40d0184413e9755829121217be9aff3a70bb752db3f0
issue-collapse:ab6078ac4fa726be,Review thread: 1 issue + 1 PR - pr #185690: Optimize AOT input view mutations,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,152346,185690,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:152346:pr:185690,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185690; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#152346 compile generates inefficient code for mutations on small slices of inputs,0e038e687b7127f915106868d2f6b3120813b21f2f882c9a26a1933081d6460b
issue-collapse:2111160f8a42126a,Review thread: 1 issue + 1 PR - pr #185632: Make AOTAutograd metadata repr subclass-safe,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,152183,185632,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:152183:pr:185632,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#152183 write a custom ViewAndMutationmeta.__repr__,33e959b997b97610cb6a5ebf82fc21e9398b609de16f1489c24e3b1f4855eeba
issue-collapse:494757eb29c12000,Review thread: 1 issue + 1 PR - pr #185624: Fix subclass static input remapping for SymInt args,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,152343,185624,,,"module: __torch_dispatch__, module: cuda graphs, module: dynamic shapes, oncall: pt2","module: __torch_dispatch__, module: cuda graphs, module: dynamic shapes, oncall: pt2, triaged",thread:issue:152343:pr:185624,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#152343 cudagraphs: `static_input_indices` incorrectly including SymInt graph args when using tensor subclasses + dynamic shapes,155680990696eb7f130f12c8ad8f9df99c8e6d78eb1e1f26be66055a33ae843c
issue-collapse:59bf53df7f705ac7,Review thread: 1 issue + 1 PR - pr #185591: Canonicalize avg_pool2d singleton lists during export,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,153149,185591,,,"module: docs, module: nn, module: pooling, oncall: export, oncall: pt2","actionable, module: docs, module: nn, module: pooling, oncall: export, oncall: pt2, triaged",thread:issue:153149:pr:185591,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153149 Behavior of kernel_size parameter of torch.nn.functional.avg_pool2d does not match with documentation,2186811d2d23c39c0653a0b2fa9b1536d89c2f1bd5fe8af91b4e3dc89e96bd13
issue-collapse:1ea2cec300b73e39,Review thread: 1 issue + 1 PR - pr #185588: [inductor] Codegen backed SymInt input runtime asserts,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,153175,185588,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:153175:pr:185588,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153175 prefer_deferred_runtime_asserts_over_guards does not work correctly e2e with compile. insert_deferred_runtime_asserts and inductor codegen skips them,cbfe20d0c08dc0207e671973b5064a5a074391f54b38a4062326700e4de1a29a
issue-collapse:6b51f38638129ea1,Review thread: 1 issue + 1 PR - pr #185585: Fix mark_unbacked stride recompiles,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,153204,185585,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:153204:pr:185585,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185585; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#153204 mark_unbacked for strides.,cc129d04595a2f1cd30ee40d6caa70b4843d7cf697ed1da522013400ed4ca490
issue-collapse:e3178a15a735cc28,Review thread: 1 issue + 1 PR - pr #185569: Handle symbolic bool short-circuit with unbacked terms,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,153227,185569,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:153227:pr:185569,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153227 eval should handle (unhinted: (s77 > 3) | (u0 > 200)) when s77 has hint =5,ee3231b9db6f42c84455dece95e8f683390be21c543d458b2d2d58d0309133e7
issue-collapse:26e155f2a155faa8,Review thread: 1 issue + 1 PR - pr #185536: Teach Dynamo PGO about sparse tensor strides,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,154161,185536,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:154161:pr:185536,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154161 PGO errors out in dynamo main path for sparse tensors,523facd5a0036a7cd6885e759abe60de885749b66144b5c57435b1d69bba78cd
issue-collapse:552f8973730951fc,Review thread: 1 issue + 1 PR - pr #185526: Fix Dynamo tracing of LazyGraphModule,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,154282,185526,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:154282:pr:185526,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#154282 LazyGraphModule causes graph breaks,05ade3115dcb0a974551dfc0cc23cdc3b7e7ee5fdb10535e6d137fb5b86c2c07
issue-collapse:d860caa6738c7b82,Review thread: 1 issue + 1 PR - pr #185508: [dynamo] Preserve observed exception stack for reraises,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,154454,185508,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:154454:pr:185508,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154454 [Dynamo] Confusing re-raise exception handling graph break message,94b0d94a7a12149a2d39d351384412b904c82b04cde644ca5594348a8c743463
issue-collapse:2e7ab01e5d288cc3,Review thread: 1 issue + 1 PR - pr #185483: Add fake support for pad_packed_sequence,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,155238,185483,,,"module: dynamo, module: pt2-dispatcher, module: rnn, oncall: pt2","module: dynamo, module: pt2-dispatcher, module: rnn, oncall: pt2, triaged",thread:issue:155238:pr:185483,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155238 Dynamo Fails on pad_packed_sequence Due to Fake Tensor Allocation Issue During FX Graph Tracing,535b1f4c209a7c9cbfd6da23fdddf8b7ac3a9572df624e600181b69dac275e50
issue-collapse:c8c9c6a83ce1aed3,Review thread: 1 issue + 1 PR - pr #185474: Fix nested tensor tangent metadata coercion,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,155421,185474,,,"module: nestedtensor, oncall: pt2","module: nestedtensor, oncall: pt2, triaged",thread:issue:155421:pr:185474,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155421 Backward fails with compiled attention on nested tensors on RTX6000,3947becc77433a5964b7e2cfcf77186e137877b803bce18d6522a883da4b5b3a
issue-collapse:f8dd8773e9f1c9ef,Review thread: 1 issue + 1 PR - pr #185471: Handle Dynamo tracing of torch._C._infer_size,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,155800,185471,,,"module: dynamo, oncall: pt2","dynamo-triage-jan2025, module: dynamo, oncall: pt2, triaged",thread:issue:155800:pr:185471,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155800 torch.distributions.kl_divergence Fails with MultivariateNormal in Dynamo Due to _infer_size Type Error,9ceee2feddf4a5ce600871e683e14ebc6373bfc1f140b662ca5ebf5bb577b5bc
issue-collapse:b280bd4600ed5aba,Review thread: 1 issue + 1 PR - pr #185469: Fix stable TMA descriptor reconstruction in Dynamo,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,155856,185469,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:155856:pr:185469,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155856 [user triton][dynamo Fix TMA Descriptor reconstruction,cb5aeb152bf1a71add0ebf175dfd979cfd5a0488506f42eb26758e7d43efbd9c
issue-collapse:137d8517ae490944,Review thread: 1 issue + 1 PR - pr #185467: Keep tensors alive for Dynamo pointer reconstruction,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,156059,185467,,,"module: correctness (silent), module: dynamo, oncall: pt2","module: correctness (silent), module: dynamo, oncall: pt2, triaged",thread:issue:156059:pr:185467,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156059 Dynamo trace an incorrect result on torch._C._storage_Use_Count,09b7b02e412d6a4fd1a6663a72b326b5abdaea40b26a2725fe6b8430ce48b2af
issue-collapse:b564a41eba7effde,Review thread: 1 issue + 1 PR - pr #185466: [dynamo] Show carets in graph break user stacks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,156127,185466,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:156127:pr:185466,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156127 [dynamo] Show carets in graph break stack traces,ad3db52e093e2912e0122f22268ac2ce4adcbf8c81305c6d6ef6b826f85c3bfb
issue-collapse:9e849add2e4ac697,Review thread: 1 issue + 1 PR - pr #185461: [dynamo] Handle module-level logging ignores,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,156191,185461,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:156191:pr:185461,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156191 Dynamo does not know how to trace method `__len__` of class `<unknown type>` with torch.logging calls,236425e1a148ad68d93cfe0cfea0736716154f7e66d4b6a567b0b07f8dc5127d
issue-collapse:9338505df27b03b3,Review thread: 1 issue + 1 PR - pr #185456: Fix Dynamo tracing for Tensor.random_,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,156614,185456,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:156614:pr:185456,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156614 [dynamo] `torch.Tensor.random_` causes graph break,cff4df870aba3251a4fd95fddc86442fbfb1e1025dbef301bffd49503903a691
issue-collapse:48d3255fe6b877a6,Review thread: 1 issue + 1 PR - pr #185451: Wrap internal Dynamo assertion errors,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,156640,185451,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:156640:pr:185451,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156640 [dynamo] Wrap asserts (and other Dynamo errors) more clearly,2928c1ebba075d07aea81192ee619ad196b5cf4106631b9b69c675f469c299ec
issue-collapse:a0eac578b490716c,Review thread: 1 issue + 1 PR - pr #185438: Scope distribution validation disabling to Dynamo tracing,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,157926,185438,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:157926:pr:185438,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#157926 torch.compile() disables torch.distributions parameter validation globally,e26024dce9c8490b9189a7411b0f4d6ef47c0285198b16765595f0079279286e
issue-collapse:b330e1b10a4bec1d,Review thread: 1 issue + 1 PR - pr #185436: Preserve eager exception assertions for eager_noexcept,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,158087,185436,,,"module: dynamo, module: nans and infs, oncall: pt2","module: NaNs and Infs, module: dynamo, oncall: pt2, triaged",thread:issue:158087:pr:185436,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158087 `torch.compile` ignores nan check in `int(torch.tensor(torch.nan))`,908a9c2a6cf0ef137c39e482ad9f95113a0b1188bfb4993fc6a11abdf7a1657f
issue-collapse:32c01bd52678fc18,Review thread: 1 issue + 1 PR - pr #185419: [dynamo] avoid stale verbose hint for backend failures,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,158384,185419,,,"module: compile ux, module: dynamo, oncall: pt2","module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:158384:pr:185419,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158384 [dynamo] full internal stack trace being shown with `torch._dynamo.exc.BackendCompilerFailed` with suggestion to show more,10dcc26dc5d11327e4e6f1e3febe2985f36fe5fb2a0eaa79608a98f447a924e2
issue-collapse:8e3e998c99bc07e2,Review thread: 1 issue + 1 PR - pr #185408: Fix torch logging to avoid closed pytest stderr,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,159160,185408,,,"module: dynamo, oncall: pt2","dynamo-logging, module: dynamo, oncall: pt2, triaged",thread:issue:159160:pr:185408,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#159160 [dynamo, test] Logging error spam at the end of pytest",2113ecc9c7cd1466426bdc8bc006dba67fdfe596496aac139479ae5b2939a4ad
issue-collapse:d5469fbb97126bb2,Review thread: 1 issue + 1 PR - pr #185373: Fix assume_constant_result tensor constants after graph breaks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,159457,185373,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:159457:pr:185373,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#159457 `@assume_constant_result` Causes `NameError` When Method Contains Tensor Operations and Dynamic Attribute Assignment During Compilation,44ce16d53933897c54d30ab39e312fcdb797d94d58c7f596b405a23344f3df93
issue-collapse:10be9b975518e592,Review thread: 1 issue + 1 PR - pr #185365: Fix FX subgraph rewrite of adjacent aggregate matches,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,159613,185365,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",thread:issue:159613:pr:185365,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#159613 [Torch FX] subgraph_rewriter.replace_pattern does not handle back-to-back matches correctly.,3d67ec7989b79962a369bac80c9e1315c82abba07c0caeb4c6756fd70e25a9d1
issue-collapse:40f68142127cdece,Review thread: 1 issue + 1 PR - pr #185357: Forward container methods on OptimizedModule,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,159831,185357,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:159831:pr:185357,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#159831 ModuleDict subscription no longer works after compile().,46f7df22262d23c64881666a127a712aae1dbafd01f444e9640e8dea5d6b3e02
issue-collapse:ab4959db426776c4,Review thread: 1 issue + 1 PR - pr #185336: [dynamo] Add regression test for prepended forward pre-hooks,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,160083,185336,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:160083:pr:185336,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160083 compile regression on forward pre hook with prepend=True,171488f41cc6f91fe96ce39a7892346675bcc4eaa1999df4aa1143696f4ddca5
issue-collapse:47a496bb83605f20,Review thread: 1 issue + 1 PR - pr #185333: Graph break existing list tensor assignments in Dynamo,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,160247,185333,,,"module: dynamo, module: memory usage, oncall: pt2","module: dynamo, module: memory usage, oncall: pt2, triaged",thread:issue:160247:pr:185333,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#160247 wan2.1 vae take more gpu memory after compile,7b1eadb08c7a53a198544e2e07094c8d5616cbb3844d36172bb59f5d097f395a
issue-collapse:11e9d2276b741033,Review thread: 1 issue + 1 PR - pr #185320: Improve scan HOP direct call error,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,160544,185320,,,"module: dynamo, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: dynamo, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:160544:pr:185320,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160544 Error with torch.ops.higher_order.scan + torch.compile [KeyError: combine_fn],6b7f2c6d82f42dd251405ffad57a75fba9fc4830eb9e8f03e064d44e045742c1
issue-collapse:7f817caa8d1de707,Review thread: 1 issue + 1 PR - pr #185299: Run CSE in AOT inference graphs,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,161067,185299,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:161067:pr:185299,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185299; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#161067 torch.compile not eliminating common quantization nodes,bc35ba087dc2a602b491dfdfd2823971dc87cda6521ae32b0993a5e1d178d88d
issue-collapse:e304746fcc5d7053,Review thread: 1 issue + 1 PR - pr #185293: Require bool conditions for torch.where,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,161088,185293,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",thread:issue:161088:pr:185293,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161088 torch.where should throw an exception when invoked with non-bool tensor,a1248a44c4cd0b6a00a41386994b6491c9a1ec64040f0df9db0e23a0f8619725
issue-collapse:0107d43efb95325f,Review thread: 1 issue + 1 PR - pr #185274: Fix ci expected accuracy updates without ClickHouse credentials,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,161289,185274,,,oncall: pt2,"better-engineering, oncall: pt2, triaged",thread:issue:161289:pr:185274,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185274; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#161289 automatically updating ci benchmark expected csv fails,dee1de3291ba379d480506513f206d564951862ca83a3913cc21cb19d1c9fc5d
issue-collapse:d15fdf47b3416e9f,Review thread: 1 issue + 1 PR - pr #185191: Support wrapper subclass storages in HOP alias checks,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,161456,185191,,,"module: dynamo, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: dynamo, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:161456:pr:185191,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161456 [dynamo][hop] speculate_subgraph doesn't work with subclass inputs,afe30ce7ebc15061340693af71b1e8a0e19afaa1475c918b9714ac2c6a0a98f4
issue-collapse:9a932d60c0e11b84,Review thread: 1 issue + 1 PR - pr #185171: Document activation memory budget config,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,161650,185171,,,"module: functorch, oncall: pt2","module: functorch, oncall: pt2, triaged",thread:issue:161650:pr:185171,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185171; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#161650 config.activation_memory_budget,b3566f13142e97efb91c735a1a8f355b611dc75fe414891d912869fa7a5bd58e
issue-collapse:02fedff7c5bd85cb,Review thread: 1 issue + 1 PR - pr #185166: Fix export slice bounds across dynamic dim,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,161705,185166,,,"module: dynamic shapes, oncall: export, oncall: pt2","module: dynamic shapes, oncall: export, oncall: pt2, triaged",thread:issue:161705:pr:185166,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161705 `aten.slice` adds guard on backed symint which causes inconsistent behavior between eager and trace modes,3780c263d10737e060fd1659f59bb51a0697821429beaf39f7b3153fb2b12848
issue-collapse:7ad01a43c88116ba,Review thread: 1 issue + 1 PR - pr #185159: Inductor: gate max-autotune 1x1 conv GEMM lowering,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,161764,185159,,,"module: performance, oncall: pt2","module: performance, oncall: pt2, triaged",thread:issue:161764:pr:185159,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#161764 The ""max-autotune"" mode in torch.compile can lead to longer execution times relative to eager execution.",0392c9a71c588b5d9932ffed4898c1085596a625466c74cbdaaf8169a4b5ad26
issue-collapse:4232a9aefa9df39d,Review thread: 1 issue + 1 PR - pr #185155: Avoid warning spam for backend fallback graph breaks,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,161796,185155,,,"module: compile ux, module: dynamo, module: logging, oncall: pt2","module: compile ux, module: dynamo, module: logging, oncall: pt2, triaged",thread:issue:161796:pr:185155,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161796 [dynamo] Graph break logspam,e49f69d8c1d2bd14f6c60e6fc0368a69d6d908d4eb712ebb6c4dd32e29647a98
issue-collapse:8e01dd0e9f2134b7,Review thread: 1 issue + 1 PR - pr #185147: Fix Z3 translation validation for torch.sym_not,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,162287,185147,,,"module: dynamic shapes, oncall: pt2","module: dynamic shapes, oncall: pt2, triaged",thread:issue:162287:pr:185147,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162287 using torch.sym_not cause Z3Exception: Symbolic expressions cannot be cast to concrete Boolean values.,09731b1f28a37b4d031f0f119835ce0a13744b5dbe15a06049e3d2f52e92e94e
issue-collapse:1577bc45423f2199,Review thread: 1 issue + 1 PR - pr #185138: [dynamo] Enforce one compiled frame for fullgraph skipped frames,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,162415,185138,,,"module: compile ux, module: dynamo, oncall: pt2","dynamo-triage-dec2025, module: compile ux, module: dynamo, oncall: pt2, triaged",thread:issue:162415:pr:185138,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162415 [dynamo] Weird interaction between fullgraph=True and skipped frames,0ccb244d1f1fd41c0ecb5a67ae5153ad3395e571dec307270d054544dde6d0ed
issue-collapse:db57678cf88e94fc,Review thread: 1 issue + 1 PR - pr #185126: Optimize GB_REGISTRY linter scanning,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,162973,185126,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:162973:pr:185126,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162973 [dynamo] GB_REGISTRY linter taking a long time,c068a61987633a6fddfcf02a51342ac8a5107a1828aba172c4f98fe80ccee755
issue-collapse:97e704ecc63e3d52,Review thread: 1 issue + 1 PR - pr #185121: Raise TypeError for invalid torch.export.load inputs,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,163040,185121,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",thread:issue:163040:pr:185121,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163040 [torch.export.load] Passing an `ExportedProgram` raises `AttributeError: 'ExportedProgram' object has no attribute 'tell'` instead of a clear `TypeError`,ef0b829165aadd6439b2fdd05b20362d4c6e72966c556084d6e9529104335a84
issue-collapse:d07ad6067d677f59,Review thread: 1 issue + 1 PR - pr #185103: Fix unbacked scalar equality reasoning through div/mod,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,163641,185103,,,"module: dynamic shapes, oncall: pt2","data dependent error, module: dynamic shapes, oncall: pt2, triaged",thread:issue:163641:pr:185103,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163641 u0==u1 but u0//u1==1 throw DDE,be4834198f11f76d5a01e9c435a2d9db8b5f7d6e5d980caaea47df73b1f0bfe2
issue-collapse:a89b7f9f19adcd9f,Review thread: 1 issue + 1 PR - pr #185088: Avoid graphsafe RNG inputs in descriptor compile,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,164559,185088,,,"module: aotdispatch, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:164559:pr:185088,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#164559 fwd_rng_state show up in the aot_export_joint grpah input,6b797929ee7f6bab10b08bbdc37dbb2270ea7ab15f6e5e1150f7686a1b064a45
issue-collapse:d8036f73fb9f26c7,Review thread: 1 issue + 1 PR - pr #185075: Improve FakeTensorMode deepcopy error for real tensors,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,165547,185075,,,"module: faketensor, module: pt2-dispatcher, module: vllm, oncall: pt2","module: fakeTensor, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:165547:pr:185075,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185075; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#165547 error message for deepcopy a cuda tensor under FakeTensorMode is pretty cryptic,9807c49e76cd95fb1a61679916f57514465ecc114c5b0fd20831e2c49c30ba3e
issue-collapse:21ef8a250f656891,Review thread: 1 issue + 1 PR - pr #185064: Add unroll option to scan,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,166516,185064,,,"module: autograd, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: autograd, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:166516:pr:185064,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185064; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#166516 Performance issue of torch._higher_order_ops.scan,92c3571bf440f80cb80d983288cad2c39fd566ea27e413c0f99f62d82ceebf29
issue-collapse:16d15109916a32ae,Review thread: 1 issue + 1 PR - pr #185063: Handle scalar carries in Inductor while_loop,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,166525,185063,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:166525:pr:185063,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#166525 Inductor Fails with `AttributeError: 'int' object has no attribute 'meta'` for `torch.while_loop` Containing Scalar Values in Loop State,feabca0d33935e310f43a0059a8288d40be96fec99da6ec363a22c8df187707c
issue-collapse:677e2e07dcab3bf8,Review thread: 1 issue + 1 PR - pr #185051: Fix FX GraphModule serialization of string annotations,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,167117,185051,,,oncall: pt2,"oncall: pt2, triaged",thread:issue:167117:pr:185051,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167117 `fx.GraphModule` serialization fails when type annotations are `str`s,f625dbb8e869e00bc75c3cc85c2f6917b4e372c95da176187adbef04d282b3d9
issue-collapse:0885be41f6d4b6d5,Review thread: 1 issue + 1 PR - pr #185042: Avoid value guards for residual lazy constants,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,167596,185042,,,"module: dynamo, oncall: pt2","enhancement, module: dynamo, oncall: pt2, triaged",thread:issue:167596:pr:185042,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#167596 [dynamo][feature] Guard on constants only if graph is specialized and not bytecode,442480fda1cc7f3d7adc9469192456dfb99f66d8f5696c4f3052dd411df1de1c
issue-collapse:4bfb09eb2a60b287,Review thread: 1 issue + 1 PR - pr #185035: Make symbolic formatting guard data-dependent scalars,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,167719,185035,,,"oncall: export, oncall: pt2","export-triage-review, oncall: export, oncall: pt2, triaged",thread:issue:167719:pr:185035,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185035; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#167719 KeyError when exporting PyTorch model with dynamic string keys in ModuleDict using torch.export,be39029a254a61d579c85e3e12fc31aca897be5c1c922dbfdd9556d4db5dc24a
issue-collapse:0ea69020f609cb8f,Review thread: 1 issue + 1 PR - pr #185017: Fix OpOverloadPacket tracing during initialization,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,168136,185017,,,"module: library, module: pt2-dispatcher, oncall: pt2","module: library, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:168136:pr:185017,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #185017; keep related issue/PR context attached.,3,Represents 1 open issue; raw issue rows remain available for audit.,#168136 tracing torch _ops.py like viztracer and cpython settrace will case stuck in an infinite loop.,7cf16afd699a7f75ac9073cd61261dd6039a43f02b0895bce5c8ffd45f1c0524
issue-collapse:0af1397cb9ad4fe2,Review thread: 1 issue + 1 PR - pr #185009: Add fuzzy DebugMode hash mismatch comparison,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,168974,185009,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:168974:pr:185009,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#168974 Fuzzy mode for DebugMode.check_hash_mismatches,bbcc378aaf9a882a073a2af6e2f4338452d602b9c55e28209c2e23861ac34dba
issue-collapse:dd6677a6eb1d87e1,Review thread: 1 issue + 1 PR - pr #185008: Add DebugMode tensor hash export,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,168975,185008,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:168975:pr:185008,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#168975 No way to dump DebugMode hashes to file for further processing / diffing,705ed9cab9e33888e32e3fb470a28428caaf668756a0df3f2a8b7de3e0dd5053
issue-collapse:79ac09aabfe078ea,Review thread: 1 issue + 1 PR - pr #184988: Allow HOPs to pass through unchanged input tensors,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,169769,184988,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:169769:pr:184988,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#169769 pytorch while_loop / cond overly strict about memory aliasing,f7d418c4e854ca46c1d809d48b6a423401b4d0e377aa0bbc3111b0293eae7aa5
issue-collapse:d90725cdc271001c,Review thread: 1 issue + 1 PR - pr #184436: Fix Triton async compile cache keys,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,112788,184436,,,"module: inductor, oncall: pt2","enhancement, low priority, module: inductor, oncall: pt2, triaged",thread:issue:112788:pr:184436,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#112788 async_compile.triton is unable to cache generated triton kernels due to meta containing kernel names,8a8eba1e042ce3c3abf5bb285decf81ab6fd7e3cce4e1750ff54390a27c4fcd0
issue-collapse:a1bbcd49320bb6c9,Review thread: 1 issue + 1 PR - pr #184432: Serialize cudagraph tree runtime,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,113809,184432,,,"module: cuda graphs, module: ddp, module: inductor, oncall: distributed, oncall: pt2","module: cuda graphs, module: ddp, module: inductor, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, triaged",thread:issue:113809:pr:184432,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#113809 Error during DDP, torch.compile, and cudagraph_trees",30a1631eb462116201941df7babbccdf9974e282a931cf1fc51003f44f1f1fa7
issue-collapse:951800d53f7d57c4,Review thread: 1 issue + 1 PR - pr #184380: Avoid fp32 materialization in checkpoint recompute,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,120381,184380,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:120381:pr:184380,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184380; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#120381 torch.compile has significantly higher peak memory for a simple checkpointed layer of matmul + layernorm + gelu,59a4f203f7b6275a90db232d611d37cd246c8ebf843045cda058205394468560
issue-collapse:2b2fd0605b1dac0b,Review thread: 1 issue + 1 PR - pr #184347: [inductor] Add third reduction tile for strided block pointers,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,125077,184347,,,"module: inductor, oncall: pt2","module: inductor, no-scrub, oncall: pt2, triaged",thread:issue:125077:pr:184347,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#125077 [Inductor] Generate triton block pointers for discontiguous strided tensors,5f03046c562b20b84ef85248b44551e32edeca592e40a07847ff12f31c5ff2b9
issue-collapse:2398726eb18dfd1c,Review thread: 1 issue + 1 PR - pr #184345: Use freezing compiler for traced inference graphs,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,125236,184345,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:125236:pr:184345,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184345; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#125236 torch.no_grad() is not working for dynamo inductor backend,63ec7413d54bedf49f46f2143ca47b48a12c16c3d7fe7723096b2e8089c5ddbc
issue-collapse:7e2f4297a0c57755,Review thread: 1 issue + 1 PR - pr #184343: Prune slow Hopper WGMMA autotune configs,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,125277,184343,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:125277:pr:184343,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#125277 H100 Max Autotune Compilation Significantly Slower than A100,82a52a88dc26a560d78e16a88f3edcd231f99d2796903ffcde8cbe57cda01f32
issue-collapse:b1a3c89046e7bb52,Review thread: 1 issue + 1 PR - pr #184323: Share foreach combo kernel bodies,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,127863,184323,,,"module: inductor, module: pt2 optimizer, oncall: pt2","module: inductor, module: pt2 optimizer, oncall: pt2, triaged",thread:issue:127863:pr:184323,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#127863 Potential 7.5x compilation time improvement for foreach kernel,b36f68e87544ff1f64cc65ef5dacafa308a228ef77dc14045fa4897f0e084d73
issue-collapse:f57eddf6393667e6,Review thread: 1 issue + 1 PR - pr #184319: [inductor] Align split reductions for tiled pointwise fusion,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,128063,184319,,,"module: floatx (formerly float8), module: inductor, oncall: pt2","module: floatx (formerly float8), module: inductor, oncall: pt2, triaged",thread:issue:128063:pr:184319,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184319; keep related issue/PR context attached.,6,Represents 1 open issue; raw issue rows remain available for audit.,#128063 [Inductor] Fusion of Tiled Point-Wise and Reduction Operators,40a507b620ca925daaf32b4a33ccee5616b497d33a572f28deeac13732ed2471
issue-collapse:09b510f365b286fb,Review thread: 1 issue + 1 PR - pr #184294: Avoid unused storage size guards for indexing,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,135492,184294,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:135492:pr:184294,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#135492 inductor guards on 32 bit indexing even when generated code would be the same,3feeba58cb7c66dee245eb5d379b216892f4f7d5c532336b8cbb3aae2423596b
issue-collapse:dc6b776200d91353,Review thread: 1 issue + 1 PR - pr #184207: Release max autotune benchmark artifacts,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,140795,184207,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:140795:pr:184207,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140795 Memory usage increase post compilation for torch.compile ViT-H-14-quickgelu_dfn5b,204453acff61881c0daa98566073a70cc45dac3bf0134e6cb1428353a6c1927d
issue-collapse:8b3e25bd435936a7,Review thread: 1 issue + 1 PR - pr #184189: Make generated Triton code use runtime device properties,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,141744,184189,,,module: inductor,"module: inductor, triaged",thread:issue:141744:pr:184189,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,"#141744 ""RuntimeError: Triton Error [CUDA]: device kernel image is invalid"" when running generated Triton code",1a03d1d96d471922674b356457a3793e7de44789c41bbe28be8d757608f8c20a
issue-collapse:5f695b9b9b378af1,Review thread: 1 issue + 1 PR - pr #184119: Avoid fp8 bf16 prologue fusion on pre-sm90,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,150621,184119,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged, upstream triton",thread:issue:150621:pr:184119,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,"#150621 2.8.0 Nightly - ""Feature 'cvt with .bf16.f16' requires .target sm_90 or higher""",6e079e20e3ecd6a337afabbb563ea05e6d4f625d2f25e8c61f5f256ca1314ba0
issue-collapse:819cd7c9591252da,Review thread: 1 issue + 1 PR - pr #184103: Fix FakeTensorUpdater HOP dtype propagation,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,152548,184103,,,"module: inductor, module: vllm, oncall: pt2","inductor_pattern_match, module: inductor, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:152548:pr:184103,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,6,Represents 1 open issue; raw issue rows remain available for audit.,#152548 FakeTensorUpdater does not trace nodes correctly,7182681cc89ade5d5d644a059b12ba333f0620ec5a494226f47b82a374da21ca
issue-collapse:f0418f2f7b5ed900,Review thread: 1 issue + 1 PR - pr #184090: Preserve benchmark integer inputs for Inductor debug,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,154592,184090,,,"module: inductor, oncall: pt2","feature, module: inductor, oncall: pt2, triaged",thread:issue:154592:pr:184090,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#154592 inductor benchmark_compiled_module's random initialization will cause crash for data dependent indexing,a409de3f9f7fc61e227d55c200a024fffc110f9ecd583c074caa6a4975efa7ec
issue-collapse:69d7dfd7b246434b,Review thread: 1 issue + 1 PR - pr #184086: Fix Inductor scaled_mm scale_result lowering,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,154807,184086,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:154807:pr:184086,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#154807 FP8 scaled mm lowering ignores scale_result argument,85f185770a51ac16398efefff3a91fde4bd744f5eb556958996a0564c342832e
issue-collapse:b483ae2bc575526d,Review thread: 1 issue + 1 PR - pr #184083: Preserve user Triton allocator in Inductor,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,155584,184083,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged, upstream triton",thread:issue:155584:pr:184083,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155584 [Upstream Triton] Handle user-specified triton.set_allocator function,090994209392ee3dce0c908a1b85c3dc2680f7f07c8bd9e6f6a1f9f907737354
issue-collapse:4237e5db207ccfa0,Review thread: 1 issue + 1 PR - pr #184065: Fix static launcher 16-bit float scalars,linked work thread,ready for maintainer decision,26,high,work_thread_graph,1,1,0,0,158521,184065,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:158521:pr:184065,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158521 [static cuda launcher] scalar float support,f352b268bee5bd4e81d4104f67bf19adb87fb8a63c222baffbd3769e0d48436b
issue-collapse:e5d762cc464d24c9,Review thread: 1 issue + 1 PR - pr #184061: Fix fused moving avg observer functional meta,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,159192,184061,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:159192:pr:184061,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184061; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#159192 Invalid meta kernel for _fused_moving_avg_obs_fq_helper_functional,d11e3fbc8bf58571d36b1936139da8dd3e23e68f1807f613a2d0e5e098d8968d
issue-collapse:5204d96ae05226c7,Review thread: 1 issue + 1 PR - pr #184052: Preserve CIA ops in AOT joint export,linked work thread,waiting on contributor,26,high,work_thread_graph,1,1,0,0,160388,184052,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:160388:pr:184052,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Wait for contributor update on PR #184052; keep related issue/PR context attached.,5,Represents 1 open issue; raw issue rows remain available for audit.,#160388 Perf differences observed between AOT and JIT on RAFT model,ace032c9858ce772726aa6d5122ccd60deeeafc56ebd7a9d7679856cc9277336
issue-collapse:50e16faeb237c11d,Review thread: 1 issue + 1 PR - pr #184047: Enable cooperative reductions by default,linked work thread,PR blocked,26,high,work_thread_graph,1,1,0,0,161138,184047,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:161138:pr:184047,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,5,Represents 1 open issue; raw issue rows remain available for audit.,#161138 Fuse 3 elementwise ops and reduction into 1 hbm reads,bb2d05a55d73016f95e867b4bef0149936df36315e7d1d7cccfb51253d4b6938
issue-collapse:c75e13c915bc54cc,Review thread: 1 issue + 1 PR - pr #181308: [CPU] Fix std_mean/var_mean all-reduce ~4x slower than std + mean (#122191),linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,122191,181308,,,"module: cpu, module: performance","module: cpu, module: performance, triaged",thread:issue:122191:pr:181308,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,3,Represents 1 open issue; raw issue rows remain available for audit.,#122191 torch.std_mean slower than separate torch.mean and torch.std calls on CPU,924ec212d05028a3b4da0ce9e62f66077754db943dfc2d869c794383b0bda087
issue-collapse:5c7f0061dd4c3351,Review thread: 1 issue + 1 PR - pr #148674: Set /NODEFAULTLIB:vcomp for MSVC when linking caffe2::mkl with libiomp5md.lib,linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,113490,148674,,,"module: build, module: windows","module: build, module: windows, triaged",thread:issue:113490:pr:148674,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#113490 /nodefaultlib:vcomp doesn't seem to be set when compiling with Intel OpenMP on Windows,c19fdec38599538ee4400801bc49f239d31eb5e9be658af796a9dd33d806af4c
issue-collapse:e21fa679556eb937,Review thread: 1 issue + 1 PR - pr #127702: Define PY_SSIZE_T_CLEAN before #include <Python.h>,linked work thread,has linked issue,26,high,work_thread_graph,1,1,0,0,38019,127702,,,"module: cuda, module: nccl","module: cuda, module: nccl, triaged",thread:issue:38019:pr:127702,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#38019 torch.cuda.nccl.init_rank does not handle ""uid"" properly, causing runtime error",541a99716567ca59feb5f80216e1db0bccc2a0f17ebd3ef1096ac05998ada066
issue-collapse:40d92bae380cb21a,Review thread: 4 issues + 1 PR - pr #84843: Support different NSE in batches of CSR and CSC tensors,linked work thread,draft/noise,2,high,work_thread_graph,4,1,0,0,"104193, 105319, 105329, 109946",84843,,,module: sparse,"module: sparse, topic: new features, triaged",thread:issue:104193:issue:105319:issue:105329:issue:109946,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#104193 Conversion from strided to batched sparse compressed tensor with a non-constant number of zeros in batches fails, #105319 Implementation of torch.sparse.sampled_baddmm, #105329 Softmax doesn't support sparse tensors with the CSR layout, #109946 Indexed batch matrix multiplication to support MoEs and FFFs",82b0bc0b1047ff8ce7aea375178e8441b7ae7e1956cf093522d5b1083ca94205
issue-collapse:56b7ebeef0a684d4,Review thread: 1 issue + 1 PR - pr #139524: Make anomaly mode asynchronous,linked work thread,draft/noise,2,high,work_thread_graph,1,1,0,0,143658,139524,,,"module: autograd, module: nans and infs","actionable, module: NaNs and Infs, module: autograd, triaged",thread:issue:143658:pr:139524,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143658 Log more contextual data when nan is detected under the anomaly mode,691aae9f58d65440c6ccda68aa8372001b1df867f765fc6c187c06f9e9567abd
issue-collapse:51a15819b1175f0e,Review thread: 1 issue + 1 PR - pr #124624: [NJT] Allow construction of NJT within graph using offsets from inputs,linked work thread,draft/noise,2,high,work_thread_graph,1,1,0,0,126472,124624,,,module: nestedtensor,"module: nestedtensor, triaged",thread:issue:126472:pr:124624,"Review issues, active PRs, referenced context, checks, and review guidance together instead of reopening each issue separately.",Review active PRs and linked issues as one maintainer work thread.,2,Represents 1 open issue; raw issue rows remain available for audit.,#126472 [NestedTensor] Graph breaks with SDPA + NT constructor,ffdc4105874799b319fa29da362dce922708594d1935b603375f6bdf8d0770e9
issue-collapse:aef56db1e89da69e,Review thread: 2 issues - issue #116254: C++ API `at::quantized_max_pool2d`: Heap-buffer-overflow,related issue thread,needs triage,121,high,work_thread_graph,2,0,0,0,"116254, 162476",,,,"module: crash, oncall: quantization","module: crash, oncall: quantization",thread:issue:116254:issue:162476,"Review this module: crash queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#116254 C++ API `at::quantized_max_pool2d`: Heap-buffer-overflow, #162476 heap-buffer-overflow in torch.quantized_max_pool2d via Python API",c76fbac750606150b7b7c6fb28912e3ef37d3812a2342e82578724512d57d826
issue-collapse:df6b5e3b230c9697,Review thread: 14 issues - issue #154297: Hangs and timeouts on dist.reduce_scatter on B200 GPU,related issue thread,needs maintainer decision,115,high,work_thread_graph,14,0,0,0,"154297, 162178, 162745, 162748, 162820, 162871, 162897, 162917, 162940, 163429, 165170, 165685, 187158, 189065",,,,"module: ci, module: cuda, module: cudacachingallocator, module: data parallel, module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, module: symm_mem, module: tests, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: pt2","bot-triaged, high priority, module: CUDACachingAllocator, module: ci, module: cuda, module: data parallel, module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, module: symm_mem, module: tests, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triage review, triaged",thread:issue:154297:issue:162178:issue:162745:issue:162748,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,6,Represents 14 open issues; raw issue rows remain available for audit.,"#154297 Hangs and timeouts on dist.reduce_scatter on B200 GPU, #162178 [CUDA] Umbrella Issues/Failures on B200 Runner, #162745 [CI][CUDA][Distributed] python test/distributed/test_data_parallel.py TestDataParallel.test_data_parallel_rnn p1.allclose(p2) False, #162748 [CI][CUDA][Distributed] test_3_level_hierarchical_model_averager Works on 4-GPU runner but requires explicit WORLD_SIZE=4 on 8-GPU runner, #162820 [CI][CUDA][Distributed] test_ring_flex_attention failed on 8xB200 Runner, #162871 [CI][CUDA][Distributed] TestDistBackendWithSpawn.test_gather_object Requires Exact World Size of 4",67c8d2cdefac0baf48a192b1e6ab016a3b4ab1dfb7cf14a16199845a29baf1e7
issue-collapse:49f77a88da64130a,Review thread: 5 issues - issue #55655: Performance debugging / warning mode,related issue thread,stale/low urgency,115,high,work_thread_graph,5,0,0,0,"55655, 57118, 68768, 72948, 75725",,,,"module: cpp-extensions, module: ddp, module: logging, module: performance, module: ux, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, high priority, module: cpp-extensions, module: ddp, module: logging, module: performance, module: ux, needs design, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",thread:issue:55655:issue:57118:issue:68768:issue:72948,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#55655 Performance debugging / warning mode, #57118 How to view VLOG information, #68768 [feature request] making pytorch less noisy, #72948 Feature: a consistent Python and C++ logging facility that handles different classes of warnings, #75725 Warning originating in C10 backend does not get translated to Python warning if run from subprocess",9d7152f0a225e76fad61a8e7c435d52630aa594e4fb28208c172b94b14ef2d65
issue-collapse:79e4addcd9a50af1,Review thread: 3 issues - issue #66504: BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation,related issue thread,needs triage,115,high,work_thread_graph,3,0,0,0,"66504, 68407, 73332",,,,"module: ddp, module: docs, oncall: distributed, oncall: distributed parallelisms","better-engineering, bot-triaged, high priority, module: ddp, module: docs, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",thread:issue:66504:issue:68407:issue:73332,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",needs triage,3,Represents 3 open issues; raw issue rows remain available for audit.,"#66504 BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation, #68407 [Runtime Error in ddp_pipeline.py] RuntimeError: unsupported operation: some elements of the input tensor and the written-to tensor refer to a single memory location. Please clone() the tensor before performing the operation., #73332 BatchNorm with LSTM in DistributedDataParallel throws one of the variables needed for gradient computation has been modified by an inplace operation",04bc42bc36bcac179d2a3cceab4c43c94f1b5f595d3cdccca71612677bde4092
issue-collapse:ce4de18f167bde8c,"Review thread: 2 issues - issue #144965: RuntimeError ""global alloc not supported yet"" when using TorchScript optimization.",related issue thread,needs triage,115,high,work_thread_graph,2,0,0,0,"69078, 144965",,,,oncall: jit,"high priority, oncall: jit, triage review",thread:issue:144965:issue:69078,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69078 RuntimeError: Global alloc not supported yet in TorchScript, #144965 RuntimeError ""global alloc not supported yet"" when using TorchScript optimization.",117d560f3db09b295b0eab2207a7989b57d383f53c9d960caa465741c94f98cf
issue-collapse:3fc2676cf3745626,Review thread: 11 issues - issue #103625: DISABLED test_backward_ddp_outside (__main__.TensorPipeDdpUnderDistAutogradTest),related issue thread,needs maintainer decision,97,high,work_thread_graph,11,0,0,0,"15963, 16295, 51062, 58856, 64093, 69178, 69179, 94003, 103625, 104011, 132845",,,,"module: c10d, module: flaky-tests, module: nccl, module: rpc, oncall: distributed, oncall: distributed infra, oncall: jit","bot-triaged, high priority, module: c10d, module: flaky-tests, module: nccl, module: rpc, oncall: distributed, oncall: distributed infra, oncall: jit, ptd-bot-triaged, skipped, triage review, triaged",thread:issue:103625:issue:104011:issue:132845:issue:15963,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,6,Represents 11 open issues; raw issue rows remain available for audit.,"#15963 ProcessGroupGlooTest.test_scatter_stress_cuda is flaky, #16295 ProcessGroupGlooTest.test_gather_stress is flaky, #51062 test_profiler_with_remote_builtin (__main__.TensorPipeRpcTestWithSpawn) is flaky, #58856 NCCL multi-gpu test intermittently failing after NCCL version upgrade, #64093 test_backward_accumulate_grads (__main__.TensorPipeDistAutogradTest) is flaky, #69178 DISABLED test_remote_timeout_to_here_in_jit (__main__.FaultyJitFaultyAgentRpcTest)",89780c3cd54f88732dc99048daf52dcff894b0fb87b3c7952c7eff9e8d93be66
issue-collapse:9e9956fefd557b1d,Review thread: 3 issues - issue #64636: [docs or feature request] torchelastic: OOM recovery / skipping batches (e.g. if inf loss or nan gradients),related issue thread,stale/low urgency,97,high,work_thread_graph,3,0,0,0,"64636, 64905, 65357",,,,"module: ddp, module: docs, oncall: distributed, oncall: r2p","better-engineering, high priority, module: ddp, module: docs, oncall: distributed, oncall: r2p, triage review, triaged",thread:issue:64636:issue:64905:issue:65357,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#64636 [docs or feature request] torchelastic: OOM recovery / skipping batches (e.g. if inf loss or nan gradients), #64905 [docs] [feature request] Tutorial on implementing correctly distributed dataset evaluators, #65357 Improve DDP tutorial",f501f02122acd97340a1cab50f7b96340603e815b9e49ec4bd2484a1bbe48a4f
issue-collapse:25a894ce71ffeb02,Review thread: 3 issues - issue #58111: Static graph training fails if forward is called multiple times before backward,related issue thread,needs triage,97,high,work_thread_graph,3,0,0,0,"58111, 62998, 70865",,,,"module: ddp, module: nn, oncall: distributed","feature, high priority, module: ddp, module: nn, oncall: distributed, triage review, triaged",thread:issue:58111:issue:62998:issue:70865,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#58111 Static graph training fails if forward is called multiple times before backward, #62998 Add nn.Module full pre-backward hooks, #70865 Support hooks-based checkpointing API with DDP",71fce7e59da954f78c290c8dcfadc9b5d90765e0403b269e752e6633829143d3
issue-collapse:5ddffbbd588dd97c,Review thread: 3 issues - issue #42258: Cannot load certain function from dumped Torchscript file,related issue thread,needs triage,97,high,work_thread_graph,3,0,0,0,"42258, 46944, 46946",,,,oncall: jit,"days, high priority, oncall: jit, triage review",thread:issue:42258:issue:46944:issue:46946,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#42258 Cannot load certain function from dumped Torchscript file, #46944 Export TorchScript Classes as TorchScript code, #46946 Changing order of field assignments breaks TorchScript classes",c9a8ed2949b2d72a2e31d738675c02a9c0bc94b4ef5d0ffccccb253cabbe87c1
issue-collapse:b90ff69fcce878d0,Review thread: 3 issues - issue #104506: TImeout in NCCL doesn't work,related issue thread,stale/low urgency,97,high,work_thread_graph,3,0,0,0,"88103, 104506, 124950",,,,"module: c10d, oncall: distributed","high priority, module: c10d, oncall: distributed, triage review, triaged",thread:issue:104506:issue:124950:issue:88103,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#88103 ProcessGroupNCCL watchdog can't catch NCCL comm initialization issues, #104506 TImeout in NCCL doesn't work, #124950 [distributed] First NCCL barrier does not respect timeout",32e9f9314e142db06d03cb19f701a49ffd84465afee44a18a0c00f8a7022d254
issue-collapse:b7e837392441f04c,Review thread: 2 issues - issue #55905: Attempt to use jited `torch.isnan` hit internal assert,related issue thread,needs triage,97,high,work_thread_graph,2,0,0,0,"55905, 56090",,,,"module: ci, module: crash, oncall: jit","NNC, module: ci, module: crash, oncall: jit",thread:issue:55905:issue:56090,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#55905 Attempt to use jited `torch.isnan` hit internal assert, #56090 Converting float16->bool causes an internal compiler error with llvm",5cd5fc6c0c79a4921449ccd58e1253fd41809d6cfc66d7223a5d61d900f88ae7
issue-collapse:d4adb99d49429f32,Review thread: 2 issues - issue #47260: DDP doesn't work with retain_graph = True,related issue thread,needs triage,97,high,work_thread_graph,2,0,0,0,"47260, 80832",,,,"module: ddp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, high priority, module: ddp, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",thread:issue:47260:issue:80832,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47260 DDP doesn't work with retain_graph = True, #80832 [DDP] doesn't support multiple backwards when static_graph=True",33fe371d50245cd59f9ba32c052e544dc6599f5879b1b80dfe835e2afea1208f
issue-collapse:41edcf1648562973,Review thread: 2 issues - issue #39351: Clean up RPC request callback implementation,related issue thread,needs triage,97,high,work_thread_graph,2,0,0,0,"39351, 46564",,,,"module: rpc, oncall: distributed","better-engineering, high priority, module: rpc, oncall: distributed, triage review, triaged",thread:issue:39351:issue:46564,"Review this module: rpc queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#39351 Clean up RPC request callback implementation, #46564 Clean up request_callback_no_python.cpp",dcebc1370d72cfbcf162d7433f9c66536f7252884a81c5e0750baea16b73f7ec
issue-collapse:504e2ffe621ef6f4,Review thread: 2 issues - issue #25888: Forward/backward hooks for C++ torch::nn modules,related issue thread,stale/low urgency,97,high,work_thread_graph,2,0,0,0,"25888, 34329",,,,"module: autograd, module: cpp, oncall: jit","high priority, module: autograd, module: cpp, oncall: jit, triage review, triaged",thread:issue:25888:issue:34329,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#25888 Forward/backward hooks for C++ torch::nn modules, #34329 [jit] Hook support tracking",8f92555ab859f682e7bff3d2cb9f39d0060cf85bfcf9298f503af411b1a52c0f
issue-collapse:2a60eedd50c12cd8,Review thread: 2 issues - issue #128693: Segmentation fault (core dumped) in `torch.fused_moving_avg_obs_fake_quant`,related issue thread,needs triage,97,high,work_thread_graph,2,0,0,0,"128693, 128694",,,,"module: crash, module: edge cases, oncall: quantization","actionable, module: crash, module: edge cases, oncall: quantization, triaged",thread:issue:128693:issue:128694,"Review this module: crash queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#128693 Segmentation fault (core dumped) in `torch.fused_moving_avg_obs_fake_quant`, #128694 Segmentation fault (core dumped) in `torch._fused_moving_avg_obs_fq_helper`",bf35d2606e7e6af95ac5840610f0632e9fd0be79f22e33cd8a25ca8107e9c04c
issue-collapse:569bbc454608948e,Review thread: 2 issues - issue #121219: Torch profiler corrupted names with Python 3.11,related issue thread,needs triage,97,high,work_thread_graph,2,0,0,0,"64345, 121219",,,,oncall: profiler,"high priority, oncall: profiler, triage review",thread:issue:121219:issue:64345,"Review this oncall: profiler queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#64345 Profiler UTF-8 decode issue, #121219 Torch profiler corrupted names with Python 3.11",8896b0b9b2facc1e429b289c294a07118cc04cc7127c681f1eb34cc1a06ddf9c
issue-collapse:5f95dcfb18a256aa,"Review thread: 3 issues - issue #139500: Custom operators registered via decorator slower than ops registered via `torch.Library.{define, impl}`",related issue thread,has linked PR,94,high,work_thread_graph,3,0,0,0,"139500, 139518, 176592",,,,"module: custom-operators, module: performance, module: pt2-dispatcher, module: vllm, oncall: pt2","PT2-Bug-Bash, actionable, better-engineering, bot-triaged, dynamo-triage-jan2025, high priority, module: custom-operators, module: performance, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged, vllm-compile",thread:issue:139500:issue:139518:issue:176592,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,3,Represents 3 open issues; raw issue rows remain available for audit.,"#139500 Custom operators registered via decorator slower than ops registered via `torch.Library.{define, impl}`, #139518 OpOverloads are slow?, #176592 custom_op : custom_op overhead is much higher than autograd.Function (in eager)",864cd95a10cef4ee64891b09ccf4da6240c01fd6335315605ec0ab0f130f8acb
issue-collapse:8be11ae84ad5c4f6,Review thread: 2 issues - issue #186241: TMA descriptor loads use stale `xoffset` in mix-order reduction loop,related issue thread,has linked PR,94,high,work_thread_graph,2,0,0,0,"186241, 187175",,,,"module: correctness (silent), module: crash, module: cuda, module: inductor, module: performance, module: reductions, oncall: pt2","bot-triaged, feature, has workaround, high priority, module: correctness (silent), module: crash, module: cuda, module: inductor, module: performance, module: reductions, oncall: pt2, triaged",thread:issue:186241:issue:187175,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#186241 TMA descriptor loads use stale `xoffset` in mix-order reduction loop, #187175 [feature request] Autotune TMA and non-TMA path for pointwise and reduction kernels",d7d7ba41b66f93602f4df25bc7a8cbffcb9b7305a3a423656d3207ff7a8d9b3f
issue-collapse:acbcdcb5499c7942,Review thread: 2 issues - issue #120597: floating point exception on `torch.fmod` when x is large negative integer and y is `-1`,related issue thread,high-priority/blocker,94,high,work_thread_graph,2,0,0,0,"120597, 121343",,,,module: crash,"actionable, module: crash, triaged",thread:issue:120597:issue:121343,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#120597 floating point exception on `torch.fmod` when x is large negative integer and y is `-1`, #121343 Floating point exception on torch.lcm",65d1e9fba78d5478f7e21df7f7f0b0d0922191975a5168376dab3228264ee55e
issue-collapse:9f48eb6dbd7f0fb4,Review thread: 2 issues - issue #117826: [MPS] Probable incorrect gradient for einsum,related issue thread,stale/low urgency,94,high,work_thread_graph,2,0,0,0,"117826, 177116",,,,"module: autograd, module: mps","bot-triaged, high priority, module: autograd, module: mps, triaged",thread:issue:117826:issue:177116,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#117826 [MPS] Probable incorrect gradient for einsum, #177116 MPS: catastrophically wrong gradients in backward pass (>32K elements)",31a68e5035499b648179ca629c233154c001562cc6d40774a1427fc75884e1a1
issue-collapse:ed06571d61ae7950,"Review thread: 2 issues - issue #162700: RuntimeError: The tensor has a non-zero number of elements, but its data is not allocated yet during torch.backward with InternVL",related issue thread,stale/low urgency,89,high,work_thread_graph,2,0,0,0,"162700, 184996",,,,"oncall: export, oncall: pt2","bot-triaged, oncall: export, oncall: pt2, triaged",thread:issue:162700:issue:184996,Handle this oncall: export stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#162700 RuntimeError: The tensor has a non-zero number of elements, but its data is not allocated yet during torch.backward with InternVL, #184996 torch.export.export accepts a parameter whose storage is too small, then save/runtime fails with low-level setStorage error",2d4c93a400d91866b393e0de4cc3969dae9a8f763c3d60bc3f81785b564f5e75
issue-collapse:cb25f57feed15520,Review thread: 2 issues - issue #171804: [On Demand Profiling] Early CUDA initialization caused by `import torch` when `KINETO_USE_DAEMON=1`,related issue thread,needs maintainer decision,77,high,work_thread_graph,2,0,0,0,"171804, 172500",,,,"module: cuda, oncall: distributed, oncall: profiler","module: cuda, needs reproduction, oncall: distributed, oncall: profiler, triaged",thread:issue:171804:issue:172500,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#171804 [On Demand Profiling] Early CUDA initialization caused by `import torch` when `KINETO_USE_DAEMON=1`, #172500 `RuntimeError: device >= 0 && device < num_gpus INTERNAL ASSERT FAILED` caused by early CUDA init",b19fb897d1d66f2ebfb7feae7160c84202dbc98176d4d55fd125d734673ad9a1
issue-collapse:212774123f95754e,Review thread: 2 issues - issue #123800: add `FlopCounterMode` documentation,related issue thread,has linked PR,76,high,work_thread_graph,2,0,0,0,"123800, 145555",,,,module: docs,"high priority, medium, module: docs, triaged",thread:issue:123800:issue:145555,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#123800 add `FlopCounterMode` documentation, #145555 need to document `FlopCounterMode`",da196277770b78e1089b3bce8c31d7f2e7f656518452ca3940a5271fda723a75
issue-collapse:97243baa392f2e6f,Review thread: 2 issues - issue #159124: AOTInductor preallocate output tensors,related issue thread,stale/low urgency,71,high,work_thread_graph,2,0,0,0,"159124, 165444",,,,"module: aotinductor, oncall: export, oncall: pt2","module: aotinductor, oncall: export, oncall: pt2, triaged",thread:issue:159124:issue:165444,Handle this module: aotinductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#159124 AOTInductor preallocate output tensors, #165444 AOTInductor not updating buffers inplace",1caee34e648d74033552d672e63f81bea447e50b1d5cd8bc9616935e913dae57
issue-collapse:4733b332e6b5903e,"Review thread: 6 issues - issue #152595: ROCm, 7900 XTX (gfx1100): Pytorch FLASH_ATTENTION SDPA is 2.5x slower than MATH (fp16, head_dim 256, seqlen 4360, 12 heads)",related issue thread,stale/low urgency,70,high,work_thread_graph,6,0,0,0,"152595, 165040, 165141, 169857, 177834, 188564",,,,"module: ci, module: compile-time, module: convolution, module: correctness (silent), module: crash, module: inductor, module: performance, module: rnn, module: rocm, module: sdpa, oncall: pt2","bot-triaged, module: ci, module: compile-time, module: convolution, module: correctness (silent), module: crash, module: inductor, module: performance, module: rnn, module: rocm, module: sdpa, oncall: pt2, triaged",thread:issue:152595:issue:165040:issue:165141:issue:169857,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 6 open issues; raw issue rows remain available for audit.,"#152595 ROCm, 7900 XTX (gfx1100): Pytorch FLASH_ATTENTION SDPA is 2.5x slower than MATH (fp16, head_dim 256, seqlen 4360, 12 heads), #165040 DISABLED rocm / linux-jammy-rocm-py3_10-gfx1100 / test (default), #165141 [ROCm][RDNA3/GFX1100] MIOpen Conv3d catastrophically fails on large batches with bfloat16, causing vLLM to extremely slowdown, #169857 [ROCm] MIOpen 3D convolution kernel fails with ""invalid configuration argument"" for Qwen2.5-VL vision encoder - silent fallback causes severe performance regression, #177834 `torch.nn.LSTM.forward` produces incorrect outputs (or crashes) on ROCm, #188564 rocm-navi31 gfx1100 test (default) jobs time out after 270 min on inductor/test_torchinductor (red since 2026-05-20)",a6c6aa61faed5e5c2dd274a525d6fedd61be34dbad20b861b5a52dbc30ae5bf0
issue-collapse:da44e48913b7d5ef,Review thread: 5 issues - issue #85329: Crash in `torch.package.PackageExporter`,related issue thread,high-priority/blocker,70,high,work_thread_graph,5,0,0,0,"85329, 87960, 87961, 88148, 94594",,,,"module: crash, module: edge cases, oncall: package/deploy","imported, module: crash, module: edge cases, oncall: package/deploy, triaged",thread:issue:85329:issue:87960:issue:87961:issue:88148,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,4,Represents 5 open issues; raw issue rows remain available for audit.,"#85329 Crash in `torch.package.PackageExporter`, #87960 crash when call `torch.set_num_interop_threads` twice, #87961 crash in `torch.package.PackageExporter`, #88148 ☂️ Issues that trigger crashes due to corner-case API usages, #94594 A segment fault can be triggered in torch.histogramdd",16b1aaf100c69f35ed5329c77379bd51eaa4dfa80e807741b325a1d5485e9636
issue-collapse:6f7b9b98eef79b27,Review thread: 4 issues - issue #188420: torch.compile produces inconsistent outputs for MultiHeadAttentionModel with scaled_dot_product_attention,related issue thread,needs maintainer decision,70,high,work_thread_graph,4,0,0,0,"188420, 188425, 188430, 188432",,,,"module: correctness (silent), module: determinism, module: inductor, module: random, module: sdpa, oncall: pt2","bot-triaged, high priority, module: correctness (silent), module: determinism, module: inductor, module: random, module: sdpa, oncall: pt2, topic: fuzzer, triaged",thread:issue:188420:issue:188425:issue:188430:issue:188432,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#188420 torch.compile produces inconsistent outputs for MultiHeadAttentionModel with scaled_dot_product_attention, #188425 torch.compile produces drastically incorrect outputs for model with symbolic dimensions and random operations, #188430 torch.compile produces large absolute errors for model with LSTM, MultiheadAttention, and torch.reciprocal, #188432 torch.compile produces inconsistent outputs for Multi-Query Attention model with manually implemented attention",dc7a4e5d382fa7af7ca661081b7a10d8d3e8495ff5965165a239f017d6554769
issue-collapse:01947722bed63d61,"Review thread: 4 issues - issue #182263: libtorch_cpu.so BLAS symbols are interposable, causing crashes when another BLAS is present in the process",related issue thread,high-priority/blocker,70,high,work_thread_graph,4,0,0,0,"78475, 78489, 82098, 182263",,,,"module: binaries, module: build, module: collect_env.py, module: crash, module: cudnn, module: determinism, module: openblas","bot-triaged, enhancement, has workaround, module: binaries, module: build, module: collect_env.py, module: crash, module: cudnn, module: determinism, module: openblas, triaged",thread:issue:182263:issue:78475:issue:78489:issue:82098,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,3,Represents 4 open issues; raw issue rows remain available for audit.,"#78475 `with torch.backends.cudnn.flags(deterministic=True)` doesn't give an exception for ctc_loss backward on CUDA, #78489 [1.9.1] [collect_env] collect_env does not collect actual runtime-loaded cudnn version, #82098 [feature request] Discover actually loaded shared libraries at runtime, #182263 libtorch_cpu.so BLAS symbols are interposable, causing crashes when another BLAS is present in the process",d03b87ae575239b7ad384ca3f52ae82ee4c26699b83dafec4875f653e5735667
issue-collapse:e5cd7698b3c89475,Review thread: 4 issues - issue #112583: Nesting no_grad in autocast causes backwards graph to be (partially) lost outside of no_grad,related issue thread,high-priority/blocker,70,high,work_thread_graph,4,0,0,0,"65766, 112583, 141896, 169934",,,,"module: activation checkpointing, module: amp (automated mixed precision), module: autograd","actionable, high priority, module: activation checkpointing, module: amp (automated mixed precision), module: autograd, triaged",thread:issue:112583:issue:141896:issue:169934:issue:65766,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,3,Represents 4 open issues; raw issue rows remain available for audit.,"#65766 Temporarily disabling grad doesn't work if AMP is enabled, #112583 Nesting no_grad in autocast causes backwards graph to be (partially) lost outside of no_grad, #141896 CheckpointError with checkpoint(..., use_reentrant=False) & autocast(), #169934 Autocast cache not updated by grad enabled state",5c70fc33938d1307645924126eff27e007009aac1953d65ca0796255093a362d
issue-collapse:0d536019636cf761,"Review thread: 3 issues - issue #93880: `PYTORCH_DEBUG_MODE`, better invalid index embedding lookup error message on cuda",related issue thread,high-priority/blocker,70,high,work_thread_graph,3,0,0,0,"93880, 99640, 99918",,,,"module: amp (automated mixed precision), module: python frontend","high priority, module: amp (automated mixed precision), module: python frontend, needs design, triaged",thread:issue:93880:issue:99640:issue:99918,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,2,Represents 3 open issues; raw issue rows remain available for audit.,"#93880 `PYTORCH_DEBUG_MODE`, better invalid index embedding lookup error message on cuda, #99640 Ban GradScaler scale from being less than 1, #99918 [RFC] DebugMode",1ee0ec0bdcaf45209ad556ab2225ef02c2660b95cce00b4979e7ceda632cde76
issue-collapse:6988e3117c3803e6,Review thread: 2 issues - issue #91570: Security policy impractical / lacks contact information?,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"91570, 92226",,,,"module: binaries, module: docs","high priority, module: binaries, module: docs, security, triaged",thread:issue:91570:issue:92226,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#91570 Security policy impractical / lacks contact information?, #92226 Hijacked package names from nightly repository",1f871631826e47f7d70f52675cdaee7da93ed2086c563d84e4b71ead2d1cdf3e
issue-collapse:83eaa191c4984d8f,Review thread: 2 issues - issue #76865: Object-base collectives create tensors at unexpected devices,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"76865, 77200",,,,"module: c10d, module: pickle, module: serialization, oncall: distributed, oncall: distributed infra","better-engineering, bot-triaged, enhancement, high priority, module: c10d, module: pickle, module: serialization, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:76865:issue:77200,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76865 Object-base collectives create tensors at unexpected devices, #77200 [RFC] Allow device override during Tensor unpickling without torch.load",1ac0390f72929c9f4d39b62a403efcd7bad6ca4c50adbd3ba39f543c6f824273
issue-collapse:7efa498af81ec586,"Review thread: 2 issues - issue #47993: build pytorch from source on ubuntu, building error from fbgemm::SparseAdaGradSignature",related issue thread,stale/low urgency,70,high,work_thread_graph,2,0,0,0,"47993, 51039",,,,"module: abi, module: binaries, module: build, module: collect_env.py, module: cpp","high priority, module: abi, module: binaries, module: build, module: collect_env.py, module: cpp, needs design, triaged",thread:issue:47993:issue:51039,Handle this module: abi stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47993 build pytorch from source on ubuntu, building error from fbgemm::SparseAdaGradSignature, #51039 Status of pip wheels with _GLIBCXX_USE_CXX11_ABI=1",fdc6843927ad74ad8db88b7761fb41460ed4e2dcfd386eff7e5e6b155fd0ea4e
issue-collapse:7418ff8676ae27c2,"Review thread: 2 issues - issue #41486: LSTMs leak memory in CPU PyTorch 1.5.1, 1.6, and 1.7 on Linux",related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"41486, 42815",,,,"module: cpu, module: docs, module: memory usage, module: rnn","high priority, module: cpu, module: docs, module: memory usage, module: rnn, triaged",thread:issue:41486:issue:42815,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#41486 LSTMs leak memory in CPU PyTorch 1.5.1, 1.6, and 1.7 on Linux, #42815 Add a guide for what to do when you think there's a memory leak",963d52da859d06a145a9680e3787e675f276febe81e5121ba76aee568368450f
issue-collapse:050251ff04485028,Review thread: 2 issues - issue #30774: Natively Declarable Fast-path Functions,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"30774, 49372",,,,"module: dispatch, module: internals, oncall: jit","PT2-Bug-Bash, better-engineering, high priority, module: dispatch, module: internals, oncall: jit, triaged",thread:issue:30774:issue:49372,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30774 Natively Declarable Fast-path Functions, #49372 [jit] Support `torch.Tensor.numel()` and `torch.Size.numel()` properly in tracing",519054238eef80043ee2fdef61db25fd65182e3a300b0801745507c8b4809e24
issue-collapse:07f3e61eaf2d57df,Review thread: 2 issues - issue #188733: [vllm] [2.12.1 regression][Spec Decode] CUDA WARP_ILLEGAL_ADDRESS in FlashAttnFwdSm90 during CUDA graph capture (EAGLE),related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"45111, 188733",,,,"module: crash, module: cuda, module: cuda graphs, module: regression, module: sdpa, module: vllm, oncall: mobile","bot-triaged, module: crash, module: cuda, module: cuda graphs, module: regression, module: sdpa, module: vllm, oncall: mobile, triaged",thread:issue:188733:issue:45111,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#45111 (prototype) Graph Mode Dynamic Quantization on BERT failure on quantize_dynamic_jit(...) call, #188733 [vllm] [2.12.1 regression][Spec Decode] CUDA WARP_ILLEGAL_ADDRESS in FlashAttnFwdSm90 during CUDA graph capture (EAGLE)",ee375ac1951cef0323b65defe16031cb3080c84086487411f9e8daa8ca652ced
issue-collapse:90d0727d42c6b1a7,Review thread: 2 issues - issue #188557: operator_microbenchmark ROCm gfx950 job never completes -- runs 8h+ and is cancelled at the workflow timeout,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"188557, 189077",,,,"module: benchmark, module: ci, module: compile-time, module: regression, module: rocm, oncall: pt2","bot-triaged, module: benchmark, module: ci, module: compile-time, module: regression, module: rocm, oncall: pt2, op-bench, triaged",thread:issue:188557:issue:189077,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#188557 operator_microbenchmark ROCm gfx950 job never completes -- runs 8h+ and is cancelled at the workflow timeout, #189077 operator_microbenchmark + attention_op_microbenchmark: b200 jobs never complete -- run ~8h and are cancelled at the workflow timeout",d38ff91185c4de9e6d686b777188c1a15d17626172ba7be215509465e0a6a8fc
issue-collapse:4ebb6026ebab9c8b,"Review thread: 2 issues - issue #179891: torch.xpu.get_device_properties() segfaults on Intel ARC PRO B70 (Battlemage/BMG, device 0xe223)",related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"179891, 188953",,,,"module: crash, module: dependency bug, module: linear algebra, module: xpu","bot-triaged, has workaround, high priority, module: crash, module: dependency bug, module: linear algebra, module: xpu, triaged",thread:issue:179891:issue:188953,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#179891 torch.xpu.get_device_properties() segfaults on Intel ARC PRO B70 (Battlemage/BMG, device 0xe223), #188953 Intel Arc B580 (Battlemage): ""could not create an engine"" on XPU — torch 2.6.0+xpu",9a329df4939edaa42074763b2112ae021db328e2205433e6d82132d7523d96c0
issue-collapse:8349cf9c6b864346,Review thread: 2 issues - issue #152954: DTensor placement propagation for `as_strided` missing,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"152954, 152963",,,,"module: pt2-dispatcher, oncall: distributed, oncall: pt2","high priority, module: pt2-dispatcher, oncall: distributed, oncall: pt2, triaged",thread:issue:152954:issue:152963,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#152954 DTensor placement propagation for `as_strided` missing, #152963 DTensor support for dynamic shapes is soft",5f79aad608de59ff7ed7cecebe566132a1fa7705bea06b4526096f62acdf7782
issue-collapse:4c980eb73739181e,Review thread: 2 issues - issue #145702: PyTorch VS2022 official build Windows binary illegal instruction on AVX2(max ISA level) CPU,related issue thread,high-priority/blocker,70,high,work_thread_graph,2,0,0,0,"145702, 153961",,,,"module: build, module: cpu, module: crash, module: windows","low priority, module: build, module: cpu, module: crash, module: windows, triaged",thread:issue:145702:issue:153961,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#145702 PyTorch VS2022 official build Windows binary illegal instruction on AVX2(max ISA level) CPU, #153961 [Windows][CPU] Root cause the VS2022 illegal instruction",2985fdcc3f2d004f8f47ea87e35f690facbaeb82210ce3fc2172ef9f05a4f11e
issue-collapse:460f746efdcf243d,Review thread: 2 issues - issue #142515: [MPS] `out_channels` <= 2**16 guard for convolution is too broad,related issue thread,stale/low urgency,70,high,work_thread_graph,2,0,0,0,"142515, 142836",,,,"module: correctness (silent), module: error checking, module: mps, module: regression","high priority, module: correctness (silent), module: error checking, module: mps, module: regression, triaged",thread:issue:142515:issue:142836,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#142515 [MPS] `out_channels` <= 2**16 guard for convolution is too broad, #142836 [MPS] Incorrect output from convolution ops with large dimensions",e3e7dc07edb1d0052199cf49bc57d8bc04aae3501808ffd3d814f9475a773230
issue-collapse:26c9033634b1d67d,Review thread: 10 issues - issue #50444: [TorchScript] Rough edges discovered during type system rewrite,related issue thread,needs triage,65,high,work_thread_graph,10,0,0,0,"50444, 51135, 51136, 51137, 51138, 51140, 51563, 52920, 53017, 53022",,,,oncall: jit,"OSS contribution wanted, TSRootCause:InvalidCustomClass, TSRootCause:PyTorchParityGap, TSUsability, good first issue, oncall: jit",thread:issue:50444:issue:51135:issue:51136:issue:51137,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,6,Represents 10 open issues; raw issue rows remain available for audit.,"#50444 [TorchScript] Rough edges discovered during type system rewrite, #51135 [TorchScript Usability] Behaviors of used-defined subclasses of torch.tensor are silently ignored, #51136 [TorchScript Usability] error message when invoking __init__ of module types inside TorchScript, #51137 [TorchScript Usability] Earlier type checking needed for interface type, #51138 [TorchScript Usability] tensor printout format differs from eager mode, #51140 [TorchScript Usability] Non-intuitive error messages for user errors",23e631dc2cf1e89e897c654918d54fbcb23924071264242307877f83d6ec6a9e
issue-collapse:6bdf883f99be47be,Review thread: 10 issues - issue #147179: [FSDP2] OOM when use integer `reshard_after_forward` that smaller than DP size,related issue thread,stale/low urgency,65,high,work_thread_graph,10,0,0,0,"147179, 153302, 158793, 159013, 159014, 159015, 159017, 159018, 159019, 177815",,,,"module: c10d, module: devicemesh, module: fsdp, oncall: distributed, oncall: distributed infra","bot-triaged, distributed-backlog, feature, module: DeviceMesh, module: c10d, module: fsdp, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:147179:issue:153302:issue:158793:issue:159013,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 10 open issues; raw issue rows remain available for audit.,"#147179 [FSDP2] OOM when use integer `reshard_after_forward` that smaller than DP size, #153302 [FSDP2] avoid GPU OOM for reshard_after_forward=int with shared post_forward_mesh, #158793 [RFC]: Iterations of DeviceMesh based on recent user requests/feedback, #159013 [RFC] Support slicing of submesh and more flexible operations to one device mesh such as reshape/split, etc., #159014 [RFC] Make it easy to subclass DeviceMesh and also working in multi-threading setup, #159015 [RFC] Support Fake PG mode for the device mesh and provide an option to reuse PG under the hood",5d6ef3dd458a30afb3f0ffe5a0b9c64c3ade9a9940eed6f82feee4f47c7c0348
issue-collapse:5219a03342c412f2,Review thread: 8 issues - issue #146326: Tracking issue for Improvements for DeepSeek,related issue thread,needs maintainer decision,65,high,work_thread_graph,8,0,0,0,"146326, 146328, 146329, 146330, 146331, 146332, 146368, 146542",,,,"module: floatx (formerly float8), module: linear algebra, module: nestedtensor, module: optimizer, module: pipelining, module: sdpa, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: quantization","bot-triaged, feature, module: floatx (formerly float8), module: linear algebra, module: nestedtensor, module: optimizer, module: pipelining, module: sdpa, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: quantization, ptd-bot-triaged, triage review, triaged",thread:issue:146326:issue:146328:issue:146329:issue:146330,"Review this module: floatx (formerly float8) queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,6,Represents 8 open issues; raw issue rows remain available for audit.,"#146326 Tracking issue for Improvements for DeepSeek, #146328 DeepSeek: Grouped/Ragged gemm with offsets on the GPU, #146329 DeepSeek: a2a communication with metadata on the GPU, #146330 DeepSeek: MLA attention, #146331 DeepSeek: hierarchical a2a, #146332 DeepSeek: fine-grained overlap",1b549f9a879210c03d1ba3629947c3c01fb7f7d0ef5d26c759aa98d399653c29
issue-collapse:ca683c90aa05ed72,Review thread: 7 issues - issue #113045: [Feature][DTensor] Manage additional `_padded_local_tensor` attribute,related issue thread,needs maintainer decision,65,high,work_thread_graph,7,0,0,0,"113045, 114299, 120003, 147662, 158657, 174960, 187615",,,,"module: activation checkpointing, module: autograd, module: dtensor, module: fsdp, module: serialization, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms","bot-triaged, feature, module: activation checkpointing, module: autograd, module: dtensor, module: fsdp, module: serialization, needs design, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:113045:issue:114299:issue:120003:issue:147662,"Review this module: activation checkpointing queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,6,Represents 7 open issues; raw issue rows remain available for audit.,"#113045 [Feature][DTensor] Manage additional `_padded_local_tensor` attribute, #114299 [RFC] Per-Parameter-Sharding FSDP, #120003 [FSDP2] Eager-Mode Execution Tracker, #147662 GPUDirect Storage dcp.save/dcp.load torch.save/torch.load, #158657 extend torch.util.checkpoint to support offload to cpu/nvme, #174960 Feature Request: Activation offloading with async prefetch in FSDP",17db23951c92f27bebb7ec410eab56977dc6a46d6f0495b00a3e38a4107257a1
issue-collapse:110d60fd6daf8976,"Review thread: 5 issues - issue #44511: LR scheduler throws warning when using scaler.step instead of optimizer.step, and when saving optimizer state",related issue thread,stale/low urgency,65,high,work_thread_graph,5,0,0,0,"44511, 55585, 67589, 67590, 76282",,,,"module: amp (automated mixed precision), module: optimizer, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: amp (automated mixed precision), module: optimizer, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",thread:issue:44511:issue:55585:issue:67589:issue:67590,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 5 open issues; raw issue rows remain available for audit.,"#44511 LR scheduler throws warning when using scaler.step instead of optimizer.step, and when saving optimizer state, #55585 [amp scaler] unable to prevent ""scheduler before optimizer step"" warning, #67589 Expose the `OptState` in `torch.cuda.amp.GradScaler`, #67590 Add `OptState.UPDATED` to `torch.cuda.amp.GradScaler`, #76282 `torch.cuda.amp.GradScaler` may skip parameter synchronization required by post localSGD optimizer",1dd80c86bb665f8bf86e468ed14ba23bad613ec7ff95253ddad45d0a811fac03
issue-collapse:6a430e2a5b1e457a,Review thread: 4 issues - issue #174469: [RFC] Overview of Test Case Refactoring in PyTorch,related issue thread,needs maintainer decision,65,high,work_thread_graph,4,0,0,0,"174469, 176877, 185142, 185590",,,,"module: backend, module: ci, module: openreg, module: privateuse1, module: tests, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, feature, large, module: PrivateUse1, module: backend, module: ci, module: openreg, module: tests, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:174469:issue:176877:issue:185142:issue:185590,"Review this module: backend queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#174469 [RFC] Overview of Test Case Refactoring in PyTorch, #176877 [RFC] Adding Distributed Support to OpenReg, #185142 RFC Test class classification, #185590 [Test] PyTorch Test Case Refactoring and Track",0b4d51d6d9083e89c640b694db1cca89ac7a8d82332746119be44a7d2bc95700
issue-collapse:126e0af7ee46e027,"Review thread: 4 issues - issue #166291: ""FSDP does not support non-contiguous parameters yet"" prevents using channels-last convolutions",related issue thread,needs maintainer decision,65,high,work_thread_graph,4,0,0,0,"166291, 177427, 186601, 187239",,,,"module: c10d, module: fsdp, module: nccl, module: performance, module: pipelining, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: c10d, module: fsdp, module: nccl, module: performance, module: pipelining, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:166291:issue:177427:issue:186601:issue:187239,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#166291 ""FSDP does not support non-contiguous parameters yet"" prevents using channels-last convolutions, #177427 [RFC] Communication primitives for non-contiguous tensors, #186601 [RFC]SDMA AllGather and Zero-Copy Output for PyTorch FSDP2, #187239 [NCCL][FSDP2] Support parameter-contiguous all-gather output to skip copy-out",03ff0550b4695d0d07704e550a650d1e3b90e7bfd824c5bf572194cf35ca3a7e
issue-collapse:3e94e59318f5868d,Review thread: 4 issues - issue #107298: dist.destroy_process_group did not destroy the process group well,related issue thread,stale/low urgency,65,high,work_thread_graph,4,0,0,0,"75097, 107298, 115388, 119215",,,,"module: c10d, module: cuda graphs, oncall: distributed, oncall: distributed infra","PT2-Bug-Bash, bot-triaged, module: c10d, module: cuda graphs, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",thread:issue:107298:issue:115388:issue:119215:issue:75097,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#75097 destroy_process_group() has a certain probability of hangs, #107298 dist.destroy_process_group did not destroy the process group well, #115388 `torch.distributed.destroy_process_group()` hangs after CUDA graph capture of NCCL operations, #119215 [c10d] Workaround the ncclCommAbort() hang issues",357d60a43b299470b748bce64544bce130603c27637dc78de64e6b1ca1674462
issue-collapse:42fbb16c29891610,Review thread: 3 issues - issue #180284: Unused jvp function on autograd function causes inscrutable shape error,related issue thread,needs maintainer decision,65,high,work_thread_graph,3,0,0,0,"180284, 182641, 182661",,,,"module: compile ux, module: compiled autograd, module: ddp, module: fsdp, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, dynamo-autograd-function, has workaround, module: compile ux, module: compiled autograd, module: ddp, module: fsdp, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triaged",thread:issue:180284:issue:182641:issue:182661,"Review this module: compile ux queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#180284 Unused jvp function on autograd function causes inscrutable shape error, #182641 [v2.11.0] Cannot jvp through FSDP2, #182661 [v2.11.0] Cannot jvp through DDP",5a10149d7d399aeaa1b76e228cb192a00252f2e8123f3c8e2c216c114801d557
issue-collapse:9889b0232b5a3ccf,Review thread: 3 issues - issue #172568: Control flow operators issue tracker,related issue thread,needs triage,65,high,work_thread_graph,3,0,0,0,"172568, 186107, 186465",,,,"module: dynamo, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: export, oncall: pt2","bot-triaged, enhancement, large, module: dynamo, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:172568:issue:186107:issue:186465,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#172568 Control flow operators issue tracker, #186107 torch.scan gaps, #186465 [RFC] Fix Control flow operators consistency",9fde2a46ecc6b8dcb3e7354c712bcfa641fa11f8a1d5b68710012aa0a4d034dc
issue-collapse:2c5bb0c1b128f3a7,Review thread: 3 issues - issue #165230: _int_mm is slow with row-major rhs matrix,related issue thread,needs triage,65,high,work_thread_graph,3,0,0,0,"165230, 165231, 187094",,,,"module: cuda, module: floatx (formerly float8), module: linear algebra, oncall: quantization","bot-triaged, matrix multiplication, module: cuda, module: floatx (formerly float8), module: linear algebra, oncall: quantization, triaged",thread:issue:165230:issue:165231:issue:187094,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#165230 _int_mm is slow with row-major rhs matrix, #165231 _scaled_mm errors with row-major rhs matrix, #187094 torch._int_mm dispatches Ampere-era cutlass_80 kernels on sm_103 (GB300): 6.4-13.4x SLOWER than the fp32 tf32 matmul of the same shape",9d7e5512ee2405c0b1038b254f14762774902abb5961ac2c4bfd4cb9f224a11a
issue-collapse:c982b844ae3248be,Review thread: 3 issues - issue #145949: [CUDA][Blackwell] Blackwell Tracking Issue,related issue thread,stale/low urgency,65,high,work_thread_graph,3,0,0,0,"145949, 147692, 165727",,,,"module: binaries, module: build, module: cuda, module: ddp, module: nccl, oncall: distributed, oncall: distributed infra","bot-triaged, module: binaries, module: build, module: cuda, module: ddp, module: nccl, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",thread:issue:145949:issue:147692:issue:165727,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#145949 [CUDA][Blackwell] Blackwell Tracking Issue, #147692 Please generate a new version of pytorch into blackwell rtx5080 with cuda 12.8. Thank you very much., #165727 DDP hangs at first AllGather during parameter broadcast with tied weights on Blackwell GPUs (sm_120)",72683da5afb6ed681f1091adaf483922f9234d57e1eb9515ed9b260808f586dc
issue-collapse:17fdd30595c94132,Review thread: 2 issues - issue #49961: rfc: automating the switching of inputs to the device of the params,related issue thread,needs triage,65,high,work_thread_graph,2,0,0,0,"49961, 50112",,,,"module: cuda, module: docs, module: rpc, module: ux, oncall: distributed","feature, module: cuda, module: docs, module: rpc, module: ux, needs design, oncall: distributed, triaged",thread:issue:49961:issue:50112,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49961 rfc: automating the switching of inputs to the device of the params, #50112 need a clear guide for when and how to use torch.cuda.set_device()",ac6ac2ae2b5221e5b7361ad0a7cac0cf0a614784f3e62ca0b525efbcd0931399
issue-collapse:56c1e9f9412ca3d4,Review thread: 2 issues - issue #177254: AArch64 Unit Test Failure - test_aot_inductor_arrayref - RuntimeError: unknown architecure,related issue thread,needs triage,65,high,work_thread_graph,2,0,0,0,"177254, 177258",,,,"module: arm, module: tests, oncall: jit, oncall: quantization","bot-mislabeled, bot-triaged, module: arm, module: tests, oncall: jit, oncall: quantization",thread:issue:177254:issue:177258,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#177254 AArch64 Unit Test Failure - test_aot_inductor_arrayref - RuntimeError: unknown architecure, #177258 AArch64 Unit Test Failure - test/jit/test_freezing.py TestFreezing.test_module_with_shared_type_instances",271e211c27b9588b683e7f3be1c5f86ea6f598e3e51c363df7485f5cda9accd9
issue-collapse:fd1cbffd0fcbeeb8,Review thread: 2 issues - issue #173546: [vllm] [release 2.10] V1 Test e2e + engine : test_draft_model_correctness[False-args0],related issue thread,needs triage,65,high,work_thread_graph,2,0,0,0,"173546, 174081",,,,"module: inductor, module: vllm, oncall: pt2, oncall: releng","module: inductor, module: vllm, oncall: pt2, oncall: releng, triaged",thread:issue:173546:issue:174081,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#173546 [vllm] [release 2.10] V1 Test e2e + engine : test_draft_model_correctness[False-args0], #174081 [vllm] [release 2.10] V1 e2e + engine test_ngram_and_suffix_correctness[speculative_config1], test_draft_model_realistic_example",391cd0e7c341caace2f57059be68e23204cec8be12a2d1559240766f9dc8dfd3
issue-collapse:217425bbb635481a,Review thread: 2 issues - issue #167631: `jit.export` analoge for `torch.export`,related issue thread,needs triage,65,high,work_thread_graph,2,0,0,0,"167631, 167645",,,,"oncall: export, oncall: pt2","export-triage-review, oncall: export, oncall: pt2, triaged",thread:issue:167631:issue:167645,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#167631 `jit.export` analoge for `torch.export`, #167645 `torch.export` incompatible with `parametrize.cached()` - model 25x slower than expected",928b3fda40453a3a946ca11f0101de5e36e78e765875dc33328531f835f9f986
issue-collapse:ca89032a16b7a176,"Review thread: 39 issues - issue #100249: [discussion] ""TensorList"" as first-class python-world abstraction (to mirror existing dispatcher's `TensorList` and `torch::autograd::tensor_list`) and as key for dispatch for merging `torch._foreach_*` into regular `torch.*` functions",related issue thread,stale/low urgency,64,high,work_thread_graph,39,0,0,0,"1512, 4959, 12672, 22924, 24915, 24985, 25162, 25522, 25691, 25743, 26957, 27522, 27617, 28743, 31758, 32851, 33181, 33296, 35642, 35759, 36650, 38228, 38419, 39570, 41292, 42405, 43012, 43502, 46176, 65156, 70533, 83775, 85157, 91471, 91508, 96085, 100249, 105325, 150989",,,,"module: bootcamp, module: crash, module: cuda, module: dataloader, module: dependency bug, module: multiprocessing, module: nestedtensor, module: nn, module: padding, module: performance, module: sparse, module: vision, oncall: distributed, oncall: jit","Stale, better-engineering, enhancement, feature, function request, high priority, module: bootcamp, module: crash, module: cuda, module: dataloader, module: dependency bug, module: multiprocessing, module: nestedtensor, module: nn, module: padding, module: performance, module: sparse, module: vision, needs reproduction, needs research, oncall: distributed, oncall: jit, release notes: nested tensor, tensor subclass, triaged",thread:issue:100249:issue:105325:issue:12672:issue:150989,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 39 open issues; raw issue rows remain available for audit.,"#1512 [feature request] Support tensors of different sizes as batch elements in DataLoader, #4959 Speed up data loading for `TensorDataset` if the underlying dataset supports index by a list of indices, #12672 Move collate_fn functionality / responsibility into Dataset object, #22924 [data loader] Graceful data loader threads exit on KeyboardInterrupt, #24915 Shared Dataset Functionality, #24985 Data worker should fetch a sample instead of a batch.",18787e6455c288b23514a5452a5225d576e29baa97711d536248b8fa131f8a10
issue-collapse:f344373eb4f8f896,Review thread: 15 issues - issue #102078: AddressSanitizer: heap-buffer-overflow in test_comprehensive_nn_functional_embedding_bag_cpu_bfloat16,related issue thread,needs reproduction,64,high,work_thread_graph,15,0,0,0,"88, 13993, 14436, 19668, 23240, 48491, 66247, 70245, 89817, 95944, 99561, 102078, 123421, 126963, 147141",,,,"module: android, module: assert failure, module: binaries, module: build, module: cpu, module: crash, module: cuda, module: cudnn, module: deadlock, module: decompositions, module: embedding, module: internals, module: mkl, module: nccl, module: tests, module: third_party, oncall: distributed, oncall: mobile","Stale, feature, module: android, module: assert failure, module: binaries, module: build, module: cpu, module: crash, module: cuda, module: cudnn, module: deadlock, module: decompositions, module: embedding, module: internals, module: mkl, module: nccl, module: tests, module: third_party, needs reproduction, oncall: distributed, oncall: mobile, triaged",thread:issue:102078:issue:123421:issue:126963:issue:13993,"Review this module: android queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,6,Represents 15 open issues; raw issue rows remain available for audit.,"#88 expose backend selection and cudnn settings to the end user, #13993 [tracking task] FBGEMM guarding AVX2 properly, #14436 Use of STL templates in cpu/ directory (compiling with different AVX settings) is silently hazardous, #19668 Importing matlab.engine after torch causes bad_alloc, #23240 UBSAN failure in test_simple_model (__main__.TestTensorBoardNumpy): runtime error: call to function pybind11::class_<caffe2::GradientWrapper>::dealloc(pybind11::detail::value_and_holder&) through pointer to incorrect function type 'void (*)(pybind11::detail::value_and_holder &)', #48491 SegmentationFault when pytorch is installed from source.",9516fa58cbc063abc6c52f13fc5520afe9e00ea9368b81b866c531a4f4dbaabb
issue-collapse:628c4681c8c0a41b,Review thread: 3 issues - issue #26165: Memory leak in multithreading environment when loading checkpoint,related issue thread,high-priority/blocker,64,high,work_thread_graph,3,0,0,0,"26165, 64412, 64535",,,,"module: cpu, module: intel, module: memory usage, module: mkl, module: mkldnn, module: multiprocessing, module: multithreading, module: performance","high priority, module: cpu, module: intel, module: memory usage, module: mkl, module: mkldnn, module: multiprocessing, module: multithreading, module: performance, needs reproduction, triaged",thread:issue:26165:issue:64412:issue:64535,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,3,Represents 3 open issues; raw issue rows remain available for audit.,"#26165 Memory leak in multithreading environment when loading checkpoint, #64412 Memory leak in multi-thread inference, #64535 Memory Leak in MKL OpenMP on AVX2 machine",410dbc0f8edb6b4e67fcd2e9719ca23fd0a7eeedc1d27d89582a1e37d4ae1460
issue-collapse:e7a548b537538d81,Review thread: 2 issues - issue #71149: Memory leak in distributions.multivariate_normal.MultivariateNormal,related issue thread,high-priority/blocker,64,high,work_thread_graph,2,0,0,0,"71149, 99042",,,,"module: distributions, module: memory usage","high priority, module: distributions, module: memory usage, needs reproduction, triaged",thread:issue:71149:issue:99042,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71149 Memory leak in distributions.multivariate_normal.MultivariateNormal, #99042 Memory allocation issues in distributions.multivariate_normal.MultivariateNormal",deb230ce8803a51cf60b478a062c00c94c0cd9eaa85d429aba9c63fae4cb3170
issue-collapse:d5a8ecf7cd83fe23,Review thread: 2 issues - issue #109819: ValueError: only one element tensors can be converted to Python scalars,related issue thread,high-priority/blocker,64,high,work_thread_graph,2,0,0,0,"109819, 110605",,,,"module: error checking, module: numpy, module: regression","module: error checking, module: numpy, module: regression, needs reproduction, triaged",thread:issue:109819:issue:110605,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#109819 ValueError: only one element tensors can be converted to Python scalars, #110605 ValueError issued instead of TypeError when tensor is cast to a scalar",c28a7835139bddab69735ebe9f40edacf88269ac25e73c5a402099ed4df12a45
issue-collapse:4c9ed51f15e90167,"Review thread: 12 issues - issue #124931: Missing description of Transformer argument ""memory_mask"" shape in 3D (including the batch dimension) case",related issue thread,stale/low urgency,62,high,work_thread_graph,12,0,0,0,"21876, 24826, 24930, 25132, 28657, 32529, 32590, 34573, 74092, 124931, 149534, 163408",,,,"module: convolution, module: docs, module: nn","enhancement, feature, function request, module: convolution, module: docs, module: nn, needs design, proposal accepted, triaged",thread:issue:124931:issue:149534:issue:163408:issue:21876,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 12 open issues; raw issue rows remain available for audit.,"#21876 nn.TransformerLayer, #24826 Transformer Lack of Embedding Layer and Positional Encodings, #24930 Successive Layer Normalization in nn.Transformer, #25132 Feature request: Fix dimension convention for masks in transformer, #28657 [docs] Unclear input/output format for TransformerEncoderLayer, #32529 addition of attention based techniques to pytorch",c99f543f834dba3f3a1b75e6122afe1f0896891a00625149a169e1e7b5bc007e
issue-collapse:4db9c210ecfb7601,Review thread: 25 issues - issue #15386: F.grid_sample doesn't respect padding_mode when height of inputs is 1,related issue thread,stale/low urgency,52,high,work_thread_graph,25,0,0,0,"2732, 4085, 5565, 15386, 19826, 21457, 24470, 24823, 24870, 25014, 25039, 30565, 36107, 36108, 48389, 50334, 50335, 50336, 50337, 50338, 50339, 50340, 50341, 61528, 81868",,,,"module: cpu, module: crash, module: cuda, module: cudnn, module: docs, module: error checking, module: interpolation, module: nans and infs, module: nn, module: numpy, module: performance, module: vision","enhancement, function request, hackamonth, high priority, module: NaNs and Infs, module: cpu, module: crash, module: cuda, module: cudnn, module: docs, module: error checking, module: interpolation, module: nn, module: numpy, module: performance, module: vision, proposal accepted, tracker, triaged",thread:issue:15386:issue:19826:issue:21457:issue:24470,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 25 open issues; raw issue rows remain available for audit.,"#2732 support grid_sample with batch=1 but supprting batch affine parameters, #4085 GridSampler behaviours, #5565 [feature request] Different interpolation algos for 'grid_sample' function, #15386 F.grid_sample doesn't respect padding_mode when height of inputs is 1, #19826 vectorized convert_to_int_of_same_size <int64_t> can't handle nan, #21457 downsampling with grid_sample doesn't match interpolate",24119de01eb0caed477a9545cd5f765549d5c9b7635112be24d92a41cdbb32ea
issue-collapse:920c06e944dc8e43,"Review thread: 8 issues - issue #106959: `tensor.repeat` quirks: has no `torch.` variant, no `out=` variant, no inplace variant | `torch.tile` also does not have `out=` variant and uses `dims=` instead of `dim=`",related issue thread,stale/low urgency,52,high,work_thread_graph,8,0,0,0,"18095, 31458, 39007, 61585, 68616, 71210, 106959, 154613",,,,"module: doc infra, module: docs, module: internals, module: numpy, module: python array api, module: python frontend, module: ux, module: viewing and reshaping, oncall: jit","function request, high priority, module: doc infra, module: docs, module: internals, module: numpy, module: python array api, module: python frontend, module: ux, module: viewing and reshaping, needs design, oncall: jit, triaged",thread:issue:106959:issue:154613:issue:18095:issue:31458,Handle this module: doc infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#18095 torch.flip is inconsistent with np.flip and also uses `dims` arg instead of `dim`, #31458 Integrate `torch.xxx` and `Tensor.xxx`, #39007 [docs] Urls changed => forum links would become invalid, #61585 Proposal: allow using axis, axes, dim and dims interchangeably, #68616 `torch.flip`, `torch.roll`, `torch.tile` has arguments `dims=` for both dim/multidim while other functions have argument `dim=` for the same usage, #71210 axis to dim remapping is not working for flip and roll",94fe39e2b3e9cb7a3b748bc3da18d9ea55fffb3549e072b19346383694eb747c
issue-collapse:16bdcb3e50ffba91,Review thread: 7 issues - issue #118507: RuntimeError in torch.istft with center=False: Window Overlap Add Issue,related issue thread,stale/low urgency,52,high,work_thread_graph,7,0,0,0,"70073, 81428, 91309, 94718, 118507, 134323, 142392",,,,"module: complex, module: fft, module: python frontend, module: scipy compatibility","feature, high priority, module: complex, module: fft, module: python frontend, module: scipy compatibility, triaged",thread:issue:118507:issue:134323:issue:142392:issue:70073,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 7 open issues; raw issue rows remain available for audit.,"#70073 Feature request: [STFT] Add warning message if signal length is not a multiple of hop_length in torch.stft, #81428 torch.stft does not normalize non-rectangular windows correctly, #91309 iSTFT produces RuntimeError with center=False and Blackman/Bartlett/Hann windows, #94718 Option to bypass NOLA check in torch.istft, #118507 RuntimeError in torch.istft with center=False: Window Overlap Add Issue, #134323 Fix `torch.stft` and `torch.istft` when using `center=False`, non-rectangular windows and `win_length==hop_length`",8b7cc08dda15feae1e45ef1439956e522c98ad2b9efb94b5caaa3ccac0dbb422
issue-collapse:647855ce8d952823,Review thread: 6 issues - issue #113713: [cuDNN][cuDNN V8 API] cuDNN Flash-Attention Upstreaming RFC/tracking issue,related issue thread,stale/low urgency,52,high,work_thread_graph,6,0,0,0,"58414, 58858, 58859, 58860, 58862, 113713",,,,"module: convolution, module: cuda, module: cudnn, module: sdpa","high priority, module: convolution, module: cuda, module: cudnn, module: sdpa, tracker, triaged",thread:issue:113713:issue:58414:issue:58858:issue:58859,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 6 open issues; raw issue rows remain available for audit.,"#58414 cuDNN v8 API tracking issue, #58858 [cuDNN v8] Extend current cuDNN v8 API binding to support convolution backward and transposed convolution forward, #58859 [cuDNN v8] Extend current cuDNN convolution v8 API binding to support cuDNN benchmark, #58860 [cuDNN v8] Extend current cuDNN convolution v8 API binding to support conv-bias-activation fusion, #58862 [cuDNN v8] Improve cuDNN convolution v8 API error reporting, #113713 [cuDNN][cuDNN V8 API] cuDNN Flash-Attention Upstreaming RFC/tracking issue",b82ac6a281decc7f45aa9c9367b3ce05e37435ba26f50c44b8b7981f03428608
issue-collapse:7e818679af5aae2d,Review thread: 5 issues - issue #59552: test_batch_isend_irecv_nccl fails with NCCL 2.8.3,related issue thread,needs triage,52,high,work_thread_graph,5,0,0,0,"59552, 63486, 64688, 68866, 73960",,,,"module: c10d, module: correctness (silent), module: nccl, oncall: distributed, oncall: distributed infra","bot-triaged, high priority, module: c10d, module: correctness (silent), module: nccl, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:59552:issue:63486:issue:64688:issue:68866,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",needs triage,4,Represents 5 open issues; raw issue rows remain available for audit.,"#59552 test_batch_isend_irecv_nccl fails with NCCL 2.8.3, #63486 Distributed / MPI / CUDA: Incorrect messages received from isend/irecv in PyTorch 1.9, #64688 Backend capability chart for nccl GPU send/recv may be stale, #68866 wait() does not block the default stream for NCCL's asynchronous P2P operations, #73960 processes hang when executing cross-machine asynchronous P2P communication on NCCL backend",9b7b124afa222183bee0c17bfcf3492d2897937cb8ec75429ea97173f06dfc4f
issue-collapse:a440422b2d92c206,Review thread: 5 issues - issue #41950: [docs] Clarify behavior of torch.cuda.device_count() when torch compiled without CUDA (cpu-only) or when CUDA is not available,related issue thread,stale/low urgency,52,high,work_thread_graph,5,0,0,0,"41950, 52675, 62931, 92250, 99982",,,,"module: crash, module: cuda, module: docs, module: edge cases","enhancement, module: crash, module: cuda, module: docs, module: edge cases, small, triaged",thread:issue:41950:issue:52675:issue:62931:issue:92250,Handle this module: crash stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 5 open issues; raw issue rows remain available for audit.,"#41950 [docs] Clarify behavior of torch.cuda.device_count() when torch compiled without CUDA (cpu-only) or when CUDA is not available, #52675 [feature request] Introduce env var `TORCH_USE_CUDA=OFF` and some `torch.cuda.enabled = False` to faithfully/completely disable cuda (torch.cuda.is_available() <- False and torch.cuda.get_device_count() <- 0) and driver loading/interaction, #62931 GPU compatibility check without initializing CUDA, #92250 torch.cuda.is_available() returns True even if the CUDA hardware can't run pytorch, #99982 torch.cuda.is_available() crashes python in systems with disabled gpu",e51af48ea6e86466ecbcfc2e960900e02cabc8fc4a43d27d176c2c29a214bb5a
issue-collapse:9c65732dc800b390,Review thread: 5 issues - issue #155992: [feature request] Native checkpointing to/from `s3://` (DCP / `torch.load` / `torch.save`),related issue thread,stale/low urgency,52,high,work_thread_graph,5,0,0,0,"68320, 91965, 96412, 155992, 171412",,,,"module: hub, module: serialization, oncall: distributed, oncall: distributed checkpointing","actionable, bot-triaged, feature, high priority, module: hub, module: serialization, oncall: distributed, oncall: distributed checkpointing, ptd-bot-triaged, triaged",thread:issue:155992:issue:171412:issue:68320:issue:91965,Handle this module: hub stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 5 open issues; raw issue rows remain available for audit.,"#68320 Potential race conditions between multiple workers trying to download and cache the same file in torch.hub.load_state_dict_from_url and torch.hub.download_url_to_file <- duplicate dataset/model downloads across DDP workers, #91965 Proxy/cache server option/hooks for downloading model checkpoints and dataset archive files in cloud environment, #96412 Proxy Options for Pytorch Hub, #155992 [feature request] Native checkpointing to/from `s3://` (DCP / `torch.load` / `torch.save`), #171412 [Feature Request][DCP] Add ""Broadcast Load"" arg to dcp.load for HSDP storage optimization",bb0d758cfbec2ed0ec42c6fac5fef8c9e587b3d2cb4ea27db897f7e43252d421
issue-collapse:3fcaa1ce3b607e66,Review thread: 5 issues - issue #134570: histogramdd 'Cuda' backend,related issue thread,stale/low urgency,52,high,work_thread_graph,5,0,0,0,"14095, 30443, 69519, 99719, 134570",,,,"module: bootcamp, module: cuda, module: functorch, module: numpy, module: sorting and selection","actionable, enhancement, feature, function request, high priority, module: bootcamp, module: cuda, module: functorch, module: numpy, module: sorting and selection, triaged",thread:issue:134570:issue:14095:issue:30443:issue:69519,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#14095 [feature request] bincount along specified dimension(s), #30443 Expose bhistc to python, #69519 Feature Request: CUDA torch.histogram (and histogramdd), #99719 New feature requested: vmap for torch.histc, #134570 histogramdd 'Cuda' backend",355feb778eac92491fe4fa18a6754b9cc74beb60479b9ba1db40e94ee84e0251
issue-collapse:7f71b4c12b21a4ce,Review thread: 4 issues - issue #44282: PyTorch wheel's own OpenMP library clashing with system-wide OpenMP library at runtime,related issue thread,needs maintainer decision,52,high,work_thread_graph,4,0,0,0,"44282, 91989, 98836, 99625",,,,"module: binaries, module: build, module: dependency bug, module: intel, module: mkl, module: multithreading, module: openmp, module: third_party","high priority, intel priority, module: binaries, module: build, module: dependency bug, module: intel, module: mkl, module: multithreading, module: openmp, module: third_party, triaged",thread:issue:44282:issue:91989:issue:98836:issue:99625,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#44282 PyTorch wheel's own OpenMP library clashing with system-wide OpenMP library at runtime, #91989 Torch's affinity setting lead to openvino using only one core., #98836 PyTorch's packaged libgomp causes significant performance penalties on CPU when used together with other Python packages, #99625 Conda Pytorch set processor affinity to the first physical core after fork",97645f2319e9755ae1b8d47b3b15b266689dc8a47c7f124043112edb6344629f
issue-collapse:bbfbf8d1fcd9d75b,Review thread: 4 issues - issue #43567: `torch.svd()` CUDA gives incorrect results when input contains `nan`,related issue thread,stale/low urgency,52,high,work_thread_graph,4,0,0,0,"43567, 52633, 71911, 72106",,,,"module: ci, module: cuda, module: linear algebra, module: magma, module: mkl, module: nans and infs, module: openblas","high priority, module: NaNs and Infs, module: ci, module: cuda, module: linear algebra, module: magma, module: mkl, module: openblas, triaged",thread:issue:43567:issue:52633:issue:71911:issue:72106,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#43567 `torch.svd()` CUDA gives incorrect results when input contains `nan`, #52633 CUDA `linalg.norm` matrix order +/-2 and nuclear norm for extreme values may be incorrect, #71911 Functions depending on SVD are broken for inputs with non-finite values with MKL 2022+ and OpenBLAS 0.3.15+, #72106 Update MKL version used with MAGMA",116e144e58085c67203e4d40569c33808fbdac5350073651e8b74ebf47290104
issue-collapse:3889e57355adf8de,Review thread: 4 issues - issue #34058: libtorch.so file size is very large,related issue thread,high-priority/blocker,52,high,work_thread_graph,4,0,0,0,"34058, 90608, 93081, 93955",,,,"module: binaries, module: build, module: cpp, module: performance, oncall: jit, oncall: mobile, oncall: releng","high priority, module: binaries, module: build, module: cpp, module: performance, oncall: jit, oncall: mobile, oncall: releng, quansight-nack, topic: performance, triaged",thread:issue:34058:issue:90608:issue:93081:issue:93955,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,3,Represents 4 open issues; raw issue rows remain available for audit.,"#34058 libtorch.so file size is very large, #90608 Operator overload priority should not rely on static initialization order, #93081 [Releng] Optimize PyTorch packaging, #93955 Enable Link Time Optimization in PyTorch 2.0 Release Binaries - Smaller, Faster, Better Binaries",50d2d6329f83bb7b13f381bfd47e0205b9e5404aa5c8af2dca29bebbe41c36de
issue-collapse:4fbe175ee4414f2d,Review thread: 4 issues - issue #23940: Improve the performance of linear algebra operations in CUDA for small problem sizes,related issue thread,stale/low urgency,52,high,work_thread_graph,4,0,0,0,"23940, 47953, 53879, 55122",,,,"module: cuda, module: linear algebra, module: magma, module: performance","high priority, module: cuda, module: linear algebra, module: magma, module: performance, triaged",thread:issue:23940:issue:47953:issue:53879:issue:55122,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#23940 Improve the performance of linear algebra operations in CUDA for small problem sizes, #47953 Linear algebra GPU backend tracking issue [magma/cusolver/cublas], #53879 Linear algebra GPU library function bug tracking issue [magma/cusolver/cublas], #55122 Cusolver handle may decrease MAGMA performance on GPU",517ebb904c48f8059b6811a28fcdabf904ee0beadc7ae23f953f1cb267dfa81a
issue-collapse:971ba674b37d9fb3,Review thread: 4 issues - issue #132020: `torch.Tensor.to` ignores `memory_format` kwarg,related issue thread,has linked PR,52,high,work_thread_graph,4,0,0,0,"41226, 62027, 86558, 132020",,,,"module: docs, module: memory format, module: primtorch, module: python frontend","actionable, high priority, module: docs, module: memory format, module: primTorch, module: python frontend, needs design, triaged",thread:issue:132020:issue:41226:issue:62027:issue:86558,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,3,Represents 4 open issues; raw issue rows remain available for audit.,"#41226 The name ""Contiguous"" in `torch::MemoryFormat::Contiguous` can mislead users into assuming contiguous memory, #62027 `x.to(memory_format=torch.contiguous_format)` does not always return a contiguous tensor, #86558 Data conversion ops ignore `memory_format=torch.contiguous_format`, #132020 `torch.Tensor.to` ignores `memory_format` kwarg",62bdb1d485613f1267b939f7c98d5dd418275d8905c4ea40205dd1d316bfbb2b
issue-collapse:2a2fb3676e0acfd0,Review thread: 4 issues - issue #10615: Request for better memory management,related issue thread,stale/low urgency,52,high,work_thread_graph,4,0,0,0,"10615, 44380, 48559, 108858",,,,"module: cuda, module: memory usage, module: multiprocessing, module: serialization","feature, high priority, module: cuda, module: memory usage, module: multiprocessing, module: serialization, triaged",thread:issue:10615:issue:108858:issue:44380:issue:48559,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#10615 Request for better memory management, #44380 Support for NVIDIA UVM technology, #48559 [FR] torch.load should support loading directly to pinned/shared memory, #108858 Ubuntu vs Windows: torch.cuda.OutOfMemoryError only happens on Ubuntu",79f7cfd0f2aefae29ec249245a1bc21ef07eb001fde3f43d491e32114f3ed1f4
issue-collapse:b06f8a1ae31094a2,Review thread: 4 issues - issue #102517: [feature request] PyTorch support for sub-interpreters with PEP 684 accepted and release in Python 3.12,related issue thread,stale/low urgency,52,high,work_thread_graph,4,0,0,0,"10950, 23490, 102517, 105943",,,,"module: cpp, module: python frontend, oncall: pt2","feature, high priority, module: cpp, module: python frontend, months, needs reproduction, oncall: pt2, todo, triaged",thread:issue:102517:issue:105943:issue:10950:issue:23490,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#10950 Cannot run torch in different sub-interpreters, #23490 upcoming PEP 554: how much effort we need to support sub-interpreter, #102517 [feature request] PyTorch support for sub-interpreters with PEP 684 accepted and release in Python 3.12, #105943 PT2 is not thread safe",1b048dc8f2a2b094884ba9c808fedc959302a890ffb07fdfb8c6c5d9a786471a
issue-collapse:a5f1502be532de17,Review thread: 3 issues - issue #80017: RPC init fails and crashes when world_size is greater than 18,related issue thread,stale/low urgency,52,high,work_thread_graph,3,0,0,0,"80017, 80104, 85607",,,,"module: rpc, oncall: distributed","high priority, module: rpc, oncall: distributed, triaged",thread:issue:80017:issue:80104:issue:85607,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#80017 RPC init fails and crashes when world_size is greater than 18, #80104 Provide error message when thread pool is exhausted in RPC, #85607 [Distributed: RPC] Failed to initialize RPC with >18 workers",5ca0537df67067da4706099529902bbb9c49c3d17b907a6108ef76e7e31d8199
issue-collapse:2177d3135ae3c054,Review thread: 3 issues - issue #64812: PyTorch inference on tensors of a particular size cause Illegal Instruction (core dumped) on Jetson Nano,related issue thread,stale/low urgency,52,high,work_thread_graph,3,0,0,0,"64812, 68349, 70350",,,,"module: arm, module: cpu, module: crash, module: dependency bug, module: jetson, module: mkldnn","Stale, module: arm, module: cpu, module: crash, module: dependency bug, module: jetson, module: mkldnn, triaged",thread:issue:64812:issue:68349:issue:70350,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#64812 PyTorch inference on tensors of a particular size cause Illegal Instruction (core dumped) on Jetson Nano, #68349 SIGILL, Illegal Instruction from libtorch_cpu.so when callling backward() function in a VM., #70350 Core dumped with large matmul on aarch64",1bac84c2ee699994cb60465bb13673a9ccf142e62fd3cbb215abeb03c1987fc6
issue-collapse:27d9487495d639bc,Review thread: 3 issues - issue #50012: torch.split is divergent from np.split,related issue thread,high-priority/blocker,52,high,work_thread_graph,3,0,0,0,"5212, 50012, 68174",,,,"module: deprecation, module: numpy","high priority, module: deprecation, module: numpy, triaged",thread:issue:50012:issue:5212:issue:68174,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,2,Represents 3 open issues; raw issue rows remain available for audit.,"#5212 Weird error message in torch.split_size_or_sections, #50012 torch.split is divergent from np.split, #68174 RFC: “What’s in a (NumPY) name?” — Changing PyTorch Operator Names",552a038154188f116409d42ae5091ef048bf1d27f89a3b3bf701883151fbc322
issue-collapse:06ab2256e3dd2544,Review thread: 3 issues - issue #35600: Increased memory usage in repetitive torch.jit.trace calls,related issue thread,high-priority/blocker,52,high,work_thread_graph,3,0,0,0,"35600, 86537, 96726",,,,"module: memory usage, oncall: jit","high priority, module: memory usage, oncall: jit, triaged",thread:issue:35600:issue:86537:issue:96726,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,3,Represents 3 open issues; raw issue rows remain available for audit.,"#35600 Increased memory usage in repetitive torch.jit.trace calls, #86537 Running JIT trace for many times leads to OOM, #96726 Memory not release after jit.trace/freeze",346f59893bb92d542974ea7bb366496a243f78382921da1855af17104bc75128
issue-collapse:48c2f257063648ff,Review thread: 3 issues - issue #21018: free(): invalid pointer Aborted (core dumped),related issue thread,needs reproduction,52,high,work_thread_graph,3,0,0,0,"21018, 50804, 75912",,,,"module: crash, module: regression","module: crash, module: regression, needs reproduction, triaged",thread:issue:21018:issue:50804:issue:75912,"Review this module: crash queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,2,Represents 3 open issues; raw issue rows remain available for audit.,"#21018 free(): invalid pointer Aborted (core dumped), #50804 Segmentation fault encountered when using nn.MultiheadAttention with v1.7.1, #75912 interaction with psychopy during imports, script exits with: free(): invalid pointer. Aborted (core dumped)",d8737abecfddb142f23055c5f53fe9504cbe69dc90fe6acb044f50c53c17b55a
issue-collapse:71d609e59055bd12,Review thread: 3 issues - issue #126523: Summary of tests in machine readable format,related issue thread,stale/low urgency,52,high,work_thread_graph,3,0,0,0,"126523, 144563, 153938",,,,"module: ci, module: testing, module: tests","actionable, high priority, low priority, module: ci, module: testing, module: tests, triaged",thread:issue:126523:issue:144563:issue:153938,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#126523 Summary of tests in machine readable format, #144563 Multiple tests not run / run as no-ops by `run_test.py`, #153938 Subtests not properly accounted for in test reports (e.g. on CI)",010a0dd84e7565e87b4acbc04b2cca7404113df0f77b6f3800ff177705f18f89
issue-collapse:af54be75a28b95b1,Review thread: 3 issues - issue #102911: PackedSequences on MPS accelerator yields `grad_y` missing or crashes the kernel.,related issue thread,high-priority/blocker,52,high,work_thread_graph,3,0,0,0,"94691, 102911, 102971",,,,"module: correctness (silent), module: crash, module: loss, module: masked operators, module: mps","module: correctness (silent), module: crash, module: loss, module: masked operators, module: mps, triaged",thread:issue:102911:issue:102971:issue:94691,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,2,Represents 3 open issues; raw issue rows remain available for audit.,"#94691 Nan is output by GRU on mps, #102911 PackedSequences on MPS accelerator yields `grad_y` missing or crashes the kernel., #102971 Discussion and Design for Masked Loss Functions which can be used with PackedSequence training (but not exclusively)",d135ed7257cc09d2791b8f3e8976f5f0a0ce1856a414137d21bfae69682145ab
issue-collapse:b7528a9098bdd463,Review thread: 2 issues - issue #98537: Segfault when using torch.ops.* to access de-registered op,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"98537, 98541",,,,"module: crash, module: dispatch, module: internals, module: library","module: crash, module: dispatch, module: internals, module: library, triaged",thread:issue:98537:issue:98541,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#98537 Segfault when using torch.ops.* to access de-registered op, #98541 Memory corruption using torch.ops.* to access re-registered operator",3af3a9332efd856e6d38f2bd8b209ac77da462f4ae6c43fafd691b1b4eb10061
issue-collapse:1b386e32cf3d00a8,Review thread: 2 issues - issue #96275: A Segment Fault can be triggered in torch.pinverse,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"96275, 96276",,,,"module: crash, module: edge cases","module: crash, module: edge cases, triaged",thread:issue:96275:issue:96276,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#96275 A Segment Fault can be triggered in torch.pinverse, #96276 A Segment Fault can be triggered in torch.geqrf with an edge case",eca5ef90bbf9b5107930fa63e75ef6557734bf2fae38ed721bcdff459f5431d3
issue-collapse:554515d8bb06f28a,Review thread: 2 issues - issue #94668: A segment fault can be triggered in torch.lstm with edge cases,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"94668, 94808",,,,"module: crash, module: edge cases, module: rnn","module: crash, module: edge cases, module: rnn, triaged",thread:issue:94668:issue:94808,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#94668 A segment fault can be triggered in torch.lstm with edge cases, #94808 Memory Corruption in torch.lstm caused by edge cases",25df8524ecf5bf8a99fc58117de0fdeb1455737d254dc25c2f01e1e3f975705c
issue-collapse:89d32292b0f1517f,Review thread: 2 issues - issue #81541: [vulkan]compiling VulkanOpContext.cpp with some errors,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"81541, 82072",,,,"module: build, module: vulkan, oncall: mobile","high priority, module: build, module: vulkan, oncall: mobile, triaged",thread:issue:81541:issue:82072,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#81541 [vulkan]compiling VulkanOpContext.cpp with some errors, #82072 cant build with USE_VULKAN=1",eb5c789206cd2747fd715dd805c688b7b557c1997e19959a63f139cd25621382
issue-collapse:b55c4376940b0230,"Review thread: 2 issues - issue #69801: Dropout2d doesn't drop channels for (C, H, W)",related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"69801, 69960",,,,"module: batching, module: correctness (silent), module: nn","Stale, high priority, module: batching, module: correctness (silent), module: nn, triaged",thread:issue:69801:issue:69960,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69801 Dropout2d doesn't drop channels for (C, H, W), #69960 FeatureAlphaDropout doesn't drop channels for (C, D, H, W)",d9c76de974fef897a9c789bf9b670756af37d80c4eebab74674cab039b3b5b83
issue-collapse:338c2a62f38346bf,Review thread: 2 issues - issue #67448: Deadlock in test_multiprocessing_spawn.py,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"67448, 67712",,,,"module: ci, module: deadlock, module: multiprocessing","better-engineering, high priority, module: ci, module: deadlock, module: multiprocessing, triaged",thread:issue:67448:issue:67712,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#67448 Deadlock in test_multiprocessing_spawn.py, #67712 Decorate tests with a ""deadline""",74be8efa8d292883d588539605c9ff362caa1b36f6709ba7ffcae70e70dc559c
issue-collapse:c444e7504f7f6bb8,Review thread: 2 issues - issue #55375: `DataParallel` (`broadcast_coalesced`) with complex tensors yield real views,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"55375, 57534",,,,"module: complex, module: data parallel, oncall: distributed","high priority, module: complex, module: data parallel, oncall: distributed, triaged",thread:issue:55375:issue:57534,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,2,Represents 2 open issues; raw issue rows remain available for audit.,"#55375 `DataParallel` (`broadcast_coalesced`) with complex tensors yield real views, #57534 `torch.view_as_complex()` does not work when `storage_offset` is odd",3416f89a698b1b2b919765ac04226a1c60ac46f22fc09910b674c2b16fc467c2
issue-collapse:1463ab05712bb9c8,Review thread: 2 issues - issue #51872: Tensor.nonzero tries to allocate huge amount of memory for tensors on GPU with num_elements close to INT_MAX,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"51872, 80337",,,,"module: crash, module: cuda, module: dependency bug, module: edge cases, module: memory usage","module: crash, module: cuda, module: dependency bug, module: edge cases, module: memory usage, triaged",thread:issue:51872:issue:80337,Handle this module: crash stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#51872 Tensor.nonzero tries to allocate huge amount of memory for tensors on GPU with num_elements close to INT_MAX, #80337 Illegal Memory Access from nonzero method when Tensor is Too Large",572e9399008a8788b5ae8d376d439e30072bc30faabf9190c30ba0a4b6018530
issue-collapse:1f2f58ea41ba0aa1,Review thread: 2 issues - issue #49758: c10::scalar_to_tensor(...) uses should be audited for performance and type promotion impact,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"49758, 49909",,,,"module: cuda, module: performance, module: tests","high priority, module: cuda, module: performance, module: tests, triaged",thread:issue:49758:issue:49909,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49758 c10::scalar_to_tensor(...) uses should be audited for performance and type promotion impact, #49909 Test functionality to detect extra cross device synchronizations",6c7ef5eba987a2c6dddd03d4801a289b2186c4b7c8d9ebad9d67f67707b63029
issue-collapse:53c6ec9e3cb73849,Review thread: 2 issues - issue #48945: CUDA error: illegal memory access Conv3d,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"48945, 49459",,,,"module: convolution, module: crash, module: cudnn","module: convolution, module: crash, module: cudnn, triaged",thread:issue:48945:issue:49459,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#48945 CUDA error: illegal memory access Conv3d, #49459 This error occurs occasionally during the run",66b7ba5c6b18d4229f82fa4871e7e943844b942a1888b03c8210dd8642dcf010
issue-collapse:5273be1cbe168f6f,Review thread: 2 issues - issue #40137: Improve RPC test debugability,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"40137, 56297",,,,"module: pybind, module: rpc","better-engineering, high priority, module: pybind, module: rpc, triaged",thread:issue:40137:issue:56297,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40137 Improve RPC test debugability, #56297 Audit destructors of classes bound using pybind11::class_ to see if they can block; such cases can deadlock",6681115ab41ffbc80974c038e4d8a2a1c158ef175afbb50c1da7fa45ddc03f44
issue-collapse:5444ee68e18e9a9b,Review thread: 2 issues - issue #34257: CustomFromMask pruning stores a copy of the user-provided mask,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"34257, 85538",,,,"module: memory usage, module: nn, module: pruning","high priority, module: memory usage, module: nn, module: pruning, triaged",thread:issue:34257:issue:85538,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#34257 CustomFromMask pruning stores a copy of the user-provided mask, #85538 Iterative Global Pruning Cause GPU Memory Leak",56416d597d18d5288a85d76517a5a54fc9272df2f381862fbf618ffce47c63b9
issue-collapse:f2c9a79590d92abd,Review thread: 2 issues - issue #32918: Interpolate in the “bicubic” mode with the same shape outputs zeros from second sample onwards,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"32918, 34544",,,,module: nn,"high priority, module: nn, triaged",thread:issue:32918:issue:34544,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#32918 Interpolate in the “bicubic” mode with the same shape outputs zeros from second sample onwards, #34544 Strange behaviour of F.interpolate with bicubic mode.",d6ff346da02bab755ff220fefefa9407c8d88947b2c8fb5f5e47400aeea63746
issue-collapse:7da772e8e0ff7db4,Review thread: 2 issues - issue #2576: CUDA multinomial is limited to 2^24 categories,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"2576, 43115",,,,"module: 64-bit, module: cuda, module: distributions, module: numerical-stability","function request, high priority, module: 64-bit, module: cuda, module: distributions, module: numerical-stability, triaged",thread:issue:2576:issue:43115,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#2576 CUDA multinomial is limited to 2^24 categories, #43115 torch.multinomial with replacement=True produces inaccurate results for large number of categories",59bab49e7ba0aec0f4616fa1104473e7b3b5712fccf9ca7ea3b26a4b0098564c
issue-collapse:fdca1665d5d46d9b,Review thread: 2 issues - issue #17199: Deadlock with multiprocessing (using fork) and OpenMP / PyTorch should warn after OMP and fork that multithreading may be broken,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"17199, 58962",,,,"module: docs, module: multiprocessing, oncall: distributed","high priority, module: docs, module: multiprocessing, oncall: distributed, triaged",thread:issue:17199:issue:58962,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#17199 Deadlock with multiprocessing (using fork) and OpenMP / PyTorch should warn after OMP and fork that multithreading may be broken, #58962 Operating on a shared memory tensor with multiple threads hangs",6c27756b49e646931102818eca28ec1d3aeedaf80d2892915c8ef3241d8483ae
issue-collapse:2417e10bd11cc19a,Review thread: 2 issues - issue #16897: Implement `numpy.random.choice` equivalent,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"16897, 82761",,,,"module: bootcamp, module: numpy","enhancement, feature, high priority, module: bootcamp, module: numpy, triaged",thread:issue:16897:issue:82761,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#16897 Implement `numpy.random.choice` equivalent, #82761 torch.Tensor.bag() should automatically implement bagging",7ec095e241f8fcaa75c9d426cffa7c4875da762637b4d8a06e0b6ddbe3274ddd
issue-collapse:38ec06baf6c8e759,Review thread: 2 issues - issue #164402: NCCL-2.28.3 build locally is unusable on H100,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"164402, 164787",,,,"module: build, module: ci, module: cuda, module: nccl, module: regression, module: third_party","actionable, high priority, module: build, module: ci, module: cuda, module: nccl, module: regression, module: third_party, triaged",thread:issue:164402:issue:164787,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#164402 NCCL-2.28.3 build locally is unusable on H100, #164787 Add static build for NCCL in CI",efafed99b1f6543e42fc132157ece9f6bdf5640575f75ddb0875319bb9372a14
issue-collapse:130ec63bec7b8d70,Review thread: 2 issues - issue #15849: DataLoader with option to re-use worker processes,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"15849, 39443",,,,"module: dataloader, module: multiprocessing, module: performance","feature, high priority, module: dataloader, module: multiprocessing, module: performance, triaged",thread:issue:15849:issue:39443,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#15849 DataLoader with option to re-use worker processes, #39443 PyTorch multiprocessing.spawn seems slow with list of tensors",fac5ce4cb50fc1c72d70254fdb7f3671c5075c2790c657b71031a420d375b0cf
issue-collapse:3836eefea2700ab1,"Review thread: 2 issues - issue #146828: On Linux, passing torch.Generator to multiprocessing.Process crashes for forkserver and spawn start method",related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"32351, 146828",,,,"module: multiprocessing, module: random, module: serialization","high priority, module: multiprocessing, module: random, module: serialization, triaged",thread:issue:146828:issue:32351,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#32351 Problem with multiprocessing, custom __getstate__ with Tensors and forkserver, #146828 On Linux, passing torch.Generator to multiprocessing.Process crashes for forkserver and spawn start method",e5c1cf41972b3efd236ed6663d1adbb58873da29894d12dc52edb7e64cbb296c
issue-collapse:9e75a6459d2d42cb,Review thread: 2 issues - issue #141218: Floating point exception (core dumped) in `scaled_dot_product_attention`,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"141218, 141219",,,,"module: crash, module: edge cases, module: empty tensor, module: error checking, module: sdpa","module: crash, module: edge cases, module: empty tensor, module: error checking, module: sdpa, topic: fuzzer, triaged",thread:issue:141218:issue:141219,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#141218 Floating point exception (core dumped) in `scaled_dot_product_attention`, #141219 Floating point exception (core dumped) in `_scaled_dot_product_flash_attention_for_cpu`",94301f5affd28676d6f6fd1327568cbe92e390c1ce73cd6019fe70ab5f04212c
issue-collapse:f3fac9953a9d8c94,Review thread: 2 issues - issue #139298: CUDNN sdp attention causes loss explosion,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"139298, 141133",,,,"module: cuda, module: cudnn, module: sdpa","high priority, module: cuda, module: cudnn, module: sdpa, triaged",thread:issue:139298:issue:141133,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#139298 CUDNN sdp attention causes loss explosion, #141133 CuDNN SDPA Issue Tracker",c1a1b3a03f0d13e582233384192e6c1ee012e756adea14e2ecf1a72df26ae684
issue-collapse:44282ebca74201e4,Review thread: 2 issues - issue #13918: Pytorch very slow to convert list of numpy arrays into tensors,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"13918, 38703",,,,"module: numpy, module: performance","enhancement, has workaround, high priority, module: numpy, module: performance, triaged",thread:issue:13918:issue:38703,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13918 Pytorch very slow to convert list of numpy arrays into tensors, #38703 torch.as_tensor(np_array) is sometimes much faster than torch.tensor(np_array)",e9b1dbd75ddd4bc79bb1288fc6946f137d14a74a2d006de5f693de9099de9eaf
issue-collapse:fbbd7d85d3d742f5,"Review thread: 2 issues - issue #135437: Cannot build libtorch from source with NVHPC compilers - ""Could NOT find Threads""",related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"82627, 135437",,,,"module: build, module: crash, module: cuda, module: openmp, module: third_party","module: build, module: crash, module: cuda, module: openmp, module: third_party, triaged",thread:issue:135437:issue:82627,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#82627 PyTorch crashes when running with OpenACC, #135437 Cannot build libtorch from source with NVHPC compilers - ""Could NOT find Threads""",f2d45e6e4e90514b4266c0b8eaca77135a79184d7d33980f1d40d2ad0de3889d
issue-collapse:571f3abb54989d29,Review thread: 2 issues - issue #13402: batch_norm doesn't bump version counter of running stats,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"13402, 33867",,,,"module: autograd, module: cudnn, module: nn","actionable, fixathon, high priority, module: autograd, module: cudnn, module: nn, triaged",thread:issue:13402:issue:33867,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13402 batch_norm doesn't bump version counter of running stats, #33867 cuDNN batchnorm with non-contiguous running mean silently discards updates",2cf6acd47bbe3456a86c2fc1ced6c3091cf41a20452078f12bf72b3b3258262d
issue-collapse:23221027e7125c8b,Review thread: 2 issues - issue #129118: interpolate nearest get values zero when outputs over 4G elements,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"129118, 141831",,,,"module: 64-bit, module: correctness (silent), module: cuda, module: interpolation, module: nn","high priority, module: 64-bit, module: correctness (silent), module: cuda, module: interpolation, module: nn, triaged",thread:issue:129118:issue:141831,Handle this module: 64-bit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#129118 interpolate nearest get values zero when outputs over 4G elements, #141831 Nearest-neighbour upsampling with `interpolate` silently fails on CUDA when output size exceeds 2^31",047e075969969b6332753608cde66f64dc8432f651517167084517d172316c6f
issue-collapse:416dbcfeb30afb7b,Review thread: 2 issues - issue #115347: [Functorch] vmap over index_select expands the output,related issue thread,high-priority/blocker,52,high,work_thread_graph,2,0,0,0,"115347, 124291",,,,module: functorch,"actionable, high priority, module: functorch, triaged",thread:issue:115347:issue:124291,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,high-priority/blocker,1,Represents 2 open issues; raw issue rows remain available for audit.,"#115347 [Functorch] vmap over index_select expands the output, #124291 vmap over getitem indexing raises a RunTime error",e785ac80e2ba028bf97f5c2b7f779dea390a9fac0285048a01f56cb954925250
issue-collapse:2fc7b1aec2ec9ac1,"Review thread: 2 issues - issue #111693: [export] 14k models: AssertionError: graph-captured input # 2, of type <class 'torch.nn.parameter.Parameter'>, is not among original inputs of types",related issue thread,needs reproduction,52,high,work_thread_graph,2,0,0,0,"111693, 112043",,,,"oncall: export, oncall: pt2","export-triage-review, high priority, low priority, needs reproduction, oncall: export, oncall: pt2, triaged",thread:issue:111693:issue:112043,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#111693 [export] 14k models: AssertionError: graph-captured input # 2, of type <class 'torch.nn.parameter.Parameter'>, is not among original inputs of types, #112043 AssertionError: graph-captured input #1, of type <class 'torch.Tensor'>, is not among original inputs of types…",2b4e04c5e1d18981e6184c3924f8cf8d97fc0451f3ee8fd0f8920f11cb993afa
issue-collapse:caf007c6a54149e9,Review thread: 2 issues - issue #109923: Import order issue with torch and pybind11 Library Statically Linked to libstdc++,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"2575, 109923",,,,"module: abi, module: assert failure, module: build, module: crash, module: static linking","has workaround, module: abi, module: assert failure, module: build, module: crash, module: static linking, triaged",thread:issue:109923:issue:2575,Handle this module: abi stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#2575 ImportError: dlopen: cannot load any more object with static TLS, #109923 Import order issue with torch and pybind11 Library Statically Linked to libstdc++",84842b385ecbd79730812f12106805b9df3d479587f5c894e495783a0f7792a7
issue-collapse:6c257ce7254ac32e,Review thread: 2 issues - issue #105494: workaround for using vmap when .item() is being used internally,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"105494, 124423",,,,"module: functorch, module: vmap","actionable, has workaround, high priority, module: functorch, module: vmap, triaged",thread:issue:105494:issue:124423,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#105494 workaround for using vmap when .item() is being used internally, #124423 `__getitem__` fails to vmap for one dimensional tensors",c30d1ca8bd63c5d370d4fb9b32a82e15bc3f6f286af93327918ac186a5d9b40d
issue-collapse:198dd85479e53761,Review thread: 2 issues - issue #102963: torch.svd fails on large matrices,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"102963, 166004",,,,"module: cublas, module: cuda, module: linear algebra, module: regression","module: cublas, module: cuda, module: linear algebra, module: regression, triaged",thread:issue:102963:issue:166004,Handle this module: cublas stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#102963 torch.svd fails on large matrices, #166004 [CUDA] torch.linalg.eigh input size reduced since PyTorch 2.8.0",aade6ee66b3a63c04c272e1fd7f15ca0cc208f044e92d907245b176e786f3dcf
issue-collapse:ddfd954396a73596,Review thread: 2 issues - issue #102533: Segfault when running vulkan program linked against libtorch,related issue thread,stale/low urgency,52,high,work_thread_graph,2,0,0,0,"102533, 103756",,,,"module: cpp, module: crash, module: vulkan","ciflow/periodic, high priority, module: cpp, module: crash, module: vulkan, triaged",thread:issue:102533:issue:103756,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#102533 Segfault when running vulkan program linked against libtorch, #103756 libtorch > 1.9.1 produces segfault on Qt5 gui application exit",11e4fd79539b5d8ad6c92dd228200f5a4c6aa4553601d190132fb997098a536a
issue-collapse:a101965a80abcd86,"Review thread: 15 issues - issue #104247: [proposal] ""Name"" string attribute for modules, parameters, buffers, tensors for more pleasant debugging (including distributed FSDP2 logging)",related issue thread,needs design decision,47,high,work_thread_graph,15,0,0,0,"1529, 16668, 25267, 27479, 30049, 31252, 31353, 35633, 84321, 85652, 85698, 91692, 104247, 167656, 168976",,,,"module: amp (automated mixed precision), module: autograd, module: bootcamp, module: cuda, module: cudacachingallocator, module: internals, module: memory usage, module: performance, module: pt2-dispatcher, oncall: distributed, oncall: jit, oncall: profiler, oncall: pt2","actionable, feature, good first issue, has workaround, low priority, module: CUDACachingAllocator, module: amp (automated mixed precision), module: autograd, module: bootcamp, module: cuda, module: internals, module: memory usage, module: performance, module: pt2-dispatcher, needs design, oncall: distributed, oncall: jit, oncall: profiler, oncall: pt2, triaged",thread:issue:104247:issue:1529:issue:16668:issue:167656,"Review this module: amp (automated mixed precision) queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,6,Represents 15 open issues; raw issue rows remain available for audit.,"#1529 [feature request] Caching allocator diagnostics and memory allocation tracing/visualization, #16668 Reduce fragmentation with CUDA caching allocator when using many streams, #25267 JIT leaks memory when I change the max sequence length, #27479 [JIT] Figure out how to easily investigate memory usage issues issues, #30049 CPU Memory Leak for JIT ScriptModule on DataParallel, #31252 [feature request] Better handling for CUDA Out of Memory",0ac3fa14e31dc1ffb3e8f916e23066bddeb2d61a52dcd4c07570be9c0a41dcf0
issue-collapse:85e6759d6e139fba,Review thread: 10 issues - issue #103475: [Inductor] Optimize More Cases of Int32 -> Int64,related issue thread,stale/low urgency,47,high,work_thread_graph,10,0,0,0,"26889, 40373, 43501, 44026, 46948, 54982, 68204, 103475, 106596, 110175",,,,"module: aotdispatch, module: codegen, module: inductor, module: internals, module: pt2-dispatcher, module: python frontend, oncall: jit, oncall: pt2","enhancement, feature, module: aotdispatch, module: codegen, module: inductor, module: internals, module: pt2-dispatcher, module: python frontend, needs research, oncall: jit, oncall: pt2, shadow review, triaged",thread:issue:103475:issue:106596:issue:110175:issue:26889,Handle this module: aotdispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 10 open issues; raw issue rows remain available for audit.,"#26889 Statically checked tensor shapes, #40373 [discussion] Expressing tensor dimension semantics / constraints through typing / constraints blocks. Constraints block could be scripted/traced and help for tracing/script execution and codegen, #43501 [jit] Shape hints for TorchScript, #44026 [Feature] Einsum like ShapeGuard, #46948 Record shaping assertions and use them for tracing / scripting optimization and codegen, #54982 [RFC] A PyTorch Tensor Shape DSL For Symbolic Shape Inference",d9fedfcb566687130180b7fb8decf2fef92b7d789b288cf6c875f7686d4b40d2
issue-collapse:82dd62fe03b83992,Review thread: 6 issues - issue #16293: Unexpected behavior of jit.trace when PYTORCH_JIT=0,related issue thread,needs triage,47,high,work_thread_graph,6,0,0,0,"16293, 25066, 29092, 29177, 40914, 48108",,,,oncall: jit,"TSRootCause:ModuleInheritance, TSRootCause:PyTorchParityGap, TSUsability, enhancement, jit-backlog, low priority, oncall: jit, triage review, triaged, weeks",thread:issue:16293:issue:25066:issue:29092:issue:29177,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,5,Represents 6 open issues; raw issue rows remain available for audit.,"#16293 Unexpected behavior of jit.trace when PYTORCH_JIT=0, #25066 [FR][jit] torch.jit.script as a class decorator, #29092 [FR] script returns subclass of the original module class, #29177 [FR] trace_module traces both eval and train graph, #40914 Inconsistent handling of torch.Size.__add__, #48108 JIT should not force users to write ugly code",5a866af8e5bbfe9ee63d521d9691390627072fb194262d26b88362e0b238d61b
issue-collapse:dda232e478c65c26,Review thread: 4 issues - issue #24904: tensorboard add_graph error,related issue thread,stale/low urgency,47,high,work_thread_graph,4,0,0,0,"24904, 28206, 30459, 38614",,,,"module: tensorboard, oncall: visualization","module: tensorboard, oncall: visualization, triaged",thread:issue:24904:issue:28206:issue:30459:issue:38614,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#24904 tensorboard add_graph error, #28206 torch.utils.tensorboard.SummaryWriter.add_graph do not support non-tensor inputs, #30459 [TensorBoard] Graph with objects other than torch.nn.Module can not be visualized., #38614 `SummaryWriter.add_graph` borks with simple example",ea5c0e08d462be76cc4b2b9bae9b572b4f4c60d5244bcc87da7d211b4796436c
issue-collapse:94fcb1707cf80254,Review thread: 4 issues - issue #126714: move label_to_label.yml to an issue?,related issue thread,stale/low urgency,47,high,work_thread_graph,4,0,0,0,"24422, 64492, 126714, 151579",,,,"module: bootcamp, module: devx, oncall: profiler","actionable, better-engineering, module: bootcamp, module: devx, oncall: profiler, triaged",thread:issue:126714:issue:151579:issue:24422:issue:64492,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#24422 Label tracking meta-issue (edit me to get automatically CC'ed on issues! cc bot), #64492 AttributeError: Can't pickle local object 'schedule.<locals>.schedule_fn', #126714 move label_to_label.yml to an issue?, #151579 [BE] Consolidate PR labeling logic",265c8278797ae38d8001871781a2058ecb632952d6ea76268ae027e0188f62e5
issue-collapse:0d931bc9f28e3433,"Review thread: 4 issues - issue #104702: [feature request] Specialized memory layouts and wide blocked/tiled dtypes for cublasLt (`CUBLASLT_ORDER_COL32`, `CUBLASLT_ORDER_COL4_4R2_8C`, `CUBLASLT_ORDER_COL32_2R_4R4`)/onednn: e.g. torch.float16x32 / torch.int8x32 / torch.bits1x512 (akin to torch.quint2x4)",related issue thread,stale/low urgency,47,high,work_thread_graph,4,0,0,0,"69364, 104702, 127062, 127521",,,,"module: cuda, module: memory format, oncall: quantization","feature, module: cuda, module: memory format, oncall: quantization, triaged",thread:issue:104702:issue:127062:issue:127521:issue:69364,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#69364 [feature request] `quantized::linear_dynamic` on CUDA/eager, and other quantized and low-level int8 operators (matmul, gemm etc) on CUDA + integrate LLM.int8 + integrate ZeroQuant?, #104702 [feature request] Specialized memory layouts and wide blocked/tiled dtypes for cublasLt (`CUBLASLT_ORDER_COL32`, `CUBLASLT_ORDER_COL4_4R2_8C`, `CUBLASLT_ORDER_COL32_2R_4R4`)/onednn: e.g. torch.float16x32 / torch.int8x32 / torch.bits1x512 (akin to torch.quint2x4), #127062 T5 -small Dynamic quantization in graviton3, #127521 Modeling ViT does not support quantized models",86963bd4bbc3e6e5e2dadcc331617451e49c0d51bba6941cc69140e9a94c9ce6
issue-collapse:e1937f29d9ba9d5b,Review thread: 3 issues - issue #63740: DDP Gradient reduction is not triggered after calling autograd.backward(),related issue thread,needs triage,47,high,work_thread_graph,3,0,0,0,"63740, 63812, 63929",,,,"module: c10d, module: ddp, oncall: distributed","module: c10d, module: ddp, oncall: distributed, triaged",thread:issue:63740:issue:63812:issue:63929,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",needs triage,3,Represents 3 open issues; raw issue rows remain available for audit.,"#63740 DDP Gradient reduction is not triggered after calling autograd.backward(), #63812 Calling backward with create_graph on the output of a DistributedDataParallel throws error, #63929 [Poll] Support higher-order gradient computation in DDP",689c1aa05161c168c211747ec96c5256c6dcf14bc9254bbdd599cde759adc9dd
issue-collapse:f5d8d430f21c2209,Review thread: 3 issues - issue #60953: Add support for Python's optimized mode to torchscript,related issue thread,needs triage,47,high,work_thread_graph,3,0,0,0,"60953, 76659, 77869",,,,"module: docs, module: performance, module: python frontend, oncall: jit","better-engineering, module: docs, module: performance, module: python frontend, oncall: jit, triaged",thread:issue:60953:issue:76659:issue:77869,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#60953 Add support for Python's optimized mode to torchscript, #76659 Remove all docstrings when python is running in optimization mode, #77869 Throw warning if python optimise flags are enabled",7cfe706d24775a58709e3488a51ec4043c5f1f565060023541c80e3fad9df6bc
issue-collapse:302aec76d9dde916,Review thread: 3 issues - issue #44710: AMP support for libtorch/c++,related issue thread,stale/low urgency,47,high,work_thread_graph,3,0,0,0,"44710, 72295, 89241",,,,"module: amp (automated mixed precision), oncall: jit","enhancement, module: amp (automated mixed precision), oncall: jit, triaged",thread:issue:44710:issue:72295:issue:89241,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#44710 AMP support for libtorch/c++, #72295 [JIT][Autocasting] Add autocasting & constant propagation as part of freezing, #89241 Can't use JIT modules traced with AMP autocast, with Triton Server (or any C++ environment) - freeze() issue ?",15a3fab585621e7be0e89029e586b624089a0495fc4dbbf4d284cc829d63dccf
issue-collapse:c2efe28b6b103977,Review thread: 3 issues - issue #32976: [JIT] pytorch 1.4 breaks torch.jit.script(LSTM/GRU),related issue thread,needs triage,47,high,work_thread_graph,3,0,0,0,"32976, 51835, 68559",,,,oncall: jit,"has workaround, oncall: jit, triage review, triaged",thread:issue:32976:issue:51835:issue:68559,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#32976 [JIT] pytorch 1.4 breaks torch.jit.script(LSTM/GRU), #51835 methods decorated with torch.jit._overload_method are not accessible, #68559 LibTorch -> TorchScript -> PyTorch (Python) fails with `AttributeError: 'RecursiveScriptModule' object has no attribute 'forward'`",e71def264405cd173415c91007fb3f163c4d1735918cbaffabe49f1c6f6664ae
issue-collapse:feb15ba87a884d7d,Review thread: 3 issues - issue #30812: Tensorboard graph failing to nest modules,related issue thread,needs triage,47,high,work_thread_graph,3,0,0,0,"30812, 33670, 33691",,,,"module: tensorboard, oncall: visualization","module: tensorboard, oncall: visualization, triaged",thread:issue:30812:issue:33670:issue:33691,"Review this module: tensorboard queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 3 open issues; raw issue rows remain available for audit.,"#30812 Tensorboard graph failing to nest modules, #33670 Some module info is missing in nested graph for tensorboard, #33691 Some module has incorrect scope in complex naming situations in tensorboard graph export",d6dd44c7d528432f1ba0b3d4b8dff3568ad38ca8ef075f48f61cb3db1226f390
issue-collapse:81a54ddcc32e4e9e,Review thread: 3 issues - issue #24243: hasSideEffects INTERNAL ASSERT FAILED when using .split method with JIT,related issue thread,stale/low urgency,47,high,work_thread_graph,3,0,0,0,"24243, 30138, 54133",,,,oncall: jit,"oncall: jit, triaged",thread:issue:24243:issue:30138:issue:54133,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#24243 hasSideEffects INTERNAL ASSERT FAILED when using .split method with JIT, #30138 Unable to register custom JIT Operator with AliasAnalysisKind::CONSERVATIVE, #54133 [TorchScript] kind_.is_prim() INTERNAL ASSERT FAILED",f7b2d0ede9739ea40051f435b986985c4cbfe09dbfd892a6101b1392f25c79d8
issue-collapse:c631de0715f06227,Review thread: 3 issues - issue #124572: Variant of slice that doesn't perform clamp behavior,related issue thread,stale/low urgency,47,high,work_thread_graph,3,0,0,0,"41625, 43125, 124572",,,,"module: dynamic shapes, oncall: jit","enhancement, function request, module: dynamic shapes, needs research, oncall: jit, triaged",thread:issue:124572:issue:41625:issue:43125,Handle this module: dynamic shapes stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#41625 [feature request] New function `torch.slice(...)` mirroring TorchScript op signature or add step argument to `torch.narrow`, #43125 Non-conforming variable identifiers in JIT code errors / printouts: e.g. ""Tensor 0"", #124572 Variant of slice that doesn't perform clamp behavior",b33f7c9c7ad53b8d3a7d5e350e5333b497d174e9debd360a2c1dbe15ab80f440
issue-collapse:8149aeec57c13c5f,Review thread: 3 issues - issue #100741: Reading `state_dict()` in C++,related issue thread,stale/low urgency,47,high,work_thread_graph,3,0,0,0,"36577, 67902, 100741",,,,"module: cpp, module: nn, module: serialization, oncall: jit","module: cpp, module: nn, module: serialization, oncall: jit, triaged",thread:issue:100741:issue:36577:issue:67902,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#36577 Add load_state_dict and state_dict() in C++, #67902 torch::jit::pickle_load is unable to read some files saved by torch.save, #100741 Reading `state_dict()` in C++",62e787d306ea782e930396412a8647b091ab9b854a9f86ee60f7b3055ad8965d
issue-collapse:0dbaffd717737b58,Review thread: 2 issues - issue #92910: torch.jit.save() generates different contents in a file among different endian machines,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"92910, 92987",,,,"module: power, oncall: jit","module: POWER, oncall: jit",thread:issue:92910:issue:92987,"Review this module: power queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#92910 torch.jit.save() generates different contents in a file among different endian machines, #92987 Test Failure: TestUpgraders.test_aten_div_scalar_at_3 on a big-endian machine (issue in torch.jit.load())",f1cea20f7b03458262f6228adf965747dcac643d20b8d3f308262008829b6f8f
issue-collapse:cf0331a93ca8612f,Review thread: 2 issues - issue #90366: [JIT] INTERNAL ASSERT FAILED `torch.mul` with boolean primitive constant,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"90366, 90367",,,,oncall: jit,oncall: jit,thread:issue:90366:issue:90367,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#90366 [JIT] INTERNAL ASSERT FAILED `torch.mul` with boolean primitive constant, #90367 [JIT] INTERNAL ASSERT FAILED `torch.add` with boolean primitive constant",23714baf89f96fdffe3aef87f5083e6161929eab5540f0a9d5bab0832a77daf7
issue-collapse:935a2b929713ad7b,Review thread: 2 issues - issue #86791: We don't have an op for vulkan_prepack::conv2d_clamp_prepack but it isn't a special case.,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"86791, 95124",,,,"module: build, module: convolution, module: docker, module: vulkan, oncall: mobile","module: build, module: convolution, module: docker, module: vulkan, oncall: mobile, triaged",thread:issue:86791:issue:95124,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#86791 We don't have an op for vulkan_prepack::conv2d_clamp_prepack but it isn't a special case., #95124 `INTERNAL ASSERT FAILED` -When using the PyTorch docker environment released by pytorch, a Vulcan support issue occurs",7b6cc24f3db675973cf6da9ecc0c01b9059e656761ec41d13e2968e1947ba9a6
issue-collapse:7595d91b031d7104,Review thread: 2 issues - issue #85877: JIT model could return 'NaN' gradient after the first execution,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"85877, 86804",,,,oncall: jit,oncall: jit,thread:issue:85877:issue:86804,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#85877 JIT model could return 'NaN' gradient after the first execution, #86804 JIT model will have a different jacobian after the first computation",c65ee46a05a33f4cfbbabb7455d4b3bd7a939af26772fdaa94df38d05c4f2ee3
issue-collapse:4ef19b6ba068cc2d,Review thread: 2 issues - issue #80561: Cannot run scripted BERT_Pytorch,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"80561, 85792",,,,oncall: jit,oncall: jit,thread:issue:80561:issue:85792,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#80561 Cannot run scripted BERT_Pytorch, #85792 TorchScript causes range_check error after a few iterations of forward-backward passes",efff5d4420182bccc025eb61b1314921df8b7d9d434c5f72db1523a6491c2ced
issue-collapse:4f4c06a5db50fcfd,Review thread: 2 issues - issue #79715: [FX] TypeError when tracing cat taking split's output as input,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"79715, 91293",,,,"module: fx, oncall: fx","module: fx, oncall: fx, triaged",thread:issue:79715:issue:91293,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#79715 [FX] TypeError when tracing cat taking split's output as input, #91293 torch.fx tracer emits type error when tracing module that directly contains and uses the torch.cat() function",f4dd50624bf0c780560cdbc11f68f0bb67b0b618b12c675b159daab582561340
issue-collapse:93e915cd1fbd8629,Review thread: 2 issues - issue #76578: [JIT][Autocast] Batchnorm folding pass during freezing doesn't preserve types,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"76578, 88448",,,,oncall: jit,oncall: jit,thread:issue:76578:issue:88448,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76578 [JIT][Autocast] Batchnorm folding pass during freezing doesn't preserve types, #88448 linear mm weight and bias dtypes mismatch bypasses",3df88596a5bfae199569001cee4e8b8868dd95ebc68a458f14c688a67123f265
issue-collapse:715d86520e4238b1,Review thread: 2 issues - issue #75336: Libtorch Crash after Attempting to Evaluate Model after Opening an OpenGL Context,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"75336, 96060",,,,"module: binaries, oncall: jit","NNC, module: binaries, oncall: jit, triaged",thread:issue:75336:issue:96060,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#75336 Libtorch Crash after Attempting to Evaluate Model after Opening an OpenGL Context, #96060 Linking libtorch with QT5 OpenGL application using llvmpipe mesa opengl crashes",3190f78b9e6a1ac08021f549d6be03a0e6cb8d7ad56ee736a09f24d0e8594254
issue-collapse:400931495bdd133f,Review thread: 2 issues - issue #71470: torch.jit.script failed to compile nn.MultiheadAttention when specifying the kdim and vdim parameters.,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"71470, 92073",,,,oncall: jit,oncall: jit,thread:issue:71470:issue:92073,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71470 torch.jit.script failed to compile nn.MultiheadAttention when specifying the kdim and vdim parameters., #92073 compilig MultiHeadAttention",3a9ee0449c0595541b2d2bffbabf541b82726131f09668799e2e1cdf0239b84b
issue-collapse:87102af052d03552,Review thread: 2 issues - issue #69741: JIT: Support for `torch.autograd.functional.jacobian` in TorchScript,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"69741, 69744",,,,"module: autograd, oncall: jit","Stale, enhancement, module: autograd, oncall: jit, triaged",thread:issue:69741:issue:69744,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69741 JIT: Support for `torch.autograd.functional.jacobian` in TorchScript, #69744 Add ability to ignore arguments/outputs in `torch.autograd.functional.jacobian`",552885f827f144fddd05abe2245bb07c41eef38dfeadeb726f7dfc283677cc20
issue-collapse:7bf77258f6e75da2,Review thread: 2 issues - issue #69033: Lazy Tensor Core Documentation Out-of-Date,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"69033, 72112",,,,"module: docs, module: lazy, module: xla, oncall: jit","module: docs, module: lazy, module: xla, oncall: jit, triaged",thread:issue:69033:issue:72112,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69033 Lazy Tensor Core Documentation Out-of-Date, #72112 Unable to parse IR generated by Lazy Tensor Core with TorchScript Backend",6caeb87d7786f45601e05241a5f70d7c255077959f24c06ad27a57ef2dac234c
issue-collapse:4ab3939f35202457,Review thread: 2 issues - issue #68385: ImportError: cannot import name 'ProcessGroup' from 'torch.distributed',related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"68385, 83271",,,,"module: binaries, module: macos, oncall: distributed","module: binaries, module: macos, oncall: distributed, triaged",thread:issue:68385:issue:83271,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#68385 ImportError: cannot import name 'ProcessGroup' from 'torch.distributed', #83271 cannot import name 'ProcessGroup' from 'torch.distributed'",3b21f57506b78645790a6c9f45ac59acda6f43f4f9ec10d9292c4195f5af403f
issue-collapse:00f004a2933c8893,Review thread: 2 issues - issue #67955: Torch 1.9.1.post3 - Error in magma_getdevice_arch: MAGMA not initialized (call magma_init() first) or bad device,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"67955, 68340",,,,oncall: jit,"oncall: jit, triaged",thread:issue:67955:issue:68340,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#67955 Torch 1.9.1.post3 - Error in magma_getdevice_arch: MAGMA not initialized (call magma_init() first) or bad device, #68340 Torch 1.10.0 - RuntimeError: requires_grad_ is not supported on ScriptModules",e10d9e4ef59dea7aa5a3ecfa0df99cb4ed96917adf11f1893fddfc0c8988e7af
issue-collapse:782d3808857682e9,Review thread: 2 issues - issue #67570: RFC: Overlap optimizer computation with DDP/FSDP backward,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"67570, 71595",,,,"module: ddp, oncall: distributed","module: ddp, oncall: distributed, triaged",thread:issue:67570:issue:71595,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#67570 RFC: Overlap optimizer computation with DDP/FSDP backward, #71595 Optimizer Overlap: Follow up features",82f33639dbbcd651bf8691e69557fdef78a5a76828818fe3c1bcc73bb5fba565
issue-collapse:76733e3b358522db,"Review thread: 2 issues - issue #61693: Pytorch Mobile on Android, LAPACK library not found in compilation",related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"61693, 68871",,,,"module: android, module: ios, oncall: mobile","enhancement, module: android, module: ios, oncall: mobile",thread:issue:61693:issue:68871,"Review this module: android queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#61693 Pytorch Mobile on Android, LAPACK library not found in compilation, #68871 [LibTorch-Lite] Add a custom flag to build LibTorch-Lite with LAPACK included",96a159974749db9c884e86e675773d85f5bf8384f4a4541cf451117a97c90d3d
issue-collapse:240cf4c45952f3a7,Review thread: 2 issues - issue #55757: Synchronize RRef.to_here() CUDA Streams properly when the profiler is enabled,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"55757, 56244",,,,"module: rpc, module: tensorpipe, oncall: distributed","module: rpc, module: tensorpipe, oncall: distributed, triaged",thread:issue:55757:issue:56244,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#55757 Synchronize RRef.to_here() CUDA Streams properly when the profiler is enabled, #56244 Cuda RPC error when using then()",c7f781d2800f4e2c559b0eb3f92019e930ad167b29146d28621c29a7212c8986
issue-collapse:e26bcb5713afbed6,Review thread: 2 issues - issue #55571: Update TCPStore.wait() error message to be more friendly,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"55571, 66627",,,,oncall: distributed,"better-engineering, oncall: distributed, triaged",thread:issue:55571:issue:66627,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#55571 Update TCPStore.wait() error message to be more friendly, #66627 Overhaul error handling in `TCPStore`",ebc1d67aade26b027a6c7dcefaf7271ff101074a4b9f413d86fdc3956694ee74
issue-collapse:63eac9deddc8782b,Review thread: 2 issues - issue #52788: distributed/optim/test_zero_redundancy_optimizer uses unittest.main() instead of common_utils.run_tests(),related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"52788, 52800",,,,"module: ci, oncall: distributed","enhancement, module: ci, oncall: distributed, triaged",thread:issue:52788:issue:52800,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",needs triage,2,Represents 2 open issues; raw issue rows remain available for audit.,"#52788 distributed/optim/test_zero_redundancy_optimizer uses unittest.main() instead of common_utils.run_tests(), #52800 Create CI test checker to ensure # of test ran is the same as # of test reported in xmlrunner",3c1da494d9148a97da87749224b99e755b1e20feb52e080a446d8209f489f9f9
issue-collapse:09166d494d1805e7,Review thread: 2 issues - issue #51018: len of ModuleList is incorrect for jitted models,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"51018, 60333",,,,oncall: jit,oncall: jit,thread:issue:51018:issue:60333,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#51018 len of ModuleList is incorrect for jitted models, #60333 Two-element ModuleList results in error in inference_mode when jit'ed",756848a8804f4a288f67163b394b14d875c1e951134a2179aef56f033c25d4a1
issue-collapse:9cf7f1ac56eaf481,Review thread: 2 issues - issue #50533: Non-avairable nightly version for converting models to NNAPI,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"50533, 50678",,,,oncall: mobile,oncall: mobile,thread:issue:50533:issue:50678,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#50533 Non-avairable nightly version for converting models to NNAPI, #50678 [PyTorch Mobile] NNAPI prototype installer no longer available",c664a7e50acb7697942db98e63df8a972842f4a91e54c6667a05dc8277dd4a4c
issue-collapse:d808aa2cce4d1522,Review thread: 2 issues - issue #48932: Unknown builtin op: torchvision::nms when loading scripted FasterRCNN,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"48932, 71359",,,,"module: build, oncall: jit","module: build, oncall: jit, triaged",thread:issue:48932:issue:71359,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#48932 Unknown builtin op: torchvision::nms when loading scripted FasterRCNN, #71359 Torchvision Installation Logic Python vs C++ ""Mismatch""",3f73c97f532e1500c37399f2f75c7f8e19ab3b1e3032a2b871fa2bb9e585e7d3
issue-collapse:31318dd3a5b376de,Review thread: 2 issues - issue #47964: TorchScript is unable to distinguish int and ScalarType,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"47964, 90365",,,,oncall: jit,"TSRootCause:DefaultTypes, TSUsability, oncall: jit, weeks",thread:issue:47964:issue:90365,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47964 TorchScript is unable to distinguish int and ScalarType, #90365 [JIT] INTERNAL ASSERT FAILED when dispatching for `torch.Tensor.view`",d19d65a556c5a5002bdcaea8bd5bfbbbb33e1ee0ae63bc3bfd9bfbd0edc2482d
issue-collapse:5e355f040a9b01b8,Review thread: 2 issues - issue #45465: torchscript model size increases when loading and saving it again,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"45465, 46749",,,,"oncall: jit, oncall: quantization","oncall: jit, oncall: quantization, triaged",thread:issue:45465:issue:46749,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#45465 torchscript model size increases when loading and saving it again, #46749 Quantization - we need a better solution for tracking quantization backend settings in a model",6927694bc40af7f74f9cb23de790570e2bfd2917341f6bdaf3421731e8651368
issue-collapse:5205eb66ce6f9542,Review thread: 2 issues - issue #44768: [JIT] function decorated with @torch.no_grad() can not be exported when another function is called inside it.,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"44768, 45208",,,,oncall: jit,"days, oncall: jit",thread:issue:44768:issue:45208,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#44768 [JIT] function decorated with @torch.no_grad() can not be exported when another function is called inside it., #45208 Adding @torch.no_grad() to forward () causes undefined value in torch.jit.script()",3493c380ca4faea3161c2085f5ef38a5f625f814021b0ef448d3916ad0f2630e
issue-collapse:819981ffa997bc84,Review thread: 2 issues - issue #42847: Get test_jit.py below 10k lines,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"42847, 63085",,,,"module: nn, module: tests, oncall: jit","better-engineering, enhancement, module: nn, module: tests, oncall: jit, triaged",thread:issue:42847:issue:63085,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#42847 Get test_jit.py below 10k lines, #63085 Split up test_nn.py",9aaf362239b89eefa4f788a55a72b4197fa5f6b252f4deab154bf6c7b5691c97
issue-collapse:b330b432c47a1652,Review thread: 2 issues - issue #42355: nn::Sequential Link error in Android version,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"42355, 42427",,,,"module: android, oncall: mobile","feature, module: android, oncall: mobile, triaged",thread:issue:42355:issue:42427,"Review this module: android queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#42355 nn::Sequential Link error in Android version, #42427 ./scripts/build_android.sh need support nn::module nn::Functional nn::Linear",72c30c16fc7dbb33e27089466fa16aeaced4828bfd412d0ae0c3652135f78567
issue-collapse:45d7a555d742b95d,Review thread: 2 issues - issue #32651: [Tensorboard] Problem with subfolders from SummaryWriter,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"32651, 63725",,,,"module: tensorboard, oncall: visualization","module: tensorboard, oncall: visualization, triaged",thread:issue:32651:issue:63725,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#32651 [Tensorboard] Problem with subfolders from SummaryWriter, #63725 [Feature Request] add graph to hparams for tensorboard",a62ea01a05eb0648d04bcb1ccaf132cf839f153302af7537dd397acbe0cd8cfd
issue-collapse:50ac2346a8486a6f,Review thread: 2 issues - issue #31772: Support Python builtins on iterators in JIT,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"31772, 71548",,,,oncall: jit,"oncall: jit, triaged",thread:issue:31772:issue:71548,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#31772 Support Python builtins on iterators in JIT, #71548 Support freezing of models containing ModuleContainerIndex",7ecdd33e40374723930381d9eedf0ce3899ba61cf456ea11f79dd497d59b134e
issue-collapse:620bad481ab682e9,Review thread: 2 issues - issue #30786: C++ / JIT Parity for ops in `torch/functional.py` and `torch/tensor.py`,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"30786, 48549",,,,oncall: jit,"TSRootCause:PyTorchParityGap, TSUsability, days, oncall: jit, triaged",thread:issue:30786:issue:48549,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30786 C++ / JIT Parity for ops in `torch/functional.py` and `torch/tensor.py`, #48549 JIT script has wrong signature for .norm",0e7b15ab5baeabc924c5b77abeaf8ccb1fe075fd17bfd4659545bf22c6e31e8e
issue-collapse:d2b737cb13576a12,Review thread: 2 issues - issue #27504: [JIT] List Comprehensions With Ifs not Supported,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"27504, 29843",,,,oncall: jit,"feature, oncall: jit, triage review, triaged",thread:issue:27504:issue:29843,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#27504 [JIT] List Comprehensions With Ifs not Supported, #29843 torch.distributions.normal.Normal is not JIT supported",f1504500da876d44161a89e5dd7e4734e411cb777f5cbe82d8d4c49f2b649b22
issue-collapse:befedd1befd95855,Review thread: 2 issues - issue #22155: [jit] Optional type refinement on non-named expressions,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"22155, 44077",,,,oncall: jit,"TSRootCause:TypeRefinement, TSUsability, months, oncall: jit, triaged, weeks",thread:issue:22155:issue:44077,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#22155 [jit] Optional type refinement on non-named expressions, #44077 [jit] Better type refinement for class attributes which are class types",1835baf81852729b4f9ef6bd3d0304cae015857666b7a35ade5a6c79a746bb24
issue-collapse:f1f2d1fb305ac7b2,Review thread: 2 issues - issue #19053: RNN module weights not compacted,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"19053, 42350",,,,"module: cuda, module: rnn, oncall: jit","module: cuda, module: rnn, oncall: jit, triaged",thread:issue:19053:issue:42350,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#19053 RNN module weights not compacted, #42350 An error occurred deploying CRNN using libtorch",e668b91b5ac7fb62ffcf8c91cc2aec080c2c84ff4a92b5d590e6f936710f2a85
issue-collapse:695740227e17b3ea,Review thread: 2 issues - issue #162365: [Windows][CPU] 3 UT of test/inductor/test_torchinductor.py failed with 0903 nightly whl on windows,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"162365, 162451",,,,"module: aotinductor, module: dynamo, module: windows, oncall: cpu inductor, oncall: export, oncall: pt2","export-triaged, module: aotinductor, module: dynamo, module: windows, oncall: cpu inductor, oncall: export, oncall: pt2",thread:issue:162365:issue:162451,"Review this module: aotinductor queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#162365 [Windows][CPU] 3 UT of test/inductor/test_torchinductor.py failed with 0903 nightly whl on windows, #162451 Some Windows AOTI UTs regression",5ce69801f88d726a994a175fc5bfe8a47a3c45a7448041685724b4b2d034ef35
issue-collapse:c0b79f64963f2480,Review thread: 2 issues - issue #15421: JIT is not compatible with data parallel,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"15421, 30635",,,,"module: data parallel, oncall: jit","module: data parallel, oncall: jit, triaged",thread:issue:15421:issue:30635,"Review this module: data parallel queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#15421 JIT is not compatible with data parallel, #30635 Cannot use torch.jit.script with nn.DataParallel",5c951c319e43db94b5fe6af50be622e810d722a046dd9793672f8ea017680e6b
issue-collapse:426e8f495875ab21,Review thread: 2 issues - issue #148748: export deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B failed,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"148748, 148946",,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2, triaged",thread:issue:148748:issue:148946,Handle this oncall: export stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#148748 export deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B failed, #148946 export deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B failed",19d8575dd56a590358992c0ded5317f99f4f848c2bd246bb6ce115d149ef496f
issue-collapse:8fe99569f0d6f684,Review thread: 2 issues - issue #143258: Proper support for optionals in TorchScript,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"75002, 143258",,,,oncall: jit,oncall: jit,thread:issue:143258:issue:75002,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#75002 JIT autocasting fails on Optional[Tensor], #143258 Proper support for optionals in TorchScript",02407e30fa8a769c637ac4976b79e16da7794cd06dcbaf49d9f91c6c06710508
issue-collapse:b338ebb05cb74647,Review thread: 2 issues - issue #138220: Update auto_functionalized_v2 to pass view subgraphs,related issue thread,needs maintainer decision,47,high,work_thread_graph,2,0,0,0,"138220, 138221",,,,"module: functionalization, module: pt2-dispatcher, oncall: export, oncall: pt2","better-engineering, module: functionalization, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",thread:issue:138220:issue:138221,"Review this module: functionalization queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#138220 Update auto_functionalized_v2 to pass view subgraphs, #138221 auto_functionalized_v2 doesn't support export",6bb85c8f285d38480927e7aff146ed7f339422e19273887112eadf419905c4e9
issue-collapse:e7c8996eef7979d3,"Review thread: 2 issues - issue #134191: AdamW is CPU-bottlenecked with FSDP2, with slow foreach kernels",related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"134191, 159169",,,,"module: fsdp, module: optimizer, oncall: profiler","feature, module: fsdp, module: optimizer, oncall: profiler, triaged",thread:issue:134191:issue:159169,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#134191 AdamW is CPU-bottlenecked with FSDP2, with slow foreach kernels, #159169 [DTensor][FSDP2][DDP] benchmark dtensor cpu overhead for adam optimizer",21450d9bcf114cc978235051be811c47919aba3a289e23435259f72473b1cdf6
issue-collapse:66b6984b3250d4dc,"Review thread: 2 issues - issue #131452: JIT load traced model with ""Expression of type"" error",related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"131452, 137252",,,,oncall: jit,oncall: jit,thread:issue:131452:issue:137252,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#131452 JIT load traced model with ""Expression of type"" error, #137252 RuntimeError: Expression of type - cannot be used in a type expression: __torch__.transformers_modules.code-5p-110m-embedding.modeling_codet5p_embedding.___torch_mangle_1368.CodeT5pEmbeddingModel ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ <--- HERE",3e1b99bd010bf27932f1932e724b23415125500c537d22f084d633c3731b91f9
issue-collapse:d3af625e64caf7ff,Review thread: 2 issues - issue #126276: torch script not support ModuleList/Sequential ?,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"38034, 126276",,,,"module: bootcamp, oncall: jit","large, medium, module: bootcamp, oncall: jit, triaged",thread:issue:126276:issue:38034,"Review this module: bootcamp queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38034 Support Slicing of ModuleList during JIT model tracing/scripting, #126276 torch script not support ModuleList/Sequential ?",75ba36b552ad4328fa5ca9379847f26d5ec480d1d9f2d00ef5a889067933618d
issue-collapse:8fe13083fffc57e5,Review thread: 2 issues - issue #125678: Unnecessary warning when numpy not installed,related issue thread,needs maintainer decision,47,high,work_thread_graph,2,0,0,0,"60252, 125678",,,,"module: build, module: numpy, oncall: package/deploy","imported, module: build, module: numpy, oncall: package/deploy, triaged",thread:issue:125678:issue:60252,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#60252 UserWarning: Failed to initialize NumPy: No module named 'numpy', #125678 Unnecessary warning when numpy not installed",51fcc749680bff1da13655f75de301898a3cff2fd0e063d410832bf12fddac69
issue-collapse:4af7b3b9a5fce9e2,Review thread: 2 issues - issue #125245: [export] Add customization API for unflattener to control restored module types.,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"125245, 174481",,,,"module: dynamo, module: fx, oncall: export, oncall: pt2","bot-triaged, module: dynamo, module: fx, oncall: export, oncall: pt2, triaged",thread:issue:125245:issue:174481,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#125245 [export] Add customization API for unflattener to control restored module types., #174481 Modular Dynamo FX Graph in torch 2.10",bde69c60d53cd74ea91c9a15b70bc4081e5d14ff177b5626438c7bb3d2dc0db1
issue-collapse:d505fc5b2a80fb5b,Review thread: 2 issues - issue #124461: Support constant value in torch.jit.script,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"52312, 124461",,,,oncall: jit,oncall: jit,thread:issue:124461:issue:52312,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#52312 TorchScript can't wrap Python constants in current module, #124461 Support constant value in torch.jit.script",074dfd064fbb5cd1248e0543aada66abe935087d1f31dc693d6187af6b2574df
issue-collapse:9805ed06d41bdd19,Review thread: 2 issues - issue #124275: Performance Impact of `torch.autograd.profiler.record_function` in DataLoader's `_BaseDataLoaderIter` Iteration,related issue thread,needs maintainer decision,47,high,work_thread_graph,2,0,0,0,"76410, 124275",,,,"module: dataloader, module: performance, oncall: jit","module: dataloader, module: performance, oncall: jit, triaged",thread:issue:124275:issue:76410,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76410 TorchFunction handling and overload resolution very slow in `torch.ops`, #124275 Performance Impact of `torch.autograd.profiler.record_function` in DataLoader's `_BaseDataLoaderIter` Iteration",382b8e6be640d56181c9a80b946d6befef3b0d723a67bd31808bfca003407fcd
issue-collapse:1136079ee5a078f2,Review thread: 2 issues - issue #11967: Jit cannot trace autograd for certain operator,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"11967, 63145",,,,oncall: jit,oncall: jit,thread:issue:11967:issue:63145,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#11967 Jit cannot trace autograd for certain operator, #63145 torch.jit.trace does not work if there is an autograd in the function",19df22ad38a84121e6c4535a60ea5d55796db8f625f8c45d9ab108c947dd1dfe
issue-collapse:d7553d091fd565bf,Review thread: 2 issues - issue #118462: Inconsistency in Output for `torch.nn.functional.poisson_nll_loss` in JIT Compiled Model,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"118462, 118463",,,,"module: loss, oncall: jit","module: loss, oncall: jit",thread:issue:118462:issue:118463,"Review this module: loss queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#118462 Inconsistency in Output for `torch.nn.functional.poisson_nll_loss` in JIT Compiled Model, #118463 Inconsistency in Output for `torch.argmin` in JIT Compiled Model",92b0d9e490afc90b1222b507e6d4a1f961f5349d527ded473efb909f50059b34
issue-collapse:68ec96f575001319,Review thread: 2 issues - issue #111676: [export] self.buffer += 1 raises error,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"111676, 114967",,,,"module: graph breaks, oncall: export","export-triage-review, export-triaged, module: graph breaks, oncall: export, triaged",thread:issue:111676:issue:114967,Handle this module: graph breaks stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#111676 [export] self.buffer += 1 raises error, #114967 Inplace update to buffers doesn't work with `aot_compile`",2d8688c14306d103117cb3dcb7a354a8d020fd17d12b3ea9700eba2e5b402023
issue-collapse:42f5ed824d2a1561,Review thread: 2 issues - issue #106128: Torchscript optimizer incorrectly applies constant propagation to convert prim::ListConstruct() into prim::Constant,related issue thread,needs triage,47,high,work_thread_graph,2,0,0,0,"106128, 106129",,,,oncall: jit,oncall: jit,thread:issue:106128:issue:106129,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",needs triage,1,Represents 2 open issues; raw issue rows remain available for audit.,"#106128 Torchscript optimizer incorrectly applies constant propagation to convert prim::ListConstruct() into prim::Constant, #106129 Allow settingsetGraphExecutorOptimize default for all threads",45b752b08c0e4de58008dbd65e7ff43196bbf4237732b8ec2c900173dce658cd
issue-collapse:16a85bf48f6c3505,Review thread: 2 issues - issue #105582: RFC: Integrating oneDNN Graph Compiler into Inductor C++/OpenMP Backend for Enhanced Graph Fusion and Performance,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"49444, 105582",,,,"module: inductor, oncall: cpu inductor, oncall: jit, oncall: pt2","feature, module: inductor, oncall: cpu inductor, oncall: jit, oncall: pt2, triaged",thread:issue:105582:issue:49444,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49444 [RFC] Add JIT graph fuser for oneDNN Graph API, #105582 RFC: Integrating oneDNN Graph Compiler into Inductor C++/OpenMP Backend for Enhanced Graph Fusion and Performance",2a762e3df48c7fa31be8dc70baa26d9f979c22ebf5c22b93ff156bcf7a453bf6
issue-collapse:ae132ee91ddc22e2,Review thread: 2 issues - issue #10471: [Caffe2] Multithreading in Caffe2,related issue thread,stale/low urgency,47,high,work_thread_graph,2,0,0,0,"10471, 68104",,,,oncall: distributed,"caffe2, oncall: distributed",thread:issue:10471:issue:68104,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#10471 [Caffe2] Multithreading in Caffe2, #68104 DDP is not working with Pytorch Lightning",b5e3ee95d2fc7db9febfd9a33d256e2882fd27a5672141ecb62433b9cd834cad
issue-collapse:0118d4255088216c,Review thread: 5 issues - issue #158698: [fsdp2] Possible to fuse reduce-scatter (in fp32) + div + dtype conversion (to bf16) in post-backward?,related issue thread,needs maintainer decision,44,high,work_thread_graph,5,0,0,0,"69431, 92801, 158698, 158710, 179128",,,,"module: ddp, module: fsdp, module: memory usage, module: performance, oncall: distributed, oncall: distributed parallelisms","better-engineering, bot-triaged, enhancement, feature, function request, module: ddp, module: fsdp, module: memory usage, module: performance, needs research, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:158698:issue:158710:issue:179128:issue:69431,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,4,Represents 5 open issues; raw issue rows remain available for audit.,"#69431 [feature request] `torch.to(obj, device, dtype)` supporting recursive lists/dicts/tuples of tensors probably by uplifting/promoting `torch.distributed.utils._recursive_to`, #92801 [BE] move _apply_to_tensors from FSDP to torch.distributed.utils, use in _recursive_to, #158698 [fsdp2] Possible to fuse reduce-scatter (in fp32) + div + dtype conversion (to bf16) in post-backward?, #158710 [feature request] Inplace downcast dtype conversion, #179128 FSDP2 HSDP: AllReduceState holds FP32 reduce buffers across all layers during backward",9b1c5caa7e42f31146bdef22f4a92bf54e2c3aefef766d5d52fcc724f93bbed0
issue-collapse:adbc71b76e7b8906,Review thread: 4 issues - issue #145374: Memory Leak in MPS Backend During LSTM Iterations (Out of Memory Error),related issue thread,needs maintainer decision,44,high,work_thread_graph,4,0,0,0,"145374, 154329, 155060, 164299",,,,"module: memory usage, module: mps, module: rnn","module: memory usage, module: mps, module: rnn, needs reproduction, triaged",thread:issue:145374:issue:154329:issue:155060:issue:164299,"Review this module: memory usage queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,4,Represents 4 open issues; raw issue rows remain available for audit.,"#145374 Memory Leak in MPS Backend During LSTM Iterations (Out of Memory Error), #154329 MPS Memory Leak, #155060 MPS Memory Leak, #164299 MPS Memory Leaks across core MPS files",23adbacee3c4b24fc0f5c441083b0756728af65968fb2ea09331ae20b6af37af
issue-collapse:87b0caae4e383a0d,Review thread: 3 issues - issue #109706: Make standard container classes satisfy container Protocols.,related issue thread,needs design decision,44,high,work_thread_graph,3,0,0,0,"80821, 109706, 146938",,,,"module: nn, module: typing","module: nn, module: typing, needs research, triaged",thread:issue:109706:issue:146938:issue:80821,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#80821 Add typing support to ModuleList and ModuleDict, #109706 Make standard container classes satisfy container Protocols., #146938 ModuleDict has an incomplete (or wrong) typing, producing type errors when accessed with attributes",861469b5b245cf48c3194a1b65a9dd72c3fc39302d67654e99f3249fde16cd4d
issue-collapse:c3fdf9c3f62917c9,"Review thread: 2 issues - issue #3867: [Feature Request] Implement ""same"" padding for convolution operations?",related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"3867, 80301",,,,"module: convolution, module: nn, module: padding","enhancement, feature, module: convolution, module: nn, module: padding, needs design, triaged",thread:issue:3867:issue:80301,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#3867 [Feature Request] Implement ""same"" padding for convolution operations?, #80301 Need ""valid"" and ""same"" padding mode for convTranspose2d",49f00cb253e6972ab0c9ebb5bc884537c8ae9b006e0159e7865eecd260f7f064
issue-collapse:e6d4fcc809dc3cf2,Review thread: 2 issues - issue #186824: torch.distributions.Cauchy.icdf(0) returns large finite value with wrong sign in float32; Gumbel.icdf(0/1) also finite instead of ±inf,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"186824, 186826",,,,"module: correctness (silent), module: distributions, module: edge cases, module: error checking","bot-triaged, module: correctness (silent), module: distributions, module: edge cases, module: error checking, triaged",thread:issue:186824:issue:186826,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#186824 torch.distributions.Cauchy.icdf(0) returns large finite value with wrong sign in float32; Gumbel.icdf(0/1) also finite instead of ±inf, #186826 Distribution.icdf performs no argument validation even with validate_args=True, unlike cdf/log_prob",047971a556eed364b50b20fa3f64251d15e433cd861e2a1b8d93f9101fefbf8e
issue-collapse:4e2f8330ab8f46ff,Review thread: 2 issues - issue #184361: [pipelining] PP fails when trying to pipeline any HF transformers model that uses RoPE,related issue thread,needs maintainer decision,44,high,work_thread_graph,2,0,0,0,"184361, 185780",,,,"module: pipelining, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: pipelining, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:184361:issue:185780,"Review this module: pipelining queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#184361 [pipelining] PP fails when trying to pipeline any HF transformers model that uses RoPE, #185780 [pipelining] Add downstream tests to test pipelining integration with popular HF transformers architectures",d34e342d035b92d906c580bab900e087d3f621b5a6aa60314827faaf7513106d
issue-collapse:8cfb501b01e5e910,Review thread: 2 issues - issue #182817: [MPS] Switch test_output_match / test_output_grad_match to magnitude-floored relative tolerance,related issue thread,needs maintainer decision,44,high,work_thread_graph,2,0,0,0,"182817, 183621",,,,"module: half, module: mps, module: numerical-stability, module: tests","bot-triaged, enhancement, module: half, module: mps, module: numerical-stability, module: tests, triaged",thread:issue:182817:issue:183621,"Review this module: half queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#182817 [MPS] Switch test_output_match / test_output_grad_match to magnitude-floored relative tolerance, #183621 [MPS] For accumulating ops allow the correctness tolerance to scale within the envelope of the reference implementations drift",832a4a672dbce379c6f4fadff0d2afd878425ed4ed96bf17a559761113cb9738
issue-collapse:fa552eb55d90efcc,Review thread: 2 issues - issue #180762: torch.utils.cpp_extension.load should support CUDA device linking (-dlink) similar to CUDAExtension,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"44279, 180762",,,,"module: cpp-extensions, module: cuda","bot-triaged, feature, has workaround, module: cpp-extensions, module: cuda, triaged",thread:issue:180762:issue:44279,"Review this module: cpp-extensions queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#44279 How to use CUDA Dynamic Parallelism in PyTorch CPP extension?, #180762 torch.utils.cpp_extension.load should support CUDA device linking (-dlink) similar to CUDAExtension",55b33c69e1f55c676df4d52143aa3bc928b2c881652d1cad686de3e6df6f1e8a
issue-collapse:e7113c76b65e2620,Review thread: 2 issues - issue #178593: torch.gcd returns incorrect results due to automatic dtype casting between signed and unsigned integers,related issue thread,needs maintainer decision,44,high,work_thread_graph,2,0,0,0,"178593, 187338",,,,"module: correctness (silent), module: edge cases, module: type promotion, module: unsigned int","bot-triaged, module: correctness (silent), module: edge cases, module: type promotion, module: unsigned int, triaged",thread:issue:178593:issue:187338,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#178593 torch.gcd returns incorrect results due to automatic dtype casting between signed and unsigned integers, #187338 torch.gcd returns an incorrect negative result when one input contains the minimum value of torch.int8 (-128)",02e6c22fb2cfcc255601f65dc80b3547adda07c5e3e83c74963e19d0a37e8d6b
issue-collapse:3f3cc42e450415ff,Review thread: 2 issues - issue #177827: Internal Assert Failure in `torch.cuda.memory.caching_allocator_alloc` with negative allocation size,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"177827, 182277",,,,"module: cuda, module: edge cases, module: error checking","actionable, bot-mislabeled, bot-triaged, enhancement, low priority, module: cuda, module: edge cases, module: error checking, triaged",thread:issue:177827:issue:182277,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#177827 Internal Assert Failure in `torch.cuda.memory.caching_allocator_alloc` with negative allocation size, #182277 [CUDA][CUDACachingAllocator] size overflow in caching_allocator_alloc raises SystemError for size >= 2**63",9c9cb638a0c4e73380e3321a7e871fead7cbeb96784a5df859f7f55f8c743d1e
issue-collapse:3ebe316e3b410d54,Review thread: 2 issues - issue #176296: [MPS] Binary ops on uint16/uint32/uint64 produce garbage values,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"176296, 176298",,,,"module: correctness (silent), module: cpu, module: mps, module: unsigned int","actionable, bot-mislabeled, bot-triaged, module: correctness (silent), module: cpu, module: mps, module: unsigned int, triaged",thread:issue:176296:issue:176298,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#176296 [MPS] Binary ops on uint16/uint32/uint64 produce garbage values, #176298 [CPU] Binary ops (add, sub, mul) not implemented for uint16/uint32/uint64",5b094eac8f5af1f3dfcda6727ab1644c44db05c76d7d1e5fea6d036caf9bd6b8
issue-collapse:ef240e9fa7e3f435,Review thread: 2 issues - issue #175211: CUDA/ROCm/Accelerator testing should replace get_device_capability() with feature queries,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"175211, 177969",,,,"module: accelerator, module: cuda, module: rocm, module: tests","actionable, bot-triaged, enhancement, feature, module: accelerator, module: cuda, module: rocm, module: tests, triaged",thread:issue:175211:issue:177969,"Review this module: accelerator queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#175211 CUDA/ROCm/Accelerator testing should replace get_device_capability() with feature queries, #177969 test utilities should intersect `torch.accelerator.get_device_capability` when resolving dtypes used for a test",de88796a5a4ee0f8ca6ff38651e5d26b1c6c27605df75133d9c655c82b55b3f1
issue-collapse:4caf950d9a8e92aa,Review thread: 2 issues - issue #173382: Make torch.cuda.empty_cache() actually empty the GPU cache,related issue thread,needs design decision,44,high,work_thread_graph,2,0,0,0,"173382, 184084",,,,"module: cublas, module: cuda, module: cudacachingallocator","bot-triaged, function request, has workaround, module: CUDACachingAllocator, module: cublas, module: cuda, needs research, topic: new features, triaged",thread:issue:173382:issue:184084,"Review this module: cublas queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#173382 Make torch.cuda.empty_cache() actually empty the GPU cache, #184084 Expose public API for clearing cuBLAS workspaces (currently only private torch._C._cuda_clearCublasWorkspaces)",b4ac73ae76105b8cfe8b3bc26cf018e4b2520549b90161edddfdb51332dbdf93
issue-collapse:7e5380ba50dac894,Review thread: 2 issues - issue #157431: ``torch.quantile`` edge case,related issue thread,stale/low urgency,44,high,work_thread_graph,2,0,0,0,"64947, 157431",,,,"module: edge cases, module: numpy, module: performance, module: sorting and selection","actionable, module: edge cases, module: numpy, module: performance, module: sorting and selection, triaged",thread:issue:157431:issue:64947,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#64947 Quantile is limited to 16 million elements and have poor performance., #157431 ``torch.quantile`` edge case",a82f9a2170211abaf4423c759d311cc40bceeaa7f413cc5bd0617d1f43870535
issue-collapse:26b28bd5b5cef431,"Review thread: 2 issues - issue #154356: Q.size(-1) == m INTERNAL ASSERT FAILED at ""/pytorch/aten/src/ATen/native/BatchLinearAlgebra.cpp",related issue thread,needs maintainer decision,44,high,work_thread_graph,2,0,0,0,"154356, 180377",,,,"module: correctness (silent), module: edge cases, module: error checking, module: linear algebra","actionable, bot-triaged, module: correctness (silent), module: edge cases, module: error checking, module: linear algebra, triaged",thread:issue:154356:issue:180377,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#154356 Q.size(-1) == m INTERNAL ASSERT FAILED at ""/pytorch/aten/src/ATen/native/BatchLinearAlgebra.cpp, #180377 [Silent Bug] torch.linalg.qr allows the same tensor for both Q and R in the 'out' argument when the input is square",0678b3c49373de7e117e55351f1b890bfc9c756b21a338a6b5ad8133b848a459
issue-collapse:70408fa431a8510e,Review thread: 2 issues - issue #142228: Scaled Dot-Product Attention Invalid Configuration Error on Large batch size,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"142228, 162588",,,,module: sdpa,"module: sdpa, triaged",thread:issue:142228:issue:162588,"Review this module: sdpa queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#142228 Scaled Dot-Product Attention Invalid Configuration Error on Large batch size, #162588 scaled_dot_product_attention produces incorrect results with attn_mask on large sequence lengths",0e8e4947bd3d47d2aeb25372991b0b6a535562fdefbae2b296bebf3298a23390
issue-collapse:79fe85c89b683879,Review thread: 2 issues - issue #139019: Stabilizing the Kumaraswamy Distribution,related issue thread,stale/low urgency,44,high,work_thread_graph,2,0,0,0,"139019, 173912",,,,"module: distributions, module: numerical-stability","bot-triaged, module: distributions, module: numerical-stability, triaged",thread:issue:139019:issue:173912,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#139019 Stabilizing the Kumaraswamy Distribution, #173912 Wrong mode in `torch.distributions.Kumaraswamy`",c6c7d2411b2f04a52ec03567ea3433c5ebc6ef4192c24e0ece020330ff6266cd
issue-collapse:f7f761474ca01d17,Review thread: 2 issues - issue #137635: [MPS] `torch.linspace` produces different results that CPU implementation for integral types,related issue thread,has linked PR,44,high,work_thread_graph,2,0,0,0,"137635, 170168",,,,"module: correctness (silent), module: cuda, module: edge cases, module: mps, module: python frontend","module: correctness (silent), module: cuda, module: edge cases, module: mps, module: python frontend, triaged",thread:issue:137635:issue:170168,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#137635 [MPS] `torch.linspace` produces different results that CPU implementation for integral types, #170168 torch.linspace with integer dtype gives inconsistent results on CPU vs CUDA",3de17587baec8dc526d1454d37be8c2d60a85733ba33c08d84d163c72f17b071
issue-collapse:ac95045b8b0b40c4,Review thread: 2 issues - issue #127078: Wrong Flag defined for ppc64le arch,related issue thread,needs maintainer decision,44,high,work_thread_graph,2,0,0,0,"127078, 173881",,,,"module: build, module: cpu, module: power, oncall: releng","bot-triaged, module: POWER, module: build, module: cpu, oncall: releng, triaged",thread:issue:127078:issue:173881,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#127078 Wrong Flag defined for ppc64le arch, #173881 [Build][ppc64le] Fixes required for compiling PyTorch 2.5.1 on PowerPC with GCC 12",7649659d7805df45ed57f7696b77a130a140c86b46a6f64006c44d0c1d0723c2
issue-collapse:bcd9b54df65cc4a1,Review thread: 2 issues - issue #10454: [feature request] Rank-Revealing QR - Adding dgeqp3 support to torch.qr,related issue thread,stale/low urgency,44,high,work_thread_graph,2,0,0,0,"10454, 117122",,,,"module: cuda, module: linear algebra","actionable, feature, module: cuda, module: linear algebra, todo, triaged",thread:issue:10454:issue:117122,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#10454 [feature request] Rank-Revealing QR - Adding dgeqp3 support to torch.qr, #117122 Improve behaviour of `torch.linalg.lstsq` on CUDA GPU for rank defficient matrices",d16dd7dff44031a48a5e44f6a009c304d1ce413577d9326097111c85f7c32940
issue-collapse:63adbe62d0bb635f,Review thread: 2 issues - issue #102207: skipIfTorchInductor Tracking Issue,related issue thread,good volunteer slice,44,high,work_thread_graph,2,0,0,0,"102207, 108798",,,,"module: dynamo, module: inductor, oncall: pt2","dynamo-must-fix, dynamo-symbolic-analysis, good first issue, module: dynamo, module: inductor, oncall: pt2, triaged",thread:issue:102207:issue:108798,Turn this module: dynamo slice into one contributor-ready queue with shared scope and testing expectations.,good volunteer slice,1,Represents 2 open issues; raw issue rows remain available for audit.,"#102207 skipIfTorchInductor Tracking Issue, #108798 Meta kernel not raising",6ee8417400da2c3492c9651878da4c1e90dc90471046881afd79a96eb53c7ee3
issue-collapse:8361ad3de5a6a576,Review thread: 2 issues - issue #100347: nn.MultiheadAttention doesn't use efficient scaled_dot_product_attention,related issue thread,stale/low urgency,44,high,work_thread_graph,2,0,0,0,"100347, 181725",,,,"module: docs, module: mps, module: nn, module: performance","bot-triaged, module: docs, module: mps, module: nn, module: performance, triaged",thread:issue:100347:issue:181725,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#100347 nn.MultiheadAttention doesn't use efficient scaled_dot_product_attention, #181725 [MPS] nn.MultiheadAttention is ~9x slower than direct F.scaled_dot_product_attention with bit-identical output (B=1)",f21020f3b8c443aedde378068cd676f1924172e885ed658d8d580e1a0eaa4a9a
issue-collapse:33d3f8f0c4270b66,Review thread: 5 issues - issue #108158: TORCHELASTIC_RESTART_COUNT doesn't seem to be broadcasted to all worker,related issue thread,stale/low urgency,38,high,work_thread_graph,5,0,0,0,"108158, 133877, 136312, 137014, 148048",,,,"module: c10d, module: elastic, oncall: distributed, oncall: distributed infra, oncall: r2p","bot-triaged, feature, module: c10d, module: elastic, needs design, oncall: distributed, oncall: distributed infra, oncall: r2p, ptd-bot-triaged, triaged",thread:issue:108158:issue:133877:issue:136312:issue:137014,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#108158 TORCHELASTIC_RESTART_COUNT doesn't seem to be broadcasted to all worker, #133877 Torch Distributed Elastic has no way to address non-retriable errors, #136312 Fault tolerance with k8s, #137014 [RFC] Allow user to abort group (followed by custom recovery), #148048 [RFC] Customization of ElasticAgent for fault-tolerance and node-replacement in big ddp job",fb4a13c1e0a17efc52293740057f907115e74ea7ffe797ebfc52eb82ff172284
issue-collapse:06f39f80e8b5ebf9,Review thread: 3 issues - issue #47117: ctx.save_for_backward doesn't save torch.Tensor subclasses fully,related issue thread,needs design decision,38,high,work_thread_graph,3,0,0,0,"47117, 58212, 63485",,,,"module: __torch_function__, module: autograd, module: docs","actionable, enhancement, module: __torch_function__, module: autograd, module: docs, needs design, triaged",thread:issue:47117:issue:58212:issue:63485,"Review this module: __torch_function__ queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,3,Represents 3 open issues; raw issue rows remain available for audit.,"#47117 ctx.save_for_backward doesn't save torch.Tensor subclasses fully, #58212 `torch.autograd.Function` subclasses *sometimes* throw away custom subclasses, #63485 Preserve tensor subclasses when unpacking a SavedTensor",afa84ca95993a51c7c50af3ad3ee137ac806dc31a4a13adf73e6594e84dcd79c
issue-collapse:18dd28401d208b82,Review thread: 3 issues - issue #158827: [RFC] Graph generalization,related issue thread,needs design decision,38,high,work_thread_graph,3,0,0,0,"158827, 164302, 177010",,,,"module: accelerator, module: autograd, module: cuda graphs, module: privateuse1, module: random","bot-triaged, enhancement, feature, module: PrivateUse1, module: accelerator, module: autograd, module: cuda graphs, module: random, needs research, release-feature-request, triaged",thread:issue:158827:issue:164302:issue:177010,"Review this module: accelerator queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#158827 [RFC] Graph generalization, #164302 Mixing cuda graph and non cuda graph code works poorly in the presence of autograd., #177010 [RFC] Introduce the unified RNG-related API to torch.accelerator",b89afa32a809d09a49df1baa4ed28339526e31ff804eb50ee116cef2c7479f6e
issue-collapse:f345b9f0e416ae27,Review thread: 3 issues - issue #146414: MX basic dtypes in pytorch/pytorch,related issue thread,needs design decision,38,high,work_thread_graph,3,0,0,0,"146414, 146647, 146980",,,,"module: internals, module: python frontend","enhancement, module: internals, module: python frontend, needs research, triaged",thread:issue:146414:issue:146647:issue:146980,"Review this module: internals queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,3,Represents 3 open issues; raw issue rows remain available for audit.,"#146414 MX basic dtypes in pytorch/pytorch, #146647 placeholder for shell dtype improvements in pytorch/pytorch, #146980 Feature Request: Add dlfloat16 support as a new dtype",65a58dec80245c710549453ba37c78b6e5c82ef4517b4b213b7ee567790df84d
issue-collapse:45165ba4d4b71cd7,Review thread: 2 issues - issue #160230: Offer official Pytorch Vulkan backend on pytorch.org,related issue thread,needs design decision,38,high,work_thread_graph,2,0,0,0,"160230, 187698",,,,"module: autograd, module: nn, module: vulkan, module: windows","bot-triaged, enhancement, feature, module: autograd, module: nn, module: vulkan, module: windows, needs research, triaged",thread:issue:160230:issue:187698,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#160230 Offer official Pytorch Vulkan backend on pytorch.org, #187698 Vulkan nn.Linear unavailable with requires_grad=True on Windows (inference works, training path blocked)",9b9139a86fc6890760dd9fc5292411da8aab71de974078e185cac6ed013ab9fe
issue-collapse:dadf6daba6e2f3c4,Review thread: 2 issues - issue #137574: [feature request] Provide FlexAttention as a new available/selectable backend for SDPA,related issue thread,needs maintainer decision,38,high,work_thread_graph,2,0,0,0,"137574, 170839",,,,"module: flex attention, module: nn, module: sdpa, oncall: pt2","enhancement, module: flex attention, module: nn, module: sdpa, needs design, oncall: pt2, triaged",thread:issue:137574:issue:170839,"Review this module: flex attention queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#137574 [feature request] Provide FlexAttention as a new available/selectable backend for SDPA, #170839 Varlen attention API with custom varlen `attn_mask`",59a186f45c9d497b652983a420d53c0daef3ccc7ccbe6af41fe50e96637a6e72
issue-collapse:7d077699bdd8cd06,Review thread: 2 issues - issue #117844: torch.fft.rfft: Unsupported dtype BFloat16,related issue thread,needs design decision,38,high,work_thread_graph,2,0,0,0,"95100, 117844",,,,"module: bfloat16, module: complex, module: cuda, module: fft","module: bfloat16, module: complex, module: cuda, module: fft, needs design, triaged",thread:issue:117844:issue:95100,"Review this module: bfloat16 queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#95100 RuntimeError: view_as_complex is only supported for half, float and double tensors, but got a tensor of scalar type: BFloat16, #117844 torch.fft.rfft: Unsupported dtype BFloat16",9c443009d68483bab0e94829433cd8b1cf19c7792223680b4a816be4c7245429
issue-collapse:845a018ab9a6bb4b,Review thread: 2 issues - issue #108046: Enhanced Available Backend Discovery and Selection in PyTorch 2,related issue thread,stale/low urgency,38,high,work_thread_graph,2,0,0,0,"108046, 110080",,,,module: python frontend,"enhancement, feature, module: python frontend, needs design, topic: new features, triaged",thread:issue:108046:issue:110080,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#108046 Enhanced Available Backend Discovery and Selection in PyTorch 2, #110080 Devices API",4c2c558bf99640201cc61dad86c0033b88d1aaff75cf48040265e52b4fd22846
issue-collapse:7c94e4552ef67a60,Review thread: 3 issues - issue #126654: torch.nn.functional.scaled_dot_product_attention returns NaN values after backward pass.,related issue thread,needs reproduction,32,high,work_thread_graph,3,0,0,0,"126654, 138649, 178251",,,,"module: autograd, module: cuda, module: numerical-stability, module: sdpa","bot-triaged, module: autograd, module: cuda, module: numerical-stability, module: sdpa, needs reproduction, triaged",thread:issue:126654:issue:138649:issue:178251,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,2,Represents 3 open issues; raw issue rows remain available for audit.,"#126654 torch.nn.functional.scaled_dot_product_attention returns NaN values after backward pass., #138649 SDPA 2.5 Issue tracking, #178251 RuntimeError: Function 'ScaledDotProductEfficientAttentionBackward0' returned nan values in its 0th output when using attn_mask in SDPA backward",1e6704afa00d92c935384d29633d5282c9f26949a81dc0228ffedd43b3efecd1
issue-collapse:164e6f84cb1067e5,Review thread: 2 issues - issue #146941: Make HOPs more debuggable,related issue thread,needs maintainer decision,32,high,work_thread_graph,2,0,0,0,"146941, 147115",,,,"module: higher order operators, module: pt2-dispatcher, module: tests, oncall: pt2","feature, module: higher order operators, module: pt2-dispatcher, module: tests, needs reproduction, oncall: pt2, triaged",thread:issue:146941:issue:147115,"Review this module: higher order operators queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#146941 Make HOPs more debuggable, #147115 Unable to print in a branch run by torch.cond",d442e8200d67feacc119c419ee51bcc8aa52af0b7bd485554ad731ead0f1b625
issue-collapse:d3394969214202b5,Review thread: 2 issues - issue #138422: vmap with torch.autograd.grad does not work on output of compiled function,related issue thread,needs reproduction,32,high,work_thread_graph,2,0,0,0,"138422, 171604",,,,"module: functorch, module: pt2-dispatcher, module: vmap, oncall: distributed, oncall: distributed infra, oncall: pt2","bot-triaged, module: functorch, module: pt2-dispatcher, module: vmap, needs reproduction, oncall: distributed, oncall: distributed infra, oncall: pt2, ptd-bot-triaged, triaged",thread:issue:138422:issue:171604,"Review this module: functorch queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#138422 vmap with torch.autograd.grad does not work on output of compiled function, #171604 NotImplementedError when using AsyncCollectiveTensor with torch.func.jvp",dad2a93f9268d69e6221853145909bb41d564451ffe6530421121418186a0581
issue-collapse:142b7bb1f614cf2d,Review thread: 2 issues - issue #118092: torch.load with weights_only=True show error for pickle protocol 4,related issue thread,needs reproduction,32,high,work_thread_graph,2,0,0,0,"97772, 118092",,,,module: serialization,"module: serialization, needs reproduction, triaged",thread:issue:118092:issue:97772,"Review this module: serialization queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#97772 consider bumping `DEFAULT_PROTOCOL`, #118092 torch.load with weights_only=True show error for pickle protocol 4",89018fe6a81b2bfdcfc02894fabf64f6df9151bd8f0cb507fd90fbf715609cdf
issue-collapse:6e82383f7a584fbb,"Review thread: 22 issues - issue #10440: [caffe2] enforce fail at context_gpu.cu:285, cannot get GPU memory usage statistics",related issue thread,needs maintainer decision,28,high,work_thread_graph,22,0,0,0,"6549, 6785, 6979, 7060, 7362, 7374, 7490, 7491, 7569, 7614, 7835, 8442, 8533, 8595, 9349, 9707, 9867, 10440, 10464, 10818, 11869, 128959",,,,module: build,"better-engineering, caffe2, enhancement, module: build, triaged",thread:issue:10440:issue:10464:issue:10818:issue:11869,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,6,Represents 22 open issues; raw issue rows remain available for audit.,"#6549 caffe2/cuda_rtc can throw during the destructor, #6785 [caffe2] Training and inference, #6979 [Caffe2] Is there any way to load custom operators in c++?, #7060 [feature request] Provide Caffe2 CUDA dockerfile with USE_REDIS=ON, #7362 [caffe2] Can I load the gpu data to cpu in .cu file?, #7374 [Caffe2] Can't use resnet50_trainer.py through redis.",724f334a718eea898e7928ac72543c3499fd740de5da9d838ad75148faa8dd74
issue-collapse:195ec71d9351e13f,Review thread: 7 issues - issue #168756: TestModule: test/test_linalg.py::TestLinalg,related issue thread,stale/low urgency,28,high,work_thread_graph,7,0,0,0,"168756, 168766, 168767, 168768, 168771, 168772, 174913",,,,"module: linear algebra, module: rocm","module: linear algebra, module: rocm, rocm-skipped-tests, triaged",thread:issue:168756:issue:168766:issue:168767:issue:168768,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 7 open issues; raw issue rows remain available for audit.,"#168756 TestModule: test/test_linalg.py::TestLinalg, #168766 Test: TestLinalg.test__dyn_quant_pack_4bit_weight, #168767 Test: TestLinalg.test__dyn_quant_matmul_4bit, #168768 Test: TestLinalg.test_compile_dyn_quant_matmul_4bit, #168771 Test: TestLinalg.test_ldl_solve, #168772 Test: TestLinalg.test_ck_blas_library",bb4f9deeba4bf2a6275ccd9fbb673774e9cfd2a1aa5c6488b682b1951b14aa4b
issue-collapse:62aaed07af44d215,Review thread: 7 issues - issue #129627: [RFC][DTensor] DTensor Strided Sharding: A More Flexible Way To Shard Tensors,related issue thread,stale/low urgency,28,high,work_thread_graph,7,0,0,0,"55207, 72138, 88838, 89160, 89884, 129627, 140069",,,,"module: cuda, module: dtensor, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: cuda, module: dtensor, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, sharded_tensor, triaged",thread:issue:129627:issue:140069:issue:55207:issue:72138,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 7 open issues; raw issue rows remain available for audit.,"#55207 [RFC] Model Sharding for distributed training, #72138 [RFC] PyTorch Sharder for distributed training, #88838 [RFC] PyTorch DistributedTensor, #89160 Got many TestDTensorOpsCUDA.test_dtensor_op_db_X test failures, #89884 [RFC] PyTorch Tensor Parallel(TP) User API for Distributed Training, #129627 [RFC][DTensor] DTensor Strided Sharding: A More Flexible Way To Shard Tensors",e7aa63e6de5486f801855bc74a5d574ce811958f9cb5b159ed05bf1b9a171f86
issue-collapse:c6a15e1531aa6a1e,Review thread: 5 issues - issue #168871: TestModule: test/test_transformers.py::TestSDPACudaOnly,related issue thread,needs maintainer decision,28,high,work_thread_graph,5,0,0,0,"168871, 168872, 168875, 168876, 168877",,,,"module: rocm, module: sdpa","module: rocm, module: sdpa, rocm-skipped-tests, triaged",thread:issue:168871:issue:168872:issue:168875:issue:168876,"Review this module: rocm queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,4,Represents 5 open issues; raw issue rows remain available for audit.,"#168871 TestModule: test/test_transformers.py::TestSDPACudaOnly, #168872 Test: TestSDPACudaOnly.test_cudnn_attention_broken_166211, #168875 Test: TestSDPACudaOnly.test_fused_backwards_throws_determinism_warning, #168876 Test: TestSDPACudaOnly.test_fused_kernels_nested_broadcasting, #168877 Test: TestSDPACudaOnly.test_fused_kernels_nested_broadcasting_query_dense",b932c2695aa74b840d8a659a60fba3ab5b716fa381ed834e48e6cd9c078dee72
issue-collapse:d409f2803f8aad84,Review thread: 5 issues - issue #158643: torch.compile() produces inconsistent outputs with models containing dropout and in-place ReLU operations,related issue thread,stale/low urgency,28,high,work_thread_graph,5,0,0,0,"25104, 76580, 92927, 158643, 158970",,,,"module: aotdispatch, module: higher order operators, module: nn, module: pt2-dispatcher, module: random, oncall: pt2","enhancement, feature, module: aotdispatch, module: higher order operators, module: nn, module: pt2-dispatcher, module: random, oncall: pt2, triaged",thread:issue:158643:issue:158970:issue:25104:issue:76580,Handle this module: aotdispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#25104 [FR] Dropout modules/functions should take in generator=, #76580 Random Generator for Dropout, #92927 Inplace fused (leaky)relu+(leaky)dropout for memory savings (I think, can be made fully allocation-less if never fully allocating random mask in FlashAttention style and recover the mask from the output), #158643 torch.compile() produces inconsistent outputs with models containing dropout and in-place ReLU operations, #158970 Add option to assert if kernel is not fully fused in foreach_map",314ec583ceac316ff32fa74d7197f24738890537ce807f057c5f2f520235f8b1
issue-collapse:59202e2a4a2ecd46,Review thread: 5 issues - issue #105641: Turn indexing with a scalar tensor into an copy into a view and avoid a D2H synchronization.,related issue thread,stale/low urgency,28,high,work_thread_graph,5,0,0,0,"58181, 73175, 105641, 108968, 128396",,,,"module: advanced indexing, module: bc-breaking, module: bootcamp, module: cuda, module: cuda graphs, module: docs, module: numpy, module: performance","enhancement, module: advanced indexing, module: bc-breaking, module: bootcamp, module: cuda, module: cuda graphs, module: docs, module: numpy, module: performance, topic: bc breaking, triaged",thread:issue:105641:issue:108968:issue:128396:issue:58181,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#58181 torch.split_with_sizes is not documented, #73175 [feature request] Support tensor count vector argument in torch.split, #105641 Turn indexing with a scalar tensor into an copy into a view and avoid a D2H synchronization., #108968 Unnecessary cuda synchronizations that we should remove in PyTorch, #128396 torch.prod cannot be used with cudagraphs",23a73be1d2b6c9b2e6876d15b247078bc7bd3a843baa0872b8c3d3d6bfb776a9
issue-collapse:b3145b9e709ca8ef,Review thread: 4 issues - issue #189135: [Sub-RFC] PrivateUse1 device identity/capability in torch.compile (RFC 1/3),related issue thread,needs maintainer decision,28,high,work_thread_graph,4,0,0,0,"189135, 189136, 189137, 189138",,,,"module: aotinductor, module: backend, module: dynamo, module: inductor, module: privateuse1, oncall: export, oncall: pt2","bot-triaged, enhancement, feature, large, module: PrivateUse1, module: aotinductor, module: backend, module: dynamo, module: inductor, oncall: export, oncall: pt2, triaged",thread:issue:189135:issue:189136:issue:189137:issue:189138,"Review this module: aotinductor queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#189135 [Sub-RFC] PrivateUse1 device identity/capability in torch.compile (RFC 1/3), #189136 [Sub-RFC] PrivateUse1 Dynamo tracing via registry splicing (RFC 2/3), #189137 [Sub-RFC] PrivateUse1 Inductor codegen registration interfaces (RFC 3/3), #189138 [RFC] DeviceInterface: Registry Support for PrivateUse1 Backends in torch.compile",95d091b34dde5443d14a2eed86ccea38777c2381d07b82db8a34972317ebfd73
issue-collapse:e1c519639546b3fa,Review thread: 4 issues - issue #155720: [rocm] HIP Graph capture raises segmentation fault on AMD GPU but CUDA Graph capture succeeds on Nvidia GPU,related issue thread,needs maintainer decision,28,high,work_thread_graph,4,0,0,0,"65462, 65464, 65465, 155720",,,,"module: build, module: cuda, module: lazy, module: rocm, oncall: distributed","lazy, module: build, module: cuda, module: lazy, module: rocm, oncall: distributed, triaged",thread:issue:155720:issue:65462:issue:65464:issue:65465,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#65462 Why here save `store`, rather than `prefix_store`?, #65464 linking error when building on Linux, #65465 [LTC] Fail to run testcase of latest lazy_tensor_core branch, #155720 [rocm] HIP Graph capture raises segmentation fault on AMD GPU but CUDA Graph capture succeeds on Nvidia GPU",2d137160477e7c3ed9e8a3c36dd63748aa85a5854ecf1affdfa4bf3ad0cd939f
issue-collapse:4a89b98645d92880,Review thread: 4 issues - issue #105499: [FSDP] Revisit mixed-precision casting logic,related issue thread,stale/low urgency,28,high,work_thread_graph,4,0,0,0,"105499, 111317, 111552, 112164",,,,"module: ddp, module: fsdp, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, module: ddp, module: fsdp, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, pt2d-triage-nov2024, ptd-bot-triaged, triaged",thread:issue:105499:issue:111317:issue:111552:issue:112164,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#105499 [FSDP] Revisit mixed-precision casting logic, #111317 Torch 2.1 compile + FSDP (mixed precision) + LlamaForCausalLM: `RuntimeError: attempting to assign a gradient with dtype 'c10::BFloat16' to a tensor with dtype 'float'.`, #111552 [Bug]: some parameters' grad is None when using FSDP with torch2.1.0, #112164 [Tracking] torch.compile + torch.distributed + set_grad_enabled/autocast",a19443c4a07580150f7213e9b4847d6a3a4ec39cc0e818928090660e8bbb0077
issue-collapse:f66e1fa523b2a055,Review thread: 4 issues - issue #103161: Calling jacrev with LSTM and functional_call gives error,related issue thread,stale/low urgency,28,high,work_thread_graph,4,0,0,0,"103161, 105982, 112459, 134606",,,,"module: functorch, module: vmap","module: functorch, module: vmap, triaged",thread:issue:103161:issue:105982:issue:112459:issue:134606,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#103161 Calling jacrev with LSTM and functional_call gives error, #105982 vmap and rnn/lstm ""accessing '.data' under vmap transform is not allowed"", #112459 GRUCell batching rule for vmap, #134606 VMAP over GRU: Batching rule not implemented for aten::gru.input",56d62f6df7c48cb50413326bee60892796c273d4a51211c97d78f67b6f29099f
issue-collapse:472adc14ba19a35f,Review thread: 3 issues - issue #180343: Accelerator-independent API to determine if a specific backend is available?,related issue thread,needs maintainer decision,28,high,work_thread_graph,3,0,0,0,"180343, 180345, 180712",,,,"module: accelerator, module: backend, module: privateuse1, module: ux","bot-triaged, feature, function request, module: PrivateUse1, module: accelerator, module: backend, module: ux, triaged",thread:issue:180343:issue:180345:issue:180712,"Review this module: accelerator queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#180343 Accelerator-independent API to determine if a specific backend is available?, #180345 Allow registering multiple out-of-tree backends and setting current_accelerator, #180712 Add a lazy import hook for autoloading?",814a0b8e3923e4034cb036f17d302b4747809618d011fcd51b8dc241bc169023
issue-collapse:83652b8e32b16429,Review thread: 3 issues - issue #175873: [MPS][Performance] SDPA op implementation is wasting resources by returning attention weights,related issue thread,needs maintainer decision,28,high,work_thread_graph,3,0,0,0,"175873, 176730, 179294",,,,"module: mps, module: nn, module: performance, module: sdpa","bot-mislabeled, bot-triaged, enhancement, module: mps, module: nn, module: performance, module: sdpa, triaged",thread:issue:175873:issue:176730:issue:179294,"Review this module: mps queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 3 open issues; raw issue rows remain available for audit.,"#175873 [MPS][Performance] SDPA op implementation is wasting resources by returning attention weights, #176730 scaled_dot_product_attention_math usage and perf implications of returning the second value, #179294 [MPS] `scaled_dot_product_attention` (SDPA) improvements",e719777e7ada2003c83ed7f26fccbf7ab86e0d032c3fdff7284cbe095714823d
issue-collapse:d4db116417f5640a,Review thread: 3 issues - issue #168819: TestModule: test/test_scaled_matmul_cuda.py::TestFP8Matmul,related issue thread,stale/low urgency,28,high,work_thread_graph,3,0,0,0,"168819, 168820, 168823",,,,"module: floatx (formerly float8), module: linear algebra, module: rocm","module: floatx (formerly float8), module: linear algebra, module: rocm, rocm-skipped-tests, triaged",thread:issue:168819:issue:168820:issue:168823,Handle this module: floatx (formerly float8) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#168819 TestModule: test/test_scaled_matmul_cuda.py::TestFP8Matmul, #168820 Test: TestFP8Matmul.test_scaled_mm_vs_emulated_row_wise, #168823 Test: TestFP8Matmul.test_blockwise_mxfp8_nvfp4_mxfp4_numerics",b6887554e23ac9dae131f5edd2d1b42f93ac97e1ff536a50c9b7ccbf977209bf
issue-collapse:956b28b7eae00549,Review thread: 3 issues - issue #166563: [RFC] Modifying Getting started page for Experimental Wheel Variant Support,related issue thread,needs maintainer decision,28,high,work_thread_graph,3,0,0,0,"166563, 167283, 167881",,,,"module: accelerator, module: backend, module: docs","module: accelerator, module: backend, module: docs, release-feature-request, triaged",thread:issue:166563:issue:167283:issue:167881,"Review this module: accelerator queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,3,Represents 3 open issues; raw issue rows remain available for audit.,"#166563 [RFC] Modifying Getting started page for Experimental Wheel Variant Support, #167283 [RFC] Pytorch compute platform quality levels proposal, #167881 [RFC] Modifying Getting Started Page for XPU support",cfddbd64eeb4206d53f41ed15ae0eb71bf0f48653be8d267422f42cb5e6c7597
issue-collapse:807b949f311247bd,Review thread: 3 issues - issue #144824: [MPS] Indexing Returns 0 if OOB,related issue thread,stale/low urgency,28,high,work_thread_graph,3,0,0,0,"84520, 144824, 188756",,,,"module: 64-bit, module: correctness (silent), module: embedding, module: error checking, module: mps, module: scatter & gather ops","bot-triaged, feature, module: 64-bit, module: correctness (silent), module: embedding, module: error checking, module: mps, module: scatter & gather ops, triaged",thread:issue:144824:issue:188756:issue:84520,Handle this module: 64-bit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#84520 MPS backend appears to be limited to 32 bits, #144824 [MPS] Indexing Returns 0 if OOB, #188756 [MPS] index_select / F.embedding silently return all-zero rows when the source tensor exceeds 2**31 elements (fixed on main, still broken in release/2.13)",89e52ccaae9a2a8ec5f5e61cc86ec9c1a9bed424e6b79267b80dada503eb85fc
issue-collapse:07092e8deded2ffe,Review thread: 3 issues - issue #143112: FFT half precision only let CUDA pass the check,related issue thread,stale/low urgency,28,high,work_thread_graph,3,0,0,0,"143112, 165038, 169277",,,,"module: accelerator, module: backend, module: fft, module: privateuse1, module: python frontend","module: PrivateUse1, module: accelerator, module: backend, module: fft, module: python frontend, triaged",thread:issue:143112:issue:165038:issue:169277,Handle this module: accelerator stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#143112 FFT half precision only let CUDA pass the check, #165038 Print float tensors on devices that do not support fp64, #169277 [Accelerator] Hardware native support versus Software simulated support for Device Capabilties",df531b16e26723cbcc4c74d8d11901e5344277a98abecb0c80d4773929e7922e
issue-collapse:4322ce314a5630ea,Review thread: 2 issues - issue #90544: [Dispatchable Collectives] Follow up tasks,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"90544, 90848",,,,"module: c10d, module: nccl, oncall: distributed, oncall: distributed infra","bot-triaged, module: c10d, module: nccl, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:90544:issue:90848,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#90544 [Dispatchable Collectives] Follow up tasks, #90848 [Distributed] Destruction order fiasco in ProcessGroupNCCL workCleanupLoop()",df931867192ff3ddd3a01d6c73412f7078503e8a31ef6bbdd9c3cf679ba872a0
issue-collapse:545bf117669987ad,Review thread: 2 issues - issue #74528: Supports for dist.send/dist.recv sending and recving torch.shorttensor,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"74528, 74734",,,,"module: c10d, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, function request, module: c10d, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:74528:issue:74734,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 2 open issues; raw issue rows remain available for audit.,"#74528 Supports for dist.send/dist.recv sending and recving torch.shorttensor, #74734 RuntimeError: Unconvertible NCCL type Short when sending torch.cuda.ShortTensor.",0b9ad89d44785289cf723a72fad42c48b1683805870e7aff1e8175052788940a
issue-collapse:e06fb7a58ccce91c,Review thread: 2 issues - issue #70160: NewOperatorRegistrationTest.testImplNoDefGetsCaught failed.,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"70160, 76103",,,,"module: cpp, module: dispatch, module: tests, oncall: distributed, oncall: distributed infra","bot-triaged, module: cpp, module: dispatch, module: tests, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:70160:issue:76103,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#70160 NewOperatorRegistrationTest.testImplNoDefGetsCaught failed., #76103 Some test failed when running in parallel.",1c3c30c02cf1f1cbe40977688cf9baad26c67a84d4674756f22386d0367f359d
issue-collapse:98a8059fef6cf77c,Review thread: 2 issues - issue #39242: Add numerically stable log1mexp = log(1 - exp(-|x|)) function,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"39242, 72759",,,,module: numerical-stability,"enhancement, function request, module: numerical-stability, triaged",thread:issue:39242:issue:72759,"Review this module: numerical-stability queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#39242 Add numerically stable log1mexp = log(1 - exp(-|x|)) function, #72759 Add softplus inverse",dd5bb72f0613adcfb7dadea9aa9357657cd820333b260d3fc9c44b423a6a9ece
issue-collapse:94ae1ecb04926d24,Review thread: 2 issues - issue #30458: add method to make tensor constant for debug purposes,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"30458, 44027",,,,module: autograd,"module: autograd, triaged",thread:issue:30458:issue:44027,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30458 add method to make tensor constant for debug purposes, #44027 Immutable (read-only) tensors",41a8dafc99455b82d8a268e31630159e01bd24d16431f266b0a15c0d714420c0
issue-collapse:51a09c4437e184cd,Review thread: 2 issues - issue #28224: Allow to disable polling for CUDA synchronization,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"28224, 60541",,,,"module: cuda, module: performance","enhancement, module: cuda, module: performance, triaged",thread:issue:28224:issue:60541,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#28224 Allow to disable polling for CUDA synchronization, #60541 torch.cuda.Event(blocking=True) doesn't work",d0474be1e10c75772f61df392bfb8e33128a7934543b2e2ac6c9a535b7925621
issue-collapse:3308bec53a77b7f7,Review thread: 2 issues - issue #23430: [Feature request] Let DistributedSampler take a Sampler as input,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"23430, 77154",,,,"module: dataloader, module: distributed_tool, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, has workaround, module: dataloader, module: distributed_tool, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:23430:issue:77154,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 2 open issues; raw issue rows remain available for audit.,"#23430 [Feature request] Let DistributedSampler take a Sampler as input, #77154 Distributed Weighted Sampler.",f67b26718f79d3b90e6131f17edb14767dba3f752c0f9c307a7b76412b87cee1
issue-collapse:6092228ec5e9f390,Review thread: 2 issues - issue #186825: Bernoulli(probs=0).log_prob(1) returns dtype-dependent finite values instead of -inf (Geometric special-cases this and tests it),related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"72525, 186825",,,,"module: distributions, module: edge cases","bot-triaged, module: distributions, module: edge cases, triaged",thread:issue:186825:issue:72525,"Review this module: distributions queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#72525 KL divergence between two Continuous Bernoulli is negative, #186825 Bernoulli(probs=0).log_prob(1) returns dtype-dependent finite values instead of -inf (Geometric special-cases this and tests it)",ef7a93f391175d6fd687ae34828b7aade17cfd856d47546bd1b70bfb31f2da11
issue-collapse:ce87be32f4ca6df6,Review thread: 2 issues - issue #184352: Python 3.15 support for PyTorch,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"184352, 186099",,,,"module: binaries, module: ci, module: python frontend, module: python version","bot-triaged, module: binaries, module: ci, module: python frontend, module: python version, release-feature-request, triaged",thread:issue:184352:issue:186099,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#184352 Python 3.15 support for PyTorch, #186099 Triton linux nightly wheels fail to install on Python 3.15 (requires-python: <3.15)",d9b1a9990b1ca2a82363e5976db403b12794b0c6f3a1c1230b48116b7235ea6a
issue-collapse:a8cad280f1b7b424,Review thread: 2 issues - issue #180595: capture_scalar_outputs=True crashes 7/50 HuggingFace models with PendingUnbackedSymbolNotFound,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"180595, 180596",,,,"module: dynamic shapes, module: guards, oncall: pt2","bot-triaged, module: dynamic shapes, module: guards, oncall: pt2, triaged",thread:issue:180595:issue:180596,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#180595 capture_scalar_outputs=True crashes 7/50 HuggingFace models with PendingUnbackedSymbolNotFound, #180596 capture_dynamic_output_shape_ops=True crashes 8/50 HuggingFace models with GuardOnDataDependentSymNode",388ad14da01e51b5db2141c58a2a2b6c53832703a19ad650154d11db98a2aee4
issue-collapse:9b806efa00e8b5cd,Review thread: 2 issues - issue #179008: [privateuse1] Update the privateuse1 tutorial to mention registering HooksInterface,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"179008, 179010",,,,module: privateuse1,"module: PrivateUse1, triaged",thread:issue:179008:issue:179010,"Review this module: privateuse1 queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 2 open issues; raw issue rows remain available for audit.,"#179008 [privateuse1] Update the privateuse1 tutorial to mention registering HooksInterface, #179010 [privateuse1] Update privateuse1 tutorial to show the Python backend approach",8de2f22c1d9dbf75246dd82d2eca39cce6cd41cfc135a27805d86e11f7039882
issue-collapse:308f3aba99c5d4a9,Review thread: 2 issues - issue #174496: Improve performance of torch._foreach_copy_ on CUDA tensors,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"174496, 175695",,,,"module: cuda, module: mta, module: performance, oncall: pt2","bot-mislabeled, bot-triaged, enhancement, feature, module: cuda, module: mta, module: performance, oncall: pt2, triaged",thread:issue:174496:issue:175695,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#174496 Improve performance of torch._foreach_copy_ on CUDA tensors, #175695 Multi Tensor Apply ops have low occupancy due to a maximum of 320 thread blocks per kernel launch",405489c048f1d221af105ed126781795bb7c217f70cc1ef62b2fcca7d6d6f89c
issue-collapse:9e765563e6b5a447,Review thread: 2 issues - issue #173164: allgather throw exception with discontigous input tensor,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"173164, 173362",,,,"oncall: distributed, oncall: distributed infra","bot-triaged, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:173164:issue:173362,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#173164 allgather throw exception with discontigous input tensor, #173362 `dist.reduce_scatter` throws `ValueError: Tensors must be contiguous` when the output tensor is non-contiguous (e.g., from `permute()` or `transpose()`)",e2f312b454cf36b21c71892236b34951b8e5bbb05307bd6074c2e8a90129f459
issue-collapse:0821f1e63aa2da87,Review thread: 2 issues - issue #171938: Device id in distributed process group with multiple backends,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"171938, 186548",,,,"module: xpu, oncall: distributed, oncall: distributed infra","bot-triaged, module: xpu, oncall: distributed, oncall: distributed infra, triaged",thread:issue:171938:issue:186548,"Review this module: xpu queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 2 open issues; raw issue rows remain available for audit.,"#171938 Device id in distributed process group with multiple backends, #186548 `ProcessGroupXCCL` never overrides `Backend::supportsSplitting()`, breaking `DeviceMesh._unflatten` (and any `dist.split_group` caller) on XPU",ac4c8a5d9039b2064c3a707a3b7cf8ae9bb3a73b11e27c9b34de08e2b0dc6540
issue-collapse:b60cdd31f6ddb49a,Review thread: 2 issues - issue #170426: argmax over multiple axis,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"72388, 170426",,,,"module: python frontend, module: reductions, module: ux","feature, module: python frontend, module: reductions, module: ux, triaged",thread:issue:170426:issue:72388,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#72388 ArgMax for Multi Dimensional Tensor, #170426 argmax over multiple axis",7c47b15156cfef999f5e705af9e7cba3389250f27f27e755a3960aefc23b98d9
issue-collapse:c477925a1fb16cde,Review thread: 2 issues - issue #168651: TestModule: test/inductor/test_select_algorithm.py::TestSelectAlgorithm,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"168651, 168652",,,,"module: inductor, module: rocm, oncall: pt2","module: inductor, module: rocm, oncall: pt2, rocm-skipped-tests, triaged",thread:issue:168651:issue:168652,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168651 TestModule: test/inductor/test_select_algorithm.py::TestSelectAlgorithm, #168652 Test: TestSelectAlgorithm.test_mm_dropout",b2e1febc09982981a672d232787a148621b523d76708402f12dbb991daf9b15a
issue-collapse:a9e83405308187cd,Review thread: 2 issues - issue #166628: [WinError 1114] A dynamic link library (DLL) initialization routine failed,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"166628, 169429",,,,"module: binaries, module: windows","module: binaries, module: windows, triaged",thread:issue:166628:issue:169429,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#166628 [WinError 1114] A dynamic link library (DLL) initialization routine failed, #169429 Error loading `torch\lib\torch_python.dll` due to DLL initialization failure on Windows from torch `2.9.0`",5e217017c13b86a97a4f61bc2eeca4b676ca50d66bcdbfc6a4e04d891985e0ae
issue-collapse:37e87befb42ca08b,Review thread: 2 issues - issue #164281: Document Symmetric Memory,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"164281, 173514",,,,"module: docs, module: symm_mem","module: docs, module: symm_mem, triaged",thread:issue:164281:issue:173514,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,2,Represents 2 open issues; raw issue rows remain available for audit.,"#164281 Document Symmetric Memory, #173514 Improve documentation of assumptions for symmetric memory",259dc8c1afd12a1737b59495318cb5f717dbfee71dd03ced10c0b34812c545bd
issue-collapse:8e8a196534fe5eac,Review thread: 2 issues - issue #164057: The device (Intel(R) Arc(TM) A380 Graphics) doesn't support querying the available free memory,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"164057, 164429",,,,module: xpu,"module: xpu, triaged",thread:issue:164057:issue:164429,"Review this module: xpu queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#164057 The device (Intel(R) Arc(TM) A380 Graphics) doesn't support querying the available free memory, #164429 Add support for Intel GPU Memory Information",69150833db8765efb0bcac9b735846d492bb91419aff40cd306fa56d35e67b09
issue-collapse:92f35194b9317eff,Review thread: 2 issues - issue #161381: [XPU] ​​Incorrect free_memory reported by torch.xpu.mem_get_info()​,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"161381, 164966",,,,"module: memory usage, module: xpu","module: memory usage, module: xpu, triaged",thread:issue:161381:issue:164966,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#161381 [XPU] ​​Incorrect free_memory reported by torch.xpu.mem_get_info()​, #164966 XPU OOM when allocate tensor according to its reported available memory",ee4080602d0684f5d0b77a1b197b7fdda2a55010110ee81d1b2f7aa4ee6b6c34
issue-collapse:c0085a561a89537a,Review thread: 2 issues - issue #156217: [Tracker] AutoParallel's feature request to DTensor,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"156217, 157495",,,,"module: dtensor, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: dtensor, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:156217:issue:157495,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#156217 [Tracker] AutoParallel's feature request to DTensor, #157495 [DTensor] Improve `tensor_metadata` and `redistribute_cost` coverage for op strategies.",326a05226a7b3c1238ae497cbb17ebf68a5d6ea949c62ebd4bf82a81f0e1215e
issue-collapse:462baafb9aafbdd7,Review thread: 2 issues - issue #154824: torch inductor cudagraph tree could incorrectly release some input nodes during replay,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"154824, 175557",,,,"module: cuda graphs, oncall: pt2","PT2-Bug-Bash, bot-triaged, module: cuda graphs, oncall: pt2, triaged",thread:issue:154824:issue:175557,Handle this module: cuda graphs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#154824 torch inductor cudagraph tree could incorrectly release some input nodes during replay, #175557 torch.compile crashes during autoregressive decode with growing KV cache in cudagraph_trees",48e24f082d18f23b073290d78823d9a0b941c689c24d0b4dc9b08e8615034edf
issue-collapse:6b79bd2c77762348,Review thread: 2 issues - issue #152406: [DTensor] Calling .item() on DTensor with Partial placement results in local value,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"152406, 166638",,,,"module: dtensor, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: dtensor, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",thread:issue:152406:issue:166638,Handle this module: dtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#152406 [DTensor] Calling .item() on DTensor with Partial placement results in local value, #166638 [DTensor] calling item on dtensor should perform full tensor first",86e2347ac2d671c552d8eccf392825e4b997a14109add3353691b616e478a46e
issue-collapse:12ee2b545a090a2e,Review thread: 2 issues - issue #151549: multiprocessing.pool.MaybeEncodingError Reason: 'RuntimeError('_share_fd_: only available on CPU')',related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"151549, 170636",,,,"module: multiprocessing, module: xpu","module: multiprocessing, module: xpu, triaged",thread:issue:151549:issue:170636,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#151549 multiprocessing.pool.MaybeEncodingError Reason: 'RuntimeError('_share_fd_: only available on CPU')', #170636 [xpu] No XPU implementation of multiprocessing tensor reduction.",3b54eb13065e656dc637e73ba2e605eb20e34e0f35b9af18a0bd5f563bccff3a
issue-collapse:f8d1af4dfe15ec8f,Review thread: 2 issues - issue #149953: `RuntimeError: UR error` with XPU,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"149953, 153996",,,,"module: binaries, module: windows, module: xpu","module: binaries, module: windows, module: xpu, triaged",thread:issue:149953:issue:153996,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#149953 `RuntimeError: UR error` with XPU, #153996 UR Error when calling grid_sample",80bcd0d7f48bd2a1b535983115b5020115137a489fc9974c008d91a46e1408bd
issue-collapse:b63c7e4cab6f3943,Review thread: 2 issues - issue #146108: [associative_scan] Support lifted arguments for inductor,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"146108, 146111",,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","feature, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:146108:issue:146111,Handle this module: higher order operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#146108 [associative_scan] Support lifted arguments for inductor, #146111 [scan] Support lowering and lifted arguments for inductor",a7e15a75dbbf0730d57530d98f63e05f79513855358ab4aa31633fcf4ec58459
issue-collapse:13c9e08c6f6290d6,Review thread: 2 issues - issue #144040: Torch.sparse.mm failing gradient computation at half precision.,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"144040, 164666",,,,"module: bfloat16, module: half, module: sparse","actionable, module: bfloat16, module: half, module: sparse, triaged",thread:issue:144040:issue:164666,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#144040 Torch.sparse.mm failing gradient computation at half precision., #164666 sparse.mm(CSR x dense) backward fails on CUDA for bfloat16 but CSR @ dense backward works",1927dea863d720109daccbf7f1d548dcd1ce06e1674967d7ed929d3cb37bab11
issue-collapse:58225a799be48ed6,Review thread: 2 issues - issue #142315: Better fusion on triton templates,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"142315, 151705",,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged",thread:issue:142315:issue:151705,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#142315 Better fusion on triton templates, #151705 Enable TorchInductor to Generate Matmuls Natively via `tl.dot`",18ad4701c900328ced3559aba031d8a9688ab7453365e99240c204c24140fc0b
issue-collapse:d21b5c08d9520090,Review thread: 2 issues - issue #140556: DISABLED test_backward_prod_cuda_float32 (__main__.TestNestedTensorOpInfoCUDA),related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"140556, 189075",,,,"module: ci, module: flaky-tests, module: nestedtensor","bot-triaged, module: ci, module: flaky-tests, module: nestedtensor, skipped, triaged",thread:issue:140556:issue:189075,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#140556 DISABLED test_backward_prod_cuda_float32 (__main__.TestNestedTensorOpInfoCUDA), #189075 periodic cuda13.0 slow-gradcheck: default test shards time out (~270 min) before finishing -- most shards killed",7280ea1cec1c3fa9c800db52f3adefe507073f300ea0e6564965f9398c97bc5f
issue-collapse:096ae484cd9a50e2,Review thread: 2 issues - issue #137564: Implement Incremental PCA with GPU support,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"40770, 137564",,,,module: cuda,"feature, function request, module: cuda, triaged",thread:issue:137564:issue:40770,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40770 Incremental version of pca_lowrank, #137564 Implement Incremental PCA with GPU support",7b31fca8b4e5963b0ce85ee08b7f17f8cc855c0f915ab6052ead04af1be4204b
issue-collapse:e3671ef6529d255e,Review thread: 2 issues - issue #136003: [MPS] Inconsistent performance issues,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"136003, 140706",,,,"module: mps, module: performance","module: mps, module: performance, triaged",thread:issue:136003:issue:140706,"Review this module: mps queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#136003 [MPS] Inconsistent performance issues, #140706 MPS random number generation is slow, if not hanging forever",7a8900c658c29b6192635fd442b85fdde148b47c2aa3399b7e83a6add44c2072
issue-collapse:6e7c54cb58373d08,Review thread: 2 issues - issue #123447: dist.barrier() hangs after calling async_save,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"95895, 123447",,,,"module: c10d, oncall: distributed, oncall: distributed infra","bot-triaged, has workaround, module: c10d, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, release notes: distributed (checkpoint), triaged",thread:issue:123447:issue:95895,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#95895 [PTD] dist.barrier() unreliable when using collectives from multiple threads., #123447 dist.barrier() hangs after calling async_save",64c5edb4801a331e7858bf856e5dce17982aa0d387b6bdfda52a91f2755ca751
issue-collapse:b2e2a976cba0ccd0,Review thread: 2 issues - issue #122840: a log_softmax kernel get much worse perf with padding,related issue thread,has linked PR,28,high,work_thread_graph,2,0,0,0,"122840, 148402",,,,"module: inductor, oncall: pt2","feature, module: inductor, oncall: pt2, triaged",thread:issue:122840:issue:148402,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#122840 a log_softmax kernel get much worse perf with padding, #148402 Generate two reduction loops for vectorization",6a6b78100f59ea3abd49cb794c31c156e7f089ad1c80605b040732443f0c50f0
issue-collapse:2cf400261e9446fe,Review thread: 2 issues - issue #122540: [RFC] Solving Warm Start in torch.compile,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"122540, 125958",,,,"module: aotdispatch, module: compile-time, module: dynamo, module: faketensor, module: inductor, oncall: pt2","feature, module: aotdispatch, module: compile-time, module: dynamo, module: fakeTensor, module: inductor, oncall: pt2, triaged",thread:issue:122540:issue:125958,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#122540 [RFC] Solving Warm Start in torch.compile, #125958 Support AOT Autograd level Caching",925756688a845fab647723b65670d2fb7850efe87f4fd470a7026530d33060ec
issue-collapse:7f9ce5694f7febef,Review thread: 2 issues - issue #119574: Cannot configure dist_timeout when using device_mesh,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"119574, 119613",,,,"module: devicemesh, oncall: distributed","module: DeviceMesh, oncall: distributed, triaged",thread:issue:119574:issue:119613,Handle this module: devicemesh stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#119574 Cannot configure dist_timeout when using device_mesh, #119613 init_device_mesh doesn't have a timeout kwarg",baa772fd0f46c47073495d2beb9676d4b1bff4d73c9bb7b9af62fd1f72a937ea
issue-collapse:7c495fe7e6c5b531,Review thread: 2 issues - issue #118492: `torch.compile(dynamic=True)` compiles forever,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"118492, 118798",,,,"module: compile-time, module: dynamic shapes, module: performance, oncall: pt2","ezyang's list, module: compile-time, module: dynamic shapes, module: performance, oncall: pt2, triaged",thread:issue:118492:issue:118798,"Review this module: compile-time queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#118492 `torch.compile(dynamic=True)` compiles forever, #118798 Automatic dynamic shapes should give up if a symbolic variable has too many guards on it",60e1e42aeec2cbe3a2dcc221745cd502bd4743af35a82cffc7d75bebdb0db871
issue-collapse:84b121992792925b,Review thread: 2 issues - issue #115305: torchrun to support NUMA affiinity,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"115305, 152822",,,,"module: distributed_tool, oncall: distributed, oncall: distributed infra","bot-triaged, module: distributed_tool, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",thread:issue:115305:issue:152822,Handle this module: distributed_tool stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#115305 torchrun to support NUMA affiinity, #152822 TorchRun: Option to specify which GPUs to run on",ce7574407536090eb9db540734906fea0fce2bd5783f8567b1ca71eff858c478
issue-collapse:7f269b954853ce16,Review thread: 2 issues - issue #114392: [BUG][pytree] equal `dict`s do not imply equal leaves and equal treespecs,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"114392, 183060",,,,module: pytree,"bot-triaged, enhancement, module: pytree, triaged",thread:issue:114392:issue:183060,Handle this module: pytree stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#114392 [BUG][pytree] equal `dict`s do not imply equal leaves and equal treespecs, #183060 PyTorch pytree equality comparison is too strict.",ffd4c0692392e9fbb4889146e92ee4248979b896a57b245c7298955b5cafe149
issue-collapse:87a2ecccaa86e12d,Review thread: 2 issues - issue #112883: Torchbench inference failures,related issue thread,stale/low urgency,28,high,work_thread_graph,2,0,0,0,"112883, 113063",,,,"module: benchmark, module: dynamic shapes, oncall: pt2","feature, module: benchmark, module: dynamic shapes, no-scrub, oncall: pt2, triaged",thread:issue:112883:issue:113063,Handle this module: benchmark stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#112883 Torchbench inference failures, #113063 Rework Dynamic Benchmarks To Actually Vary Shapes",4427fbed79c5b646da9a91df3a65bce111c668489e6b7e14064f7430623fd454
issue-collapse:35af2b40ab33396a,Review thread: 2 issues - issue #106704: Dataloader extremely slow on in-memory datasets,related issue thread,has linked PR,28,high,work_thread_graph,2,0,0,0,"106704, 154318",,,,"module: data, module: dataloader, module: performance","module: data, module: dataloader, module: performance, triaged",thread:issue:106704:issue:154318,"Review this module: data queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#106704 Dataloader extremely slow on in-memory datasets, #154318 [BUG] DataLoader low GPU utilization and extremely slow compared to manual batching",d53baa8cfe25f433e0df2ba8acc9f473459b302e913a8900ae0bfe4cb0d21bb6
issue-collapse:1eb72bb84b27c8fb,Review thread: 2 issues - issue #101850: Unexpected modification to CPU affinity of Dataloader workers,related issue thread,needs maintainer decision,28,high,work_thread_graph,2,0,0,0,"101850, 102269",,,,"module: dependency bug, module: multiprocessing, module: openmp","module: dependency bug, module: multiprocessing, module: openmp, triaged",thread:issue:101850:issue:102269,"Review this module: dependency bug queue together so the maintainer decides category, owner, and response pattern once.",needs maintainer decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#101850 Unexpected modification to CPU affinity of Dataloader workers, #102269 Import of torch breaks standard multiprocessing",077ab6485450e5cb244e9c8f11c3d8354a47e921afc7d576ded95e69d3a1cc6b
issue-collapse:0166760e2495c019,Review thread: 8 issues - issue #113750: [docs] Add reference/decomp impl snippets to functions in online docs for hackability and educational purposes / compensate for some unclear language in existing docs,related issue thread,stale/low urgency,26,high,work_thread_graph,8,0,0,0,"51455, 71465, 72341, 75862, 76012, 77731, 78050, 113750",,,,"module: complex, module: cuda, module: docs, module: memory format, module: nn, module: norms and normalization, module: primtorch, module: python frontend","actionable, better-engineering, module: complex, module: cuda, module: docs, module: memory format, module: nn, module: norms and normalization, module: primTorch, module: python frontend, triaged",thread:issue:113750:issue:51455:issue:71465:issue:72341,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#51455 [docs] Improve documentation for LayerNorm, GroupNorm, etc (+ add python reference impl), #71465 torch.nn.LayerNorm support for arbitrary axis in order to allow NCHW application, #72341 InstanceNorm doesn't preserve memory format, #75862 LayerNorm and GroupNorm with num_groups=1 not equivalent, #76012 torch.nn.LayerNorm is very slow on GPU (much slower than a custom LayerNorm version in the ConvNext model), #77731 complex abs strides are wrong on empty tensors and tensors with 1 dimension",44a6788f1876d32b86cf78feedf758acf691ec76341b9764417c0578058a5940
issue-collapse:49cb511333698624,Review thread: 4 issues - issue #168563: TestModule: test/inductor/test_aot_inductor.py::AOTInductorTestsTemplate,related issue thread,stale/low urgency,26,high,work_thread_graph,4,0,0,0,"168563, 168566, 168567, 168568",,,,"module: aotinductor, module: rocm, oncall: export, oncall: pt2","module: aotinductor, module: rocm, oncall: export, oncall: pt2, rocm-skipped-tests, triaged",thread:issue:168563:issue:168566:issue:168567:issue:168568,Handle this module: aotinductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#168563 TestModule: test/inductor/test_aot_inductor.py::AOTInductorTestsTemplate, #168566 Test: AOTInductorTestsTemplate.test_autotune_int64_user_defined_triton_kernel, #168567 Test: AOTInductorTestsTemplate.test__weight_int4pack_mm, #168568 Test: AOTInductorTestsTemplate.test__weight_int4pack_mm_with_scales_and_zeros",e2e353a2093a380f57c3dbc4a5e9e7d0ce5132228397390cc3f193950c3d564a
issue-collapse:b29c0eb86b568d90,Review thread: 3 issues - issue #151558: 2.6: SDPA disallows masked attention under vmap,related issue thread,has linked PR,26,high,work_thread_graph,3,0,0,0,"151558, 151559, 151561",,,,"module: docs, module: functorch, module: sdpa, module: vmap","module: docs, module: functorch, module: sdpa, module: vmap, triaged",thread:issue:151558:issue:151559:issue:151561,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,3,Represents 3 open issues; raw issue rows remain available for audit.,"#151558 2.6: SDPA disallows masked attention under vmap, #151559 SDPA doesn't document requirement for head dimension in masks, #151561 2.6: vmap doesn't vectorize all args",626518364d71f80b18e5c1e1d7fccba2132b40816e5a42b2909a985c96c7dc74
issue-collapse:5f999b334160f3d6,Review thread: 2 issues - issue #38208: nn.Module Abstract Class attribute overrides Child.,related issue thread,stale/low urgency,26,high,work_thread_graph,2,0,0,0,"38208, 78618",,,,module: nn,"actionable, module: nn, triaged",thread:issue:38208:issue:78618,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38208 nn.Module Abstract Class attribute overrides Child., #78618 Module parameters/submodules can be shadowed by class attributes silently",d80a49f1a38b87f648b64d6188a4477f91392a58f2c8c7151bc4ef91089e8648
issue-collapse:74503e67dc2322cf,Review thread: 2 issues - issue #134784: Dynamo x autograd.Function x funtorch transform: intermediate sliced tensor doesn't require grad caused incorrect result,related issue thread,has linked PR,26,high,work_thread_graph,2,0,0,0,"134784, 158610",,,,"module: autograd, module: dynamo, module: functorch, module: vmap, oncall: pt2","dynamo-autograd-function, dynamo-functorch, module: autograd, module: dynamo, module: functorch, module: vmap, oncall: pt2, triaged",thread:issue:134784:issue:158610,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#134784 Dynamo x autograd.Function x funtorch transform: intermediate sliced tensor doesn't require grad caused incorrect result, #158610 vmap will always disable gradient calculation",47c6917d54a10af38730c327efeb4e3580a9d21e064402b1bf264d1ea6ee0ea1
issue-collapse:44f286ac88be228e,Review thread: 2 issues - issue #133250: [torch.compile] tighten FX graph restrictions post-functionalization,related issue thread,stale/low urgency,26,high,work_thread_graph,2,0,0,0,"133250, 134363",,,,"module: inductor, module: pt2-dispatcher, oncall: pt2","module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:133250:issue:134363,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#133250 [torch.compile] tighten FX graph restrictions post-functionalization, #134363 Improving inductor pattern matcher's replacement graph assumptions",6802f9e0bcd3f5482efbf6c3b02b8cbdcd6ad3abef1c190a0016e2451915b52e
issue-collapse:28624ef5efff565b,Review thread: 2 issues - issue #120819: Support a context manager to set torch dtype?,related issue thread,has linked PR,26,high,work_thread_graph,2,0,0,0,"120819, 144673",,,,module: python frontend,"actionable, enhancement, feature, module: python frontend, triaged",thread:issue:120819:issue:144673,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#120819 Support a context manager to set torch dtype?, #144673 Thread-safe approach on temporarily changing the `set_default_type`",65279a4578f849c5642f0383dad5456e8ab6ff1ef506044c3d336e12a7744077
issue-collapse:22f070a6a9e23f5c,Review thread: 2 issues - issue #105318: [docs] torch.sigmoid to make clear equivalence relations to other sigmoid functions,related issue thread,has linked PR,26,high,work_thread_graph,2,0,0,0,"105318, 106664",,,,module: docs,"actionable, enhancement, module: docs, triaged",thread:issue:105318:issue:106664,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",has linked PR,1,Represents 2 open issues; raw issue rows remain available for audit.,"#105318 [docs] torch.sigmoid to make clear equivalence relations to other sigmoid functions, #106664 [docs] URL and link format proposal to make function page URLs more concise",a7830f7af88cb38915d99637fee7b5d128717e3f531deaa2d92bfddde4cf9dac
issue-collapse:a1272769f68a2027,Review thread: 7 issues - issue #153854: Better padding API,related issue thread,needs design decision,20,high,work_thread_graph,7,0,0,0,"28214, 29863, 46240, 52205, 57911, 60294, 153854",,,,"module: convolution, module: nn, module: numpy, module: padding, module: tensorflow","enhancement, function request, module: convolution, module: nn, module: numpy, module: padding, module: tensorflow, needs design, triaged",thread:issue:153854:issue:28214:issue:29863:issue:46240,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,6,Represents 7 open issues; raw issue rows remain available for audit.,"#28214 reflective padding for 5D tensor, #29863 n-dimensional non-constant padding functional, #46240 Support ""symmetric"" reflection padding, #52205 Allow F.pad(mode = 'reflect') when shape == pad, #57911 Circular padding in Convolution layers should not only be wrap for once., #60294 Add a NumPy-like `pad` function",369d5c0a5f0115b65f56d3f986146cd8dfe8b2e403735c7677b31e3a76af51ad
issue-collapse:28ac9951417e59bc,Review thread: 7 issues - issue #12659: Differentiation through Module parameters updates,related issue thread,stale/low urgency,20,high,work_thread_graph,7,0,0,0,"7313, 12659, 32005, 39279, 45242, 46886, 65913",,,,"module: activation checkpointing, module: autograd, module: cpp, module: nn, module: optimizer, module: ux","enhancement, feature, module: activation checkpointing, module: autograd, module: cpp, module: nn, module: optimizer, module: ux, needs design, triaged",thread:issue:12659:issue:32005:issue:39279:issue:45242,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 7 open issues; raw issue rows remain available for audit.,"#7313 [proposal] [discussion] Refactor pruning/weight_norm using new Reparametrization functionality + actually deprecate old impl of SpectralNorm, #12659 Differentiation through Module parameters updates, #32005 gradients inside gradient checkpoint, #39279 Differentiable Optimizers, #45242 Functional interface for optimizers, #46886 Support deleting a parameter/buffer by name",77d8617fd2c528cc3e854ab57dce77caef5e180e76d76498a1e2eefdb66b45c8
issue-collapse:e442daefef45afe4,"Review thread: 6 issues - issue #36524: Drop _stacklevel from argspecs of F.softmax, F.softmin, F.log_softmax (for implicit dim has been long deprecated)",related issue thread,stale/low urgency,20,high,work_thread_graph,6,0,0,0,"36524, 64038, 71357, 83163, 84290, 84524",,,,"module: deprecation, module: docs, module: nn, module: norms and normalization, oncall: quantization","actionable, enhancement, low priority, module: deprecation, module: docs, module: nn, module: norms and normalization, needs research, oncall: quantization, triaged",thread:issue:36524:issue:64038:issue:71357:issue:83163,Handle this module: deprecation stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 6 open issues; raw issue rows remain available for audit.,"#36524 Drop _stacklevel from argspecs of F.softmax, F.softmin, F.log_softmax (for implicit dim has been long deprecated), #64038 torch.nn.functional.softmax _stacklevel undocumented, #71357 [feature request] Multidim support for softmax/log_softmax/softmin, #83163 torch.nn.functional.log_softmax parameter '_stacklevel' undocumented, #84290 Fix convert path for fixed qparam ops (sigmoid and softmax), #84524 nn.Softmax should not allow default/implicit/unset dim constructor argument",4fec7028881891a78dfa15aed8dab3dd87309100d594c443fc890410c078bc88
issue-collapse:e52ae4128d7f90da,Review thread: 4 issues - issue #51156: [Feature Request] Support tensor creation from objects that implement the __array__ interface,related issue thread,needs design decision,20,high,work_thread_graph,4,0,0,0,"51156, 54138, 54139, 58036",,,,"module: cuda, module: numba, module: numpy","enhancement, function request, module: cuda, module: numba, module: numpy, needs design, triaged",thread:issue:51156:issue:54138:issue:54139:issue:58036,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,5,Represents 4 open issues; raw issue rows remain available for audit.,"#51156 [Feature Request] Support tensor creation from objects that implement the __array__ interface, #54138 Support Array Interface (__array_interface__ attribute), #54139 as_tensor returns CPU copy of a CUDA buffer object implementing CUDA Array Interface, #58036 [feature request] torch.as_tensor to support any object that NumPy's asarray or array can consume (consume __array_interface__)",65ae994fc4ebc6d61dbb75c07069380f9010322b5ba450861683fcb640e0074d
issue-collapse:03635900f80d5628,Review thread: 4 issues - issue #40761: Sparse Convolutional support,related issue thread,stale/low urgency,20,high,work_thread_graph,4,0,0,0,"40761, 47915, 64544, 78262",,,,"module: convolution, module: sparse","feature, function request, module: convolution, module: sparse, needs research, triaged",thread:issue:40761:issue:47915:issue:64544:issue:78262,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#40761 Sparse Convolutional support, #47915 sparse filter layers (more specifically convolutions), #64544 [Feature Request] Any plan to add 'Sparse Convolution' as default nn module?, #78262 Request for adding the possibility for training on sparse tensors",2573d45ac3d79fb66396f2d88c577be64ae1482b728f3e072e7f4343d4a507ea
issue-collapse:bc41adda6394e8b3,Review thread: 4 issues - issue #18173: [docs] How to achieve high-order derivation in my .cpp?,related issue thread,stale/low urgency,20,high,work_thread_graph,4,0,0,0,"18173, 25883, 28440, 33227",,,,"module: autograd, module: cpp, module: docs, module: nn","actionable, good first issue, module: autograd, module: cpp, module: docs, module: nn, triaged",thread:issue:18173:issue:25883:issue:28440:issue:33227,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#18173 [docs] How to achieve high-order derivation in my .cpp?, #25883 Python/C++ API Parity: torch.nn modules and functional, #28440 Python/C++ API Parity: torch.optim optimizers, #33227 [v1.5] Python/C++ API parity master tracking task",66a05ff8df6e3e5c2459ed033f5518f0eaec7595e552e09cd5a20402d8482dcc
issue-collapse:9a1d8922109dca86,Review thread: 4 issues - issue #130529: Support tuple of `int`s in `torch.unsqueeze`,related issue thread,stale/low urgency,20,high,work_thread_graph,4,0,0,0,"30702, 62148, 94233, 130529",,,,"module: python frontend, module: viewing and reshaping","enhancement, feature, function request, module: python frontend, module: viewing and reshaping, needs design, triaged",thread:issue:130529:issue:30702:issue:62148:issue:94233,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#30702 [FR] multidim squeeze and flatten, #62148 [feature request] Function to broadcast right, #94233 unsqueeze a single dimension multiple times, #130529 Support tuple of `int`s in `torch.unsqueeze`",c78a50768725516af9904d66cbe55febce94cf919d79a00a3ca94a43b4bdef12
issue-collapse:4c8f85c18d56c37b,Review thread: 4 issues - issue #102730: Best practices clarification for initialization strategies,related issue thread,needs design decision,20,high,work_thread_graph,4,0,0,0,"18182, 20102, 102730, 125861",,,,"module: bc-breaking, module: initialization, module: nn, module: python frontend, module: rnn","enhancement, feature, module: bc-breaking, module: initialization, module: nn, module: python frontend, module: rnn, needs research, quansight-nack, triaged",thread:issue:102730:issue:125861:issue:18182:issue:20102,"Review this module: bc-breaking queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,3,Represents 4 open issues; raw issue rows remain available for audit.,"#18182 Update weight initialisations to current best practices, #20102 LSTM forget bias must be initialized properly, #102730 Best practices clarification for initialization strategies, #125861 `torch.nn.LazyLinear` custom weight initialization",ff10e1c9571fea39f284b2a68c1c8ecd0650a6ba7218aa9461172923580b6bf8
issue-collapse:422d3b94f89112da,Review thread: 3 issues - issue #160322: `torch.Tensor` methods return type annotation too broad for `torch.Tensor` subclasses,related issue thread,needs design decision,20,high,work_thread_graph,3,0,0,0,"160322, 161713, 163990",,,,"module: infra, module: typing","module: infra, module: typing, needs design, triaged",thread:issue:160322:issue:161713:issue:163990,"Review this module: infra queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#160322 `torch.Tensor` methods return type annotation too broad for `torch.Tensor` subclasses, #161713 [Proposal] Improving Type Signatures for Types Exported by torch, #163990 Add CI job to track typing changes made to PyTorch",f86a62b1010641f7a93ee5030956a1e5515ea2f5efdd3efe6ed56f02ca029c23
issue-collapse:39f73229a5761929,Review thread: 3 issues - issue #157547: Several `torch.*` functions raise uninformative `NotImplementedError`s when called with integer `dtype`,related issue thread,needs design decision,20,high,work_thread_graph,3,0,0,0,"31779, 35527, 157547",,,,"module: distributions, module: error checking","actionable, better-engineering, module: distributions, module: error checking, needs design, triaged",thread:issue:157547:issue:31779:issue:35527,"Review this module: distributions queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#31779 torch.poisson returns floating point tensor, #35527 `torch.bernoulli` and `torch.rand` to support `dtype=torch.bool` kwarg (and sypport operation of `tensor.uniform_` on bool tensors), #157547 Several `torch.*` functions raise uninformative `NotImplementedError`s when called with integer `dtype`",a646c79651a619bb3e5b3bd8e5092030d4ec420e64c856cfa884c3d5fad2e916
issue-collapse:dfbca7c56bb6dece,Review thread: 3 issues - issue #153378: MPS backend silently ignores dimension mismatch,related issue thread,needs design decision,20,high,work_thread_graph,3,0,0,0,"153378, 160553, 163504",,,,"module: error checking, module: mps","module: error checking, module: mps, needs design, triaged",thread:issue:153378:issue:160553:issue:163504,"Review this module: error checking queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#153378 MPS backend silently ignores dimension mismatch, #160553 [MPS] Out of bounds indexing inconsistency between MPS and CPU/CUDA, #163504 [MPS] Unexpected behavior in torch.nn.functional.one_hot when target values exceed num_classes in MPS",cd59f89f8065f5d2d11a808d3e42a2df4d1faf42403bf40ad6330568fc6b485b
issue-collapse:a3a22e6b4e740940,"Review thread: 3 issues - issue #120139: On `Dataset`, `IterableDataset` inheritance",related issue thread,stale/low urgency,20,high,work_thread_graph,3,0,0,0,"47055, 120139, 122410",,,,"module: data, module: docs","enhancement, module: data, module: docs, needs research, triaged",thread:issue:120139:issue:122410:issue:47055,Handle this module: data stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#47055 [types] torch.utils.data.{Dataset, Sampler} are not Sized, #120139 On `Dataset`, `IterableDataset` inheritance, #122410 On Lack of Default `__len__` in Python Abstract Base Classes",828d8f390944e66d27b56762c0f03fd0830c0315b1f51d8a3c5decd10b25baf9
issue-collapse:86c71452ecf857d1,Review thread: 3 issues - issue #115852: `groups` parameter for `nn.Linear`,related issue thread,needs design decision,20,high,work_thread_graph,3,0,0,0,"115852, 116506, 144012",,,,"module: convolution, module: cuda, module: inductor, module: mkldnn, module: nn, module: performance, module: xnnpack, oncall: pt2","matrix multiplication, module: convolution, module: cuda, module: inductor, module: mkldnn, module: nn, module: performance, module: xnnpack, needs research, oncall: pt2, triaged",thread:issue:115852:issue:116506:issue:144012,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,2,Represents 3 open issues; raw issue rows remain available for audit.,"#115852 `groups` parameter for `nn.Linear`, #116506 [discussion] Route pointwise Conv1d/Conv2d to matmul? (also in eager), #144012 [RFC] Add CPP Grouped GEMM Template for Inductor CPU",061240de8fa92c387479f399465d5139328a248f7552d4f9bdf84c78fac96361
issue-collapse:b1a87a7db943dd21,Review thread: 2 issues - issue #96764: Improve checkpoint thread-safety,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"96764, 96766",,,,"module: autograd, module: multithreading","actionable, module: autograd, module: multithreading, needs design, triaged",thread:issue:96764:issue:96766,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#96764 Improve checkpoint thread-safety, #96766 Follow-ups to do after adding nested checkpoint",79a06e6c22c72eff80c13af8cb116e93b32a46f9b4628a49b3b32ee91519278f
issue-collapse:efef5ee792b00469,Review thread: 2 issues - issue #96319: views created in __torch_dispatch__ share storage but not version_counter,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"96319, 97500",,,,"module: __torch_dispatch__, module: autograd, module: molly-guard","has workaround, module: __torch_dispatch__, module: autograd, module: molly-guard, needs design, triaged",thread:issue:96319:issue:97500,"Review this module: __torch_dispatch__ queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#96319 views created in __torch_dispatch__ share storage but not version_counter, #97500 .set_ operation on a view (detach()) of the view tensor changes grad_fn of the original view tensor from ViewBackward0 to AsStridedBackward0",81dd63c5dbee3abfeeebfde2efbad9a6f4e8d93b72cf4438215bdb7f9bb96ea1
issue-collapse:8332b72cf6ab7293,Review thread: 2 issues - issue #73513: Wrapper tensor level confusion,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"73513, 73537",,,,"module: __torch_dispatch__, module: functionalization","module: __torch_dispatch__, module: functionalization, needs research, tensor subclass, triaged",thread:issue:73513:issue:73537,"Review this module: __torch_dispatch__ queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73513 Wrapper tensor level confusion, #73537 Functionalization doesn't work when applied twice with wrapper tensor",98929753e2e38aaafc36599575500075e79f75f35d1e7446efc2bd1965116ca4
issue-collapse:b3a0940d9eee8d04,Review thread: 2 issues - issue #64067: Execute C++ based GTest unit tests from a python wrapper,related issue thread,good volunteer slice,20,high,work_thread_graph,2,0,0,0,"64067, 66824",,,,"module: ci, module: tests","enhancement, good first issue, module: ci, module: tests, triaged",thread:issue:64067:issue:66824,Turn this module: ci slice into one contributor-ready queue with shared scope and testing expectations.,good volunteer slice,1,Represents 2 open issues; raw issue rows remain available for audit.,"#64067 Execute C++ based GTest unit tests from a python wrapper, #66824 Migrate C++ tests to Python runner",6262073dc36994af15942e309c80e1d17edd8b394a1a8a79628c6a44ab1072c2
issue-collapse:83677087a5dbb180,Review thread: 2 issues - issue #55152: Better handling of OpInfo.sample_inputs,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"55152, 55159",,,,module: tests,"better-engineering, ezyang's list, module: tests, needs design, triaged",thread:issue:55152:issue:55159,"Review this module: tests queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#55152 Better handling of OpInfo.sample_inputs, #55159 Better syntax for OpInfo",f1c1c49d7bda494ac4eb988b6360b48a9fc687aa50334867c621da5e169b1cbd
issue-collapse:124073fac772e9b8,Review thread: 2 issues - issue #41499: Helping test example code blocks in the docs,related issue thread,stale/low urgency,20,high,work_thread_graph,2,0,0,0,"6662, 41499",,,,"module: doc infra, module: docs, module: tests","feature, good first issue, module: doc infra, module: docs, module: tests, todo, triaged",thread:issue:41499:issue:6662,Handle this module: doc infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#6662 Autogenerate code example / tutorial outputs in documentation, #41499 Helping test example code blocks in the docs",8e6f7aefdeac9647af3ce97058a84d378c9fe39e485c538b4e5c3ba46f1a62d9
issue-collapse:0b06a58e0169034a,Review thread: 2 issues - issue #40316: Refactor the adaptive avg pool code,related issue thread,good volunteer slice,20,high,work_thread_graph,2,0,0,0,"40316, 42779",,,,"module: memory format, module: performance, module: pooling, oncall: quantization","good first issue, low priority, module: memory format, module: performance, module: pooling, oncall: quantization, triaged",thread:issue:40316:issue:42779,Turn this module: memory format slice into one contributor-ready queue with shared scope and testing expectations.,good volunteer slice,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40316 Refactor the adaptive avg pool code, #42779 [quant] Quantized AdaptivePool3d is much slower for ChannelsLast3d.",9871410e59a46e57515d243973c43ee915bc099b2220f7cd25fbdebe6e519c1f
issue-collapse:5a139fecb99e6851,Review thread: 2 issues - issue #152877: Motivate Pytorch's forward mode AD APIs with training examples,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"71117, 152877",,,,"module: autograd, module: docs, module: forward ad","actionable, module: autograd, module: docs, module: forward ad, needs design, triaged",thread:issue:152877:issue:71117,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71117 Rollup: forward-mode AD operator coverage, #152877 Motivate Pytorch's forward mode AD APIs with training examples",b4555cba1dab65652a0ebbda4911c45ffe71841e92cb2c1107ce7b6f960077cb
issue-collapse:d9414fd75cfb2f86,Review thread: 2 issues - issue #13811: Feature request: von Mises-Fisher distribution,related issue thread,stale/low urgency,20,high,work_thread_graph,2,0,0,0,"13811, 147048",,,,module: distributions,"module: distributions, needs research, triaged",thread:issue:13811:issue:147048,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13811 Feature request: von Mises-Fisher distribution, #147048 Feature Request: rsample for Von Mises distribution",1bb48444edd14bbe9ce8a23a64d7e33bcf8e1abc9c757960b3408c6fcc7a6b36
issue-collapse:5775223a8cea631a,Review thread: 2 issues - issue #133566: Performance: Autodiff stores inputs of convolution layers with non-trainable weight,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"97575, 133566",,,,"module: autograd, module: memory usage, module: nestedtensor","actionable, enhancement, module: autograd, module: memory usage, module: nestedtensor, needs design, triaged",thread:issue:133566:issue:97575,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#97575 Multi-output derivative formulas can save unnecessary tensors, #133566 Performance: Autodiff stores inputs of convolution layers with non-trainable weight",c038250280772934deecbc686df870e6e3fc7de634a013d56dab83b6e5bf3f20
issue-collapse:e2dbab17ad6677c3,Review thread: 2 issues - issue #132378: per sample gradient with torch.autograd.grad,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"132378, 156616",,,,"module: autograd, module: functorch, module: vmap","module: autograd, module: functorch, module: vmap, needs research, triaged",thread:issue:132378:issue:156616,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#132378 per sample gradient with torch.autograd.grad, #156616 Add `is_outputs_batched` param to `autograd.grad`",f604e88a23554cc34b36a0be79ae3b15a60c1092715ce3b32e5a89ede9ea53e3
issue-collapse:25b6df207b2b6978,Review thread: 2 issues - issue #130513: RuntimeError Occurred In getBar1SizeOfGpu when initializing torch RPC,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"130513, 159354",,,,"module: tensorpipe, oncall: distributed","module: tensorpipe, needs research, oncall: distributed, triaged",thread:issue:130513:issue:159354,"Review this module: tensorpipe queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130513 RuntimeError Occurred In getBar1SizeOfGpu when initializing torch RPC, #159354 distributed/rpc/test_share_memory fails with timeouts on 64bit BAR systems",1fbdd64e206bc0fbae9b888994c1176848e5713c2b0caf12c668a16b5ba57dba
issue-collapse:82f359c9f756412c,Review thread: 2 issues - issue #126164: Context manager for determinism,related issue thread,stale/low urgency,20,high,work_thread_graph,2,0,0,0,"72831, 126164",,,,"module: autograd, module: determinism","module: autograd, module: determinism, needs research, triaged",thread:issue:126164:issue:72831,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#72831 Toggling deterministic mode for individual autograd backward functions, #126164 Context manager for determinism",9ac2ee7b9af02965499af38e07eb6001d574b68a903af926f14f9e5c37baf54c
issue-collapse:92fe03add265e236,Review thread: 2 issues - issue #116490: Legitimize/harden differentiable optimizers in torch.optim,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"116490, 141832",,,,"module: autograd, module: optimizer","actionable, module: autograd, module: optimizer, needs research, triaged",thread:issue:116490:issue:141832,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#116490 Legitimize/harden differentiable optimizers in torch.optim, #141832 Optimizers' `differentiable` flag doesn't work",3f231ba81ea61a11f60fb2e5f3c88e1717ecb850b71b48d28fb1de70290d40fc
issue-collapse:58efc62df7aebe52,Review thread: 2 issues - issue #112342: [pytree] Pytree node registration hygeine: deprecate global _register_pytree_node; only allow enabling registered pytree extensions locally,related issue thread,stale/low urgency,20,high,work_thread_graph,2,0,0,0,"112342, 112352",,,,module: pytree,"module: pytree, needs design, triaged",thread:issue:112342:issue:112352,Handle this module: pytree stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#112342 [pytree] Pytree node registration hygeine: deprecate global _register_pytree_node; only allow enabling registered pytree extensions locally, #112352 RFC: [pytree] node registration namespaces",2b9ecaf5b627da389df780cd40d90eb8cee38ebf97a76aa5434717d4e2d4e683
issue-collapse:2f79d7c2e9196d75,Review thread: 2 issues - issue #110295: Tests modify global state cause later tests to fail,related issue thread,needs design decision,20,high,work_thread_graph,2,0,0,0,"110295, 125239",,,,"module: ci, module: devx, module: tests","module: ci, module: devx, module: tests, needs design, triaged",thread:issue:110295:issue:125239,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",needs design decision,1,Represents 2 open issues; raw issue rows remain available for audit.,"#110295 Tests modify global state cause later tests to fail, #125239 Improved strategy for dealing with deterministically flaky tests which are order sensitive",9d67c7fd63f2ebb170ad6750d9247ee3bc44321365dcea6f1fa748d1f5ad3c3a
issue-collapse:7593fd218151d8ac,Review thread: 7 issues - issue #55267: Improve CUDA extension building experience,related issue thread,stale/low urgency,14,high,work_thread_graph,7,0,0,0,"55267, 60341, 61260, 68905, 70191, 77184, 81680",,,,"module: abi, module: binaries, module: build, module: cpp, module: cpp-extensions, module: cuda, oncall: java","enhancement, feature, module: abi, module: binaries, module: build, module: cpp, module: cpp-extensions, module: cuda, needs reproduction, oncall: java, triaged",thread:issue:55267:issue:60341:issue:61260:issue:68905,Handle this module: abi stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 7 open issues; raw issue rows remain available for audit.,"#55267 Improve CUDA extension building experience, #60341 C++ version 1.9.0 libtorch dynamic load fails -- GCC only, #61260 libtorch_cpu.so: undefined symbol: _ZN3c1010ThreadPool3runESt8functionIFvvEE, #68905 Shared `~/.cache/torch_extensions` needs to be pytorch version aware., #70191 [libtorch] Loading in Java two differente libtorch_cpu.so from different versions fails, #77184 Undefined symbol error when compiling and loading C++ extension",d15ec87fbe425321f7b38a96e3f9561e82021d21854185c93495f3a93fdbdee3
issue-collapse:4940d3ad7382e63e,Review thread: 3 issues - issue #104587: Conv1d step-by-step numerical error,related issue thread,needs reproduction,14,high,work_thread_graph,3,0,0,0,"104587, 133006, 170038",,,,"module: arm, module: convolution, module: determinism, module: macos, module: numerical-stability","module: arm, module: convolution, module: determinism, module: macos, module: numerical-stability, needs reproduction, topic: fuzzer, triaged",thread:issue:104587:issue:133006:issue:170038,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,2,Represents 3 open issues; raw issue rows remain available for audit.,"#104587 Conv1d step-by-step numerical error, #133006 Precision Differences in Using `dropout` and `linear`, #170038 Conv2D numerical mismatch vs TensorFlow",75eabaa3e743117e0d7b3529214ac9eff64e0d961e32d73cb48bd919e49b42cf
issue-collapse:0419cafe2679af0a,Review thread: 2 issues - issue #30900: CUDA error: initialization error (multiprocessing) with Python 3.7,related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"30900, 33248",,,,"module: autograd, module: cuda, module: multiprocessing","module: autograd, module: cuda, module: multiprocessing, needs reproduction, triaged",thread:issue:30900:issue:33248,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30900 CUDA error: initialization error (multiprocessing) with Python 3.7, #33248 PyTorch 1.4.0 CUDA initialization error with CPU-only (multiprocessing) on Python 3.7.5",29fc0ff989906be482bfe1093f77c253ea7d6316a5c09d8509e4bd26f2877465
issue-collapse:f2aff900a53d22e9,Review thread: 2 issues - issue #152790: [CXX11ABI] torch 2.6.0-cu126 and cu124 have different exported symbols,related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"152790, 155066",,,,"module: binaries, module: cuda, oncall: releng","module: binaries, module: cuda, needs reproduction, oncall: releng, triaged",thread:issue:152790:issue:155066,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#152790 [CXX11ABI] torch 2.6.0-cu126 and cu124 have different exported symbols, #155066 Add extensions `flash_attention` and `vllm` as test of new PyTorch releases for known issues of compat of their binaries and of possibility of compiling these from source",4b330781bc2c7c66a3b71661393a4daaf8918dd5342d53d05e5bfd5b604e6cf1
issue-collapse:371f0bf9d546d878,Review thread: 2 issues - issue #152335: pin_memory crashes for big tensors and leaks page locked memory,related issue thread,stale/low urgency,14,high,work_thread_graph,2,0,0,0,"152335, 159447",,,,"module: cuda, module: dataloader, module: memory usage","module: cuda, module: dataloader, module: memory usage, needs reproduction, triaged",thread:issue:152335:issue:159447,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#152335 pin_memory crashes for big tensors and leaks page locked memory, #159447 dataloader hits CUDA error: invalid argument in data.pin_memory(device) on RTX 3090",2cadf5b1b77f51953f0249b0f1f30959574df41915365440a66bb2fad681f083
issue-collapse:894c20f28550c6ac,Review thread: 2 issues - issue #129658: [TensorDict - compile] dynamo generator compatibility,related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"129658, 129668",,,,"module: dynamo, oncall: pt2","dynamo-ctx-manager, dynamo-dataclasses, dynamo-dicts, dynamo-nn-modules, dynamo-torch-function, low priority, module: dynamo, needs reproduction, oncall: pt2, triaged",thread:issue:129658:issue:129668,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#129658 [TensorDict - compile] dynamo generator compatibility, #129668 TensorDict - torch.compile issue tracker",0964611e4beadd7940940c2d260810072074d774c5416f9fb00e13c328b613e8
issue-collapse:b3b71b2ac0f650d3,Review thread: 2 issues - issue #128564: internal asset failure,related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"128564, 128778",,,,module: linear algebra,"module: linear algebra, needs reproduction, triaged",thread:issue:128564:issue:128778,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#128564 internal asset failure, #128778 False INTERNAL ASSERT FAILED bug whilst training Neural Network",9fbc79cfa91dbc4437d4479cef870693531336cb0f884fd3c6785c58f7cb0de4
issue-collapse:4226a7369a5163b1,Review thread: 2 issues - issue #120237: FFT operations on MPS are less accurate than on CPU,related issue thread,stale/low urgency,14,high,work_thread_graph,2,0,0,0,"120237, 151740",,,,"module: correctness (silent), module: fft, module: mps","module: correctness (silent), module: fft, module: mps, needs reproduction, triaged",thread:issue:120237:issue:151740,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#120237 FFT operations on MPS are less accurate than on CPU, #151740 mps and cpu backends produce different training results with FFT and Adam",5ef3a10689055834e2d0dd21454a9f3a1e2bb3122ee837cb31e5e7e68d03f2b8
issue-collapse:b745ebd394af93ba,Review thread: 2 issues - issue #116567: [MPS] `sort` incorrectly handles 'negative' NaNs,related issue thread,stale/low urgency,14,high,work_thread_graph,2,0,0,0,"116567, 117617",,,,"module: mps, module: python frontend","module: mps, module: python frontend, needs reproduction, triaged",thread:issue:116567:issue:117617,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#116567 [MPS] `sort` incorrectly handles 'negative' NaNs, #117617 torch.topk, when the input contains +nan and -nan, the result is not sure; sometimes -nan is treated greater than normal number, sometimes -nan is the least;",02de179d42fb131539f7c4c5e00d5b64061340ba34904f4b368d6f552fa70b65
issue-collapse:e5ba48724494ffb3,Review thread: 2 issues - issue #114465: Installation error: 'CMake Error at third_party/benchmark/CMakeLists.txt:304 (message): Failed to determine the source files for the regular expression backend',related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"30896, 114465",,,,"module: build, module: third_party","module: build, module: third_party, needs reproduction, triaged",thread:issue:114465:issue:30896,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30896 FATAL_ERROR ""Failed to determine the source files for the regular expression backend"", #114465 Installation error: 'CMake Error at third_party/benchmark/CMakeLists.txt:304 (message): Failed to determine the source files for the regular expression backend'",ac2db616bbfc457a2b75148daceef4ae06aa20ea54c87ae5701e8d8608fe830d
issue-collapse:9b48c0a2528cb1b9,Review thread: 2 issues - issue #114455: Torch Cpu Memory Leak with FastApi uvicorn.,related issue thread,needs reproduction,14,high,work_thread_graph,2,0,0,0,"68114, 114455",,,,module: memory usage,"module: memory usage, needs reproduction, triaged",thread:issue:114455:issue:68114,"Review this module: memory usage queue together so the maintainer decides category, owner, and response pattern once.",needs reproduction,1,Represents 2 open issues; raw issue rows remain available for audit.,"#68114 CPU Memory Deallocation, #114455 Torch Cpu Memory Leak with FastApi uvicorn.",294c72b86ea9b3a61fd2bcb6e446c7c0208dc38eb3b6d71671338d87be321898
issue-collapse:4e8ad65e8ddde701,Review thread: 33 issues - issue #62540: [Meta] Change default branch name to `main` for repos in `pytorch` project,related issue thread,stale/low urgency,2,high,work_thread_graph,33,0,0,0,"62540, 71806, 71810, 71812, 71813, 71814, 71815, 71816, 71817, 71818, 71819, 71820, 71821, 71822, 71823, 71824, 71825, 71826, 71828, 71830, 71831, 71832, 71833, 71834, 71835, 71836, 71837, 71838, 71839, 71840, 71841, 71842, 71843",,,,"module: bootcamp, module: infra","module: bootcamp, module: infra, triaged",thread:issue:62540:issue:71806:issue:71810:issue:71812,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 33 open issues; raw issue rows remain available for audit.,"#62540 [Meta] Change default branch name to `main` for repos in `pytorch` project, #71806 Migrate master to main: https://github.com/pytorch/pytorch, #71810 Migrate master to main: https://github.com/pytorch/labeler-github-action, #71812 Migrate master to main: https://github.com/pytorch/pytorch-ci-dockerfiles, #71813 Migrate master to main: https://github.com/pytorch/pytorch-integration-testing, #71814 Migrate master to main: https://github.com/pytorch/ossci-job-dsl",b7409127ec681b6c2b22641b7f7ba7f261fb157e517d18995ee77219e6a174b1
issue-collapse:fb37db14ca0ba00b,Review thread: 11 issues - issue #168835: TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredTraining,related issue thread,stale/low urgency,2,high,work_thread_graph,11,0,0,0,"168835, 168836, 168837, 168838, 168839, 168840, 168841, 168842, 168843, 168844, 168845",,,,"module: rocm, module: sparse","module: rocm, module: sparse, rocm-skipped-tests, triaged",thread:issue:168835:issue:168836:issue:168837:issue:168838,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 11 open issues; raw issue rows remain available for audit.,"#168835 TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredTraining, #168836 Test: TestSparseSemiStructuredTraining.test_prune_dense_static_sort, #168837 Test: TestSparseSemiStructuredTraining.test_pruning_algo_largest_abs_values_greedy, #168838 Test: TestSparseSemiStructuredTraining.test_pack_both_ways_meta_correctness, #168839 Test: TestSparseSemiStructuredTraining.test_pack_both_ways_id, #168840 Test: TestSparseSemiStructuredTraining.test_pack_both_ways_edge_case1",aa6c34bc97d83197399227c3808a6bc7d7845b96ae29072a1b50d04611df8a7d
issue-collapse:89aecec1e2094c97,Review thread: 8 issues - issue #69687: Performance improvement in Autograd Forward AD using ZeroTensors,related issue thread,stale/low urgency,2,high,work_thread_graph,8,0,0,0,"69687, 72345, 72347, 72358, 72359, 72361, 72362, 75586",,,,"module: autograd, module: forward ad, module: performance","ZeroTensor, actionable, module: autograd, module: forward ad, module: performance, triaged",thread:issue:69687:issue:72345:issue:72347:issue:72358,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#69687 Performance improvement in Autograd Forward AD using ZeroTensors, #72345 Add ZeroTensor fastpath for torch.addmm, #72347 Add ZeroTensor fastpath for torch.addcmul, #72358 Add ZeroTensor fastpath for torch.addbmm, #72359 Add ZeroTensor fastpath for torch.addmv, #72361 Add ZeroTensor fastpath for torch.baddbmm",2a70c3eeb9a596872b79fbc084b2ecc527f8ddf5e643d9822ce7dc1473ef144b
issue-collapse:e3b4e2df6ff95905,Review thread: 8 issues - issue #168522: TestModule: test/distributed/test_symmetric_memory.py::AsyncTPTest,related issue thread,stale/low urgency,2,high,work_thread_graph,8,0,0,0,"168522, 168523, 168524, 168525, 168526, 168527, 168528, 168529",,,,"module: rocm, module: symm_mem","module: rocm, module: symm_mem, rocm-skipped-tests, triaged",thread:issue:168522:issue:168523:issue:168524:issue:168525,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#168522 TestModule: test/distributed/test_symmetric_memory.py::AsyncTPTest, #168523 Test: AsyncTPTest.test_fused_all_gather_matmul, #168524 Test: AsyncTPTest.test_fused_all_gather_matmul_native, #168525 Test: AsyncTPTest.test_multimem_all_gather_matmul, #168526 Test: AsyncTPTest.test_fused_all_gather_scaled_matmul, #168527 Test: AsyncTPTest.test_fused_matmul_reduce_scatter",5ddbdfd60ce8bbb87ba26253ca0906883002f03869a09e1d39d1766220350290
issue-collapse:c299d4ac5417ab03,Review thread: 8 issues - issue #168453: TestModule: test/distributed/test_c10d_gloo.py::ProcessGroupGlooTest,related issue thread,stale/low urgency,2,high,work_thread_graph,8,0,0,0,"168453, 168454, 168455, 168456, 168457, 168458, 168459, 168460",,,,"module: c10d, module: rocm","module: c10d, module: rocm, rocm-skipped-tests, triaged",thread:issue:168453:issue:168454:issue:168455:issue:168456,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#168453 TestModule: test/distributed/test_c10d_gloo.py::ProcessGroupGlooTest, #168454 Test: ProcessGroupGlooTest.test_broadcast_stress_cuda, #168455 Test: ProcessGroupGlooTest.test_allreduce_stress_cuda, #168456 Test: ProcessGroupGlooTest.test_scatter_stress_cuda, #168457 Test: ProcessGroupGlooTest.test_gather_stress_cuda, #168458 Test: ProcessGroupGlooTest.test_allgather_stress_cuda",d9d766dcf49ce20c5f6126c539f16a4df64862096d0c24b4635de5bd8b75c688
issue-collapse:1cfe7b4a196e3daf,Review thread: 8 issues - issue #168424: TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorFromLocalShards,related issue thread,stale/low urgency,2,high,work_thread_graph,8,0,0,0,"168424, 168425, 168426, 168427, 168428, 168429, 168430, 168431",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168424:issue:168425:issue:168426:issue:168427,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 8 open issues; raw issue rows remain available for audit.,"#168424 TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorFromLocalShards, #168425 Test: TestShardedTensorFromLocalShards.test_init_from_local_shards, #168426 Test: TestShardedTensorFromLocalShards.test_recalc_for_metadata, #168427 Test: TestShardedTensorFromLocalShards.test_init_from_local_shards_with_different_glb_size, #168428 Test: TestShardedTensorFromLocalShards.test_non_rw_sharded_recalc_for_metadata, #168429 Test: TestShardedTensorFromLocalShards.test_init_from_local_shards_and_global_metadata_with_all_zeros",dd13028f58e0e7c145effc48f3203041f6dcdbb477dd4d25e9839977e68c9706
issue-collapse:51e76692524f970a,Review thread: 6 issues - issue #44316: [FX] Schema normalization tooling,related issue thread,stale/low urgency,2,high,work_thread_graph,6,0,0,0,"44316, 45565, 53678, 53937, 55095, 55104",,,,"module: __torch_function__, module: fx, module: internals","feature, module: __torch_function__, module: fx, module: internals, triaged",thread:issue:44316:issue:45565:issue:53678:issue:53937,Handle this module: __torch_function__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 6 open issues; raw issue rows remain available for audit.,"#44316 [FX] Schema normalization tooling, #45565 Resolving overloads and default arguments from within __torch_function__, #53678 [FX] Regression from 1.8: FX can no longer trace functions where the first element of an int list is a Proxy, #53937 Only tensor can be proxied in builtin operators with `__torch_function__`, #55095 Python dispatch key and backend fallback, #55104 Alternative backend fallback like mechanism which has higher precedence than operator-specific composite implementations",927a5a7c3f4c79715db31ca44c57c85ac432947f37944ab5b148aa6670bf09be
issue-collapse:648ba51416ed0498,Review thread: 6 issues - issue #168503: TestModule: test/distributed/test_nvshmem_triton.py::NVSHMEMTritonTest,related issue thread,stale/low urgency,2,high,work_thread_graph,6,0,0,0,"168503, 168515, 168516, 168517, 168518, 168519",,,,"module: rocm, module: symm_mem","module: rocm, module: symm_mem, rocm-skipped-tests, triaged",thread:issue:168503:issue:168515:issue:168516:issue:168517,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 6 open issues; raw issue rows remain available for audit.,"#168503 TestModule: test/distributed/test_nvshmem_triton.py::NVSHMEMTritonTest, #168515 Test: NVSHMEMTritonTest.test_triton_alltoall, #168516 Test: NVSHMEMTritonTest.test_triton_broadcast, #168517 Test: NVSHMEMTritonTest.test_triton_sum_reduce, #168518 Test: NVSHMEMTritonTest.test_triton_minmax_reduce, #168519 Test: NVSHMEMTritonTest.test_triton_prod_reduce",39b792190d8c6f2db07505aca7b208007835e9c43dca9723d8fa24a8695bf8eb
issue-collapse:e67dd87512b3f0b5,Review thread: 6 issues - issue #168416: TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorEnumerable,related issue thread,stale/low urgency,2,high,work_thread_graph,6,0,0,0,"168416, 168417, 168418, 168419, 168420, 168421",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168416:issue:168417:issue:168418:issue:168419,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 6 open issues; raw issue rows remain available for audit.,"#168416 TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorEnumerable, #168417 Test: TestShardedTensorEnumerable.test_grid_sharding, #168418 Test: TestShardedTensorEnumerable.test_partial_world_size, #168419 Test: TestShardedTensorEnumerable.test_new_group, #168420 Test: TestShardedTensorEnumerable.test_multiple_local_shards, #168421 Test: TestShardedTensorEnumerable.test_with_rpc_names",d0b05b63fa3c7ccd248ec72b89839a8b90039a6b0288a2e3f3bfbc04d83f14c9
issue-collapse:892d6bff1adda43e,Review thread: 6 issues - issue #138696: [CI/CD] Deprecating PyTorch’s official Anaconda channel,related issue thread,stale/low urgency,2,high,work_thread_graph,6,0,0,0,"138696, 141764, 148335, 148336, 148338, 148339",,,,module: ci,"better-engineering, enhancement, module: ci, topic: docs, triaged",thread:issue:138696:issue:141764:issue:148335:issue:148336,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 6 open issues; raw issue rows remain available for audit.,"#138696 [CI/CD] Deprecating PyTorch’s official Anaconda channel, #141764 Replace Miniconda with minforge as option, #148335 [CI] [anaconda] Docker files have conda environment installed, #148336 [CI] [anaconda] CI Build and Test scripts Linux, #148338 [CI] [anaconda] CI Build and Test scripts Windows, #148339 [Docs] [anaconda] Review and update",8f659e9521b835380438539e4390011fec5bc9449983a05180c2e9226ee37425
issue-collapse:eb8f2d91678dc93d,Review thread: 6 issues - issue #103580: Support ByteTensor and ShortTensor for nn.Embedding and nn.EmbeddingBag,related issue thread,stale/low urgency,2,high,work_thread_graph,6,0,0,0,"52743, 61819, 81413, 103580, 147628, 162918",,,,"module: advanced indexing, module: docs, module: embedding, module: nn, module: numpy, module: python frontend, module: scatter & gather ops, module: sorting and selection, module: type promotion","actionable, enhancement, module: advanced indexing, module: docs, module: embedding, module: nn, module: numpy, module: python frontend, module: scatter & gather ops, module: sorting and selection, module: type promotion, topic: improvements, triaged",thread:issue:103580:issue:147628:issue:162918:issue:52743,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 6 open issues; raw issue rows remain available for audit.,"#52743 torch.searchsorted issues, #61819 [discussion] Support other index dtypes for indexing, take, put, scatter, scatter_reduce and other indexing functions in addition to int64: int8, uint8, int16, int32, uint16 etc (without casting copies/reallocations), #81413 torch.searchsorted error message and documentation is unclear, #103580 Support ByteTensor and ShortTensor for nn.Embedding and nn.EmbeddingBag, #147628 Optimize memory usage of torch.sort significantly, with dynamic dtype indices, #162918 `torch.embedding` should accept `indices` of dtype `torch.uint8`",62e59359a0564f6f8997c8c6db25b1a89febab27de654b528b715ca0946f4beb
issue-collapse:119edeb75a34ba2c,Review thread: 5 issues - issue #64327: RFC: multi-part `torch.load`/`torch.save` to support huge models and/or low CPU memory,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"64327, 75242, 79967, 87033, 97196",,,,"module: lazy, module: memory usage, module: nn, module: pickle, module: serialization","feature, module: lazy, module: memory usage, module: nn, module: pickle, module: serialization, onnx-needs-info, onnx-triaged, triaged",thread:issue:64327:issue:75242:issue:79967:issue:87033,Handle this module: lazy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#64327 RFC: multi-part `torch.load`/`torch.save` to support huge models and/or low CPU memory, #75242 large model, low memory: need `torch.load` that loads one submodule at a time, #79967 [feature request] LazyTensor that provides/loads/computes its contents only upon request to be returned from torch.load, #87033 Saving and loading from physical storage, #97196 Sequential/Partial unpickling and loading of models",3848ae61c699c1ad20951dfe27250b80e45d1da58ac8603005585102e3597bb6
issue-collapse:208c754f6854de21,"Review thread: 5 issues - issue #44989: Support unfold for integral types (long, byte etc) tensors",related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"44989, 44991, 45009, 45851, 60466",,,,"module: nn, module: performance, module: pooling, module: sorting and selection, module: ux","actionable, feature, module: nn, module: performance, module: pooling, module: sorting and selection, module: ux, triaged",thread:issue:44989:issue:44991:issue:45009:issue:45851,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#44989 Support unfold for integral types (long, byte etc) tensors, #44991 Unite/unify tensor.unfold with F.unfold and make them more performant (zero-copy or little-copy with stride tricks, as in NumPy), #45009 Median / quantile / mode / rank / percentile pooling, #45851 The same function can have different signatures in the torch and torch.nn.functional namespaces, #60466 PyTorch unfold could be faster",52e8c93c0636f6a375a9487a70e41f0c22000c134b01946a9aaa62006c2a8bec
issue-collapse:efe8de69731f4540,Review thread: 5 issues - issue #23110: [RFC] RPC Based Distributed Model Parallel,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"23110, 26759, 27099, 28245, 37529",,,,module: rpc,"feature, module: rpc, triaged",thread:issue:23110:issue:26759:issue:27099:issue:28245,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 5 open issues; raw issue rows remain available for audit.,"#23110 [RFC] RPC Based Distributed Model Parallel, #26759 [RFC] RRef Protocol, #27099 Support implicit RRef type conversion, #28245 PyTorch RPC should expose critical metrics to the application., #37529 Implement generic function scheduler in c10/util",0aeffcf28ee8e339fbcff32206fdbf8a5f1ea83d03b4177dab659bbc7de6daf9
issue-collapse:b0dbbb6a3ad4b709,Review thread: 5 issues - issue #168846: TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredCUTLASS,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"168846, 168847, 168848, 168849, 168850",,,,"module: rocm, module: sparse","module: rocm, module: sparse, rocm-skipped-tests, triaged",thread:issue:168846:issue:168847:issue:168848:issue:168849,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#168846 TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredCUTLASS, #168847 Test: TestSparseSemiStructuredCUTLASS.test_linear_cutlass, #168848 Test: TestSparseSemiStructuredCUTLASS.test_sparse_semi_structured_ops_cutlass, #168849 Test: TestSparseSemiStructuredCUTLASS.test_conversions, #168850 Test: TestSparseSemiStructuredCUTLASS.test_conversions_all_patterns",a36c9de6116a14160c0b38e69a03fc7b6c45ecedd6248176108aeb45b717a9e8
issue-collapse:bed29fdba6df7540,Review thread: 5 issues - issue #168784: TestModule: test/test_nestedtensor.py::TestNestedTensorSubclass,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"168784, 168788, 168789, 168791, 168792",,,,"module: rocm, module: sdpa","module: rocm, module: sdpa, rocm-skipped-tests, triaged",thread:issue:168784:issue:168788:issue:168789:issue:168791,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#168784 TestModule: test/test_nestedtensor.py::TestNestedTensorSubclass, #168788 Test: TestNestedTensorSubclass.test_sdpa_backwards, #168789 Test: TestNestedTensorSubclass.test_sdpa_autocast, #168791 Test: TestNestedTensorSubclass.test_dummy_mha_with_nt, #168792 Test: TestNestedTensorSubclass.test_compile_preserves_metadata_cache",42dc18d70a653422d91c740b55e48a6097011275ca86149deded614068a080a6
issue-collapse:0a6f1f82ab79188e,Review thread: 5 issues - issue #168773: TestModule: test/test_matmul_cuda.py::TestMatmulCuda,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"168773, 168774, 168776, 168777, 168778",,,,"module: linear algebra, module: rocm","module: linear algebra, module: rocm, rocm-skipped-tests, triaged",thread:issue:168773:issue:168774:issue:168776:issue:168777,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#168773 TestModule: test/test_matmul_cuda.py::TestMatmulCuda, #168774 Test: TestMatmulCuda.test_cublas_addmm, #168776 Test: TestMatmulCuda.test_cublas_batch_invariance_blackwell, #168777 Test: TestMatmulCuda.test_grouped_gemm_compiled, #168778 Test: TestMatmulCuda.test_mm_bmm_dtype_overload",c67b62cd3c2183572920fb1a10aa682c93f36c171d2476b3152edd57c894f121
issue-collapse:eb2e8bd41538cd1e,Review thread: 5 issues - issue #168411: TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorChunked,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"168411, 168412, 168413, 168414, 168415",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168411:issue:168412:issue:168413:issue:168414,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#168411 TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorChunked, #168412 Test: TestShardedTensorChunked.test_complete_world_size, #168413 Test: TestShardedTensorChunked.test_partial_world_size, #168414 Test: TestShardedTensorChunked.test_new_group, #168415 Test: TestShardedTensorChunked.test_multiple_local_shards",6f15d4ea107fab81eefa6b3c4be3a363b5b4ae4797852039e5bf2ac6514be8e3
issue-collapse:6f5a7743dee04dfa,Review thread: 5 issues - issue #121596: Memory Usage of Conv2d Rapidly Increases when Params >1G,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"26950, 52439, 94869, 121596, 123157",,,,"module: convolution, module: cublas, module: cuda, module: cudnn, module: dependency bug, module: memory usage","enhancement, feature, matrix multiplication, module: convolution, module: cublas, module: cuda, module: cudnn, module: dependency bug, module: memory usage, triaged",thread:issue:121596:issue:123157:issue:26950:issue:52439,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#26950 conv2d Memory usage is too large； pytorch 1.1.0, #52439 [proposal] Way to export/display the convolution algorithm found with cudnn.benchmark = True + allow to override cudnn algo choice with conv/matmul module/function algo/hint arguments, #94869 aten::cudnn_convolution chooses different conv implementation given the same inputs., #121596 Memory Usage of Conv2d Rapidly Increases when Params >1G, #123157 Add support for CUBLAS_COMPUTE_16F for GEMM opeartions in cudaBLAS",db61760a73106732ed1aea4d1332e40fe059850bd63ceb0bb76305af448a6d27
issue-collapse:93f526b3dbef4cbc,Review thread: 5 issues - issue #116175: Performance for `TransformerEncoderLayer` and `TransformerDecoderLayer` drops severely if pytorch version changed from `pt1.10` to `pt2.0`,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"94428, 116175, 118972, 121193, 122660",,,,"module: nn, module: performance","module: nn, module: performance, triaged",thread:issue:116175:issue:118972:issue:121193:issue:122660,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#94428 nn.TransformerEncoderLayer fastpath (BetterTransformer) is much slower with src_key_padding_mask, #116175 Performance for `TransformerEncoderLayer` and `TransformerDecoderLayer` drops severely if pytorch version changed from `pt1.10` to `pt2.0`, #118972 Inconsistent usage of `attn_mask` and `is_causal` arguments, #121193 Detailed documentation with an example for the causal mask in nn.TransformerEncoder, #122660 Deprecate `torch.nn.MultiHeadAttention` and `torch.nn.Transformer`-related modules",1809c157d2a397a7355a58d5acad780abed0a5cdd3f02e21f9a6a90006a44d26
issue-collapse:71f588560def5ba2,Review thread: 5 issues - issue #107112: from_blob python api,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"30387, 34646, 34649, 45901, 107112",,,,"module: docs, module: memory usage, module: numpy, module: python frontend, module: serialization","enhancement, module: docs, module: memory usage, module: numpy, module: python frontend, module: serialization, triaged",thread:issue:107112:issue:30387:issue:34646:issue:34649,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 5 open issues; raw issue rows remain available for audit.,"#30387 [docs] Missing docs for Storage.from_buffer, #34646 Support creating a CPU tensor from ctypes pointer in Python / from_blob(ptr, shape, strides, dtype), #34649 Confusing error message of tensor constructor when passing a storage, #45901 Add support for reading the whole file in from_file, #107112 from_blob python api",b60e9fdbe86d73de5a08808f0b5eb151ae27daf9d4af4c9732b2fcbff0675b8c
issue-collapse:645f5c15b5c5667b,Review thread: 5 issues - issue #103761: Merge type stubs for `torch.nn.functional`,related issue thread,stale/low urgency,2,high,work_thread_graph,5,0,0,0,"103761, 103787, 103798, 103800, 103803",,,,"module: typing, oncall: jit","enhancement, module: typing, oncall: jit, release notes: devx, triaged",thread:issue:103761:issue:103787:issue:103798:issue:103800,Handle this module: typing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 5 open issues; raw issue rows remain available for audit.,"#103761 Merge type stubs for `torch.nn.functional`, #103787 Generate complete annotations for `torch._C._nn`, #103798 Type misalignments in `nn.functional` (also JIT related), #103800 Mis-annotated return for `F._no_grad_embedding_renorm_` (also JIT related), #103803 Support `Sequence` type in JIT",152bb626c20d2a6b1b3a67333f9a7856d4b52ee50fa46553f19c30d2a8cdd1be
issue-collapse:108cc4edc0c604d3,Review thread: 4 issues - issue #62614: LazyTensor recomputes intermediate tensor in fragmented computation,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"62614, 62616, 62619, 62675",,,,module: lazy,"LazyTensor_nvfuser_integration, module: lazy, triaged",thread:issue:62614:issue:62616:issue:62619:issue:62675,Handle this module: lazy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#62614 LazyTensor recomputes intermediate tensor in fragmented computation, #62616 LazyTensor cuda fallback, #62619 LazyTensor support for fuser integration investigation, #62675 dynamic shapes",6192e47bb0d00bc73d4f861781dee841ad4b6224c545a186a4a76a6a3be2abba
issue-collapse:3bb321184228036b,Review thread: 4 issues - issue #49016: Discrete Cosine Transform,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"49016, 73479, 90560, 93859",,,,"module: nestedtensor, module: serialization, module: sparse, module: vision","feature, module: nestedtensor, module: serialization, module: sparse, module: vision, triaged",thread:issue:49016:issue:73479:issue:90560:issue:93859,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#49016 Discrete Cosine Transform, #73479 Strange case of empty non-coalesced sparse tensor, #90560 [discussion, idea] Batched, vectorized base64 decoding / encoding + maybe RLE decoding / encoding, #93859 Bitwise-perfect method for (de)serializing tensors in base64",6dac0f5f50f5a481d1f66ab74ccd560ebcddcdf223c8f4c969e9bd4102f8fea4
issue-collapse:22bbd856d2bb9798,Review thread: 4 issues - issue #46374: Complex Number support for torch.nn modules,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"46374, 46642, 47052, 81749",,,,"module: complex, module: nn, module: primtorch","complex_autograd, function request, module: complex, module: nn, module: primTorch, triaged",thread:issue:46374:issue:46642:issue:47052:issue:81749,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#46374 Complex Number support for torch.nn modules, #46642 Loss functions for complex tensors, #47052 Activation functions for complex tensors, #81749 BatchNorm for complex tensor",5ba83db82cd795b80842db04e8c6ee3a38e35fb11b4c92a5e5b617d3a3d74f6a
issue-collapse:0e088e5d28aeeaba,Review thread: 4 issues - issue #44901: for loop can't be symbolically traced,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"44901, 51803, 54602, 78487",,,,module: fx,"module: fx, triaged",thread:issue:44901:issue:51803:issue:54602:issue:78487,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#44901 for loop can't be symbolically traced, #51803 torch.fx.symbolic_trace fails on torch.arange with input-dependent size, #54602 if cannot be symbolically traced, #78487 torch.fx: symbolic_trace: ones() received an invalid combination of arguments",2424965f9aa5467b24489f9fe07425764cacdb17268302485c34cd9435235b4b
issue-collapse:6a081a1c64131ab5,Review thread: 4 issues - issue #35041: Make operator registrations truly commutative using priority,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"35041, 62032, 64407, 65151",,,,"module: autograd, module: bc-breaking, module: custom-operators, module: dispatch, module: internals, module: molly-guard, module: pt2-dispatcher, oncall: pt2","better-engineering, feature, module: autograd, module: bc-breaking, module: custom-operators, module: dispatch, module: internals, module: molly-guard, module: pt2-dispatcher, oncall: pt2, topic: bc breaking, triaged",thread:issue:35041:issue:62032:issue:64407:issue:65151,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#35041 Make operator registrations truly commutative using priority, #62032 Create a boxed fallback / template recipe for Autograd that forwards, but errors on backwards, #64407 Create a boxed/templated ADInplaceOrView kernel, #65151 Registering a global fallback for all operators that defaults us to assuming that autograd is not implemented",032ff74da5640e91f23939a8a2ec202ecffaa76a69dbedb3539d4988de21a233
issue-collapse:0e910e68b785ede7,Review thread: 4 issues - issue #31394: torch.scatter_logsumexp,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"31394, 80439, 83980, 94083",,,,module: scatter & gather ops,"enhancement, function request, module: scatter & gather ops, triaged",thread:issue:31394:issue:80439:issue:83980:issue:94083,Handle this module: scatter & gather ops stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#31394 torch.scatter_logsumexp, #80439 scatter_reduce choosed indices, #83980 Scatter min/max reduce operation that returns the corresponding indices, #94083 Tracker for `scatter_reduce` additional reduction options requests",7c456395f2b6c4c786ffd8dcb18a536bb15ebd7ff0d11bc59fdb75d375e2b8fd
issue-collapse:28e52d1992fc4242,Review thread: 4 issues - issue #30421: [jit] Module references aren't preserved,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"30421, 31129, 32463, 33034",,,,oncall: jit,"low priority, oncall: jit, triaged",thread:issue:30421:issue:31129:issue:32463:issue:33034,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#30421 [jit] Module references aren't preserved, #31129 [jit] Python objects as arguments are not mutated, #32463 [JIT] Make `torch.jit.script` work on all objects which we can represent as IValues, #33034 Scripting fails to preserve attribute aliasing",88e929fe154a860dadfcb2f8562d201439d579ae6bc59fd9dad9341d920d0150
issue-collapse:8b8265513e3bb0a0,Review thread: 4 issues - issue #21760: [FR] faster reduce sum on expanded/unfolded tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"21760, 57121, 60295, 88002",,,,"module: linear algebra, module: numpy, module: performance","function request, module: linear algebra, module: numpy, module: performance, triaged",thread:issue:21760:issue:57121:issue:60295:issue:88002,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,5,Represents 4 open issues; raw issue rows remain available for audit.,"#21760 [FR] faster reduce sum on expanded/unfolded tensors, #57121 einsum ""jk,ijkl->il"" is ~16x slower than numpy, #60295 Optimize torch.einsum, #88002 Einsum Optimization Tracker",b5970817e48ba464763fa0cf519f82c5f9def10865d011599bc4b561e079797d
issue-collapse:aeb597e66acd044b,Review thread: 4 issues - issue #19177: 8 tests in test_c10d fail when running all tests in one command,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"19177, 41337, 53623, 55340",,,,"module: tests, oncall: distributed","better-engineering, module: tests, oncall: distributed, triaged",thread:issue:19177:issue:41337:issue:53623:issue:55340,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,6,Represents 4 open issues; raw issue rows remain available for audit.,"#19177 8 tests in test_c10d fail when running all tests in one command, #41337 Distributed tests fail with pytest, #53623 test/distributed/test_c10d.py::DistributedDataParallelTest only passes if run before other test cases, #55340 [RFC] Plan to deduplicate test_c10d and distributed_test",5b8b4848720c6061495c5951a1e7d3d5ed9cc566c26d0d259d57a94b60cd9cda
issue-collapse:f122f221cabf96fd,Review thread: 4 issues - issue #19092: [RFC] Memory format (aka layout aka NHWC) support,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"19092, 24090, 32078, 37837",,,,"module: internals, module: memory format, module: mkldnn","enhancement, module: internals, module: memory format, module: mkldnn, triaged",thread:issue:19092:issue:24090:issue:32078:issue:37837,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#19092 [RFC] Memory format (aka layout aka NHWC) support, #24090 `suggest_memory_format` has ambiguity & cannot represent intended layout format for corner cases, #32078 Reduce degrees of freedom in strides so that they unambiguously specify layout permutation, #37837 [DRAFT] Channels Last + AMP support plan for 1.6 release",230252a9c382ac3e29915b0e004ba71f87f8f76611b60987ede6317dacb339e2
issue-collapse:e23d4361b2f1a405,Review thread: 4 issues - issue #168851: TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredCUSPARSELT,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"168851, 168852, 168853, 168854",,,,"module: rocm, module: sparse","module: rocm, module: sparse, rocm-skipped-tests, triaged",thread:issue:168851:issue:168852:issue:168853:issue:168854,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#168851 TestModule: test/test_sparse_semi_structured.py::TestSparseSemiStructuredCUSPARSELT, #168852 Test: TestSparseSemiStructuredCUSPARSELT.test_cslt_sparse_mm_alpha, #168853 Test: TestSparseSemiStructuredCUSPARSELT.test_cslt_sparse_mm_alpha_compile_autotune, #168854 Test: TestSparseSemiStructuredCUSPARSELT.test_cslt_sparse_mm_alpha_mixed_dtype",0eca3aa6d1878a2e4349d7177ed560ef3466069ad1c6698f8b26ab5e7abb3155
issue-collapse:bf57b0b4d1237a8b,Review thread: 4 issues - issue #168831: TestModule: test/test_sparse_semi_structured.py::SparseSemiStructuredTensorCompileTest,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"168831, 168832, 168833, 168834",,,,"module: rocm, module: sparse","module: rocm, module: sparse, rocm-skipped-tests, triaged",thread:issue:168831:issue:168832:issue:168833:issue:168834,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#168831 TestModule: test/test_sparse_semi_structured.py::SparseSemiStructuredTensorCompileTest, #168832 Test: SparseSemiStructuredTensorCompileTest.test_mlp_contiguous_relu_compile_cusparselt, #168833 Test: SparseSemiStructuredTensorCompileTest.test_mlp_contiguous_relu_compile_cutlass, #168834 Test: SparseSemiStructuredTensorCompileTest.test_sp24_compile",7c8eb17ae8df33d4c81d49b7f89364abd88ec08369e7678c76595196bbb3ddf4
issue-collapse:69ba65896074ddeb,Review thread: 4 issues - issue #168706: TestModule: test/test_cpp_extensions_aot.py::TestCppExtensionAOT,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"168706, 168707, 168708, 168709",,,,"module: aotinductor, module: rocm, oncall: export, oncall: pt2","module: aotinductor, module: rocm, oncall: export, oncall: pt2, rocm-skipped-tests, triaged",thread:issue:168706:issue:168707:issue:168708:issue:168709,Handle this module: aotinductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#168706 TestModule: test/test_cpp_extensions_aot.py::TestCppExtensionAOT, #168707 Test: TestCppExtensionAOT.test_cublas_extension, #168708 Test: TestCppExtensionAOT.test_cusolver_extension, #168709 Test: TestCppExtensionAOT.test_cuda_dlink_libs",0ade9cd1de179b4952622adca10601bb329373062eae95b75678db7e6dbb076c
issue-collapse:8db74e88b2c72a9e,Review thread: 4 issues - issue #168532: TestModule: test/distributed/test_symmetric_memory.py::SymmMemNegativeTest,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"168532, 168533, 168534, 168535",,,,"module: rocm, module: symm_mem","module: rocm, module: symm_mem, rocm-skipped-tests, triaged",thread:issue:168532:issue:168533:issue:168534:issue:168535,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#168532 TestModule: test/distributed/test_symmetric_memory.py::SymmMemNegativeTest, #168533 Test: SymmMemNegativeTest.test_barrier_timeout, #168534 Test: SymmMemNegativeTest.test_put_signal_timeout, #168535 Test: SymmMemNegativeTest.test_wait_signal_timeout",528c8ce2139147f4cc0cd96865ab49195864d333f4822e723f4e2164b5fb3bf1
issue-collapse:bf2e5b24f4dd2f50,Review thread: 4 issues - issue #117394: [RFC] Dynamo Single Step Graph,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"117394, 120814, 133230, 138217",,,,"module: autograd, module: compiled autograd, module: dynamo, module: optimizer, module: pt2 optimizer, oncall: pt2","feature, module: autograd, module: compiled autograd, module: dynamo, module: optimizer, module: pt2 optimizer, oncall: pt2, triaged",thread:issue:117394:issue:120814:issue:133230:issue:138217,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#117394 [RFC] Dynamo Single Step Graph, #120814 torch.compile() and optimizer UX issues, #133230 [dynamo] torch.compile does not support compiling a module that contains autograd within it., #138217 [PT2] Don't graph break on loss.backward()",d0149e7aafe452c8cce39fbc7ddb0df24bb2fee128093dae2af289aab222b001
issue-collapse:19adfd72d0fff8d0,Review thread: 4 issues - issue #115821: [BUG] missing automatic dtype promotion for out-of-place `torch.scatter_*()` functions,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"5405, 5740, 56975, 115821",,,,"module: docs, module: numpy, module: scatter & gather ops","function request, module: docs, module: numpy, module: scatter & gather ops, triaged",thread:issue:115821:issue:5405:issue:56975:issue:5740,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#5405 scatter_add_ should support scalar source (including Python scalar), #5740 [feature request] scalar input to scatter_add_ like scatter_, #56975 scatter does not accept scalar src=, #115821 [BUG] missing automatic dtype promotion for out-of-place `torch.scatter_*()` functions",09bf3d79aafe6c5ab0f86f29c30f2a94fe46baf73ac99506a241bfdb15854533
issue-collapse:0a9b56ab040929b1,Review thread: 4 issues - issue #114153: [Tracker] Inconsistencies between CPU and GPU computation,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"114153, 132041, 132045, 132047",,,,"module: cuda, module: numerical-stability, module: python frontend","module: cuda, module: numerical-stability, module: python frontend, triaged",thread:issue:114153:issue:132041:issue:132045:issue:132047,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#114153 [Tracker] Inconsistencies between CPU and GPU computation, #132041 `torch.nn.functional.max_unpool2d`'s behavior is different on cpu and gpu on torch 2.5.0.dev20240708+cu121, #132045 `torch.scatter`'s behavior is different on cpu and gpu on torch 2.5.0.dev20240708+cu121, #132047 Inconsistent behavior of gumbel_softmax on cpu vs. gpu",6ac4be993d496b2f5b3fac04a00f86e0e32dd18ab229d0cb8de831d5b6ff2058
issue-collapse:627ba9b5ab164844,Review thread: 4 issues - issue #10386: [feature request] batch_first option in torch.utils.data,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"10386, 49440, 53982, 64881",,,,module: dataloader,"enhancement, feature, module: dataloader, triaged",thread:issue:10386:issue:49440:issue:53982:issue:64881,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 4 open issues; raw issue rows remain available for audit.,"#10386 [feature request] batch_first option in torch.utils.data, #49440 [RFC] DataLoader architecture updates and TarDataset implementation, #53982 [Feature Request] Optionally specify batch dimension in DataLoader and collate_fn, #64881 Add support for GPU based data loading and transformation",0d6e7d3ed089d7bff78477cd029985ceca7b2946b4971fd7ea7963a806efe537
issue-collapse:bc1c88721037ef3a,Review thread: 4 issues - issue #10007: Upgrade Caffe2 Hypothesis,related issue thread,stale/low urgency,2,high,work_thread_graph,4,0,0,0,"9853, 9854, 10007, 10254",,,,,caffe2,thread:issue:10007:issue:10254:issue:9853:issue:9854,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,4,Represents 4 open issues; raw issue rows remain available for audit.,"#9853 TestSequenceOps.test_gather_padding failing, #9854 TestConvolution.test_conv_separate_stride_pad_gradients failing, #10007 Upgrade Caffe2 Hypothesis, #10254 [Caffe2] I don't knw whether GPU support is right, is there something wrong with Hypothesis?",82ed0099ac1b051e5d6950deabd72747f7b3b7bb05cc5a6381279e3308967470
issue-collapse:2080191d0421be4b,Review thread: 3 issues - issue #88144: AvgPool2D output shapes are inconsistent when ceil_mode=True,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"88144, 88464, 95116",,,,"module: bc-breaking, module: correctness (silent), module: nn, module: pooling, module: shape checking","module: bc-breaking, module: correctness (silent), module: nn, module: pooling, module: shape checking, topic: bc breaking, triaged",thread:issue:88144:issue:88464:issue:95116,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#88144 AvgPool2D output shapes are inconsistent when ceil_mode=True, #88464 MaxPool1D output shapes can be negative when ceil_mode=True, #95116 When using `ceil_mode=True`, `torch.nn.AvgPool1d` could get negative shape.",ba46441ebd2fc1ab848b7dd27cdb91cdc39272d84a9973fa0901ab95833466d9
issue-collapse:f3a21308dc4c5d78,Review thread: 3 issues - issue #72960: Refactor/Cleanup lazy IR codegen,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"72960, 73710, 73711",,,,module: lazy,"module: lazy, triaged",thread:issue:72960:issue:73710:issue:73711,Handle this module: lazy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#72960 Refactor/Cleanup lazy IR codegen, #73710 Refactor/Cleanup Lazy Tensor Core, #73711 Refactor/Cleanup LazyTensor, LTCTensorImpl, Data",6866d3af674d0ad6ce3c01ad8308e1ed5b7b4e46e79f2eb11aee4fdd2da49c6e
issue-collapse:8fc2335eb282bec0,Review thread: 3 issues - issue #69991: Composite Compliance Problems Tracker,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"69991, 72911, 84565",,,,"module: __torch_dispatch__, module: autograd, module: internals, module: linear algebra, module: vmap","module: __torch_dispatch__, module: autograd, module: internals, module: linear algebra, module: vmap, tensor subclass, triaged",thread:issue:69991:issue:72911:issue:84565,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#69991 Composite Compliance Problems Tracker, #72911 Vectorized Jacobian and Hessian errors with ffts, #84565 Tensor Subclass that doesn't require grad may wrap a Tensor subclass that requires grad",84c0caf84e3b66025fbd4c7072739e0454ade05a785860eaf7eee3c64a9e5b63
issue-collapse:f1b594e43a817a05,Review thread: 3 issues - issue #68287: [Lazy] Add torch.autograd.Function wrappers for the following ops in order to support dynamic shapes,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"68287, 68288, 68298",,,,,"lazy, triaged",thread:issue:68287:issue:68288:issue:68298,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#68287 [Lazy] Add torch.autograd.Function wrappers for the following ops in order to support dynamic shapes, #68288 List of issues to unblock dynamic shapes in Lazy Tensor, #68298 [Lazy] Add torch.autograd.Function wrappers for the following ops in order to support dynamic shapes",ddf1fc6da90b623584cfed66c147e52dfaf5f554662fb30e9371b4bb0c3396ca
issue-collapse:a075ecf478404e0c,Review thread: 3 issues - issue #65868: torch.mul is not consistent with torch.multiply,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"65868, 66491, 76555",,,,"module: numerical-stability, module: numpy, module: python frontend, module: tests","better-engineering, feature, module: numerical-stability, module: numpy, module: python frontend, module: tests, triaged",thread:issue:65868:issue:66491:issue:76555,Handle this module: numerical-stability stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#65868 torch.mul is not consistent with torch.multiply, #66491 Normalize handling of scalar arguments, #76555 [numpy] Missing Tensor-Scalar support for multiple binary ops",976f93de57554c750b30c4424b498c90fdc35017e452606436cc7dbac9294bcd
issue-collapse:6f2cf6cf46d87ba7,Review thread: 3 issues - issue #65760: Codegen issues with Tensor(a!)? arguments in native_functions.yaml schemas,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"65760, 65788, 73050",,,,"module: amp (automated mixed precision), module: codegen, module: intel","module: amp (automated mixed precision), module: codegen, module: intel, triaged",thread:issue:65760:issue:65788:issue:73050,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#65760 Codegen issues with Tensor(a!)? arguments in native_functions.yaml schemas, #65788 Autocast sometimes discards changes to mutable arguments, #73050 aten::batch_norm schema does not mention mutation of running_mean/running_var",6b571741aad639a0aee93365461100eeb40e05297be5a7d72294d414c4aea121
issue-collapse:1233457578669657,Review thread: 3 issues - issue #65049: Setting a random seed or passing a generator to Modules for deterministic sampling.,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"65049, 73640, 98200",,,,"module: determinism, module: nn, module: random","actionable, enhancement, module: determinism, module: nn, module: random, triaged",thread:issue:65049:issue:73640:issue:98200,Handle this module: determinism stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#65049 Setting a random seed or passing a generator to Modules for deterministic sampling., #73640 [Feature Request] Allow torch.Generator to be passed to torch.nn.init functions, #98200 torch.nn.init functions with `generator` argument",12d3b2294785187f80b5bd4a5e4bd442d13a12612ba24c6c7ea77df83973d9a9
issue-collapse:20fc062dd74d6496,Review thread: 3 issues - issue #53712: Poor support of `Optimizer.add_param_group`,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"53712, 62475, 79888",,,,"module: lrscheduler, module: optimizer","module: LrScheduler, module: optimizer, triaged",thread:issue:53712:issue:62475:issue:79888,Handle this module: lrscheduler stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#53712 Poor support of `Optimizer.add_param_group`, #62475 Learning rate scheduler list index out of range, #79888 `CosineAnnealingWarmRestarts` does not update parameters added with `add_param_group`",8a54f41224b17980d316e537600032615f9cfc3dc5cc2a5684f48feb7ec9188a
issue-collapse:0c0fe3037f92a770,Review thread: 3 issues - issue #52973: Add numerically stable methods to torch.distributions (e.g. logcdf),related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"52973, 69467, 69469",,,,module: distributions,"module: distributions, triaged",thread:issue:52973:issue:69467:issue:69469,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#52973 Add numerically stable methods to torch.distributions (e.g. logcdf), #69467 Distribution `cross_entropy` method, #69469 Distribution `covariance` property",7acf0ec97893ec1ebc0dcc1de39d0488b3be9764d09a8e3deed2e9f11b441d48
issue-collapse:41d5e6c39f6e640f,Review thread: 3 issues - issue #50097: Cannot print 32-bit complex tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"50097, 67324, 74537",,,,"module: complex, module: cuda, module: docs, module: half, module: printing","feature, module: complex, module: cuda, module: docs, module: half, module: printing, triaged",thread:issue:50097:issue:67324:issue:74537,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#50097 Cannot print 32-bit complex tensors, #67324 torch.stft - fill_cuda not implemented for ComplexHalf, #74537 ComplexHalf Coverage Tracker",c6b6ed595b7e6fe9b0c030fd80d72282e25accc8b2bc477f29760d0d0caaab6e
issue-collapse:c838737649cb868d,Review thread: 3 issues - issue #49666: Use rlibm for faster and more accurate floating point operations,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"49666, 52310, 54622",,,,"module: bfloat16, module: complex, module: numerical-stability, module: numpy","enhancement, module: bfloat16, module: complex, module: numerical-stability, module: numpy, triaged",thread:issue:49666:issue:52310:issue:54622,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#49666 Use rlibm for faster and more accurate floating point operations, #52310 Implementation of many complex functions is fast but inaccurate in libc++, #54622 torch.pow returns incorrect value for 0^0j",22294613c8e1f0e45c5083a41c955967d6ddd629651453c5d04220098b04de01
issue-collapse:baf6dccb34fdf416,Review thread: 3 issues - issue #49477: Callbacks may not be automatically synchronized in a single NCCL future chain,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"49477, 49716, 74148",,,,"module: ddp, oncall: distributed","module: ddp, oncall: distributed, topic: performance, triaged",thread:issue:49477:issue:49716:issue:74148,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#49477 Callbacks may not be automatically synchronized in a single NCCL future chain, #49716 Two independent future chains in PowerSGD cannot be kicked off asynchronously, #74148 powersgd can not get linear growth due to extra 2 times allreduce",bb80ab93d3793ef775656e3774c7eca2debf5625e03309a29c501bd984dc7208
issue-collapse:fbbb141c1ad7a619,Review thread: 3 issues - issue #43947: torch.cuda.synchronize Influence distributed training,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"43947, 44103, 63618",,,,"module: cuda, module: cudnn, module: data parallel, module: performance","module: cuda, module: cudnn, module: data parallel, module: performance, triaged",thread:issue:43947:issue:44103:issue:63618,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#43947 torch.cuda.synchronize Influence distributed training, #44103 Item seems to affect the backward process of DataParallel, #63618 C++ torch::cuda::synchronize speeds up training",f0a2dba6742fc4464b13e1abc1db9aeffd3f0b4e12ab4d166290a10a2013bd8b
issue-collapse:4db1d2fd636f66ce,Review thread: 3 issues - issue #42502: [proposal] batch mode for randperm,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"42502, 71409, 77140",,,,"module: multithreading, module: performance, module: random","enhancement, feature, module: multithreading, module: performance, module: random, triaged",thread:issue:42502:issue:71409:issue:77140,Handle this module: multithreading stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#42502 [proposal] batch mode for randperm, #71409 Random Shuffle along Axis, #77140 torch.randperm uses too much cpu, but not efficient.",930f56a3cf6c8a639006f4f9007e0fd54f58aa024ab3546617c2f4d8d4623b87
issue-collapse:716d473c77672fd7,Review thread: 3 issues - issue #41186: test_autograd failures on Power,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"41186, 41400, 57617",,,,"module: autograd, module: complex, module: cuda, module: linear algebra, module: power, module: tests, oncall: jit","complex_autograd, module: POWER, module: autograd, module: complex, module: cuda, module: linear algebra, module: tests, oncall: jit, triaged",thread:issue:41186:issue:41400:issue:57617,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#41186 test_autograd failures on Power, #41400 test_jit.py fails on Power, #57617 complex128 autograd failures on PPC",525988337d5fc9050783ea7fd81a011c6739a77a6e55467923f831b37d352ee3
issue-collapse:e3e292c448bdcea5,Review thread: 3 issues - issue #30929: How to set not to build libtorch_cpu.so and libmkl_*.so dependencies?,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"30929, 32407, 59218",,,,"module: build, module: docs, module: linear algebra, module: mkl","module: build, module: docs, module: linear algebra, module: mkl, triaged",thread:issue:30929:issue:32407:issue:59218,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#30929 How to set not to build libtorch_cpu.so and libmkl_*.so dependencies?, #32407 Build without MKL is not possible when MKL is installed, #59218 Clarify BLAS configuration option",171f7062a2ca0211b5b1343c87d24af5bad4de5a3a00670960004889950fcafc
issue-collapse:b4cc589b84617c47,Review thread: 3 issues - issue #30532: Version 1.3 no longer supporting Tesla K40m?,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"30532, 31007, 61411",,,,"module: binaries, module: cuda, module: docs","module: binaries, module: cuda, module: docs, triaged",thread:issue:30532:issue:31007:issue:61411,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#30532 Version 1.3 no longer supporting Tesla K40m?, #31007 GPU version of minimal example for libtorch fails with ""no kernel image is available..."", #61411 I do have the GPU, but the print(torch.cuda.is_available()) always shows False",4cf96d0ba3ccc7b54bd7b5503f083d658a2ec469010db661b821a149ad83dec3
issue-collapse:30e54b0b0e90a533,Review thread: 3 issues - issue #29554: [feature request] Print some measure of fragmentation at CUDA out-of-memory,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"29554, 36990, 40989",,,,"module: cuda, module: docs, module: memory usage","enhancement, module: cuda, module: docs, module: memory usage, triaged",thread:issue:29554:issue:36990:issue:40989,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#29554 [feature request] Print some measure of fragmentation at CUDA out-of-memory, #36990 [docs] Explain active_bytes in torch.cuda.memory_stats and Cuda Memory Management, #40989 Difference between allocated and reserved CUDA memory",db63176f93d38db9f29dea82db33c802f38cf509d041e35ebd491fe294996d77
issue-collapse:7bcfce90e9e61ef6,Review thread: 3 issues - issue #29093: error: 'SO_REUSEPORT' was not declared in this scope,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"29093, 29739, 35363",,,,"module: build, module: dependency bug, module: third_party","module: build, module: dependency bug, module: third_party, triaged",thread:issue:29093:issue:29739:issue:35363,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#29093 error: 'SO_REUSEPORT' was not declared in this scope, #29739 [build] gcc 7.4 needs CMAKE_CXX_FLAGS=""-std=gnu++11"", #35363 NCCL version upgrade for PyTorch",a53494d4aa1958e7a4531232d157b2bb581c638d21d6ce27c9e5229fb19a495d
issue-collapse:5eb595f489801baa,Review thread: 3 issues - issue #27610: [FR][RFC] Build a serving framework to host and serve trained PyTorch models,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"27610, 30373, 33354",,,,"module: hub, oncall: jit","enhancement, feature, module: hub, oncall: jit, triaged",thread:issue:27610:issue:30373:issue:33354,Handle this module: hub stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#27610 [FR][RFC] Build a serving framework to host and serve trained PyTorch models, #30373 Refactor (a bit) `torch.hub(.load)`, #33354 Long torchscript warmup times can be problematic for production serving",b6d4596c13c69834e7d05a7766b15af7e8cfb17c2b4c8b51c82aa5ada486e178
issue-collapse:d1812a966b4b982a,Review thread: 3 issues - issue #25032: [FYI] NestedTensor Project Progress,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"25032, 25310, 59629",,,,"module: docs, module: nestedtensor, module: nn, module: rnn","feature, module: docs, module: nestedtensor, module: nn, module: rnn, triaged",thread:issue:25032:issue:25310:issue:59629,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#25032 [FYI] NestedTensor Project Progress, #25310 Handling of packed_sequence by activation functions and linear layers, #59629 Potentially misleading note in documentation for PackedSequence",fc0e01627cd552d0909a41a3769488ced4ba445e61e0f2844dca224c18664ad0
issue-collapse:f17baf70cefd1e44,Review thread: 3 issues - issue #22281: Box constraints for optimizers,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"6564, 22281, 30386",,,,module: optimizer,"feature, module: optimizer, proposal accepted, todo, triaged",thread:issue:22281:issue:30386:issue:6564,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#6564 [Feature Request] Optimization with constraint (L-BFGS-B), #22281 Box constraints for optimizers, #30386 Exponentiated gradient descent",1ebb8da0ab81136652e8217f155fd8178aa1483f7ac7e0bd2f65ce1d8b99901a
issue-collapse:40b83b0473a7e463,Review thread: 3 issues - issue #20117: [FR] [RFC] add Sequential.append & .extend,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"20117, 71249, 71329",,,,"module: docs, module: nn","feature, module: docs, module: nn, triaged",thread:issue:20117:issue:71249:issue:71329,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#20117 [FR] [RFC] add Sequential.append & .extend, #71249 [docs] nn.Sequential docs should list member functions, #71329 [RFC] Implement array methods (extend, insert, pop, etc) for container classes",8f9b8f2fe793a4a4e876a7cf747375a8dea9260648eebd30934ee6681e1b0b6d
issue-collapse:2c2d143fab1225cd,Review thread: 3 issues - issue #168801: TestModule: test/test_nn.py::TestNNDeviceType,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"168801, 168806, 168807",,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168801:issue:168806:issue:168807,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#168801 TestModule: test/test_nn.py::TestNNDeviceType, #168806 Test: TestNNDeviceType.test_upsampling_64bit_indexing_channels_last, #168807 Test: TestNNDeviceType.test_upsamplingNearest2d_launch_fail",464da208c3f6d60c47b270a77063fdbb080bfbeaa284bdc9614ac9b2ebcaac2f
issue-collapse:59d2d4a889756c16,Review thread: 3 issues - issue #168720: TestModule: test/test_cuda.py::TestCudaMallocAsync,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"168720, 168721, 168722",,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168720:issue:168721:issue:168722,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#168720 TestModule: test/test_cuda.py::TestCudaMallocAsync, #168721 Test: TestCudaMallocAsync.test_memory_profiler_viz, #168722 Test: TestCudaMallocAsync.test_device_memory_used",0448fb24861e1e944741893abc66aa284623db09752f2829547571aa0c9c19da
issue-collapse:07d86a6d7c4a2595,Review thread: 3 issues - issue #168644: TestModule: test/inductor/test_multi_kernel.py::MultiKernelTest,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"168644, 168645, 168646",,,,"module: inductor, module: rocm, oncall: pt2","module: inductor, module: rocm, oncall: pt2, rocm-skipped-tests, triaged",thread:issue:168644:issue:168645:issue:168646,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#168644 TestModule: test/inductor/test_multi_kernel.py::MultiKernelTest, #168645 Test: MultiKernelTest.test_triton_gemm, #168646 Test: MultiKernelTest.test_triton_relu_fused_gemm",558cb9320e751484c7c05a0f201e424530744ca8ef786dc13359ebf21eb806db
issue-collapse:8d2facfb28b1ec88,Review thread: 3 issues - issue #168475: TestModule: test/distributed/test_composability.py::ComposabilityTest,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"168475, 168476, 168477",,,,"module: ddp, module: fsdp, module: rocm","module: ddp, module: fsdp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168475:issue:168476:issue:168477,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#168475 TestModule: test/distributed/test_composability.py::ComposabilityTest, #168476 Test: ComposabilityTest.test_pp_fsdp, #168477 Test: ComposabilityTest.test_pp_fsdp_unshard_reshard_runtime",ae742f025ff57002b5b01e2209cddc3f3c410b256dbead033b9a6c2fdcb92204
issue-collapse:b952db6594521468,Review thread: 3 issues - issue #16717: Performance regression on CPU from 0.4.1 to 1.0.0 on ResNet inference,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"5388, 16717, 20367",,,,"module: cpu, module: cuda, module: internals, module: performance, module: single threaded","in progress, module: cpu, module: cuda, module: internals, module: performance, module: single threaded, quansight-nack, triaged",thread:issue:16717:issue:20367:issue:5388,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#5388 Perf regression: indexing 1-d tensor, #16717 Performance regression on CPU from 0.4.1 to 1.0.0 on ResNet inference, #20367 Overhead performance regression over time umbrella issue.",412a3e5c53debb887c0175c90406b67fae612f028b036744ef5ff535211a5c4d
issue-collapse:c90dcf498c768c12,Review thread: 3 issues - issue #151936: fully_shard() for huggingface 72B model: pytorch caches too much GPU memory,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"151936, 154984, 155171",,,,"module: activation checkpointing, module: dtensor, module: fsdp, oncall: distributed","module: activation checkpointing, module: dtensor, module: fsdp, oncall: distributed, triaged",thread:issue:151936:issue:154984:issue:155171,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#151936 fully_shard() for huggingface 72B model: pytorch caches too much GPU memory, #154984 [FSDP2] Slower Convergence with fully_shard() Compared to DDP during Qwen2-VL Fine-Tuning, #155171 [AC] torch.utils.checkpoint.CheckpointError from HF qwen2",4100fedea669d2576bd9a7db0eb9c31cfefef7b40d1e38227ebd3c73d2f2e91f
issue-collapse:fc00e23dea646ba3,Review thread: 3 issues - issue #15070: In-place operations on `.data` or `.detach()` of sparse tensor doesn't update the original tensor,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"15070, 29814, 29816",,,,"module: optimizer, module: sparse","module: optimizer, module: sparse, triaged",thread:issue:15070:issue:29814:issue:29816,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#15070 In-place operations on `.data` or `.detach()` of sparse tensor doesn't update the original tensor, #29814 SGD fails on sparse matrix, #29816 AdamSparse fails to run",536363aebab9c659934326a309d4de60c2d0504d6ff13882537d7b53a3ab3f60
issue-collapse:c891e3eb6f316d1a,Review thread: 3 issues - issue #135860: Lobpcg for complex gives wrong results.,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"135860, 137767, 156152",,,,"module: complex, module: correctness (silent), module: cpu, module: edge cases, module: linear algebra, module: nans and infs","low priority, module: NaNs and Infs, module: complex, module: correctness (silent), module: cpu, module: edge cases, module: linear algebra, triaged",thread:issue:135860:issue:137767:issue:156152,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#135860 Lobpcg for complex gives wrong results., #137767 [CPU]Tracking complex datatype issues, #156152 Inconsistent `torch.rsqrt` results on complex128 between CPU and CUDA",54513b63d8de8a214fcd0d42a9a39acec33daa969d42be769093db10e9f1fd2b
issue-collapse:1d7938df3f78c986,Review thread: 3 issues - issue #13222: Memory inefficient in batched matmul when requiring gradients,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"13222, 56891, 61650",,,,"module: linear algebra, module: performance","enhancement, module: linear algebra, module: performance, triaged",thread:issue:13222:issue:56891:issue:61650,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#13222 Memory inefficient in batched matmul when requiring gradients, #56891 Batched SVD_LOWRANK being much slower than loop implementation (both CPU and GPU), #61650 Move `{svd,pca}_lowrank` into `torch.linalg`",0c8284df72bbad087c2d466dba03462897159ca146317a9fb5984b22491987d6
issue-collapse:6f3fc375fe4b29bd,Review thread: 3 issues - issue #13207: [pytorch] [feature request] Error out if the needed GPU device capability is absent in runtime,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"4716, 13207, 17966",,,,"module: build, module: cuda, module: error checking, module: ux","module: build, module: cuda, module: error checking, module: ux, triaged",thread:issue:13207:issue:17966:issue:4716,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#4716 Compilation issue: problem with GPU capability check, #13207 [pytorch] [feature request] Error out if the needed GPU device capability is absent in runtime, #17966 torch.cuda.is_available() returns misleading value",7dc884273232d1119e8c58530a945bc2571f697f384b893b33c85202e5b06ea2
issue-collapse:338b0dbcee514d3f,"Review thread: 3 issues - issue #131312: Pytorch Error: ImportError: /home/dp/anaconda3/envs/ComfyUI/lib/python3.11/site-packages/torch/lib/../../nvidia/cusparse/lib/libcusparse.so.12: undefined symbol: __nvJitLinkComplete_12_4, version libnvJitLink.so.12",related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"131312, 138460, 141014",,,,"module: binaries, module: cuda, module: rocm","module: binaries, module: cuda, module: rocm, triaged",thread:issue:131312:issue:138460:issue:141014,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#131312 Pytorch Error: ImportError: /home/dp/anaconda3/envs/ComfyUI/lib/python3.11/site-packages/torch/lib/../../nvidia/cusparse/lib/libcusparse.so.12: undefined symbol: __nvJitLinkComplete_12_4, version libnvJitLink.so.12, #138460 CUDA Binary dependency chain is wrong, leading to bad binary packaging, #141014 ROCm: libamdhip64.so is linked twice when building a torch C++ program",cc128ce4cc1a08953a0ac780816a7b3592642c416cfa1bb39a95efc37f95b5d7
issue-collapse:0f46f41767758c1e,Review thread: 3 issues - issue #12983: Fail to Run Caffe2 with Successful Build: This caffe2 python run does not have GPU support,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"9484, 9604, 12983",,,,,caffe2,thread:issue:12983:issue:9484:issue:9604,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#9484 WARNING:root:This caffe2 python run does not have GPU support. Will run in CPU only mode. WARNING:root:Debug message: libcurand.so.9.0: cannot open shared object file: No such file or directory Segmentation fault (core dumped), #9604 WARNING:root:This caffe2 python run does not have GPU support. Will run in CPU only mode. CRITICAL:root:Cannot load caffe2.python. Error: libcaffe2.so: cannot open shared object file: No such file or directory, #12983 Fail to Run Caffe2 with Successful Build: This caffe2 python run does not have GPU support",dec807539b805036d195eafdcdaf3693b386982958cb9fa0dabbb0cbd85c1078
issue-collapse:3611245a1cb3ee68,Review thread: 3 issues - issue #124016: Expand stale management to Issues,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"124016, 124222, 125626",,,,module: devx,"module: devx, triaged",thread:issue:124016:issue:124222:issue:125626,Handle this module: devx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#124016 Expand stale management to Issues, #124222 Bot to recomment on high pri issues if no activity, #125626 [BE] Autoclose issues that haven't been updated in X months",f056b9a96926f3ababd31a81a5b9248d48c22ec9a07e776bac6d8b05c6a154df
issue-collapse:d5150ef2a4a2084f,Review thread: 3 issues - issue #122094: Investigate torch.compile Windows support.,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"122094, 124245, 135927",,,,"module: inductor, module: windows, oncall: cpu inductor, oncall: pt2","intel, module: inductor, module: windows, oncall: cpu inductor, oncall: pt2, triaged, upstream triton",thread:issue:122094:issue:124245:issue:135927,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#122094 Investigate torch.compile Windows support., #124245 [RFC] Add new CPP builder for inductor on pytorch Windows, #135927 [Discuss] Enable Windows inductor UTs and fix its timeout issue.",5ea59a13ae7775f67b6064ac0c5cea95154ef577d68a18eec53074661ff408e6
issue-collapse:2ee805f672894877,Review thread: 3 issues - issue #117736: Illegal memory access for nn.Conv2d when intermediate output has size larger than 2**31,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"95024, 117736, 155218",,,,"module: 64-bit, module: convolution, module: cuda, module: cudacachingallocator, module: cudnn, module: edge cases, module: performance","module: 64-bit, module: CUDACachingAllocator, module: convolution, module: cuda, module: cudnn, module: edge cases, module: performance, triaged",thread:issue:117736:issue:155218:issue:95024,Handle this module: 64-bit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#95024 cuDNN doesn't support convolutions with more than `INT_MAX` elements and native kernel uses too much memory, #117736 Illegal memory access for nn.Conv2d when intermediate output has size larger than 2**31, #155218 3D Conv Slow with Large Tensors (> 2**31) Elements",00039f2b4067e117bdc9fd6f00020f0114bac7fa3403c1d29fcfebcb28bcaf29
issue-collapse:efccf0f564acea4c,Review thread: 3 issues - issue #11578: Request to import pytest in test/*.py,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"11578, 52984, 54006",,,,"module: ci, module: tests","module: ci, module: tests, tracker, triaged",thread:issue:11578:issue:52984:issue:54006,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#11578 Request to import pytest in test/*.py, #52984 Tracker: pytest-related test improvements, #54006 enable parallel test execution for GPU tests",61d1430b6602d863bf8e27fb8d189cd42821a9176aff2ba38e9be2152ed7e069
issue-collapse:760ea2369f739894,Review thread: 3 issues - issue #11389: [distributions] Torch distribution samplers slow on expanded parameters,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"11389, 18906, 30968",,,,module: distributions,"module: distributions, todo, triaged",thread:issue:11389:issue:18906:issue:30968,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#11389 [distributions] Torch distribution samplers slow on expanded parameters, #18906 Provide option to use alias method in Categorical.sample(), #30968 Categorical.sample too slow",782e78ade905f10d318f161b7f8c99e2d4c7686eb32dfe19d0c896e57b97a9cf
issue-collapse:6e55197722ef4e56,Review thread: 3 issues - issue #11202: [pytorch][feature request] Cosine distance / simialrity between samples of own tensor or two tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"11202, 28119, 46169",,,,"module: distance functions, module: nn, module: norms and normalization, module: numpy","function request, module: distance functions, module: nn, module: norms and normalization, module: numpy, triaged",thread:issue:11202:issue:28119:issue:46169,Handle this module: distance functions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#11202 [pytorch][feature request] Cosine distance / simialrity between samples of own tensor or two tensors, #28119 Provide function similar to cdist that returns dist_p^p, #46169 Squared 2-norm pdist (as available in SciPy / Faiss)",4decde6823c9c1ecd7c4ddb118f15769d77b0a6bd2d5ad1d54bdcd502f443c2b
issue-collapse:7549ab2616637fda,Review thread: 3 issues - issue #11071: undefined reference to 'caffe2::Caffe2FlagsRegistry[abi:cxx11]()',related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"11071, 15004, 39947",,,,module: build,"caffe2, module: build, triaged",thread:issue:11071:issue:15004:issue:39947,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#11071 undefined reference to 'caffe2::Caffe2FlagsRegistry[abi:cxx11]()', #15004 [Caffe2] How to link Caffe2 in cmake file for C++ compilation?, #39947 [Caffe2] compilation in c++ undefined reference to `caffe2",eb74f5dc5aeda6ecf0fcf5a830d4e62a38fe897068ea801dd1ebc7029c44426b
issue-collapse:5ea78bb3209e11b3,Review thread: 3 issues - issue #108174: FlashAttentionV2 will OOM when building on ci/cd with default settings,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"108174, 108175, 109770",,,,"module: ci, module: cuda, module: sdpa, module: windows","module: ci, module: cuda, module: sdpa, module: windows, triaged",thread:issue:108174:issue:108175:issue:109770,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#108174 FlashAttentionV2 will OOM when building on ci/cd with default settings, #108175 Enable FlashAttentionV2 on Windows, #109770 Slow performance when running torch.jit traced model with Flash Attention using libtorch on Windows",270a66587253768453a0144db99e733d8ff9731e5d38463fddd545fa78fd0e3d
issue-collapse:d40dc4895fca7f50,Review thread: 3 issues - issue #108108: [BC BREAKING] Change default behavior of scaled_dot_product_attention's causal masking alignment,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"108108, 110681, 110702",,,,"module: bc-breaking, module: nn, module: python frontend, module: sdpa","feature, module: bc-breaking, module: nn, module: python frontend, module: sdpa, topic: bc breaking, triaged",thread:issue:108108:issue:110681:issue:110702,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#108108 [BC BREAKING] Change default behavior of scaled_dot_product_attention's causal masking alignment, #110681 [RFC] Scaled Dot Product Attention API Changes, #110702 sliding_window attention in scaled_dot_product",e7ebeef5d9a2edd8635b63f10e67c78e01a615b2ffcc252e194192456f37ed10
issue-collapse:7b2da2dd1cb4c934,Review thread: 3 issues - issue #107909: Provide a `reset_parameters()` method for MultiheadAttention to support FSDP meta device initializtion,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"66335, 71404, 107909",,,,"module: fsdp, module: fx, module: nn","module: fsdp, module: fx, module: nn, triaged",thread:issue:107909:issue:66335:issue:71404,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#66335 `torch.fx.symbolic_trace()` loses module class information, #71404 [feature request] [discussion] Generalize / recommend behavior of reset_parameters (and potentially rename), #107909 Provide a `reset_parameters()` method for MultiheadAttention to support FSDP meta device initializtion",7f5bf378c5e1930823d485ce955dfa9592618b81338d05c64717f1a6dc6734d9
issue-collapse:d9d0def952f15f4c,Review thread: 3 issues - issue #106713: Dev Container Support for PyTorch,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"106713, 106717, 106718",,,,,"better-engineering, triaged",thread:issue:106713:issue:106717:issue:106718,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#106713 Dev Container Support for PyTorch, #106717 Provide .devcontainer PyTorch - MPS environment, #106718 Add AMD image to the .devcontainer spec",1f1182409f512709547d966b7e752b4d9a31a5668b1d589ce54a7d43a93f54bb
issue-collapse:25e7f8234ee4eb56,Review thread: 3 issues - issue #106469: Extreme slowdown of torch.mm for certain sizes and strides with bfloat16,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"106469, 106485, 146817",,,,"module: bfloat16, module: cublas, module: cuda, module: cudnn, module: linear algebra, module: nn, module: performance","matrix multiplication, module: bfloat16, module: cublas, module: cuda, module: cudnn, module: linear algebra, module: nn, module: performance, triaged",thread:issue:106469:issue:106485:issue:146817,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 3 open issues; raw issue rows remain available for audit.,"#106469 Extreme slowdown of torch.mm for certain sizes and strides with bfloat16, #106485 Increasing batch size makes network forward 1000 times slower, #146817 BF16 linear(matmul) operator 100x slower on odd matrix dimension sizes on A100",a075fcd51c94262d134fd4ad9a0d3ffd0acd4a47404e15c1a7a4401499e587dd
issue-collapse:36fbe6de58581c03,Review thread: 3 issues - issue #104472: Investigate numerical stability of forward-mode AD of some foreach functions,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"58833, 68041, 104472",,,,"module: forward ad, module: mta, module: optimizer, module: performance","module: forward ad, module: mta, module: optimizer, module: performance, tracker, triaged",thread:issue:104472:issue:58833:issue:68041,Handle this module: forward ad stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#58833 Foreach Functions Tracking Issue, #68041 [RFC] APEX style fused optimizers in PyTorch, #104472 Investigate numerical stability of forward-mode AD of some foreach functions",7dc9a3949f5fc69d3d313deee75900b064f3bd1064c1c17150f299edff51bdfa
issue-collapse:ea178e68d001e28b,Review thread: 3 issues - issue #104113: Documentation building fails due to torchgen,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"70608, 104113, 144206",,,,"module: build, module: docs","actionable, module: build, module: docs, triaged",thread:issue:104113:issue:144206:issue:70608,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#70608 Cannot compile C++ documentation: Sphynx assertion, #104113 Documentation building fails due to torchgen, #144206 Error when compiling document",2da44d948bfcf285dcdce4786e97a9ae2fceb69b1750c25a7b4eb2fbb7e35b99
issue-collapse:2f1ce2ed3149acfb,Review thread: 3 issues - issue #102977: raise `RuntimeError` faster when loading an object with a torch CUDA tensor on a CPU-only machine,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"16797, 43369, 102977",,,,"module: nn, module: pickle, module: serialization","actionable, feature, module: nn, module: pickle, module: serialization, todo, triaged",thread:issue:102977:issue:16797:issue:43369,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#16797 Should be a way to unpickle an object with a torch cuda tensor on a CPU-only machine when using plain ""pickle"", #43369 torch.load(.., map_location='cpu') fails when unserializing cuda tensors on a cpu-only device serialized with pickle, #102977 raise `RuntimeError` faster when loading an object with a torch CUDA tensor on a CPU-only machine",66f63704267b162c8c4fef4b70a1ff6358b20e3ac4e3a971aa214fe6c7daf46d
issue-collapse:47486371aacc8bc6,Review thread: 3 issues - issue #101314: Shared library loading logic breaks when CUDA packages are installed in a non-standard location,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"101314, 117350, 121207",,,,"module: bazel, module: build, module: cuda","module: bazel, module: build, module: cuda, topic: build, triaged",thread:issue:101314:issue:117350:issue:121207,Handle this module: bazel stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#101314 Shared library loading logic breaks when CUDA packages are installed in a non-standard location, #117350 CUDA deps cannot be preloaded under Bazel, #121207 can not find all the deps when install the native CUDA-toolkit and torch in Redhat-like systems",5d9663e669a3ec1c5a1984cb2d7f4fe3e0b36cdc9928b47741c7c80ae3da733d
issue-collapse:db4e4734d2656a19,Review thread: 3 issues - issue #100904: Revise glossary,related issue thread,stale/low urgency,2,high,work_thread_graph,3,0,0,0,"100904, 104623, 105457",,,,module: docs,"module: docs, topic: docs, triaged",thread:issue:100904:issue:104623:issue:105457,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 3 open issues; raw issue rows remain available for audit.,"#100904 Revise glossary, #104623 I propose a new overview section in the documentation, #105457 Top level Glossary for users (not contributers)",442f38f739f07a18021962779fe63b6ecf0df6bdef98f83cd8bcf59579f49168
issue-collapse:592c0e2281ea8e0f,Review thread: 2 issues - issue #99265: the error message of torch.addcmul is wrong,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"99265, 99351",,,,"module: error checking, module: python array api, module: python frontend","actionable, feature, module: error checking, module: python array api, module: python frontend, triaged",thread:issue:99265:issue:99351,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#99265 the error message of torch.addcmul is wrong, #99351 we should make semantically meaningless positional arguments positional only in our operator API",419501a77d58f1e01e32c297edd89b69896f2ad59691be737a562fb52e16011d
issue-collapse:a55a823a525a802f,Review thread: 2 issues - issue #98861: Sparse Tensor not working for `torch.cat`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"98861, 98947",,,,module: sparse,"feature, module: sparse, triaged",thread:issue:98861:issue:98947,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#98861 Sparse Tensor not working for `torch.cat`, #98947 Add `torch.cat` support for torch native sparse tensors. (Need for PyG)",520ed476556a66afb8b288af31a2e565ea302985583883d2c4fff8759eab383d
issue-collapse:f7a304e86e3590d5,Review thread: 2 issues - issue #98073: Add PrivateUse1 folder in aten/src/ATen,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"98073, 98406",,,,"module: backend, module: internals","module: backend, module: internals, triaged",thread:issue:98073:issue:98406,Handle this module: backend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#98073 Add PrivateUse1 folder in aten/src/ATen, #98406 Contribute to the privateuse1 backend.",7e6aee7641b8987c3808db8b4efee4c7009e47cb32029e6ea136a1d5df589c21
issue-collapse:4fdfeb0d2dccca6c,Review thread: 2 issues - issue #9204: /usr/bin/ld: cannot find -lpthreads,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"9204, 9310",,,,module: build,"caffe2, has workaround, module: build, triaged",thread:issue:9204:issue:9310,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#9204 /usr/bin/ld: cannot find -lpthreads, #9310 /usr/bin/ld: cannot find -lpthreads",53240761f7737b816c9614673c422c09483c89454ccf47181fe2838753c8e3ee
issue-collapse:27fefe4f46e29112,Review thread: 2 issues - issue #91156: `quantile` fails for `float16`/`half` inputs,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"91156, 91374",,,,"module: cpu, module: cuda, module: half","enhancement, module: cpu, module: cuda, module: half, triaged",thread:issue:91156:issue:91374,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#91156 `quantile` fails for `float16`/`half` inputs, #91374 [RFC] `quantile` should work for `float16`/`half` on the GPU",4314cfb5a90b0169c7eb051117f219c6c7d1dfa56aabe54f6a166eb6857d89ea
issue-collapse:e42d5345889d210e,Review thread: 2 issues - issue #89491: build: cmake: functorch.so not installed at expected location,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"89491, 91524",,,,"module: build, module: functorch","module: build, module: functorch, triaged",thread:issue:89491:issue:91524,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#89491 build: cmake: functorch.so not installed at expected location, #91524 functorch.so is installed back into the source directory",58d9c578b4e33b33d2744bdbe14e013b739b85771640fb19823d34078dbb87c9
issue-collapse:1c0f8591dd80288f,Review thread: 2 issues - issue #89105: Bad string in GLSL shader,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"89105, 89601",,,,module: vulkan,"module: vulkan, topic: build, triaged",thread:issue:89105:issue:89601,Handle this module: vulkan stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#89105 Bad string in GLSL shader, #89601 Building PyTorch with Vulkan backend fails (1.13 and master)",20b4c789140366457cfdd1b784b402e16d8203f9a290a0416c815f5f6e409343
issue-collapse:a5526c8a8b04b8f6,"Review thread: 2 issues - issue #88658: nn.Linear allocate too many space which lead to CPUAllocator ""allocate memory failure"" if it's BF16. good for FP32.",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"88658, 95108",,,,"module: cpu, module: nn","intel, module: cpu, module: nn, triaged",thread:issue:88658:issue:95108,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#88658 nn.Linear allocate too many space which lead to CPUAllocator ""allocate memory failure"" if it's BF16. good for FP32., #95108 `torch.nn.LazyLinear` crash when using torch.bfloat16 dtype in pytorch 1.12.0 and 1.13.0",2737e110fea18e6f604e0f2b61557e068b7c2e82dea7dee5013b50c9ccc2de55
issue-collapse:d17a740e4b75509d,Review thread: 2 issues - issue #84625: Beta distribution behaves incorrectly for small parameters,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"84625, 90663",,,,"module: distributions, module: edge cases","feature, module: distributions, module: edge cases, triaged",thread:issue:84625:issue:90663,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#84625 Beta distribution behaves incorrectly for small parameters, #90663 [Feature Request] An alternative sampling routine for Dirichlet to fix Dirichlet and Beta sampling bugs",3ad00faad697a9f00188048d629f27f26cc5a19e85f453b4fb4be3ec58be41e6
issue-collapse:1704b07c27572f40,"Review thread: 2 issues - issue #82926: Slice operation on ""ragged"" dimension in NestedTensor",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"82926, 86890",,,,"module: advanced indexing, module: nestedtensor","enhancement, feature, module: advanced indexing, module: nestedtensor, triaged",thread:issue:82926:issue:86890,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#82926 Slice operation on ""ragged"" dimension in NestedTensor, #86890 View-based advanced indexing (Integer array/LongTensor indexing) of nested_tensor",4490cc1b5585a6c2d9f7a3b942aa04c8974d9284b0cfe0b234983adcb415b487
issue-collapse:95c58a8e94e6bb04,Review thread: 2 issues - issue #82886: CUDA graph capturing fails for nn.Embedding and large batch sizes,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"82886, 97210",,,,"module: cuda, module: cuda graphs, module: embedding","module: cuda, module: cuda graphs, module: embedding, triaged",thread:issue:82886:issue:97210,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#82886 CUDA graph capturing fails for nn.Embedding and large batch sizes, #97210 INTERNAL ASSERT FAILED at ""../c10/cuda/CUDAGraphsC10Utils.h"":73, please report a bug to PyTorch. Unknown CUDA graph CaptureStatus32729",7a9d2e4fee1167cfc239bda1663a8cec71252bfe0cf5b356e24c26f84196814a
issue-collapse:1dacc7f9740fc7c9,Review thread: 2 issues - issue #82139: Re-enable DynamicQuantModule in iOS simulator tests,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"82139, 82443",,,,"module: ci, module: ios, oncall: quantization","low priority, module: ci, module: ios, oncall: quantization, triaged",thread:issue:82139:issue:82443,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,3,Represents 2 open issues; raw issue rows remain available for audit.,"#82139 Re-enable DynamicQuantModule in iOS simulator tests, #82443 Quantization issue in transformers",e9ad86963b019a5a78dd742f741e18141d9236c757e85a628605540a969a7346
issue-collapse:46675c79f632d84c,Review thread: 2 issues - issue #81692: Failed to static link latest cuDNN while compiling,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"81692, 90547",,,,"module: build, module: cudnn, module: static linking","module: build, module: cudnn, module: static linking, triaged",thread:issue:81692:issue:90547,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#81692 Failed to static link latest cuDNN while compiling, #90547 Unable to link LibTorch against CUDA and CUDNN statically",8e64af5e66305237adbc7ae3ffd1817fd97bea7d8c70a2ca0941294f14722b68
issue-collapse:a0389bbc32ab759c,Review thread: 2 issues - issue #80549: Allow parameterization of Layouts,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"80549, 81104",,,,"module: python frontend, module: sparse","feature, module: python frontend, module: sparse, triaged",thread:issue:80549:issue:81104,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#80549 Allow parameterization of Layouts, #81104 Resize/reshape of sparse compressed tensors - design",d2500de894150e8d7758d660ab5f2a9fd21dde79bc2133bc53dcdd18bbfc3938
issue-collapse:a413dc83365460e1,Review thread: 2 issues - issue #80187: [NVFuser] Investigate modules with bad performance relative to eager,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"80187, 80966",,,,"module: nvfuser, module: primtorch","module: nvfuser, module: primTorch, triaged",thread:issue:80187:issue:80966,Handle this module: nvfuser stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#80187 [NVFuser] Investigate modules with bad performance relative to eager, #80966 [Prims+NVFuser] Aten2Aten decomp hurting performance",975001618f85c5c1aa2fb37422f1c67bb8d371553a980187d591af05d4ad5cd0
issue-collapse:1fdef2ab3f0bdbe6,Review thread: 2 issues - issue #79039: Add Autograd Support for Nested Tensor,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"79039, 83773",,,,"module: autograd, module: nestedtensor","module: autograd, module: nestedtensor, triaged",thread:issue:79039:issue:83773,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#79039 Add Autograd Support for Nested Tensor, #83773 [Nested Tensor] view + inplace for autograd.",b0cf39ab3daabc2768055723926b28229e0fe552953c610a2eb83559ae018958
issue-collapse:2a557ed8a93ccd02,Review thread: 2 issues - issue #78878: Improving clarity in the docs of different losses,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"78878, 80208",,,,"module: docs, module: loss, module: nn, module: python frontend","module: docs, module: loss, module: nn, module: python frontend, triaged",thread:issue:78878:issue:80208,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#78878 Improving clarity in the docs of different losses, #80208 F.binary_cross_entropy_with_logits unexpected behaviour",f39e5d546d3492a8b737ee18b9b503aace755bcca20876d5d60d5426808dc597
issue-collapse:56936fd376dc6dc3,"Review thread: 2 issues - issue #78681: Unable to install Preview (Nightly) on M1 macOS: ""Symbol not found""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"78681, 79337",,,,"module: binaries, module: macos","module: binaries, module: macos, triaged",thread:issue:78681:issue:79337,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#78681 Unable to install Preview (Nightly) on M1 macOS: ""Symbol not found"", #79337 Conda install from pytorch-nightly channel does not install the expected version on macOS",cdc9e237b85ddb7fbcec8bf03be103f12cf67be300a426ff600c15241496d0da
issue-collapse:613a89a9df17ae72,Review thread: 2 issues - issue #78071: [primTorch] item prim can't return a bool properly,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"78071, 85834",,,,module: primtorch,"module: primTorch, triaged",thread:issue:78071:issue:85834,Handle this module: primtorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#78071 [primTorch] item prim can't return a bool properly, #85834 [primTorch] Need to update data-dependent check policy",0348350efaf7a2d02391d7d3039d1623644ae2493ed03a6c9604b9f34f52d7c6
issue-collapse:3bf26b1a5e19d80a,Review thread: 2 issues - issue #77415: Suboptimal error message - nn.Linear with double argument,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"77415, 95590",,,,"module: error checking, module: lazy","module: error checking, module: lazy, triaged",thread:issue:77415:issue:95590,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#77415 Suboptimal error message - nn.Linear with double argument, #95590 Confusing error messages from `torch.nn.LazyLinear` in different versions.",0a5e1ae3af4344f6b1be4568f3d0e5a78e88090b876c8a29536a555eb135ed3b
issue-collapse:d3b4e8f26f9b31a4,Review thread: 2 issues - issue #77223: torch.ops.aten.ceil(1.5) returns Tensor rather than scalar,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"77223, 77332",,,,"module: __torch_dispatch__, module: type promotion","module: __torch_dispatch__, module: type promotion, triaged",thread:issue:77223:issue:77332,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#77223 torch.ops.aten.ceil(1.5) returns Tensor rather than scalar, #77332 wrong overload resolved for `torch.mul(x, 4)` in `__torch_dispatch__`",5ec18f46c1905b8ffd51e2b7d754a84989136c1b4a7cfc50fe4b91544c0c957c
issue-collapse:7a70afad211dc79a,"Review thread: 2 issues - issue #76785: `addr, baddmm, dist, l1_loss` will backward fail when input tensors have different dtypes",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"76785, 95434",,,,"module: autograd, module: complex, module: linear algebra, module: type promotion","actionable, complex_autograd, module: autograd, module: complex, module: linear algebra, module: type promotion, triaged",thread:issue:76785:issue:95434,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76785 `addr, baddmm, dist, l1_loss` will backward fail when input tensors have different dtypes, #95434 It seems that `torch.Tensor.addmv` and `torch.Tensor.addr` will check some inputs' dtype if and only if in `backward()`",361ea6bf96aa6152febd8df1fef62197be655324126024aa5a4208ffca8fdb00
issue-collapse:4b2e2e899f5902b1,Review thread: 2 issues - issue #76654: `torch.sort` does not exploit parallelization when invoked without the `dim` parameter.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"76654, 76655",,,,"module: cpu, module: multithreading, module: performance, module: sorting and selection","feature, module: cpu, module: multithreading, module: performance, module: sorting and selection, triaged",thread:issue:76654:issue:76655,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#76654 `torch.sort` does not exploit parallelization when invoked without the `dim` parameter., #76655 [feature request] no-param sort to exploit parallelization",9269b4323a15c7d69189a47cc0ab549be30da8c7b2e6ccc435d1a05376abb580
issue-collapse:cb71af573e3959f6,Review thread: 2 issues - issue #76483: Expand pow and float_pow sampling function for more coverage,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"76483, 76853",,,,"module: complex, module: nans and infs, module: tests","module: NaNs and Infs, module: complex, module: tests, triaged",thread:issue:76483:issue:76853,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76483 Expand pow and float_pow sampling function for more coverage, #76853 Inconsistent results between Pow and Float Pow with their numpy references for complex types",128741d05f718ea84c526522fdbc6e62c0adbd707b1501e1984aa9a8dd9d603d
issue-collapse:6a2eab6eec2b61cb,Review thread: 2 issues - issue #76433: Remove _log_softmax/_softmax in favor of log_softmax and softmax respectively.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"76433, 76514",,,,module: internals,"module: internals, triaged",thread:issue:76433:issue:76514,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#76433 Remove _log_softmax/_softmax in favor of log_softmax and softmax respectively., #76514 Clean up PyTorch's private operators",b59c8a92126bf073cddc3e1688adce648bafbf7366f54dabb3b91663106f8bdd
issue-collapse:b38e609e6c983ee7,Review thread: 2 issues - issue #75363: conv3d has numerical issue where same input produces output that are not bit-wise identical,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"75363, 76354",,,,"module: convolution, module: determinism, module: nvfuser","module: convolution, module: determinism, module: nvfuser, triaged",thread:issue:75363:issue:76354,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#75363 conv3d has numerical issue where same input produces output that are not bit-wise identical, #76354 NVFuser failing extremal opinfos",16357f3af94b914b1785c0a45e6cb18212b1bdd4467eed69ad5a7f84b1763f68
issue-collapse:22c95c95cbc3a2fc,"Review thread: 2 issues - issue #74616: __rpow__(self, other) OpInfo should not test the case where `other` is a Tensor",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"74616, 76806",,,,"module: primtorch, module: tests, module: type promotion","module: primTorch, module: tests, module: type promotion, triaged",thread:issue:74616:issue:76806,Handle this module: primtorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#74616 __rpow__(self, other) OpInfo should not test the case where `other` is a Tensor, #76806 torch.Tensor.__rdiv__ long x scalar float type promotion is incorrect",5976d7e64ae111e48a90e13917bed60bdd8c143cc45051958f82bb98522bbef1
issue-collapse:163e35981056c014,"Review thread: 2 issues - issue #74236: ""histogram_cpu"" not implemented for 'Int'",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"74236, 74420",,,,"module: numpy, module: sorting and selection","feature, module: numpy, module: sorting and selection, triaged",thread:issue:74236:issue:74420,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#74236 ""histogram_cpu"" not implemented for 'Int', #74420 `torch.histogram` has wrong output dtype and doesn't support integer inputs",926c82addcd93df61ae25d82c7788517f94630131bb3a71948497605aa132c7b
issue-collapse:363fb101ddca6a52,Review thread: 2 issues - issue #73487: PyTorch not recognizing GPU on WSL - installed cudnn and cuda,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73487, 85773",,,,"module: build, module: cuda, module: wsl","module: build, module: cuda, module: wsl, triaged",thread:issue:73487:issue:85773,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73487 PyTorch not recognizing GPU on WSL - installed cudnn and cuda, #85773 Conda Pytorch (Pytorch channel) in WSL2 Ubuntu can't find libcudnn shared objects",6168cc33270971b26d1a6a4efb2dd0ec7410ed5bbe2fb3f154e3403c43b9c250
issue-collapse:049ca23206adfc8c,Review thread: 2 issues - issue #73190: Segmentation fault in max_pool1d,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73190, 73870",,,,"module: correctness (silent), module: edge cases, module: error checking, module: nn, module: pooling","module: correctness (silent), module: edge cases, module: error checking, module: nn, module: pooling, triaged",thread:issue:73190:issue:73870,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73190 Segmentation fault in max_pool1d, #73870 max_pool1d() returns when given invalid large `kernel_size` inputs",a9b70b84f92f6dfee58cc9f892f7a71c72be27a4b396f6096cc01810d05ea775
issue-collapse:81a628000e34e715,"Review thread: 2 issues - issue #73176: Cuda lacks checking of ""out of bound""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73176, 99410",,,,"module: cuda, module: error checking, module: loss, module: nn","module: cuda, module: error checking, module: loss, module: nn, triaged",thread:issue:73176:issue:99410,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73176 Cuda lacks checking of ""out of bound"", #99410 torch.nn.functional.multilabel_margin_loss cuda lacks checking of ""out of bound""",534e49a01e614a27366e16d8528ad2bff972162df0e7dc7f606b2e4c93426e7c
issue-collapse:836b553c69ae60e9,Review thread: 2 issues - issue #73113: PyTorch fails to compile on gcc 11.2 due to breakpad,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73113, 73412",,,,module: build,"module: build, triaged",thread:issue:73113:issue:73412,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73113 PyTorch fails to compile on gcc 11.2 due to breakpad, #73412 Build from source failed",7309b966b9a4362adf4f5274866708612ece149d2a708745c12933991abe2b7f
issue-collapse:fe4d0e100a67a463,"Review thread: 2 issues - issue #72556: Clarify test dependencies (e.g., into a test-requirements.txt file)",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"72556, 72558",,,,"module: bootcamp, module: ci","module: bootcamp, module: ci, triaged",thread:issue:72556:issue:72558,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#72556 Clarify test dependencies (e.g., into a test-requirements.txt file), #72558 Pin dependencies + expand the current linter",ae210bb7713aba347fd9b3e89ccfaa06ac60b67d917cdacacc308197b3d88d1d
issue-collapse:5b0094860fae535c,Review thread: 2 issues - issue #71636: `torch.median` will return -2147483648 when input is an empty tensor,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"71636, 95756",,,,module: error checking,"module: error checking, triaged",thread:issue:71636:issue:95756,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71636 `torch.median` will return -2147483648 when input is an empty tensor, #95756 `torch.nanmedian` return a negative value when input is empty",47eb8016f033fd3a9781fb5dd9d784f68716c66c1a2e8f2bc40b1e55f20a1263
issue-collapse:82b3398db52710a6,Review thread: 2 issues - issue #71446: Discussion of TorchQuantum and QuantumNAS,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"71446, 82565",,,,"module: complex, module: scientific computing","complex_autograd, feature, function request, module: complex, module: scientific computing, triaged",thread:issue:71446:issue:82565,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71446 Discussion of TorchQuantum and QuantumNAS, #82565 PyTorch for quantum mechanics",2f1d9adfff8b07bd80f9710474c2ffe029eeaa293fe736a9018cb20b1115ea45
issue-collapse:fbfd9720c8195cfa,Review thread: 2 issues - issue #70242: `TensorIterator`: provide a two-argument version of `set_output`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"70242, 70678",,,,"module: structured kernels, module: tensoriterator","module: TensorIterator, module: structured kernels, triaged",thread:issue:70242:issue:70678,Handle this module: structured kernels stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#70242 `TensorIterator`: provide a two-argument version of `set_output`, #70678 Decouple `TensorIteratorBase` output from structured kernel outputs.",1827904e7af08c4ba30a31d582eb6518329c1541256c6deeed6e72524c087bf3
issue-collapse:abb6e988d2fdda0a,Review thread: 2 issues - issue #69822: Implement torch.*_like tensor creation functions on sparse inputs,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"69822, 69892",,,,"module: numpy, module: sparse, module: tensor creation","feature, module: numpy, module: sparse, module: tensor creation, triaged",thread:issue:69822:issue:69892,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69822 Implement torch.*_like tensor creation functions on sparse inputs, #69892 [feature request] Support `like=` argument in tensor factory methods",4d9f665c4a81bbd387585d710e9a17f988cd8c695331d69b601902ede843d635
issue-collapse:751686887093bc7a,Review thread: 2 issues - issue #69786: Implement aten::equal for sparse tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"69786, 95776",,,,module: sparse,"Stale, module: sparse, triaged",thread:issue:69786:issue:95776,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69786 Implement aten::equal for sparse tensors, #95776 `torch.Tensor.is_set_to` raises `NotImplementedError` when inputs contain sparse tensor",9441dcd8493b623d60a73300e6c91e5b719428ffb3d095ebd3422220de9ec739
issue-collapse:287dddc00f335190,"Review thread: 2 issues - issue #69531: `(svd|pca)_lowrank`: backward is unstable when for a matrix `A`, the parameter `q` is set to a value `q > rank(A)`.",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"69531, 69532",,,,"module: autograd, module: linear algebra, module: numerical-stability","Stale, feature, module: autograd, module: linear algebra, module: numerical-stability, triaged",thread:issue:69531:issue:69532,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69531 `(svd|pca)_lowrank`: backward is unstable when for a matrix `A`, the parameter `q` is set to a value `q > rank(A)`., #69532 [feature request] A rank-revealing SVD for better stability in backward.",490d636911288a54643e68d4375c202a7f85db221d96e7bcfd9404edf45658b7
issue-collapse:87d101889bcde667,Review thread: 2 issues - issue #69353: Pruning `torch.nn.MultiheadAttention` causes RuntimeError,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"69353, 78109",,,,module: pruning,"Stale, module: pruning, triaged",thread:issue:69353:issue:78109,Handle this module: pruning stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#69353 Pruning `torch.nn.MultiheadAttention` causes RuntimeError, #78109 Doesn't work when register hook to torch.nn.MultiheadAttention.out_proj",3e5f874088f1f87b24d2b69e84f39b60f21426cb6f912a1b9f10a56b87c305d1
issue-collapse:25979a2d4d7ed601,Review thread: 2 issues - issue #67970: Error when using torch.fx on bert,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"67970, 70485",,,,module: fx,"Stale, module: fx, triaged",thread:issue:67970:issue:70485,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#67970 Error when using torch.fx on bert, #70485 Cannot run FX tracer on a vision transformer model",e240dbc4a84ab88cadef5ed9ad97fe49308e1f69f9db938f687e5fc78f65c27b
issue-collapse:1701e14b7a46682c,Review thread: 2 issues - issue #67153: Build not working with cuda 11.5,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"67153, 70931",,,,"module: build, module: cuda","module: build, module: cuda, triaged",thread:issue:67153:issue:70931,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#67153 Build not working with cuda 11.5, #70931 Redefinition of `cub` namespace misses Debug",bebe74a2b135a37ef533136d5d1e888437ec95af0d4ac86696c65fcbc0d5849a
issue-collapse:056345ea5ac829ca,Review thread: 2 issues - issue #66750: torch.nn.EmbeddingBag behave differently in two cases of cpu and cuda,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"66750, 66751",,,,"module: embedding, module: error checking","module: embedding, module: error checking, triaged",thread:issue:66750:issue:66751,Handle this module: embedding stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#66750 torch.nn.EmbeddingBag behave differently in two cases of cpu and cuda, #66751 torch.nn.functional.embedding behave differently in two cases of cpu and cuda",aa2365de899114655936bf6198c7b3175b8cb2b892a37031cd38f0e9ffa78f56
issue-collapse:9d186a427368517b,Review thread: 2 issues - issue #65131: Adding a device variable to a `torch.Tensor` subclass fails,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"65131, 65132",,,,module: xla,"module: xla, triaged",thread:issue:65131:issue:65132,Handle this module: xla stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#65131 Adding a device variable to a `torch.Tensor` subclass fails, #65132 How to reference a tensor variable from a superclass of `torch.Tensor`?",c9a56d3668213894e4bf3be4b58f83007b9f655b8755ebb9da14f8c5bddac41e
issue-collapse:b2e5d757543910c6,Review thread: 2 issues - issue #64162: [feature request] torch.Generator constructor to accept seed directly,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"64162, 89054",,,,module: random,"module: random, triaged",thread:issue:64162:issue:89054,Handle this module: random stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#64162 [feature request] torch.Generator constructor to accept seed directly, #89054 Add a `device` keyword argument to `torch.manual_seed`",4160d325f5a88888437b6b422f1efe9800243fcb02db5c8f52932d4518bffee0
issue-collapse:a3b575294ae7ea97,"Review thread: 2 issues - issue #63079: PyTorch 1.9.0, test_optim fails on Nvidia A10.",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"63079, 98414",,,,"module: cuda, module: optimizer, module: tests","module: cuda, module: optimizer, module: tests, triaged",thread:issue:63079:issue:98414,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#63079 PyTorch 1.9.0, test_optim fails on Nvidia A10., #98414 PyTorch 1.12, high failure rate for test_optim/test_nadam",b0bafe9b032cead6ceed1091a84b1ca7689d9f3656756d6b42d12f8e095b5350
issue-collapse:cfc60a71c8838016,Review thread: 2 issues - issue #63034: [bug] nn.functional.pad (circular) ubsan failure,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"63034, 74519",,,,"module: nn, module: padding","lazy, module: nn, module: padding, triaged",thread:issue:63034:issue:74519,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#63034 [bug] nn.functional.pad (circular) ubsan failure, #74519 Lazy TS Test Failures",8a025c1f9463e3f658935d328b54e11ee85a68742de1a1361475a6d20f2f3862
issue-collapse:b221b28eb26c5b60,Review thread: 2 issues - issue #6273: Multithreading Scaling Issue with MKL,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"6273, 7127",,,,,caffe2,thread:issue:6273:issue:7127,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#6273 Multithreading Scaling Issue with MKL, #7127 how can i use openMP for caffe2? why caffe2 not work in multi-threads mode?",8e442711351b1eea278bdce5854171057027871d801a51b1de1321f17ffb9baa
issue-collapse:e89674d888a71989,Review thread: 2 issues - issue #62451: torch.Generator ignores the device argument?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"62451, 84234",,,,"module: determinism, module: random","module: determinism, module: random, triaged",thread:issue:62451:issue:84234,Handle this module: determinism stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#62451 torch.Generator ignores the device argument?, #84234 Randomness should be consistent across devices with use_deterministic_algorithms",dd18645ad57f5c77c70c41bf98e4861057f905d499d4e1b2306150bf90e1306b
issue-collapse:d2529f27669935fb,"Review thread: 2 issues - issue #62396: Make interpolation output size compatible with opencv, scikit-image and scipy for floating scale factor",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"62396, 70559",,,,"module: interpolation, module: nn","enhancement, module: interpolation, module: nn, triaged",thread:issue:62396:issue:70559,Handle this module: interpolation stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#62396 Make interpolation output size compatible with opencv, scikit-image and scipy for floating scale factor, #70559 Behavior of torch.nn.functional.interpolate with unchanged output size and recompute_scale_factor=False",3216dea5c03f5ab1356394f9a327cf1de6bd87329dde9ead7899053c566d6156
issue-collapse:b69ddb62a25688db,Review thread: 2 issues - issue #61658: Move `torch.cholesky_solve` into `torch.linalg`.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"61658, 77166",,,,module: linear algebra,"enhancement, module: linear algebra, triaged",thread:issue:61658:issue:77166,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#61658 Move `torch.cholesky_solve` into `torch.linalg`., #77166 torch.cholesky has been deprecated in favour of torch.linalg.cholesky. However, torch.cholesky_inverse remains as is. It should also be moved to torch.linalg",e074900443cc0ce0159cda016f699e2030804888e482685a1df29e48e26ab3c9
issue-collapse:ff7b182e73906633,Review thread: 2 issues - issue #61211: [feature request] Do zero-padding in high-frequency modes in `ifft`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"61211, 61213",,,,module: fft,"enhancement, module: fft, triaged",thread:issue:61211:issue:61213,Handle this module: fft stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#61211 [feature request] Do zero-padding in high-frequency modes in `ifft`, #61213 [feature request] Low pass filtering of FFT results",a2d0508bbc5ec8451b62e0d6e47900a01e054b1b292de2036b3d0a0451e38fbe
issue-collapse:a18c2b73b8989b1b,Review thread: 2 issues - issue #60832: State of Torch Named Tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"60832, 94586",,,,module: named tensor,"module: named tensor, triaged",thread:issue:60832:issue:94586,Handle this module: named tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#60832 State of Torch Named Tensors, #94586 `permute` for named tensors",24a1d35ec4ed74252f9f805c63743721fbc5a4736e44b602bdb79db29d0c1ac4
issue-collapse:d7852d9845dcb69a,Review thread: 2 issues - issue #60149: backward compatibility - need a way to find out when a certain API was added/modified/etc.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"60149, 60153",,,,"module: codegen, module: deprecation, module: doc infra, module: docs, module: internals","better-engineering, module: codegen, module: deprecation, module: doc infra, module: docs, module: internals, triaged",thread:issue:60149:issue:60153,Handle this module: codegen stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#60149 backward compatibility - need a way to find out when a certain API was added/modified/etc., #60153 [oneline docs] adding a source link for non-python functions",e2e8cc6d32260fc58ebdc038cec6dd86f5bf9b1d9db3927f3dea0a2f9837656d
issue-collapse:26abaa762fb09348,Review thread: 2 issues - issue #59855: pytorch test failed,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"59855, 63295",,,,"module: autograd, module: build, module: cuda, module: linear algebra, module: tests","module: autograd, module: build, module: cuda, module: linear algebra, module: tests, triaged",thread:issue:59855:issue:63295,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#59855 pytorch test failed, #63295 pytorch 1.8.2 cuda test errors",bc98f37e2f748ebd8eecd45e269d7b74043ce23502dbb6532507382e38912e5c
issue-collapse:6bda322c8fbb3014,Review thread: 2 issues - issue #59682: Add a lint rule to recommend TORCH_CHECK(false) over throw std::runtime_error,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"59682, 61079",,,,,"better-engineering, triaged",thread:issue:59682:issue:61079,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#59682 Add a lint rule to recommend TORCH_CHECK(false) over throw std::runtime_error, #61079 [clang-tidy] Add a custom rule for AT_ERROR -> TORCH_CHECK(false, ...)",c782edfe5e58647ca587c18d01770c57de79055281df0072cdccf0dcb3e15b7e
issue-collapse:29d8527c237109a6,Review thread: 2 issues - issue #58779: C++ Extensions that use ninja should enable colors by default,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"58779, 59251",,,,module: cpp-extensions,"enhancement, module: cpp-extensions, triaged",thread:issue:58779:issue:59251,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#58779 C++ Extensions that use ninja should enable colors by default, #59251 Ahead of time C++ Extensions should have an option to build with verbose=False",eb6a87c1b9b913c5926d43f93c912fd7adc46f514b78ea73232fafc1fa2ac407
issue-collapse:d2a0022c2f5fd1c5,Review thread: 2 issues - issue #58466: `set_per_process_memory_fraction()` does not ensure max used GPU memory below fraction,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"58466, 69688",,,,"module: cuda, module: docs, module: memory usage","module: cuda, module: docs, module: memory usage, triaged",thread:issue:58466:issue:69688,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#58466 `set_per_process_memory_fraction()` does not ensure max used GPU memory below fraction, #69688 `torch.cuda.set_per_process_memory_fraction()` does not perform VRAM isolation",b52f5bf87d1ecd6aa41a01e7670e1b9e3c5b23693bef3fdefd0f22a2b2b470ee
issue-collapse:d25a2ed7abaca93a,Review thread: 2 issues - issue #58128: the same input with different batchsize got different precision output,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"58128, 60234",,,,"module: batching, module: cuda, module: nn, module: numerical-reproducibility","module: batching, module: cuda, module: nn, module: numerical-reproducibility, triaged",thread:issue:58128:issue:60234,Handle this module: batching stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#58128 the same input with different batchsize got different precision output, #60234 Numerical-reproducibility issue in torch.matmul",43bb2cd4339d75f8f970daf465c971131d0e52dc8369b09b17e325bb537fc340
issue-collapse:abb523cd322fb812,Review thread: 2 issues - issue #56794: No tensor operations allowed inside at::parallel_for,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"56794, 70241",,,,module: multithreading,"module: multithreading, triaged",thread:issue:56794:issue:70241,Handle this module: multithreading stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#56794 No tensor operations allowed inside at::parallel_for, #70241 CPU parallelization across batch has random faulty behavior on backward",a37db2959bd1e8a8f1e34cceb5835684829115643aaeeb1b1134da0f6be8b077
issue-collapse:18b986c560889bd4,Review thread: 2 issues - issue #56480: shared torch.tensor with multiprocesses using python Queue cause coredump,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"56480, 56921",,,,module: multiprocessing,"module: multiprocessing, shadow review, triaged",thread:issue:56480:issue:56921,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#56480 shared torch.tensor with multiprocesses using python Queue cause coredump, #56921 torch.multiprocessing implement SyncManager",4d6c5ccb82233430d56a48f38a92a744d56843840e3d338e1da0ed563ff0d7f2
issue-collapse:c1edf2c11fe6ef6e,Review thread: 2 issues - issue #56440: Raise exception when torch.clamp min value underflows the input tensor's dtype,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"56440, 59526",,,,"module: docs, module: error checking, module: numpy","module: docs, module: error checking, module: numpy, triaged",thread:issue:56440:issue:59526,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#56440 Raise exception when torch.clamp min value underflows the input tensor's dtype, #59526 Documentation for torch.finfo doesn't match implementation",698724291e9637984f04782eb37f9d1d4ebb614d5a0e2b2ddb6395c140f24de8
issue-collapse:cf68001965890dc6,Review thread: 2 issues - issue #55296: Build errors with USE_VULKAN=ON when cross-compiling for Android,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"55296, 55297",,,,"module: android, module: build","module: android, module: build, triaged",thread:issue:55296:issue:55297,Handle this module: android stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#55296 Build errors with USE_VULKAN=ON when cross-compiling for Android, #55297 SIGSEGV at at::is_vulkan_available() invocation on Android",4d8825d7293fb3945fbb1e6fd6dd0ab2ea1b72303c644a3abe5abd3afe4109ea
issue-collapse:278b67e9062fda3e,Review thread: 2 issues - issue #53534: [FX] Ability to wrap functions in other modules for symbolic tracing,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"53534, 62021",,,,module: fx,"module: fx, triaged",thread:issue:53534:issue:62021,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#53534 [FX] Ability to wrap functions in other modules for symbolic tracing, #62021 [FX] Ability to provide autowrap function to Tracer init",61e79d7bb6a4e097a407e7cd3a4f92f72863e87951e09dc11840b4f34b556174
issue-collapse:6046b5ab9a7b27b1,Review thread: 2 issues - issue #52648: Fix `torch.linalg.vector_norm`'s backward to avoid needing `at::abs()` call in forward for ord=+/-inf,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"52648, 52717",,,,"module: autograd, module: linear algebra","module: autograd, module: linear algebra, triaged",thread:issue:52648:issue:52717,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#52648 Fix `torch.linalg.vector_norm`'s backward to avoid needing `at::abs()` call in forward for ord=+/-inf, #52717 Fix `torch.norm`'s backward to match forward for ord=+/-inf",138abf741cc03f4a24b320df48088bcd6a9069f2a985f18c1cb1829e8b93d138
issue-collapse:6a5f3136468e6fe5,Review thread: 2 issues - issue #52332: Reciprocals of complex tensors with infinities are different from NumPy.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"52332, 76558",,,,"module: complex, module: jiterator, module: nans and infs, module: numpy, module: performance","module: NaNs and Infs, module: complex, module: jiterator, module: numpy, module: performance, triaged",thread:issue:52332:issue:76558,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#52332 Reciprocals of complex tensors with infinities are different from NumPy., #76558 [jiterator] perf regression when jiterating few ops for complex dtype",1daa2570e6f54672b6dc87e5e9c016de351af16c5d6f18973ef242196367c189
issue-collapse:dcf3223404ec4031,Review thread: 2 issues - issue #52265: 'TypeError: expected CPU (got CUDA)' when placing a CUDA tensor on a class that is inheriting torch.Tensor,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"52265, 62147",,,,"module: boolean tensor, module: tensor creation","module: boolean tensor, module: tensor creation, triaged",thread:issue:52265:issue:62147,Handle this module: boolean tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#52265 'TypeError: expected CPU (got CUDA)' when placing a CUDA tensor on a class that is inheriting torch.Tensor, #62147 Tensor subclassing does not support bool input",fc2eb2d062b0fbb98360ec21f4c61fccc57bc106772b53120c39942b4211858a
issue-collapse:34bc3ed773492419,Review thread: 2 issues - issue #52260: Unlucky batch->device split by DataParallel causes an exception,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"52260, 52289",,,,module: data parallel,"module: data parallel, triaged",thread:issue:52260:issue:52289,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#52260 Unlucky batch->device split by DataParallel causes an exception, #52289 DataParallel module fails to handle data with size not divisible by number of GPUs",1ebff375dfa789182d8ec2e4788de139271507b2eb22b40fd60a580f3c61386f
issue-collapse:0f225acdf384c05c,Review thread: 2 issues - issue #51782: [discussion] [python] Generic way for generating module-wrappers for stateless parametrized functions,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"51782, 51896",,,,module: nn,"better-engineering, feature, module: nn, triaged",thread:issue:51782:issue:51896,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#51782 [discussion] [python] Generic way for generating module-wrappers for stateless parametrized functions, #51896 Can't reassign a child module instance to a lambda",63a02e27e61a7fc2be2d5eb94c3388b50570207c0646239a28946b3dc90ccdb7
issue-collapse:8fd42eb112dea1ad,Review thread: 2 issues - issue #51471: Make TorchBind object appear in state_dict,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"51471, 62094",,,,"module: nn, module: serialization","enhancement, feature, module: nn, module: serialization, triaged",thread:issue:51471:issue:62094,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#51471 Make TorchBind object appear in state_dict, #62094 [FR] Support for non-Parameter, non-Buffer objects in state_dict",f3818e4863d088e11977874d588287d681a18767732163059b24540f87838da6
issue-collapse:dd094dbdb81fe628,Review thread: 2 issues - issue #50689: RandomSampler is very slow with huge dataset,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"50689, 50940",,,,module: dataloader,"module: dataloader, triaged",thread:issue:50689:issue:50940,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#50689 RandomSampler is very slow with huge dataset, #50940 SequentialSampler getting slower as time passing by",5547876c227359918e2d8da0a1af780e61f2d7f2866474aca46095791eb2775d
issue-collapse:948221af97a312bf,Review thread: 2 issues - issue #50034: Multinomial without replacement produces samples that have zero probability,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"50034, 74771",,,,"module: distributions, module: numpy, module: random, module: ux","module: distributions, module: numpy, module: random, module: ux, topic: docs, triaged",thread:issue:50034:issue:74771,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#50034 Multinomial without replacement produces samples that have zero probability, #74771 [bug] `torch.multinomial` should throw error as documented",bbd863c97d41851c6cea9f61834b45f3baa65c8799f93c3c12e27a38e923eec3
issue-collapse:7e90ef8bc34e6438,"Review thread: 2 issues - issue #49992: [docs] nn.modules pages should mention corresponding functional versions, e.g. for nn.Hardshrink",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49992, 59566",,,,"module: docs, module: nn","module: docs, module: nn, small, triaged",thread:issue:49992:issue:59566,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49992 [docs] nn.modules pages should mention corresponding functional versions, e.g. for nn.Hardshrink, #59566 [discussion] Make docs consistent for torch.nn Modules",f89f3155f9b21fa462ab1c62d634cad2785ce22de24753ccf3d919954676b1b3
issue-collapse:1be05644ec0bc3bf,"Review thread: 2 issues - issue #49949: RFC: Add a ""see also"" section to operators' documentation",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49949, 50802",,,,module: docs,"module: docs, triaged",thread:issue:49949:issue:50802,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49949 RFC: Add a ""see also"" section to operators' documentation, #50802 Link to `torch.einsum` in `torch.tensordot`",100eacab46e3705a3f8a75fbe064527d08408532165f39d9ebd8b20ac6b03f5c
issue-collapse:5c3a6fa691cdcda5,Review thread: 2 issues - issue #49726: Wrong error is raised for property of nn.Module (again),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49726, 50292",,,,"module: bc-breaking, module: nn","module: bc-breaking, module: nn, topic: bc breaking, triaged",thread:issue:49726:issue:50292,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49726 Wrong error is raised for property of nn.Module (again), #50292 Implement dunder getattribute instead of dunder getattr for nn.Module",ede5ec2c05fe596ff6155527361c4f80ed6c475ede4d0daaf08716d42135a55e
issue-collapse:f98c7c032da0f497,Review thread: 2 issues - issue #49642: Turn deprecation warnings into errors in CI,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49642, 75419",,,,"module: ci, module: tests","better-engineering, module: ci, module: tests, triaged",thread:issue:49642:issue:75419,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49642 Turn deprecation warnings into errors in CI, #75419 Require PyTorch test suite to be warnings clean",562693c88caa86c000dfd414ba19090c83dbd27a86edc266be07d46c62327ae6
issue-collapse:2e62a57590d0470c,Review thread: 2 issues - issue #49480: Internal symbols are leaking from torch.nn.functional,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49480, 64185",,,,,"better-engineering, triaged",thread:issue:49480:issue:64185,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49480 Internal symbols are leaking from torch.nn.functional, #64185 PyTorch public API cleanup",e8a0ecc97cf177cadbdf420ed9ee48ef5b32ec1ef3670838fc31bad803b55f29
issue-collapse:e4085259aec5fca2,Review thread: 2 issues - issue #49252: Methods that solve systems of linear equations are memory-inefficient (batch-wise broadcasting),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"49252, 52915",,,,"module: linear algebra, module: memory usage","module: linear algebra, module: memory usage, triaged",thread:issue:49252:issue:52915,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#49252 Methods that solve systems of linear equations are memory-inefficient (batch-wise broadcasting), #52915 Broadcasting behaviour for linear algebra solvers",39ea9aa3d74efb1e42c31b70ee517fbf6e64a3c804cfae1e8f3147bbe2f84864
issue-collapse:f23250651998d85f,Review thread: 2 issues - issue #48353: [FR] bool tensor should support basic arithmetics,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"48353, 63023",,,,"module: boolean tensor, module: numpy, module: type promotion","module: boolean tensor, module: numpy, module: type promotion, triaged",thread:issue:48353:issue:63023,Handle this module: boolean tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#48353 [FR] bool tensor should support basic arithmetics, #63023 [feature request] Type promotions for Boolean tensors with sub operation + Numpy compatability",426f8cb500ffffc030b2d2b15bd4e18a6c892265c90f523c4d8c474a4e88b5de
issue-collapse:f226756ce0aa0ede,Review thread: 2 issues - issue #48246: [RFC] XPU device for PyTorch,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"48246, 53707",,,,module: intel,"feature, module: intel, triaged",thread:issue:48246:issue:53707,Handle this module: intel stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#48246 [RFC] XPU device for PyTorch, #53707 [RFC] Add common device runtime abstraction",88801415efb2837372e7707360a0a9a204c9016a0c709dd24861adada31f046b
issue-collapse:cee3665e0deb58e1,Review thread: 2 issues - issue #47908: The speed of pytorch with cudatoolkit 11.0 is slower than cudatoolkit 10.2,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"47908, 54408",,,,"module: cuda, module: performance","module: cuda, module: performance, triaged",thread:issue:47908:issue:54408,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#47908 The speed of pytorch with cudatoolkit 11.0 is slower than cudatoolkit 10.2, #54408 It is strange that PyTorch is slow on RTX 3090",dad969635f3499664c5ded3922bbe169e538c685c489faab9a7d121047a9c6c0
issue-collapse:50cdf66f33b7aedc,Review thread: 2 issues - issue #47688: ARM Mac 16-core Neural Engine,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"47688, 47702",,,,module: performance,"feature, module: performance, triaged",thread:issue:47688:issue:47702,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47688 ARM Mac 16-core Neural Engine, #47702 GPU acceleration for Apple's M1 chip?",7826e0b14d3d7110ea8d66ad6e910164bebfa006f3f56956ba99bb2c8d8a6be5
issue-collapse:ebc0e72e7fd05af2,Review thread: 2 issues - issue #47669: Pytorch 1.7.0 with cuda 11.1.1 and cudnn 8.0.5,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"47669, 48878",,,,"module: ci, module: cuda, oncall: releng","enhancement, module: ci, module: cuda, oncall: releng, triaged",thread:issue:47669:issue:48878,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47669 Pytorch 1.7.0 with cuda 11.1.1 and cudnn 8.0.5, #48878 Libtorch: nvrtc: error: invalid value for --gpu-architecture (-arch)",7e85249159aef06fe7b6e335940a830cc3b4bc088d4003065bfa1c59e3325136
issue-collapse:2cc1b67f952ddcfd,Review thread: 2 issues - issue #47149: Allow all torch.nn modules to accept arbitrary batch dimensions,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"47149, 62171",,,,"module: batching, module: convolution, module: nn, module: ux","enhancement, feature, module: batching, module: convolution, module: nn, module: ux, triaged",thread:issue:47149:issue:62171,Handle this module: batching stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#47149 Allow all torch.nn modules to accept arbitrary batch dimensions, #62171 [feature request] torch.nn.Conv3d on tensors having more than 5 dimensions",7585cc64040baadee113d840cb554021f5b5ac74ad6233c50fe15cb46cd84fd7
issue-collapse:392edb2669036b77,Review thread: 2 issues - issue #46902: How to use clang as a cuda compiler instead of nvcc?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"46902, 47649",,,,"module: build, module: cuda","enhancement, module: build, module: cuda, triaged",thread:issue:46902:issue:47649,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#46902 How to use clang as a cuda compiler instead of nvcc?, #47649 error: reference to __host__ function 'parallel_for<thrust::cuda_cub::for_each_f...' in __host__ __device__ function",8086e31787a67db3f594d4affe2c6be703baf9ec20c3c3f023d4faec85d22663
issue-collapse:b6ef3b64bd5e0a26,Review thread: 2 issues - issue #46559: Exception raised in rpc_async context is silently handled,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"46559, 46653",,,,"module: bootcamp, module: rpc, oncall: distributed","enhancement, module: bootcamp, module: rpc, oncall: distributed, triaged",thread:issue:46559:issue:46653,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#46559 Exception raised in rpc_async context is silently handled, #46653 Future returned by RPC should print a warning message on destruction if it's not waited",4ee58cefbb260b79a49ca70ebdcbd39cad34d030f2f9c5324bf839442ba27678
issue-collapse:71fe7dac73a422a3,Review thread: 2 issues - issue #4632: CUDNN_STATUS_INTERNAL_ERROR when training with conv3d,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"4632, 52211",,,,"module: convolution, module: cudnn","module: convolution, module: cudnn, triaged",thread:issue:4632:issue:52211,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#4632 CUDNN_STATUS_INTERNAL_ERROR when training with conv3d, #52211 Conv3D error : CUDNN_STATUS_INTERNAL_ERROR",edd817efd1d2ef8f885ccc22f2672dfe5af33aff2d383edea88c810259541ad0
issue-collapse:c3cbcbfe90b57bec,Review thread: 2 issues - issue #45682: [FX] Allow customization of the behavior of Proxy,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"45682, 45685",,,,module: fx,"module: fx, triaged",thread:issue:45682:issue:45685,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#45682 [FX] Allow customization of the behavior of Proxy, #45685 [fx] symbolic trace ""is None"" and ""is not None"" checks",b014bbe6dc936dcb4a424be9d847da1760d1881ae3eb8d8b336d5567fd7b8a1a
issue-collapse:0568baa9d7890d4c,Review thread: 2 issues - issue #44533: “doxygenfunction: Unable to resolve multiple matches for function ...” in C++ documentation,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"44533, 48548",,,,"module: cpp, module: docs","module: cpp, module: docs, triaged",thread:issue:44533:issue:48548,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#44533 “doxygenfunction: Unable to resolve multiple matches for function ...” in C++ documentation, #48548 at::size documentation conflict",895563de7a3fe98139522c30b7e714ca6780ea1a69fe32d3a178f403aaa94a90
issue-collapse:f30ebe9bb0cff2d7,"Review thread: 2 issues - issue #44159: [Feature Request] Add an option to run GPU tests only, and skip all CPU tests",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"44159, 55174",,,,module: tests,"actionable, feature, module: tests, triaged",thread:issue:44159:issue:55174,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#44159 [Feature Request] Add an option to run GPU tests only, and skip all CPU tests, #55174 Isolate CPU tests from GPU tests",dcb2548c5db7b40a425a656fd18c1862036baf6b166c07a90461e9c8ab315baf
issue-collapse:14e69235434818d0,Review thread: 2 issues - issue #43127: `onnxifi_load` symbol undefined in `libtorch_cpu.a` library built from source,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"43127, 66640",,,,module: build,"module: build, triaged",thread:issue:43127:issue:66640,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#43127 `onnxifi_load` symbol undefined in `libtorch_cpu.a` library built from source, #66640 Static Build of Libtorch not linking correctly",adbc7c7ffe7468e16d3532b70949e2e0d996febace40e1cb3c54761dcbaea5b6
issue-collapse:1e4c5fff394358b4,Review thread: 2 issues - issue #42812: checkpoint_sequential breaks backpropagation,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"7617, 42812",,,,"module: activation checkpointing, module: autograd","module: activation checkpointing, module: autograd, triaged",thread:issue:42812:issue:7617,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#7617 checkpoint(function, *args) should have the same requires_grad as function(*args), #42812 checkpoint_sequential breaks backpropagation",d8c15bbf9febfa39f49b4fa3d3996c0735dbf2353e01c31769186d5b2a446091
issue-collapse:5ee92ee3c2afc11d,Review thread: 2 issues - issue #42295: KeyError: 'track_running_stats' in batchnorm.extra_repr,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"42295, 42316",,,,"module: nn, module: printing, module: serialization, module: tests","module: nn, module: printing, module: serialization, module: tests, triaged",thread:issue:42295:issue:42316,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#42295 KeyError: 'track_running_stats' in batchnorm.extra_repr, #42316 Proper testing of `nn.Module` loading backward compatibility",010784f57d4162c855ffa521756684a5b32d306342442b285abc993a185f53f5
issue-collapse:4900f1f513d206c3,Review thread: 2 issues - issue #42109: Slow `index_add_` on `torch.long` tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"42109, 58136",,,,"module: cuda, module: docs, module: performance, module: scatter & gather ops","module: cuda, module: docs, module: performance, module: scatter & gather ops, triaged",thread:issue:42109:issue:58136,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#42109 Slow `index_add_` on `torch.long` tensors, #58136 scatter_add_ 6000-times slower with int64 compared to int32",f528683781370fa0a9a341efd7b15cbb4c057b86eb7e692abe41c9123dc0b370
issue-collapse:0c070e7860fec618,Review thread: 2 issues - issue #41530: Parallel computation of the diagonal of a Jacobian,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"41530, 53903",,,,"module: autograd, module: vmap","enhancement, feature, module: autograd, module: vmap, triaged",thread:issue:41530:issue:53903,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#41530 Parallel computation of the diagonal of a Jacobian, #53903 [torch.vmap] Support second order grads",d1b637a671c6bc7a055c2458de776e2b6dc16cc46cc7cc67f60ba5c047be49c1
issue-collapse:7f37721712143ced,Review thread: 2 issues - issue #4145: [Proposal] Consistent `batch_first` effect for RNN modules,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"4145, 4181",,,,"module: cudnn, module: docs, module: nn, module: rnn","module: cudnn, module: docs, module: nn, module: rnn, triaged",thread:issue:4145:issue:4181,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#4145 [Proposal] Consistent `batch_first` effect for RNN modules, #4181 Fused RNN refactor plan",89d84be47540ab8e4c6f2db506a2749baefb0b5beeaf0d61ead1427d67975b4d
issue-collapse:5c3f2564be67f33a,Review thread: 2 issues - issue #40581: Libtorch C++ multiple GPU performance slower than single GPU,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"40581, 48959",,,,"module: cpp, module: data parallel, module: ddp, module: performance, oncall: distributed","feature, module: cpp, module: data parallel, module: ddp, module: performance, oncall: distributed, triaged",thread:issue:40581:issue:48959,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40581 Libtorch C++ multiple GPU performance slower than single GPU, #48959 [Poll] Support DistributedDataParallel (DDP) in PyTorch C++ API (libtorch)",b99d32f2f9a01d52d8f751b9d301536bc5df0fdbb8515914bf9145b0f53cdb05
issue-collapse:b25e5d767e042927,Review thread: 2 issues - issue #39224: Negative stride values in `as_strided`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"39224, 53391",,,,"module: error checking, module: internals, module: memory format","module: error checking, module: internals, module: memory format, triaged",thread:issue:39224:issue:53391,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#39224 Negative stride values in `as_strided`, #53391 torch.empty_strided doesn't test if strides are negative",38de006c865947846dc45d18d360d69896262e6812d9e3a705ab8063b18236f1
issue-collapse:4d18071a713399f1,Review thread: 2 issues - issue #38767: weight_decay in Adam is not an L2 Penalty,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38767, 42843",,,,"module: docs, module: optimizer","module: docs, module: optimizer, triaged",thread:issue:38767:issue:42843,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38767 weight_decay in Adam is not an L2 Penalty, #42843 Documentation mistake of Adam in v1.6.0?",bf0b15930e7ea6fb10b5a2c3c4c98fee53410ae3e5f079a3a1eb62dde0eae5ab
issue-collapse:6f1e588e8e4cae6f,Review thread: 2 issues - issue #38642: Synchronization problem in torch.distributed with MPI on CUDA,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38642, 50092",,,,oncall: distributed,"oncall: distributed, triaged",thread:issue:38642:issue:50092,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38642 Synchronization problem in torch.distributed with MPI on CUDA, #50092 batch_isend_irecv: the receiving end cannot receive large tensors from the sending end correctly",8e36010bf7d5b5faeaf68fcc064fb6cae8d4ec5800c6502ef31687545b89d953
issue-collapse:d536d9ef3cffc335,Review thread: 2 issues - issue #38622: Clarification for usage of negative loss with optim.lr_scheduler.ReduceLROnPlateau,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38622, 47513",,,,"module: docs, module: optimizer","module: docs, module: optimizer, triaged",thread:issue:38622:issue:47513,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38622 Clarification for usage of negative loss with optim.lr_scheduler.ReduceLROnPlateau, #47513 ReduceLROnPlateau fails for negative input",b63bd465402ea0b1000c64e46cee937a15f396c45a818ea18622a73bce4dfc0d
issue-collapse:4a4022bd67ce1495,Review thread: 2 issues - issue #38310: Compiling errors when trying to cross-compile the C++ API for RTOS (QNX),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38310, 50503",,,,"module: binaries, module: build, module: cpp","module: binaries, module: build, module: cpp, triaged",thread:issue:38310:issue:50503,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38310 Compiling errors when trying to cross-compile the C++ API for RTOS (QNX), #50503 Huge file sizes for libtorch and dependencies?",bc2a230819938d2c828aa0293abd066a3f259b2b5325e1ceb64853b2a82e7922
issue-collapse:43034df0a22c2edd,Review thread: 2 issues - issue #38185: Feature Request: Sampled softmax loss,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"634, 38185",,,,"module: loss, module: nn, module: primtorch","Stale, feature, module: loss, module: nn, module: primTorch, triaged",thread:issue:38185:issue:634,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#634 Feature Request: NegativeSampling and HierarchicalSoftmax loss functions, #38185 Feature Request: Sampled softmax loss",d8d0bea862295f9a52dbf5cd59e2adb0d595fe93bfa5763eb7d5b02e41b87896
issue-collapse:498bd5b2405376a8,Review thread: 2 issues - issue #38010: Split more pages into sub-topics,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38010, 39836",,,,"module: doc infra, module: docs","module: doc infra, module: docs, open source, triaged",thread:issue:38010:issue:39836,Handle this module: doc infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38010 Split more pages into sub-topics, #39836 Feedback to split certain doc pages into sub-topics",e47425d370dd3273dc8d3d2327d9b7c8b916a4069d1e2bb95a8514f7f3845f7a
issue-collapse:f675f66da1013c61,Review thread: 2 issues - issue #37160: Enhanced operator context when reporting errors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"37160, 78738",,,,"module: primtorch, module: shape checking","enhancement, module: primTorch, module: shape checking, triaged",thread:issue:37160:issue:78738,Handle this module: primtorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#37160 Enhanced operator context when reporting errors, #78738 [primTorch] Sensible Error Messages",665c9f1619149ebe8f6ef062ff86a094b3ae09d5faa3e5cb66658bdc4ed0123f
issue-collapse:14aeaa64754236ed,Review thread: 2 issues - issue #37002: [RFC] Modularize DistributedDataParallel,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"37002, 42849",,,,oncall: distributed,"feature, oncall: distributed, triaged",thread:issue:37002:issue:42849,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#37002 [RFC] Modularize DistributedDataParallel, #42849 [RFC] DeepSpeed + PT Distributed Integration",03cd7b8a1beb3a594346416edd74f96ff356ce057f7fbe2a384f719b1b3ae7bf
issue-collapse:7a309dd342be262d,Review thread: 2 issues - issue #3667: Exposing CuDNN benchmark strategy selection,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"3667, 4392",,,,"module: bootcamp, module: cudnn","feature, module: bootcamp, module: cudnn, triaged",thread:issue:3667:issue:4392,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#3667 Exposing CuDNN benchmark strategy selection, #4392 Cache CuDNN benchmark selection, turn it on by default, use it across PyTorch runs",cd5ccedb491686481c6a2bc1c54d73364e2de0906580983505b3bcf7a531b2e6
issue-collapse:c64d329d5045d373,Review thread: 2 issues - issue #36444: Comparison ops for Complex Tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"36444, 55661",,,,"module: complex, module: numpy","enhancement, module: complex, module: numpy, triaged",thread:issue:36444:issue:55661,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#36444 Comparison ops for Complex Tensors, #55661 RuntimeError: ""max_cuda"" not implemented for 'ComplexFloat'",cd3c3a5e6eec91059df87060aed4a9b2bbb41a32ceb2bfb9e8680b45e7047b7e
issue-collapse:c046c1cfaef5f71c,Review thread: 2 issues - issue #36333: Illegal memory access / CUDNN_STATUS_MAPPING_ERROR on Quadro 8000,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"36333, 45769",,,,module: cudnn,"module: cudnn, triaged",thread:issue:36333:issue:45769,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#36333 Illegal memory access / CUDNN_STATUS_MAPPING_ERROR on Quadro 8000, #45769 F.conv2d() causes RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR",5d7c513bb8fa82454c081965395d1688004b64ec2a52cac3abff07629893a00f
issue-collapse:49e3c1bb2acc21b7,Review thread: 2 issues - issue #3619: Multiprocessing with torch.solve hangs,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"3619, 4358",,,,"module: linear algebra, module: mkl, module: mkldnn, module: multiprocessing","module: linear algebra, module: mkl, module: mkldnn, module: multiprocessing, triaged",thread:issue:3619:issue:4358,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#3619 Multiprocessing with torch.solve hangs, #4358 Invoking MKL in multiprocessing with importing torch causes blocking",4db031d7d9915524d3efe8bb58c9210751c17d669a6903e0d7666c9905e20961
issue-collapse:64ed00fdc02352c5,Review thread: 2 issues - issue #35678: USE_AVX/USE_AVX2 does not affect __AVX2__ macro defition,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"35678, 36999",,,,"module: build warnings, module: ci, module: tests, oncall: distributed","internals, module: build warnings, module: ci, module: tests, oncall: distributed, triaged",thread:issue:35678:issue:36999,Handle this module: build warnings stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#35678 USE_AVX/USE_AVX2 does not affect __AVX2__ macro defition, #36999 Add a CI configuration to test USE_DISTRIBUTED=0",7c2747e68420f151d8fd136d3d1f8abe102204b993ae59eb9318295b11c195c8
issue-collapse:b0f373789a73f8ac,Review thread: 2 issues - issue #34606: Multiple CMake target errors ever since commit 0e52627358,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"34606, 37488",,,,"module: build, module: xnnpack","module: build, module: xnnpack, triaged",thread:issue:34606:issue:37488,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#34606 Multiple CMake target errors ever since commit 0e52627358, #37488 undefined reference to pthreadpool_compute*",337d9770b5aa1a11da76d0e1ec02699ac8137e9d95d6e392ab1aa4ba280058e8
issue-collapse:3d0e9bbb4efff0f6,Review thread: 2 issues - issue #34361: [JIT legacy executor] device propagation regression,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"34361, 35446",,,,oncall: jit,"oncall: jit, triaged",thread:issue:34361:issue:35446,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#34361 [JIT legacy executor] device propagation regression, #35446 [JIT] dropout fails on legacy mode",815b34e61f63be4cf2cc17014161d07c00c32d2a71160352398ba6ee36b9c380
issue-collapse:af06104b413ef08d,Review thread: 2 issues - issue #34306: [docs] Missing docs online for conv_tbc,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"34306, 50962",,,,"module: docs, module: memory format","feature, module: docs, module: memory format, triaged",thread:issue:34306:issue:50962,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#34306 [docs] Missing docs online for conv_tbc, #50962 Add option for Convolutions to operate over NTC and NHWC tensors",911ac9343a76d6bc0d0b3c86ca9bfc368c08e8a2ecde6984e34dd07d5056dac9
issue-collapse:307de35aa31fdbed,Review thread: 2 issues - issue #32306: [FR] bincount along arbitrary dimension,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"32306, 43867",,,,module: sorting and selection,"enhancement, function request, module: sorting and selection, triaged",thread:issue:32306:issue:43867,Handle this module: sorting and selection stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#32306 [FR] bincount along arbitrary dimension, #43867 torch.bincount beyond 1d arrays",1668dde38fe1d53b9a839e47607751dd4a613975e9e26d8e9c20def9d8985485
issue-collapse:135893f781739dc1,Review thread: 2 issues - issue #32300: [jit] Various problems calling `@staticmethod`s in 1.4.0,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"32300, 32365",,,,oncall: jit,"oncall: jit, triaged",thread:issue:32300:issue:32365,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#32300 [jit] Various problems calling `@staticmethod`s in 1.4.0, #32365 [jit] Calling `torch.jit.script` on `@staticmethod`s which in turn call other `@staticmethod`s results in a `TypeError`",1f8dce52addb9726eea94b2325a0b5ca782d5fb7973c9054bdae455ec4124918
issue-collapse:182d379600e5aebc,Review thread: 2 issues - issue #31822: Libtorch's files conflict with glog's file?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"31822, 88980",,,,module: build,"module: build, triaged",thread:issue:31822:issue:88980,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#31822 Libtorch's files conflict with glog's file?, #88980 Glog macro redefinition problem when including headers from both libtorch and glog",ab14088f54a183d1a68fb5c178c2438d5fd7d2f0797ac84cbca957fc12e12d02
issue-collapse:4f172f4b1148a6c1,Review thread: 2 issues - issue #31528: cuCtxGetDevice error and seg fault with DDP and OpenMPI,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"31528, 43453",,,,"module: cuda, oncall: distributed","module: cuda, oncall: distributed, triaged",thread:issue:31528:issue:43453,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#31528 cuCtxGetDevice error and seg fault with DDP and OpenMPI, #43453 CUDAGuard might not create CUcontext",e3ffd9d4539bf7be285c1cb809337eecb139a0029ab121726aab3be462b5899f
issue-collapse:addca9f7f85ca0f2,Review thread: 2 issues - issue #31461: Make RRef.to_here() non-blocking,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"31461, 96036",,,,"module: multithreading, module: rpc","module: multithreading, module: rpc, triaged",thread:issue:31461:issue:96036,Handle this module: multithreading stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#31461 Make RRef.to_here() non-blocking, #96036 Questions and Possible Features: Pytorch RPC 'future.wait()' will not release GIL which will block other thread's execution when using multithreading.",09e2eb41d54dc0a06d2248381057a2e114315b932985ac1465610733629acdec
issue-collapse:f1a6826314284426,Review thread: 2 issues - issue #30965: JIT breaks with postponed annotations,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"30965, 32544",,,,oncall: jit,"oncall: jit, triaged",thread:issue:30965:issue:32544,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30965 JIT breaks with postponed annotations, #32544 [jit] Use `typing.get_type_hints` instead of parsing types manually",fe68e6b64549a9e0bed7febd68bee8c044ac4a21da9cea77e852d04eec491667
issue-collapse:f3d7c8c7e1edc1a7,Review thread: 2 issues - issue #30903: Don't ship protoc in wheels,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"30903, 31907",,,,"module: binaries, module: cpp","module: binaries, module: cpp, triaged",thread:issue:30903:issue:31907,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30903 Don't ship protoc in wheels, #31907 Compile libtorch by source code failed.",7d29aa0c29a965042092be0480479c150e99617c1fe826cc7c098255f26e06ae
issue-collapse:054529a10e7e7dbb,Review thread: 2 issues - issue #30803: Pytorch openmp thread number tuning option for CPU trainning,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"30803, 31356",,,,"module: cpu, module: multithreading, module: performance","enhancement, feature, module: cpu, module: multithreading, module: performance, triaged",thread:issue:30803:issue:31356,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30803 Pytorch openmp thread number tuning option for CPU trainning, #31356 Optimizing DLRM for CPU",a5c91889cb2d636351933e38601366cc9482d6eb6d50456d02d2b772381ff083
issue-collapse:67348804a1ca8493,Review thread: 2 issues - issue #30633: Remove RPC internal helper that overrides the default pickler,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"30633, 31228",,,,module: rpc,"module: rpc, triaged",thread:issue:30633:issue:31228,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30633 Remove RPC internal helper that overrides the default pickler, #31228 [RPC] Support nn.Module pickling with share memory",4eb863317a2d9566584b16061af48db074d656b954d7b6e261b25036735fb098
issue-collapse:1decd150951de67c,Review thread: 2 issues - issue #30574: `index_select` with multidimensional `index`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"30574, 31913",,,,module: advanced indexing,"feature, function request, module: advanced indexing, triaged",thread:issue:30574:issue:31913,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#30574 `index_select` with multidimensional `index`, #31913 `index_add_` with multidimensional index",f32031a963de0e001fb8d489f962c3fc278a12106e01ebb6b05234e45bd260d1
issue-collapse:5e792db490cbadc2,Review thread: 2 issues - issue #29981: [feature request] Multivariate normal CDF,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"29981, 72258",,,,module: distributions,"feature, module: distributions, triaged",thread:issue:29981:issue:72258,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#29981 [feature request] Multivariate normal CDF, #72258 NotImplementedError in torch.distributions",529d47188e923f82d247ae9573c923abc4843bc420054317d39118fbd8c557dc
issue-collapse:4b7c0b4bfe0e3ef8,Review thread: 2 issues - issue #29548: Dispatch key reorganization,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"29548, 38975",,,,module: internals,"module: internals, triaged",thread:issue:29548:issue:38975,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#29548 Dispatch key reorganization, #38975 Use general ATen dispatch mechanism",73face814366ddfa24ab0e728be8aa5a999d0e0d7e0a958ca66cbc83c5ab246f
issue-collapse:b0707fcde6394374,Review thread: 2 issues - issue #29028: Inconsistent documentation for in-place functions at https://pytorch.org/docs/stable/torch.html,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"29028, 29116",,,,"module: docs, module: ux","feature, module: docs, module: ux, triaged",thread:issue:29028:issue:29116,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#29028 Inconsistent documentation for in-place functions at https://pytorch.org/docs/stable/torch.html, #29116 torch.masked_fill missing out argument",02f0c9337613fdaaa3831d34b14e71d1f91124ec5c6fbbbdc7f68290db552d8d
issue-collapse:45995c66edb1ab32,Review thread: 2 issues - issue #28733: CPU MaxPool2d is very slow,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"28733, 34675",,,,"module: bootcamp, module: cpu, module: mkldnn, module: nn, module: performance, module: pooling","function request, module: bootcamp, module: cpu, module: mkldnn, module: nn, module: performance, module: pooling, triaged",thread:issue:28733:issue:34675,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#28733 CPU MaxPool2d is very slow, #34675 [feature request]Support dilation parameter for unfold2d_* function (slow cpu maxpool2d #28733)",4bde776ecc4f8377ebd227d213f90980b6b838b0e1645b0318cda85413e66f7d
issue-collapse:cc0b27ddc57ae8ab,Review thread: 2 issues - issue #28619: Channels Last (NHWC) support plan.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"28619, 59168",,,,"module: cuda, module: memory format","module: cuda, module: memory format, triaged",thread:issue:28619:issue:59168,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#28619 Channels Last (NHWC) support plan., #59168 [CUDA] Add channels_last_3d support for commonly used modules",acc313a7e49b30ff7a8e9415e9723b1c5e0213c3d448b8d0052e9e5c982adaf9
issue-collapse:ef37da08c0cc7407,Review thread: 2 issues - issue #28594: [BUG] Can't Deepcopy module with weightnorm,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"28594, 57815",,,,module: nn,"module: nn, triaged",thread:issue:28594:issue:57815,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#28594 [BUG] Can't Deepcopy module with weightnorm, #57815 copy.deepcopy fails on spectral-norm layers after the first forward",2216ed73ed0585d6d87bc18b1c5aa8d13ea0081bb425edb4c9a9b7d96884261e
issue-collapse:e4d19c216c849235,Review thread: 2 issues - issue #28417: Patch: Fix for using `clang` to compile CUDA,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"28417, 29042",,,,"module: build, module: cuda","enhancement, module: build, module: cuda, triaged",thread:issue:28417:issue:29042,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#28417 Patch: Fix for using `clang` to compile CUDA, #29042 Support clang+cuda builds",29e8b48ef112cb1169a6b01b087878627790d58a72526f84ce7d08d872418172
issue-collapse:eaf1f66bf678a284,Review thread: 2 issues - issue #27647: 🚀 Graceful RPCAgent termination in multi-driver scenario,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"27647, 40107",,,,"module: bootcamp, module: rpc","better-engineering, feature, module: bootcamp, module: rpc, triaged",thread:issue:27647:issue:40107,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#27647 🚀 Graceful RPCAgent termination in multi-driver scenario, #40107 [RFC] Add a RPC context manager that collects/waits for all RPC futures created in scope",302d2d5c345e978c8735a6f4a40eccf950619b6ed8a8969078c34d43afca383e
issue-collapse:604475a49459a926,Review thread: 2 issues - issue #27144: torch::jit::script::Module has no zero_grad(),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"27144, 27475",,,,oncall: jit,"jit-backlog, oncall: jit, triaged",thread:issue:27144:issue:27475,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#27144 torch::jit::script::Module has no zero_grad(), #27475 [JIT] script::Module API parity with nnmodule",e49c47cb1abfb5dfa883cf44bba81a68f15e3503e08de887ce253056c5f9262c
issue-collapse:63a8e7253773ae45,Review thread: 2 issues - issue #25591: IValue pickle does not work properly if an empty tensor table is not provided,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"25591, 29987",,,,oncall: jit,"jit-backlog, oncall: jit, triaged",thread:issue:25591:issue:29987,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#25591 IValue pickle does not work properly if an empty tensor table is not provided, #29987 Move the attributes of a module to the given device",f1e1d7092c8672a5b62979653f464a685e67443a8003991e1ab06feeb6952c02
issue-collapse:4c488decb20e8c0a,Review thread: 2 issues - issue #25478: Delete TensorOptions::operator==,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"25478, 68070",,,,"module: correctness (silent), module: cpp, module: internals","enhancement, module: correctness (silent), module: cpp, module: internals, small, triaged",thread:issue:25478:issue:68070,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#25478 Delete TensorOptions::operator==, #68070 C++ at::Tensor's pinned_memory status is not printing out correctly.",9fb29154063489e2114dbc048c1856a4843275a633f50ae7d83f7a700f72cce0
issue-collapse:8ef5e87a4492039d,Review thread: 2 issues - issue #25416: Add GPU support to c10d allreduce_coalesced,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"25416, 38995",,,,oncall: distributed,"feature, oncall: distributed, triaged",thread:issue:25416:issue:38995,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#25416 Add GPU support to c10d allreduce_coalesced, #38995 [FR] [distributed] coalesced primitives",d0abe35d9c7d304ffedeb6b9650add0c0dc58f45f6494cb4ea759c466dc00468
issue-collapse:3ed4073b1940a4e9,Review thread: 2 issues - issue #25150: Problems with install python from source,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"25150, 46724",,,,module: build,"module: build, triaged",thread:issue:25150:issue:46724,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#25150 Problems with install python from source, #46724 PROBLEM WITH INSTALLATION PYTORCH FROM SOURCE",99160889c5fe6bd70d9e2456dd92b765fde72836f037722da77013492d86536c
issue-collapse:4d155a56ecb9251a,Review thread: 2 issues - issue #24834: PyTorch 1.2 'module' object has no attribute 'BFloat16StorageBase',related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"24834, 56126",,,,"module: build, module: undefined reference, module: vision","module: build, module: undefined reference, module: vision, triaged",thread:issue:24834:issue:56126,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#24834 PyTorch 1.2 'module' object has no attribute 'BFloat16StorageBase', #56126 AttributeError: module 'torch._C' has no attribute 'ComplexDoubleStorageBase'",300feafb08c3a3280445284c0deb8fe2c6cf09025be54247ecaf8bf229e70dc7
issue-collapse:c9899f0090cc6bdc,Review thread: 2 issues - issue #2478: ReduceLROnPlateau with a naive Backtracking,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"2478, 30873",,,,module: optimizer,"feature, module: optimizer, triaged",thread:issue:2478:issue:30873,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#2478 ReduceLROnPlateau with a naive Backtracking, #30873 ReduceLROnPlateau detects a plateau during a steady decrease after a spike",e1cce3a2c8ae38718e4d70b0a53126e3bc0e3572c2c78d4df677e0201831874a
issue-collapse:2563c98551e372fd,Review thread: 2 issues - issue #24478: Port CPU_tensor_apply functions to TensorIterator (umbrella issue),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"24478, 24498",,,,module: cpu_tensor_apply,"better-engineering, module: CPU_tensor_apply, triaged",thread:issue:24478:issue:24498,Handle this module: cpu_tensor_apply stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#24478 Port CPU_tensor_apply functions to TensorIterator (umbrella issue), #24498 Migrate CPU_tensor_apply to TensorIterator in aten/src/ATen/native/TensorCompare.cpp:30",7e347b2a6381d7b45aeb119bc4dcf4effa77e90b9b3b91c188950218c68df61a
issue-collapse:1d0abe72e079b0c6,Review thread: 2 issues - issue #24234: torch.utils.tensorboard.SummaryWriter fails to flush at program exit,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"24234, 91887",,,,module: tensorboard,"module: tensorboard, triaged",thread:issue:24234:issue:91887,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#24234 torch.utils.tensorboard.SummaryWriter fails to flush at program exit, #91887 Only the first logged trace in a given log dir is visible in tensorboard.",894df291d0af4f0636580e467633f7b53a4dec5dfbed5057eca45f643b506373
issue-collapse:2986eabb93f4a4a0,Review thread: 2 issues - issue #23425: Hanging on when one gpu node return zero as loss in the context of distributed data parallel training,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"23425, 49321",,,,"module: ddp, oncall: distributed","module: ddp, oncall: distributed, triaged",thread:issue:23425:issue:49321,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#23425 Hanging on when one gpu node return zero as loss in the context of distributed data parallel training, #49321 Backward hangs with DDP during training.",710c8f0a0afe469665885a9d8095d012132aba17afd3884c8634ed250c46c28e
issue-collapse:0132a509729dc32d,Review thread: 2 issues - issue #22343: New Weight Scheduler Concept for Weight Decay,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"3790, 22343",,,,module: optimizer,"feature, module: optimizer, triaged",thread:issue:22343:issue:3790,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#3790 Add SGDR, SGDW, AdamW and AdamWR, #22343 New Weight Scheduler Concept for Weight Decay",80fad2ff68134a51b996e8fe723e89aebe6f695333b7a6a59f3390601d9ac6e6
issue-collapse:f0083b3d27a1bef3,Review thread: 2 issues - issue #21731: Improve multithreaded random number generation (RNG),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"21731, 22414",,,,"module: cpu, module: multithreading, module: performance, module: random","module: cpu, module: multithreading, module: performance, module: random, triaged",thread:issue:21731:issue:22414,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#21731 Improve multithreaded random number generation (RNG), #22414 CPU random number generator is slow",e1d89ad8bbbafa083610716c2c229641cee0e045b459d95991cabfe932578e6d
issue-collapse:f131344ab597d4da,Review thread: 2 issues - issue #21700: (LLD 8.0.0) ld: error: can't create dynamic relocation R_X86_64_DTPOFF32 against symbol: ideep::utils::computation_cache,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"21700, 43459",,,,"module: build, module: mkldnn, module: static linking","module: build, module: mkldnn, module: static linking, triaged",thread:issue:21700:issue:43459,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#21700 (LLD 8.0.0) ld: error: can't create dynamic relocation R_X86_64_DTPOFF32 against symbol: ideep::utils::computation_cache, #43459 Trying to build PyTorch from source with LLD 8 fails",017da06e01e4d945afab011c56d0b98ff51e3e58206868eca3e423f716435c82
issue-collapse:e50e974f6667bbd3,Review thread: 2 issues - issue #20323: Support size to `torch.normal`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"20323, 69386",,,,"module: numpy, module: structured kernels","function request, module: numpy, module: structured kernels, triaged",thread:issue:20323:issue:69386,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#20323 Support size to `torch.normal`, #69386 Port `normal` to structured kernel",e07cf776f570f46a8d1830295ca97be3923742bfc663c7d4d6ea24f6312ed1d0
issue-collapse:a01cc1aca50d19d8,Review thread: 2 issues - issue #2001: Implement similar PyTorch function as model.summary() in keras?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"2001, 62009",,,,module: nn,"feature, function request, module: nn, triaged",thread:issue:2001:issue:62009,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#2001 Implement similar PyTorch function as model.summary() in keras?, #62009 multi-input and multi-output support for nn.Module",fafa9f351891a4142afe55551d512a84a93d2ef13cc70ab18805b0b5dcff4402
issue-collapse:aa9bf3114e14563f,"Review thread: 2 issues - issue #19637: [jit] Traced {zeros,empty}()/{zeros,empty}_like() calls do not respect device args.",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"19637, 19685",,,,"module: pybind, module: tensor creation, oncall: jit","module: pybind, module: tensor creation, oncall: jit, triaged",thread:issue:19637:issue:19685,Handle this module: pybind stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#19637 [jit] Traced {zeros,empty}()/{zeros,empty}_like() calls do not respect device args., #19685 Bad overload order for zeros_like",5106c7ce78b082d7c65be8d4f68f297e6ca5924e6a7c31b35b4f1ce8ba008821
issue-collapse:65a50c1d5f4126fd,Review thread: 2 issues - issue #19248: add stable distribution in torch.distributions,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"19248, 91300",,,,module: distributions,"feature, low priority, module: distributions, triaged",thread:issue:19248:issue:91300,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#19248 add stable distribution in torch.distributions, #91300 [🚀 Feature Request] pdf and sampling from Alpha-stable distribution",a3b2876b4a5c6dffca9606dd1c5fc57f435bb121ea7f0795c0e994d8a181055f
issue-collapse:5bba0197ce04850a,Review thread: 2 issues - issue #18634: Speed-up torch.cat on CPU,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"18634, 31474",,,,"module: cpu, module: cuda, module: norms and normalization, module: performance","module: cpu, module: cuda, module: norms and normalization, module: performance, triaged",thread:issue:18634:issue:31474,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#18634 Speed-up torch.cat on CPU, #31474 Slow clip_grad_norm_ because of .item() calls when run on device",d6c009a71c67dee1310b392397b72636de32d525a14a556589af1ba015eb38b0
issue-collapse:1c9feeb634e6fcb7,Review thread: 2 issues - issue #17901: should disable AVX on 32bit x86 / refine AVX availability tests,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"17901, 40988",,,,module: build,"module: build, triaged",thread:issue:17901:issue:40988,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#17901 should disable AVX on 32bit x86 / refine AVX availability tests, #40988 '_mm256_extract_epi64' was not declared in this scope when compiling on Debian 32-bit",db27f4240c08030cbc4258735ea44c69e5a3d583ad48fb89e6c5edc1d9c724ca
issue-collapse:de511b5da67da5a4,Review thread: 2 issues - issue #17897: CUDA large matrix-vector product (torch.mv) causes illegal memory access,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"17897, 31708",,,,"module: 64-bit, module: cublas, module: cuda, module: dependency bug","module: 64-bit, module: cublas, module: cuda, module: dependency bug, triaged",thread:issue:17897:issue:31708,Handle this module: 64-bit stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#17897 CUDA large matrix-vector product (torch.mv) causes illegal memory access, #31708 Computing dot product of columns sliced from large matrix causes illegal memory access in CUDA",06d96df6e85e3bf5330d407d8c7437ede9829a30aa58860cf23e9041d005154f
issue-collapse:315d235998446591,Review thread: 2 issues - issue #17774: RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED when batch size is too large,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"17774, 21688",,,,"module: batching, module: convolution, module: cudnn","function request, module: batching, module: convolution, module: cudnn, triaged",thread:issue:17774:issue:21688,Handle this module: batching stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#17774 RuntimeError: cuDNN error: CUDNN_STATUS_EXECUTION_FAILED when batch size is too large, #21688 Batched Conv2d for sequence data",6c2dc7c1bf6439d29571cee624ffefc569b450f7b95274789c25da7dc1f00f33
issue-collapse:aa39af23b7ab98a0,"Review thread: 2 issues - issue #17425: improved assert message in the case of ""CUDA error: device-side assert triggered""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"17425, 19120",,,,"module: bootcamp, module: cuda, module: error checking, module: molly-guard","module: bootcamp, module: cuda, module: error checking, module: molly-guard, triaged",thread:issue:17425:issue:19120,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#17425 improved assert message in the case of ""CUDA error: device-side assert triggered"", #19120 Embedding layer does not check input range",b182856fd38f993a247fd9a4c3ae91b5644ee43ee5028094b7e026c02dc57ff3
issue-collapse:4f09cbbc952e16ea,Review thread: 2 issues - issue #173225: [RFC] Renaming MPS backend as Metal backend,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"86076, 173225",,,,"module: arm, module: build, module: mps","enhancement, low priority, module: arm, module: build, module: mps, triaged",thread:issue:173225:issue:86076,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#86076 Figure out the future of Metal backend given the existence of MPS, #173225 [RFC] Renaming MPS backend as Metal backend",ebf9be265e80e278dca9c8ad6c07f6bbe41cf83c322e57ca5c39770547935c2e
issue-collapse:2d5de4228035f779,Review thread: 2 issues - issue #17234: [feature request] build and move distributions w/ device and/or dtype,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"7795, 17234",,,,"module: build, module: distributions","enhancement, module: build, module: distributions, triaged",thread:issue:17234:issue:7795,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#7795 [feature request] Add cudaification API for distributions, #17234 [feature request] build and move distributions w/ device and/or dtype",d87e9a52a804121ce927265e1bae90c97ac9070e3282573cd3ca8e66cc27b237
issue-collapse:1bec8cbfc9a54e8d,Review thread: 2 issues - issue #172034: Become compatible with CMAKE_CUDA_ARCHITECTURES,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"71379, 172034",,,,"module: build, module: cpp, module: cpp-extensions, module: cuda","module: build, module: cpp, module: cpp-extensions, module: cuda, triaged",thread:issue:172034:issue:71379,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71379 Leaky cmake cuda compile options, #172034 Become compatible with CMAKE_CUDA_ARCHITECTURES",6dd445145d2b9163d1f265831e234c616502b6e436757e5da7ddc78cb751f0a9
issue-collapse:47a47ddadf149e11,Review thread: 2 issues - issue #168855: TestModule: test/test_testing.py::TestTesting,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168855, 168858",,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168855:issue:168858,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168855 TestModule: test/test_testing.py::TestTesting, #168858 Test: TestTesting.test_cuda_assert_should_not_stop_common_distributed_test_suite",fb35a51166a36d49d1e2a7b6aa15bd28e56c0b6e88bdda931108e3aa175b9487
issue-collapse:362d21ab7caa35fd,Review thread: 2 issues - issue #168798: TestModule: test/test_nn.py::TestNN,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168798, 168800",,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168798:issue:168800,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168798 TestModule: test/test_nn.py::TestNN, #168800 Test: TestNN.test_batchnorm",e5d1451ce4a132f004d1f5d15e4eb8b41977a8b58c951aebea0a86c9f97183e9
issue-collapse:229556fcb5707f2d,Review thread: 2 issues - issue #168780: TestModule: test/test_matmul_cuda.py::TestMixedDtypesLinearCuda,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168780, 168781",,,,"module: linear algebra, module: rocm","module: linear algebra, module: rocm, rocm-skipped-tests, triaged",thread:issue:168780:issue:168781,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168780 TestModule: test/test_matmul_cuda.py::TestMixedDtypesLinearCuda, #168781 Test: TestMixedDtypesLinearCuda.test_mixed_dtypes_linear",5a7959cb8b3b93b06256cb1a470f47efd61f0e8cfbc93db9d3235f35e6f10a1e
issue-collapse:cf18b1c536447a4f,Review thread: 2 issues - issue #168742: TestModule: test/test_cuda_primary_ctx.py::TestCudaPrimaryCtx,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168742, 168743",,,,module: rocm,"module: rocm, rocm-skipped-tests, triaged",thread:issue:168742:issue:168743,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168742 TestModule: test/test_cuda_primary_ctx.py::TestCudaPrimaryCtx, #168743 Test: TestCudaPrimaryCtx.test_set_device_0",9c2ff3e2a97f0ec6975a23cd2976bcad8fb3b8bca120b8c25023d57bdbb0b171
issue-collapse:4c275b8e9a98a1ea,Review thread: 2 issues - issue #168692: TestModule: test/nn/test_convolution.py::TestConvolutionNNDeviceType,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168692, 168696",,,,"module: convolution, module: rocm","module: convolution, module: rocm, rocm-skipped-tests, triaged",thread:issue:168692:issue:168696,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168692 TestModule: test/nn/test_convolution.py::TestConvolutionNNDeviceType, #168696 Test: TestConvolutionNNDeviceType.test_conv3d_cudnn_broken",1d47b33ab5b5b1b9b1e647063266822ee807df56882d70c5f19ad826a906d155
issue-collapse:88657f06d0c3d245,Review thread: 2 issues - issue #168623: TestModule: test/inductor/test_minifier_isolate.py::MinifierIsolateTests,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168623, 168624",,,,"module: inductor, module: rocm, oncall: pt2","module: inductor, module: rocm, oncall: pt2, rocm-skipped-tests, triaged",thread:issue:168623:issue:168624,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168623 TestModule: test/inductor/test_minifier_isolate.py::MinifierIsolateTests, #168624 Test: MinifierIsolateTests.test_after_aot_gpu_runtime_error",9144158c1104e3953b7edbad81eec9c415b552031dfdbec7fa8e8bdf0cf59e15
issue-collapse:ca3a8410b4e39d5c,Review thread: 2 issues - issue #168536: TestModule: test/distributed/test_symmetric_memory.py::LoweringTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168536, 168537",,,,"module: rocm, module: symm_mem","module: rocm, module: symm_mem, rocm-skipped-tests, triaged",thread:issue:168536:issue:168537,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168536 TestModule: test/distributed/test_symmetric_memory.py::LoweringTest, #168537 Test: LoweringTest.test_lowering_one_shot_all_reduce",1360afdc2acdba4be9cbd6699ad567366164deae8491491e0ff7942885af3f43
issue-collapse:02c7dece9039e3d4,Review thread: 2 issues - issue #168480: TestModule: test/distributed/test_nccl.py::NCCLSymmetricMemoryTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168480, 168481",,,,"module: nccl, module: rocm, module: symm_mem","module: nccl, module: rocm, module: symm_mem, rocm-skipped-tests, triaged",thread:issue:168480:issue:168481,Handle this module: nccl stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168480 TestModule: test/distributed/test_nccl.py::NCCLSymmetricMemoryTest, #168481 Test: NCCLSymmetricMemoryTest.test_nccl_symmem_alloc",7bb3d2824bc2ae0985dead795fc660c9a424f2c06b501aeb66fa399667ba308d
issue-collapse:818ece31d013f4ca,Review thread: 2 issues - issue #168451: TestModule: test/distributed/test_aten_comm_compute_reordering.py::TestComputeCommReorderingMultiProc,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168451, 168452",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168451:issue:168452,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168451 TestModule: test/distributed/test_aten_comm_compute_reordering.py::TestComputeCommReorderingMultiProc, #168452 Test: TestComputeCommReorderingMultiProc.test_grouped_scheduler_node",ccf252cc2fcda0d0d3873c3df4faa98a8b865ca4e4508faf2ace248c0308bce4
issue-collapse:db18ff501f70f9fe,Review thread: 2 issues - issue #168449: TestModule: test/distributed/tensor/test_tensor_ops.py::DistTensorOpsTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168449, 168450",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168449:issue:168450,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168449 TestModule: test/distributed/tensor/test_tensor_ops.py::DistTensorOpsTest, #168450 Test: DistTensorOpsTest.test_index",02bb449a8931392115d7e0a74a40ccbba7e2d746ffa2e5785269052583200365
issue-collapse:0e7a778c3b1688d3,Review thread: 2 issues - issue #168447: TestModule: test/distributed/tensor/test_matrix_ops.py::DistMatrixOpsTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168447, 168448",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168447:issue:168448,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168447 TestModule: test/distributed/tensor/test_matrix_ops.py::DistMatrixOpsTest, #168448 Test: DistMatrixOpsTest.test_grouped_mm",5d061fde4bd382fb5e708a56cce0cddeb4d5ad932da792c5dd579bc0db5cdcfa
issue-collapse:8db82fb4b1cb7389,Review thread: 2 issues - issue #168445: TestModule: test/distributed/tensor/test_attention.py::RingAttentionTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168445, 168446",,,,"module: ddp, module: rocm, module: sdpa","module: ddp, module: rocm, module: sdpa, rocm-skipped-tests, triaged",thread:issue:168445:issue:168446,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168445 TestModule: test/distributed/tensor/test_attention.py::RingAttentionTest, #168446 Test: RingAttentionTest.test_ring_attention_sdpa",13a7b7fd5cb549ae15d8d97698079b3bdf8ef2fdbe22cd79b7820bc4b319dfb3
issue-collapse:af4075fb2ffa9451,Review thread: 2 issues - issue #168434: TestModule: test/distributed/elastic/utils/distributed_test.py::DistributedUtilTest,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168434, 168435",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168434:issue:168435,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168434 TestModule: test/distributed/elastic/utils/distributed_test.py::DistributedUtilTest, #168435 Test: DistributedUtilTest.test_port_already_in_use_on_worker",9cd8a4b4d1ccc245d9551f9b269b5f9f101ba0625c5b0930fed3e0a81015fffc
issue-collapse:2fc730061d1c3c4f,Review thread: 2 issues - issue #168422: TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorFromLocalTensor,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168422, 168423",,,,"module: ddp, module: rocm","module: ddp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168422:issue:168423,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168422 TestModule: test/distributed/_shard/sharded_tensor/test_sharded_tensor.py::TestShardedTensorFromLocalTensor, #168423 Test: TestShardedTensorFromLocalTensor.test_init_from_local_tensor",edca80ab8cf967180e025f94d1abae5ffbcdd411ed9834dcc5ba420c7dec850f
issue-collapse:076940794071e8a2,Review thread: 2 issues - issue #168406: TestModule: test/distributed/_composable/test_composability/test_2d_composability.py::TestFullyShard2DTraining,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168406, 168407",,,,"module: fsdp, module: rocm","module: fsdp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168406:issue:168407,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168406 TestModule: test/distributed/_composable/test_composability/test_2d_composability.py::TestFullyShard2DTraining, #168407 Test: TestFullyShard2DTraining.test_train_parity_2d_mlp",090c26fe953daf9da0b5482f4ea0edfcb81bddbbd4e784ee06667e0488cbb537
issue-collapse:0630bfcf91a5668b,Review thread: 2 issues - issue #168404: TestModule: test/distributed/_composable/fsdp/test_fully_shard_training.py::TestFullyShardNDTraining,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168404, 168405",,,,"module: fsdp, module: rocm","module: fsdp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168404:issue:168405,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168404 TestModule: test/distributed/_composable/fsdp/test_fully_shard_training.py::TestFullyShardNDTraining, #168405 Test: TestFullyShardNDTraining.test_2d_mlp_with_nd_mesh",7c5f804b5be2713366d2c4d617cb1d2fe04e68e72c212901450bbc098d3bdb10
issue-collapse:e7a6b8e18ea11925,Review thread: 2 issues - issue #168402: TestModule: test/distributed/_composable/fsdp/test_fully_shard_overlap.py::TestFullyShardOverlap,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"168402, 168403",,,,"module: fsdp, module: rocm","module: fsdp, module: rocm, rocm-skipped-tests, triaged",thread:issue:168402:issue:168403,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#168402 TestModule: test/distributed/_composable/fsdp/test_fully_shard_overlap.py::TestFullyShardOverlap, #168403 Test: TestFullyShardOverlap.test_fully_shard_training_overlap",d68928ae302073a93435e477acb7ae8210329af89fcf75ca73211447a06e56a5
issue-collapse:e0682ecd9b1e03f0,Review thread: 2 issues - issue #165244: torch.tensor slicing with negative step raises ValueError,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"38271, 165244",,,,"module: numpy, module: python array api, module: python frontend","module: numpy, module: python array api, module: python frontend, triaged",thread:issue:165244:issue:38271,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#38271 PyTorch's flip returns a new tensor, but NumPy's flip returns a view, #165244 torch.tensor slicing with negative step raises ValueError",e9ecfafb74f0f222f09ff77f0bebf1fdfe7c6a24125a10944297f9df769d56cc
issue-collapse:910a3b09c0eb03d2,Review thread: 2 issues - issue #159906: New at::HostAllocator interface prevents using more than one allocator implementation for a device type,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"159906, 160420",,,,module: cudacachingallocator,"module: CUDACachingAllocator, triaged",thread:issue:159906:issue:160420,Handle this module: cudacachingallocator stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#159906 New at::HostAllocator interface prevents using more than one allocator implementation for a device type, #160420 CachingHostAllocator does not take into account NUMA affinity of GPU's when reusing freed blocks",12b871271c6c00f002de4f408153d695ebcd396f232214cd07f5a8a3d7f675b3
issue-collapse:4114c10b85f7eb73,Review thread: 2 issues - issue #159380: `torch.nested.nested_tensor_from_jagged` does not automatically compute `max_seqlen`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"159380, 168253",,,,module: nestedtensor,"module: nestedtensor, triaged",thread:issue:159380:issue:168253,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#159380 `torch.nested.nested_tensor_from_jagged` does not automatically compute `max_seqlen`, #168253 nestedtensor inconsistency in `torch.masked_select`",a3bb25568f8309f85d14951c3bf21285ab20d081bc0a304ce4c381ca30416de5
issue-collapse:139310ff9fb9e125,Review thread: 2 issues - issue #159353: torch.cond triggers shape mismatch with linear layer when nested in custom model,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"159353, 159852",,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",thread:issue:159353:issue:159852,Handle this module: higher order operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#159353 torch.cond triggers shape mismatch with linear layer when nested in custom model, #159852 `torch.cond()` behaves inconsistently when using symbolic predicate",dd40c954487774895bb454c929912eb249a7109db1b2aedbcef68108e22181e7
issue-collapse:19134579a62f95a3,Review thread: 2 issues - issue #159229: [Precompile] Assertion error when loading package in Nanogpt training with caching precompile fails after second load,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"159229, 159230",,,,,triaged,thread:issue:159229:issue:159230,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#159229 [Precompile] Assertion error when loading package in Nanogpt training with caching precompile fails after second load, #159230 [Precompile] Investigate recompile in NanoGPT training",4edb849451b77daea131e0f160b6472ac4d32fbeafee3f03026775e123667b81
issue-collapse:5da9f444db7c4c2d,Review thread: 2 issues - issue #15617: Feature request: transposed locally connected layer,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"499, 15617",,,,module: nn,"Stale, enhancement, feature, module: nn, proposal accepted, todo, triaged",thread:issue:15617:issue:499,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#499 Feature Request: Locally Connected Layer, #15617 Feature request: transposed locally connected layer",a06aa5dbe52b4c9c678e1f9f356bb9293584e873be012a9fe4dd83524bfd6ee0
issue-collapse:9dd40443827e52c9,Review thread: 2 issues - issue #155325: Excessively restrictive dependencies,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"97990, 155325",,,,"module: binaries, module: cuda, oncall: releng","module: binaries, module: cuda, oncall: releng, triaged",thread:issue:155325:issue:97990,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#97990 Relax version dependencies on CUDA pip wheels?, #155325 Excessively restrictive dependencies",c1707c1d189eee155d2faeb72b6542327db9c760967caf48b15366a0a07e1d7f
issue-collapse:034f4b0d3973ad7d,Review thread: 2 issues - issue #15457: Improve one_hot,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"15457, 86162",,,,module: nn,"function request, module: nn, triaged",thread:issue:15457:issue:86162,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#15457 Improve one_hot, #86162 torch.nn.functional.one_hot only works for int64",599ddd926dbab22a097150612409a3dd4d8f6acb4ea3f8b45442b4d93c5f6470
issue-collapse:7676b239c884dc0f,Review thread: 2 issues - issue #152032: RFC: The State of Custom CUDA extensions in PyTorch,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"152032, 163142",,,,"module: cpp-extensions, module: cuda","module: cpp-extensions, module: cuda, tracker, triaged",thread:issue:152032:issue:163142,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#152032 RFC: The State of Custom CUDA extensions in PyTorch, #163142 `compile_kernel`: fast inline compilation with nvrtc tracker",d801fce25977e47126f3c41c910ea44248d5bdef2649b31baf166677598a85c5
issue-collapse:22ca0b6040c91ed6,"Review thread: 2 issues - issue #150943: [RFC][TCPStore] advanced store operations (queues, pub/sub)",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"150943, 151960",,,,"module: c10d, oncall: distributed","feature, module: c10d, oncall: distributed, triaged",thread:issue:150943:issue:151960,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#150943 [RFC][TCPStore] advanced store operations (queues, pub/sub), #151960 [rfc][c10d] First Class Object-Oriented ProcessGroup and Store APIs",a7733276410230e38a43056ac36ad6af1dc8142a04742badf4c150dff42eab55
issue-collapse:4dd3e3da319ea3d2,"Review thread: 2 issues - issue #150277: Sparse tensor indexing not implemented, but partially supported by using index_select",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"150277, 152226",,,,module: sparse,"module: sparse, triaged",thread:issue:150277:issue:152226,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#150277 Sparse tensor indexing not implemented, but partially supported by using index_select, #152226 NotImplementedError: Could not run 'aten::index.Tensor' with arguments from the 'SparseCUDA' backend.",616fc1a91ca19140790bf67fbd76219ca81fe59b3035cf0a8600d514da03ed9e
issue-collapse:b2a11d03f6c55eaa,Review thread: 2 issues - issue #147044: Clean up backend_type_map from distributed_c10d,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"147044, 152285",,,,"module: build, module: c10d, oncall: distributed","better-engineering, module: build, module: c10d, oncall: distributed, triaged",thread:issue:147044:issue:152285,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#147044 Clean up backend_type_map from distributed_c10d, #152285 Error after successful build: No module named 'torch._C._distributed_c10d'",066c3ed5eeecbc7b666c40c85de356ea1c09cf6ee269226c2217c95e3bdb993c
issue-collapse:cfdf00628a92faa6,Review thread: 2 issues - issue #146898: [RFC] Test Cases Enabling for Accelerators,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"146898, 149734",,,,"module: inductor, module: testing, module: tests, oncall: pt2","feature, module: inductor, module: testing, module: tests, oncall: pt2, triaged",thread:issue:146898:issue:149734,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#146898 [RFC] Test Cases Enabling for Accelerators, #149734 [RFC] Multi-backend, multi-device test class instantiation for Inductor",6cb2eb5976b44c36e65a8390702c3b67392387a0d83abdd2cb1f9dd4d688ff62
issue-collapse:25d9d3ca1246fb75,Review thread: 2 issues - issue #145710: LBFGS-B with cuda implementation (and CPU one too),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"145710, 146201",,,,module: optimizer,"module: optimizer, triaged",thread:issue:145710:issue:146201,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#145710 LBFGS-B with cuda implementation (and CPU one too), #146201 L-BFGS-B support",eb86f1e7098aac1bd53de26fde60d338535d9c1d49700b1b3dfcbee7f438938c
issue-collapse:d738d1ef68546659,Review thread: 2 issues - issue #145511: Activation Checkpointing composability with split backward computation,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"145511, 155490",,,,"module: activation checkpointing, module: pipelining, oncall: distributed","module: activation checkpointing, module: pipelining, oncall: distributed, triaged",thread:issue:145511:issue:155490,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#145511 Activation Checkpointing composability with split backward computation, #155490 PP activation offloading",f1c6288a5220ec26d072ee05f1338960799d8d741f273d37d315a040603cc486
issue-collapse:2d66b9667fae6d17,Review thread: 2 issues - issue #145233: Not using set_num_threads results in very slow .all(),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"90760, 145233",,,,"module: cpu, module: linear algebra, module: performance","module: cpu, module: linear algebra, module: performance, triaged",thread:issue:145233:issue:90760,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#90760 Large slow down by not calling `torch.set_num_threads`, #145233 Not using set_num_threads results in very slow .all()",d1589f49ddafb035d9fdd47a88268742bd8eea1ad95c3c96efef215612ed5a02
issue-collapse:f0cda0ac64267bbc,"Review thread: 2 issues - issue #14489: Batch matmul with sparse matrix, dense vector",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"14489, 98675",,,,module: sparse,"module: sparse, todo, triaged",thread:issue:14489:issue:98675,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#14489 Batch matmul with sparse matrix, dense vector, #98675 torch.matmul with batched CSR matrix",064f1953c3509cae862b09323b5c91c6871c429e3a8d1da3e134c24b43ce28a4
issue-collapse:87a7fb969f13fbed,Review thread: 2 issues - issue #142703: We don't have test coverage for aarch64 (ARM64) bfloat16 feature (__ARM_FEATURE_BF16),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"142703, 143913",,,,"module: binaries, module: build, module: ci","enhancement, module: binaries, module: build, module: ci, triaged",thread:issue:142703:issue:143913,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#142703 We don't have test coverage for aarch64 (ARM64) bfloat16 feature (__ARM_FEATURE_BF16), #143913 Set up Mac builds with clang >= 17 even though Xcode only has at most clang 16",7876100ce7bf583f61838e5808ba2634d4f7549b5d9a841f392d4f10ec6a7e81
issue-collapse:f21286009912bdd5,Review thread: 2 issues - issue #141177: `CrossEntropyLoss` much slower with class probabilities vs. class indices target,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"141177, 141822",,,,"module: loss, module: nn, module: performance","module: loss, module: nn, module: performance, triaged",thread:issue:141177:issue:141822,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#141177 `CrossEntropyLoss` much slower with class probabilities vs. class indices target, #141822 F.cross_entropy unexpectedly slower than F.log_softmax + torch.gather",8c7bdbc1d70aaa34061957a942f61c3b5ceab7353b0853e452f3ddc43426f4aa
issue-collapse:aec9df9baf6ee137,Review thread: 2 issues - issue #140787: Apple Silicon Unified Memory usage,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"140787, 145693",,,,"module: mps, module: rocm","enhancement, feature, module: mps, module: rocm, topic: new features, triaged",thread:issue:140787:issue:145693,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#140787 Apple Silicon Unified Memory usage, #145693 AMD MI300A Unified Memory Support",9fa680ff13c827ea65a62a18e02523094ff148dc659300b5ddd67d0267a5d8a4
issue-collapse:393ff1f2e4a16350,Review thread: 2 issues - issue #140563: NCCL hangs cause timeout,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"140563, 140879",,,,oncall: distributed,"oncall: distributed, triaged",thread:issue:140563:issue:140879,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#140563 NCCL hangs cause timeout, #140879 fr_trace usability issues",15f0ba5b05d6c104e18165b7326566320b429de7616bf165ee46c9eaafcfccac
issue-collapse:e3a3db28e1cad968,Review thread: 2 issues - issue #140222: inconsistency in ```torch.nn.functional.log_softmax``` on CPU and GPU,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"140222, 140252",,,,"module: bfloat16, module: cuda, module: numerical-stability, module: special","module: bfloat16, module: cuda, module: numerical-stability, module: special, triaged",thread:issue:140222:issue:140252,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#140222 inconsistency in ```torch.nn.functional.log_softmax``` on CPU and GPU, #140252 inconsistency in ```torch.special.logit``` on CPU and GPU",0e1f966525e7c383127d62e2cc6b308c89fe973feef241de2321cdef0c362f95
issue-collapse:adcd88b5127ca875,Review thread: 2 issues - issue #139643: [triton 3.2] test_convolution2,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"139643, 141079",,,,"module: inductor, oncall: pt2","module: inductor, oncall: pt2, triaged, upstream triton",thread:issue:139643:issue:141079,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#139643 [triton 3.2] test_convolution2, #141079 [triton 3.2] test_convolution_as_mm failure on A100",8054af1b3716ce110911c1cbcd5148d2ed445526e21b8b9333a110d46b90f1a9
issue-collapse:1deba6649ff2c68c,Review thread: 2 issues - issue #138800: Batching rule not defined for `aten::_make_dual`.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"138800, 160508",,,,"module: dynamo, module: functorch, module: vmap, oncall: pt2","actionable, inference mode, module: dynamo, module: functorch, module: vmap, oncall: pt2, triaged",thread:issue:138800:issue:160508,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#138800 Batching rule not defined for `aten::_make_dual`., #160508 `torch.compile` + `torch.vmap` + `torch.func.jacfwd` results in an error",e1f451f4e8c65327f579a6da41178b8a3ebb7574b873e9719b1f163f758d9ccf
issue-collapse:4d109e3bc8044d3e,Review thread: 2 issues - issue #138786: Support nn.Module arguments for the function to torch.compiler.allow_in_graph,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"138786, 140437",,,,"module: dynamo, oncall: pt2","module: dynamo, oncall: pt2, triaged",thread:issue:138786:issue:140437,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#138786 Support nn.Module arguments for the function to torch.compiler.allow_in_graph, #140437 Let torch.compiler.allow_in_graph work in more situations",6d6779e77b7d26a6d580a7ceecfac3e0e6c5154e55d65727f185d7dbb4c3f8ee
issue-collapse:0d18ae61864b54e7,Review thread: 2 issues - issue #138396: `out=` meta device support.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"138396, 138399",,,,"module: meta tensors, module: type promotion","module: meta tensors, module: type promotion, triaged",thread:issue:138396:issue:138399,Handle this module: meta tensors stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#138396 `out=` meta device support., #138399 `dtype` promotion of `out=` functions on meta inputs not consistent.",4cb825027e2dc0c14935554bc8b0cca44a2845242ed5a634ab6b96565367a6ae
issue-collapse:11e577c3c9036e06,"Review thread: 2 issues - issue #137482: `torch.ops.aten.divide_` leads to ""RuntimeError: result type Float can't be cast to the desired output type Int""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"42246, 137482",,,,"module: core aten, module: type promotion","actionable, low priority, module: core aten, module: type promotion, triaged",thread:issue:137482:issue:42246,Handle this module: core aten stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#42246 `torch.tensor([1, 2], dtype=torch.int).fmod(torch.tensor(0, dtype=torch.float))` leads to ""RuntimeError: result type Float can't be cast to the desired output type Int"", #137482 `torch.ops.aten.divide_` leads to ""RuntimeError: result type Float can't be cast to the desired output type Int""",f7ab834249dee758b0f526c18e83e8c7da399f02db275b3b1146bf1422dc5016
issue-collapse:d1269b07c4151a30,"Review thread: 2 issues - issue #137367: `lr`, `momentum`, `weight_decay` and `dampening` parameter of `optim.SGD()` work with `bool` values",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"137367, 137373",,,,module: optimizer,"module: optimizer, triaged",thread:issue:137367:issue:137373,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#137367 `lr`, `momentum`, `weight_decay` and `dampening` parameter of `optim.SGD()` work with `bool` values, #137373 `dampening`, `maximize`, `foreach`, `differentiable` and `fused` parameter of `optim.SGD()` work with `str` values",d3ef185343ca00395ed7046b9e89faff16edb2da233a6647c80474fe4e01304c
issue-collapse:c44b46cb4d452ab8,Review thread: 2 issues - issue #13447: Support gathering nested lists in DataParallel,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"13447, 62466",,,,module: data parallel,"enhancement, module: data parallel, triaged",thread:issue:13447:issue:62466,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13447 Support gathering nested lists in DataParallel, #62466 Support gathering string in DataParallel",3b32607607e87c4ffc6cc26355b72203f3b8747da7828ac3add58ddf770f45e4
issue-collapse:ab38305d12c7136f,Review thread: 2 issues - issue #133823: Issues compiling `torch` with `mkl`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73008, 133823",,,,"module: binaries, module: build, module: cpp, module: intel, module: mkl","module: binaries, module: build, module: cpp, module: intel, module: mkl, triaged",thread:issue:133823:issue:73008,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73008 Followup requires for MKL link issue / cannot find -lmkl_core, #133823 Issues compiling `torch` with `mkl`",26c0dcd38fa830fdb32a3e9189b8fb9b3f6f8913d5a340d8c38dd7478f64c5cb
issue-collapse:f560b7d97a625a8e,Review thread: 2 issues - issue #13218: Support for integer interpolation (torch.nn.functional.interpolate),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"5580, 13218",,,,module: nn,"function request, module: nn, todo, triaged",thread:issue:13218:issue:5580,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#5580 [feature request] F.interpolate to support integral data types: bool, int8, int32, int16, int64 ||| support uint8 on CUDA, #13218 Support for integer interpolation (torch.nn.functional.interpolate)",6da552ed3df94a4233c411e66f1d94b6bdf8a1aade6fa4d7e7e99e745b95b362
issue-collapse:ff6b517b0b0121aa,Review thread: 2 issues - issue #13188: cudnn explicit paths and GCC multilib suffixes prevents detection of good cudnn headers,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"13188, 16708",,,,module: build,"module: build, triaged",thread:issue:13188:issue:16708,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13188 cudnn explicit paths and GCC multilib suffixes prevents detection of good cudnn headers, #16708 Better include path when compiling mkldnn",42a7e1aa335a35c30a22897faec086c4b207a40240902304b71ae7f5c467e61e
issue-collapse:c12535da8b5f8db3,"Review thread: 2 issues - issue #131650: Pytorch 2.4, Cuda 12.4: RuntimeError: device >= 0 && device < num_gpus INTERNAL ASSERT FAILED at ""/opt/conda/conda-bld/pytorch_1720538439675/work/aten/src/ATen/cuda/CUDAContext.cpp"":49, please report a bug to PyTorch. device=^A, num_gpus=^A",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"131650, 131667",,,,module: cuda,"module: cuda, triaged",thread:issue:131650:issue:131667,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#131650 Pytorch 2.4, Cuda 12.4: RuntimeError: device >= 0 && device < num_gpus INTERNAL ASSERT FAILED at ""/opt/conda/conda-bld/pytorch_1720538439675/work/aten/src/ATen/cuda/CUDAContext.cpp"":49, please report a bug to PyTorch. device=^A, num_gpus=^A, #131667 [RFE] Improve the error reporting if MIG is incorrectly used with NVML",523bd488c9dc8971d218dc2e984f76633c953c3819968798aa4059c7843821ed
issue-collapse:b3111390f40361e3,"Review thread: 2 issues - issue #13130: arm64 port for PyTorch, libtorch",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"13130, 166512",,,,"module: arm, module: ci, module: power","enhancement, module: POWER, module: arm, module: ci, triaged",thread:issue:13130:issue:166512,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#13130 arm64 port for PyTorch, libtorch, #166512 Enable GitHub Actions app to support ppc64le CI jobs using IBM-hosted runners",acec04b7c5e5477c1377c87eba6fbad13d9c81e8c3bb3180d96b587948f1def5
issue-collapse:c384aa0e1a92f379,Review thread: 2 issues - issue #13023: Python dataloader Improvements,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"7359, 13023",,,,module: dataloader,"feature, module: dataloader, triaged",thread:issue:13023:issue:7359,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#7359 [feature request] [PyTorch] Dynamic Samplers., #13023 Python dataloader Improvements",b229f3cfdb85802ad1b811a6126ad35447bec1e0af89a79ae6902f39b66d5274
issue-collapse:bea26a8e3e3fa5b0,Review thread: 2 issues - issue #130217: `unique()` is not intuitively understandable at all.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"130217, 148743",,,,module: python frontend,"module: python frontend, triaged",thread:issue:130217:issue:148743,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130217 `unique()` is not intuitively understandable at all., #148743 torch.unique(t, dim=1, return_counts=True) does not work",fb66d07f57b34398f0d7a9c685e558c7718ea607ad682cc3906ba1df0d6ccce5
issue-collapse:7cf2c61e4583aee4,Review thread: 2 issues - issue #130055: [pipelining] Non-first stage only takes tensors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"130055, 130058",,,,module: pipelining,"module: pipelining, triaged",thread:issue:130055:issue:130058,Handle this module: pipelining stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130055 [pipelining] Non-first stage only takes tensors, #130058 [pipelining] Expose a transform function to inject additional metadata into model forward",0749a745a2e5aea99924483d544ed202b59b3a3595fcd9c0ec852d9133eb3887
issue-collapse:09d3fd60dd236c3c,Review thread: 2 issues - issue #129657: Implementation of a numerically stable log(1 - softmax) function in PyTorch,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"129657, 135359",,,,"module: bc-breaking, module: edge cases, module: loss, module: nans and infs, module: nn, module: numerical-stability, module: python frontend","module: NaNs and Infs, module: bc-breaking, module: edge cases, module: loss, module: nn, module: numerical-stability, module: python frontend, triaged",thread:issue:129657:issue:135359,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#129657 Implementation of a numerically stable log(1 - softmax) function in PyTorch, #135359 binary_cross_entropy_with_logits outputs NaN on -inf, which is inconsistent with documentation",cc624f95aa14d7c405c3f20ddf524ec1a3315ab60d1edeff814c843bd4b57c79
issue-collapse:cf7fd7e09cdd5c33,Review thread: 2 issues - issue #129442: Volatile GPU-Util keeps 100% when using torch.distributed.isend/irecv,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"129442, 129749",,,,oncall: distributed,"oncall: distributed, triaged",thread:issue:129442:issue:129749,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#129442 Volatile GPU-Util keeps 100% when using torch.distributed.isend/irecv, #129749 nccl deadlock",ec2b905cce7c1cd191d25051395d04ad22d4c2c05f3ce29411c6d3a3e9b65a14
issue-collapse:a4d16716c9e5b591,Review thread: 2 issues - issue #129140: [RFC][C10D] Avoid creating new nccl communicator for each P2P pair,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"129140, 152220",,,,"module: nccl, oncall: distributed","module: nccl, oncall: distributed, triaged",thread:issue:129140:issue:152220,Handle this module: nccl stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#129140 [RFC][C10D] Avoid creating new nccl communicator for each P2P pair, #152220 [C10D] Allow NCCL single P2P ops to use parent/collective communicator",853c109014db456e2d1221e11b38d7894565f4331bd1cff7a6699ff69d7c7c68
issue-collapse:40819b9358f32aaf,Review thread: 2 issues - issue #128641: Backward is not supported by `MaskedTensor`'s `torch.maximum`,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"128641, 129272",,,,"module: autograd, module: masked operators","module: autograd, module: masked operators, triaged",thread:issue:128641:issue:129272,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#128641 Backward is not supported by `MaskedTensor`'s `torch.maximum`, #129272 RecursionError for MaskedTensor.where",638bf9821fed8bda81e253aa6d12ef4d7e7247eb0623b157f683211ca43c63cd
issue-collapse:704d9197009cf476,Review thread: 2 issues - issue #128071: [user empathy day 2][based] torch.compile issues,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"128071, 143670",,,,"module: compile-time, module: dynamo, oncall: pt2","dynamo-triage-june2024, dynamo-user-empathy-day, empathy-day, module: compile-time, module: dynamo, oncall: pt2, triaged",thread:issue:128071:issue:143670,Handle this module: compile-time stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#128071 [user empathy day 2][based] torch.compile issues, #143670 Potential rooms for fewer recompilations by introducing higher-level guards",579baf3d1fad0e61fe72c746abe9a1697c808cdd5474f2cd7183144fd60fe396
issue-collapse:391efa3023ff2da1,Review thread: 2 issues - issue #127164: randn generates different output for 4x4 tensor size sliced to match shape of direct 2x4 or 4x2 and compare output,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"79703, 127164",,,,module: random,"module: random, triaged",thread:issue:127164:issue:79703,Handle this module: random stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#79703 Float and double tensors randomly initialized with the same seed get different values for size >= 16, #127164 randn generates different output for 4x4 tensor size sliced to match shape of direct 2x4 or 4x2 and compare output",6d7ed24f5d25ede0c4c881f01f923b7eb1d709f3d9dff562d54a4a816036e097
issue-collapse:9d00929d039df024,Review thread: 2 issues - issue #127075: What is the processing principle when the complex64 input tensor contains nan or inf for addition?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"82510, 127075",,,,"module: complex, module: nans and infs","module: NaNs and Infs, module: complex, triaged",thread:issue:127075:issue:82510,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#82510 Complex addition result in NaN when it shouldn't, #127075 What is the processing principle when the complex64 input tensor contains nan or inf for addition?",23d92f6cb95b11828d32223b2651e7f85a529bd00aa87af102e327dc8574463d
issue-collapse:73d65fb5f52db38b,Review thread: 2 issues - issue #126921: [ONNX] view(dtype=dtype) is not supported by both onnx.export and onnx.dynamo_export,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"126921, 139429",,,,module: onnx,"actionable, module: onnx, triaged",thread:issue:126921:issue:139429,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#126921 [ONNX] view(dtype=dtype) is not supported by both onnx.export and onnx.dynamo_export, #139429 Add 'no_type_conversion' flag to register_buffer()",6bcd78f612cc19484a248d221416a0f508a4dcb5ad06b3502d47e0b7c0ec5794
issue-collapse:18b7d95f4ceb36b7,Review thread: 2 issues - issue #126037: A scalar with `input` argument of `pow()` doesn't work,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"126037, 126039",,,,module: python frontend,"has workaround, module: python frontend, triaged",thread:issue:126037:issue:126039,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#126037 A scalar with `input` argument of `pow()` doesn't work, #126039 A scalar with `input` argument of `float_power()` doesn't work",ffe43628705bd600569bbcc627057d4f05667de8a0d0f482ac5b9297f4ba443c
issue-collapse:c87a4d7946577cdd,Review thread: 2 issues - issue #125837: Support for PyTorch on iOS Following PEP 730,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"125837, 140090",,,,"module: android, module: binaries, module: ios","feature, module: android, module: binaries, module: ios, triaged",thread:issue:125837:issue:140090,Handle this module: android stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#125837 Support for PyTorch on iOS Following PEP 730, #140090 Add Android Support (Python Wheel)",e62b64acdb6f171a8dd0926cbcf6884a559acfe8c6488e1f2786a2fea370452a
issue-collapse:8ea77c362482588a,Review thread: 2 issues - issue #1249: Dice Loss PR,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"1249, 41142",,,,module: loss,"Stale, enhancement, feature, module: loss, triaged",thread:issue:1249:issue:41142,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#1249 Dice Loss PR, #41142 Support for More Loss Functions",cf52b0ab8accdc82d7de475bc7449242c8506aeb2c228f9a1f9351c508a140df
issue-collapse:935c215e8c0f0902,Review thread: 2 issues - issue #12484: CuDNN convolution on some CUDA devices will not preserve NaN weights (upstream bug),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"12484, 18536",,,,"module: cudnn, module: dependency bug, module: logging","low priority, module: cudnn, module: dependency bug, module: logging, triaged",thread:issue:12484:issue:18536,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#12484 CuDNN convolution on some CUDA devices will not preserve NaN weights (upstream bug), #18536 Make it easier to figure out what CuDNN convolution algorithm we actually chose",9b7d2cfca6dadacc54737beacdd105aae915eebc6b8dd0030f1866c5d9e85e09
issue-collapse:aea1957873e2566d,Review thread: 2 issues - issue #124788: torch.nn.checkpoint.checkpoint ignores default device in backward() call,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"84864, 124788",,,,"module: __torch_dispatch__, module: activation checkpointing","actionable, module: __torch_dispatch__, module: activation checkpointing, tensor subclass, triaged",thread:issue:124788:issue:84864,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#84864 torch.utils.checkpoint (with use_reentrant=False) doesn't work with all PyTorch features that set TLS, #124788 torch.nn.checkpoint.checkpoint ignores default device in backward() call",f19876e8d384f019f4eace66dd10afd4a76ad915f03db0c5317e5d990602be1d
issue-collapse:0910e4bd0e728196,"Review thread: 2 issues - issue #12449: Could not find a package configuration file provided by ""Torch"" with any of the following names:",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"12449, 37766",,,,"module: cpp-extensions, module: docs","module: cpp-extensions, module: docs, triaged",thread:issue:12449:issue:37766,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#12449 Could not find a package configuration file provided by ""Torch"" with any of the following names:, #37766 CMake Documentation Issue",dca74542313435fd810fb2e5ad0946a73e08475bd767507e89ae88bbbc2266b2
issue-collapse:f711dd4c7064bfa6,Review thread: 2 issues - issue #124262: CUBLAS_STATUS_EXECUTION_FAILED when calling cublasGemmEx,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"94294, 124262",,,,"module: cublas, module: cuda","module: cublas, module: cuda, triaged",thread:issue:124262:issue:94294,Handle this module: cublas stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#94294 CUBLAS_STATUS_NOT_SUPPORTED when calling cublasDgemv, #124262 CUBLAS_STATUS_EXECUTION_FAILED when calling cublasGemmEx",48d05c86f60f88c55611d2cd3258a2195ccac3d4a4477dcd8343dd20565ef8b2
issue-collapse:59dd5acae72770a3,Review thread: 2 issues - issue #123835: Package manager install on Nvidia Grace Hopper does not make cuda available,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"123835, 134117",,,,"module: arm, module: binaries, module: cuda","module: arm, module: binaries, module: cuda, triaged",thread:issue:123835:issue:134117,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#123835 Package manager install on Nvidia Grace Hopper does not make cuda available, #134117 Add aarch64 and aarch64 cuda version to conda",8410f7c4d04a4890c661ca04f25f4a2b1025949df6fa7c61cc7fd45955cb259b
issue-collapse:ae28b58f3cb047bf,Review thread: 2 issues - issue #122641: [vulkan] Vulkan backend fails creating tensor on aarch64 (Raspberry Pi 5),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"72775, 122641",,,,module: vulkan,"module: vulkan, triaged",thread:issue:122641:issue:72775,Handle this module: vulkan stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#72775 [vulkan] Vulkan backend fails creating tensor on x86_64 Linux, #122641 [vulkan] Vulkan backend fails creating tensor on aarch64 (Raspberry Pi 5)",3256075894f2e380c07bf9f68bb218c05e560a40f6f3c62a1ffd861fb00cbea7
issue-collapse:9b2573ff4d677606,Review thread: 2 issues - issue #121894: TORCH_USE_CUDA_DSA compiler flag not working,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"121894, 121969",,,,"module: cuda, module: docs","module: cuda, module: docs, triaged",thread:issue:121894:issue:121969,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#121894 TORCH_USE_CUDA_DSA compiler flag not working, #121969 which Pythorch suports ""TORCH_USE_CUDA_DSA=1"" from the shell environment?",036941673b72b36f9a3fb3202044a40178943b1bf7146cbe0136df9b28a57306
issue-collapse:4d7432b81e19ea08,"Review thread: 2 issues - issue #12117: Error: Internal Compiler Error (codegen): ""there was an error in verifying the lgenfe output!""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"12117, 15630",,,,"module: build, module: cuda","caffe2, internals, module: build, module: cuda, triaged",thread:issue:12117:issue:15630,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#12117 Error: Internal Compiler Error (codegen): ""there was an error in verifying the lgenfe output!"", #15630 [Caffe2] Internal compiler error for CUDA.",d73266f8267b3c48fa021ac83bb290d2158cd07c7bd011c8bb770f34e9b34a6e
issue-collapse:6dc02afa6775ed90,Review thread: 2 issues - issue #120930: make_graphed_callables doesn't turn off autocast for backward,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"71631, 120930",,,,"module: amp (automated mixed precision), module: cuda, module: cuda graphs","module: amp (automated mixed precision), module: cuda, module: cuda graphs, triaged",thread:issue:120930:issue:71631,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#71631 Wrapping make_graphed_callables with autocast issue, #120930 make_graphed_callables doesn't turn off autocast for backward",a53f4395d1c8f276397a32e5e6350202234379dd57ea0278c0544c4b342e8a3c
issue-collapse:72a0920bbfa2cc29,Review thread: 2 issues - issue #11982: [feature request] Publish wheels with debug symbols,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"11982, 56397",,,,module: binaries,"module: binaries, triaged",thread:issue:11982:issue:56397,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#11982 [feature request] Publish wheels with debug symbols, #56397 Symbolicate crashes from releases",d1b78d8718182cfbeb0f6a3edbb64676b336095725968d3228e6b5618b41e0b1
issue-collapse:80d7b9f203d6f378,Review thread: 2 issues - issue #118736: [typing] `torch.Tensor` subclass types (eg `nn.Parameter`) broken by 2.2 release,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"87371, 118736",,,,"module: python frontend, module: typing","actionable, module: python frontend, module: typing, tensor subclass, triaged",thread:issue:118736:issue:87371,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#87371 Better type annotations for `torch.Tensor` subclasses, #118736 [typing] `torch.Tensor` subclass types (eg `nn.Parameter`) broken by 2.2 release",b46661ea6baf935832dc63e6b727bdd7dfe2af77ccae6d913c3cc09c65ea9022
issue-collapse:10bffefb410996bc,Review thread: 2 issues - issue #118093: torch.where only accepts boolean tensor as condition deviating from np.where where you can also use binary input condition of dtype float,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"56300, 118093",,,,"module: python array api, module: type promotion","module: python array api, module: type promotion, triaged",thread:issue:118093:issue:56300,Handle this module: python array api stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#56300 torch.where with input Tensor and other Scalar raises type mismatch error, #118093 torch.where only accepts boolean tensor as condition deviating from np.where where you can also use binary input condition of dtype float",232c6c1011e9809914c6864cbc04c261f4fdbe7003f3beead8e51b18460fc73f
issue-collapse:c863fbaafd6a30f5,Review thread: 2 issues - issue #118018: Enable all optimizers (except SparseAdam + LBFGS) to be capturable,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"118018, 118715",,,,"module: cuda graphs, module: optimizer, oncall: pt2","actionable, feature, module: cuda graphs, module: optimizer, oncall: pt2, triaged",thread:issue:118018:issue:118715,Handle this module: cuda graphs stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#118018 Enable all optimizers (except SparseAdam + LBFGS) to be capturable, #118715 Add capturable Adagrad implementation",1882026c1704e327630ddc10b98fff4b6abb929ea451074e2971bf14d10cc518
issue-collapse:0b8f0e2d52ee08bc,Review thread: 2 issues - issue #117748: [NCCL 2.19.3] Performance reduction when using NVLSTree algorithm when running nccl-tests all-reduce,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"117748, 118421",,,,"module: nccl, oncall: distributed","module: nccl, oncall: distributed, triaged",thread:issue:117748:issue:118421,Handle this module: nccl stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#117748 [NCCL 2.19.3] Performance reduction when using NVLSTree algorithm when running nccl-tests all-reduce, #118421 Increase default bucket_cap_mb value to from 25MB to a more optimal value",41e4b7ac963f773284454a74be4905d587c7b84b109e69cfe7e5ae8b70082900
issue-collapse:f222741679bafb05,Review thread: 2 issues - issue #116626: `test_modules.py` does not test for half precision or complex types,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"116626, 122287",,,,"module: ci, module: nn, module: testing","module: ci, module: nn, module: testing, triaged",thread:issue:116626:issue:122287,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#116626 `test_modules.py` does not test for half precision or complex types, #122287 Tolerance setting for test_modules.py -- TestModuleCUDA.test_cpu_cuda_parity",0572a6c546784b6813540b630f181cf34e9aeace5f091c72f42a86643701117f
issue-collapse:22ff6d9cfe969d50,Review thread: 2 issues - issue #116461: Numpy like Masked/where operations,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"69359, 116461",,,,"module: masked operators, module: sparse","enhancement, module: masked operators, module: sparse, triaged",thread:issue:116461:issue:69359,Handle this module: masked operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#69359 [feature request] More masked reductions: amin/amax, argmin/argmax, quantile, mean/var/std/std_mean/var_mean, #116461 Numpy like Masked/where operations",6afb7e33fb2d113c973cf32b93ce69ae0bba3db518f98b0a351aebcdcec1832b
issue-collapse:e8322cc09fb62535,Review thread: 2 issues - issue #115434: torch.kthvalue should take a tensor of `k` values.,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"75599, 115434",,,,"module: performance, module: sorting and selection","module: performance, module: sorting and selection, triaged",thread:issue:115434:issue:75599,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#75599 kthvalue 20x slower than sort, #115434 torch.kthvalue should take a tensor of `k` values.",20e1c38f0473628cf8d3a01019d692c456ad961b65ceb5bf5bfa0ef9ad56468d
issue-collapse:230aceb36ae0bd24,Review thread: 2 issues - issue #11514: [feature request] - Allow sequences lengths to be 0 in PackSequence,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"11514, 50192",,,,"module: nn, module: rnn","enhancement, module: nn, module: rnn, triaged",thread:issue:11514:issue:50192,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#11514 [feature request] - Allow sequences lengths to be 0 in PackSequence, #50192 GRU and LSTM fail for seq_len = 0",1966551d246de587756ac1e82a0e909503bb441ad225f86d256060851289fd78
issue-collapse:bd4f972d6c359037,Review thread: 2 issues - issue #114723: [RFC] Intel GPU Upstreaming,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"114723, 114835",,,,"module: inductor, module: intel, oncall: pt2","feature, intel, module: inductor, module: intel, oncall: pt2, triaged",thread:issue:114723:issue:114835,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#114723 [RFC] Intel GPU Upstreaming, #114835 [RFC] Building system for SYCL and limited number of SYCL kernels for ATen fallbacks of TorchInductor",d6446dbbfe9c3e80d402d3785f34246adbb0d87b05c17d65dd9dca49a89056d2
issue-collapse:de7cd50a3026e458,Review thread: 2 issues - issue #114569: inconsistency on torch.clamp,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"114569, 162235",,,,"module: cuda, module: edge cases, module: nans and infs","module: NaNs and Infs, module: cuda, module: edge cases, triaged",thread:issue:114569:issue:162235,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#114569 inconsistency on torch.clamp, #162235 Inconsistent handling of `-0.0` between CPU and CUDA for many operators",99c470f0fe7fd6a7afa6fc065ba37bc757f06047e319e825227ec03e7943e939
issue-collapse:93a4929e1e551241,Review thread: 2 issues - issue #111384: [quant][pt2] Default batchnorm aten op has poor numerics during QAT,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"111384, 114859",,,,"module: pt2 accuracy, oncall: pt2, oncall: quantization","low priority, module: pt2 accuracy, oncall: pt2, oncall: quantization, triaged",thread:issue:111384:issue:114859,Handle this module: pt2 accuracy stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#111384 [quant][pt2] Default batchnorm aten op has poor numerics during QAT, #114859 PT2 QAT flow fails to get reasonable accuracy with mobilenet_v3_large",db60621a0d7e0467027dda18a9b8b8ffc6470d9146208f5d358259daf477b82d
issue-collapse:5dc7dcadc9df51ca,Review thread: 2 issues - issue #110737: pytorch consuming all cpu cores 100% on ARM,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"110737, 144633",,,,"module: arm, module: binaries, module: openmp, module: performance","module: arm, module: binaries, module: openmp, module: performance, triaged",thread:issue:110737:issue:144633,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#110737 pytorch consuming all cpu cores 100% on ARM, #144633 144x less efficient CPU usage when training NN past a certain width",2e2edb94553a6ad0f9ba48157915e67385dd733f0203e79f8fc6445ac86c14ae
issue-collapse:4909ff0d81b1246b,Review thread: 2 issues - issue #108934: PPC64le: vsx_helpers.h errors,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"108934, 108984",,,,module: build,"module: build, topic: build, triaged",thread:issue:108934:issue:108984,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#108934 PPC64le: vsx_helpers.h errors, #108984 PPC64le: GCC 11.2.1 Linker Error in bin/torch_shm_manager",65d5463c803dd0f70cc4c5baf6752a63d027383382058aeb9c90caa6b9c91460
issue-collapse:971d0522575a52ae,"Review thread: 2 issues - issue #108739: DDP Elastic ""master_addr"" resolution error in environment variables.",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"73656, 108739",,,,"module: c10d, module: elastic, oncall: distributed","module: c10d, module: elastic, oncall: distributed, triaged",thread:issue:108739:issue:73656,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#73656 torchrun: Hostname/endpoint mismatch not handled, #108739 DDP Elastic ""master_addr"" resolution error in environment variables.",161b084a663674594f0bf3d66bd47ff62718263f6d71b823a15f86abd8f75710
issue-collapse:89a673875a9b133d,Review thread: 2 issues - issue #107896: [feature request] [ux proposal] Min-max linear normalization to be supported in F.normalize (or in a new function),related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"107896, 128785",,,,"module: nn, module: python frontend","module: nn, module: python frontend, topic: new features, triaged",thread:issue:107896:issue:128785,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#107896 [feature request] [ux proposal] Min-max linear normalization to be supported in F.normalize (or in a new function), #128785 Any plans for a ""torch.minmax"" (min-max normalization) function?",4a6162262ef14afcd42c5301650fb508d531d43725d8dbb7caa8dec062a5b929
issue-collapse:2ed3eb44d72945f2,Review thread: 2 issues - issue #106951: stride of gradient is not same as the corresponding tensor,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"106951, 110758",,,,"module: autograd, module: optimizer","actionable, module: autograd, module: optimizer, triaged",thread:issue:106951:issue:110758,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#106951 stride of gradient is not same as the corresponding tensor, #110758 Fused Adamw RuntimeError: params, grads, exp_avgs, and exp_avg_sqs must have same dtype, device, and layout",bcd1f4a473df5bbc7e811eeb42c3426bd270d5ea4ff2d4c69fa51f03febd4592
issue-collapse:ffcdf615ab411c8a,"Review thread: 2 issues - issue #106455: vmap, jacrev, jacfwd, hessian, etc., in libTorch",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"40208, 106455",,,,"module: autograd, module: cpp, module: functorch, module: vmap","module: autograd, module: cpp, module: functorch, module: vmap, triaged",thread:issue:106455:issue:40208,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40208 C++ API for torch.autograd.functional.jacobian, #106455 vmap, jacrev, jacfwd, hessian, etc., in libTorch",ef56c929c64db176db22a61edec888b42dba63b6a4475901c1da81372950ccae
issue-collapse:6e340f5051a4e1d2,Review thread: 2 issues - issue #106206: Build failure due to C++ version mismatch,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"106206, 106297",,,,module: build,"module: build, triaged",thread:issue:106206:issue:106297,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#106206 Build failure due to C++ version mismatch, #106297 Can't build with non-static protobuf",66468d9be4342daaaffe2dc7378895a938055cab47cafdc7a682caafef8c164c
issue-collapse:b198a7958feb5e44,Review thread: 2 issues - issue #103820: Upgrading SpGEMM algorithm to resolve Cusparse SpGEMM insufficient resources problem,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"103820, 150285",,,,"module: cuda, module: sparse","module: cuda, module: sparse, triaged",thread:issue:103820:issue:150285,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#103820 Upgrading SpGEMM algorithm to resolve Cusparse SpGEMM insufficient resources problem, #150285 CuSparse doesn't work on sparse tensor",b06391e48b6954fac739c2bbbfe746cf8b38fdf52c941e7d1c33c64abd248e1a
issue-collapse:c1c8691b1835906c,Review thread: 2 issues - issue #103375: Inplace binary ops on tensor subclasses can cause mypy error,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"103375, 103393",,,,module: typing,"module: typing, triaged",thread:issue:103375:issue:103393,Handle this module: typing stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#103375 Inplace binary ops on tensor subclasses can cause mypy error, #103393 Typing missing on arithmetic ops on `Tensor`",90f162a66566adba331611b9821ba85735c848cd135f425b312eb767bf0efad5
issue-collapse:6e6e44f55d6c9325,Review thread: 2 issues - issue #102673: Fix dynamo-related debug Python 3.11 failures,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"99176, 102673",,,,"module: devx, module: python version","module: devx, module: python version, release notes: dynamo, triaged",thread:issue:102673:issue:99176,Handle this module: devx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#99176 Add Debug builds for python with pydebug, #102673 Fix dynamo-related debug Python 3.11 failures",0ce708580e7febceffe6dbcf4245bcad45c2cfda483ab1db93f1f5ca3f827a26
issue-collapse:ded3ad7ff3e15ee2,Review thread: 2 issues - issue #101359: RuntimeError in Scaled Dot Product Attention Tutorial Code,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"97068, 101359",,,,"module: cpu, module: half","feature, module: cpu, module: half, triaged",thread:issue:101359:issue:97068,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#97068 [RFC] CPU float16 performance optimization on eager mode., #101359 RuntimeError in Scaled Dot Product Attention Tutorial Code",cb9cbd6f09bed2a874265514a5ac1b1f5790ae98e0340c54bc444fb66883b086
issue-collapse:ef0260c83e0f9ed0,"Review thread: 2 issues - issue #10119: [Caffe2] MNIST Tutorial LMDB Error ""Cannot open db""",related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"7807, 10119",,,,,caffe2,thread:issue:10119:issue:7807,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#7807 LMDB read error for Mnist, #10119 [Caffe2] MNIST Tutorial LMDB Error ""Cannot open db""",ad317a8ec0ed5b336a69b37c23767d7f1fcaf885a8d0b3abe1095c65586fa897
issue-collapse:5b4ef0faf1411ccc,Review thread: 2 issues - issue #101159: Inconsistency between GPU memory usage in torch.cuda.memory_summary and nvidia-smi,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"37250, 101159",,,,"module: cuda, module: memory usage","module: cuda, module: memory usage, triaged",thread:issue:101159:issue:37250,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,2,Represents 2 open issues; raw issue rows remain available for audit.,"#37250 Inconsistency between GPU memory usage in torch.cuda.memory_summary and nvidia-smi, #101159 Inconsistency between GPU memory usage in torch.cuda.memory_summary and nvidia-smi",445fc47c4882638a066f7915f2ac298e6bded42250a6380d8870ac9f4363eb96
issue-collapse:76ecac2f6f16a965,Review thread: 2 issues - issue #101110: Tensorboard graph tracing with torch fx API,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"37334, 101110",,,,"module: fx, module: tensorboard, oncall: visualization","feature, module: fx, module: tensorboard, oncall: visualization, triaged",thread:issue:101110:issue:37334,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#37334 Why do we not use TorchScript to build graph for tensorboard, #101110 Tensorboard graph tracing with torch fx API",47f9eee3f41c9c30202f581fad15b0c58b5f6e3ead16cd7a416915a671dbab0b
issue-collapse:945b00d21b4739a2,Review thread: 2 issues - issue #100626: Will Deep Implicit Models ever become first class citizens in PyTorch?,related issue thread,stale/low urgency,2,high,work_thread_graph,2,0,0,0,"35749, 100626",,,,"module: autograd, module: functorch, oncall: jit","feature, module: autograd, module: functorch, oncall: jit, triaged",thread:issue:100626:issue:35749,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,stale/low urgency,1,Represents 2 open issues; raw issue rows remain available for audit.,"#35749 Cannot JIT functions with custom backwards (e.g. swish), #100626 Will Deep Implicit Models ever become first class citizens in PyTorch?",73eb5ccdaeb247690f110637748965a9c6083185fe6ad11649cd789904c6d039
issue-collapse:9e3746bb61371e03,module: arm has linked PR queue (2 issues),duplicate/related candidate,has linked PR,121,medium,title_component_similarity,2,0,0,0,"161783, 162422",,,,"module: arm, module: crash, module: dynamo, module: pt2-dispatcher, module: third_party, module: vectorization, oncall: cpu inductor, oncall: pt2","PT2-Bug-Bash, actionable, better-engineering, high priority, module: arm, module: crash, module: dynamo, module: pt2-dispatcher, module: third_party, module: vectorization, months, oncall: cpu inductor, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#161783 torch.compile has runtime overhead on small graphs, #162422 Runtime failure when running torch.compile() & using GCC 11.5.0 on Neoverse V1",cd93e1c2e5e31df3461f8055534c6994f1dedf4c92c517b41fac24fc3a0a6e90
issue-collapse:f8f95b0f1ab979c7,module: amp (automated mixed precision) high-priority/blocker queue (49 issues),duplicate/related candidate,high-priority/blocker,115,medium,title_component_similarity,49,0,0,0,"30651, 55669, 66782, 78829, 91633, 96779, 97086, 111901, 116423, 117494, 118088, 119845, 121101, 122308, 122478, 124644, 124656, 125664, 135835, 136712, 136728, 136819, 142454, 142457, 142459, 142834, 143484, 143487, 146788, 150098, 152102, 154014, 155146, 155147, 156563, 162343, 162797, 169210, 169741, 171838, 173495, 173574, 177227, 179030, 181853, 182279, 183251, 185247, 187659",,,,"module: amp (automated mixed precision), module: binaries, module: convolution, module: cpp, module: cpu, module: crash, module: cuda, module: dataloader, module: ddp, module: edge cases, module: elastic, module: empty tensor, module: error checking, module: fft, module: intel, module: linear algebra, module: macos, module: mkl, module: mkldnn, module: nn, module: openmp, module: pooling, module: random, module: regression, module: sparse, module: wsl, module: xpu, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: package/deploy, oncall: profiler, oncall: quantization, oncall: visualization","actionable, bot-mislabeled, bot-triaged, has workaround, high priority, module: amp (automated mixed precision), module: binaries, module: convolution, module: cpp, module: cpu, module: crash, module: cuda, module: dataloader, module: ddp, module: edge cases, module: elastic, module: empty tensor, module: error checking, module: fft, module: intel, module: linear algebra, module: macos, module: mkl, module: mkldnn, module: nn, module: openmp, module: pooling, module: random, module: regression, module: sparse, module: wsl, module: xpu, needs reproduction, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: package/deploy, oncall: profiler, oncall: quantization, oncall: visualization, ptd-bot-triaged, topic: binaries, topic: fuzzer, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 49 open issues; raw issue rows remain available for audit.,"#30651 [TensorBoard] The different order of import SummaryWriter may cause Segmentation fault, #55669 Segmentation fault when loss.backward(), #66782 Segmentation Fault when importing Torch, 2021 version, #78829 Certain import order triggers segmentation fault, #91633 Segmentation fault when running torch.nn.functional.fractional_max_pool3d on torch 1.13.1, #96779 Segmentation fault (core dumped) during Torch finetuning (at random step)",07f9c1a5f07b9f9e267e5ab1f233c2ce26dac2563c371e852e42448cf70ac9fb
issue-collapse:3ba61ea98999847d,module: __torch_function__ needs triage queue (24 issues),duplicate/related candidate,needs triage,115,medium,title_component_similarity,24,0,0,0,"119346, 119357, 119364, 119490, 119514, 119536, 119537, 119576, 167693, 185267, 185268, 186344, 186400, 186587, 186637, 186773, 186823, 188674, 188678, 188691, 188836, 188896, 189322, 189323",,,,"module: __torch_function__, module: autograd, module: convolution, module: cuda, module: flaky-tests, module: half, module: inductor, module: regression, module: rocm, module: tests, oncall: distributed, oncall: export, oncall: jit, oncall: profiler, oncall: pt2, oncall: r2p","bot-triaged, high priority, module: __torch_function__, module: autograd, module: convolution, module: cuda, module: flaky-tests, module: half, module: inductor, module: regression, module: rocm, module: tests, oncall: distributed, oncall: export, oncall: jit, oncall: profiler, oncall: pt2, oncall: r2p, rocm-skipped-tests, skipped, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 24 open issues; raw issue rows remain available for audit.,"#119346 DISABLED test_source_multithreaded_basic_work_in_main_thread_False (__main__.TestProfiler), #119357 DISABLED test_source_multithreaded_basic_work_in_main_thread_True (__main__.TestProfiler), #119364 DISABLED test_source_multithreaded_close_in_scope_work_in_main_thread_False (__main__.TestProfiler), #119490 DISABLED test_source_multithreaded_complex_work_in_main_thread_False (__main__.TestProfiler), #119514 DISABLED test_source_multithreaded_close_in_scope_work_in_main_thread_True (__main__.TestProfiler), #119536 DISABLED test_source_multithreaded_complex_work_in_main_thread_True (__main__.TestProfiler)",7b216723ca2e70b5650e7f9a1b23d493a2063749f2bcd5d1d33a01e5d9e5a1bd
issue-collapse:deb2de6e44e8c8b4,module: ddp needs triage queue (7 issues),duplicate/related candidate,needs triage,115,medium,title_component_similarity,7,0,0,0,"67389, 82308, 91173, 102596, 153779, 178766, 183459",,,,"module: ddp, module: docs, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms","better-engineering, bot-triaged, high priority, module: ddp, module: docs, module: nccl, needs reproduction, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, pipeline parallelism, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#67389 Missing doc for torch.distributed functions, #82308 torch.einsum gets wrong results randomly when training with multi-gpu, #91173 not able to import pipelines as torch.distributed is missing, #102596 Jetson NX with torch 1.12.0 :cannot import name 'ProcessGroup' from 'torch.distributed'., #153779 NCCL timeout while training on multi-gpu, #178766 torch.distributed._functional_collectives.reduce_scatter_tensor gets a wrong result",12cdf9dc05f6ae0cf4e7d04253292675cccc66ed028f19ff2b627f1bd3b1f027
issue-collapse:620100bd7847686d,module: autograd needs triage queue (4 issues),duplicate/related candidate,needs triage,115,medium,title_component_similarity,4,0,0,0,"189280, 189281, 189282, 189283",,,,"module: autograd, module: cpu, module: embedding, module: performance, module: regression, oncall: quantization","bot-triaged, module: autograd, module: cpu, module: embedding, module: performance, module: regression, oncall: quantization, op-bench, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,3,Represents 4 open issues; raw issue rows remain available for audit.,"#189280 [CPU][quant] operator_benchmark: qcat ~3-4x slower (likely) and fake_quantize ~700x slower (possible) since ~Nov 2025, #189281 [CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue), #189282 [CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact), #189283 [CPU] operator_benchmark: isolated ~2x regressions (index_add, baddbmm) since ~Nov 2025",35dce33288afc00519e3c3b69040f8945f2ddaf386f4f800dd2396020da7ea30
issue-collapse:1bd353a736c121ca,module: activation checkpointing needs triage queue (2 issues),duplicate/related candidate,needs triage,115,medium,title_component_similarity,2,0,0,0,"85088, 183207",,,,"module: activation checkpointing, module: dtensor, module: fsdp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, high priority, module: activation checkpointing, module: dtensor, module: fsdp, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#85088 reentrant torch.utils.checkpoint does not work with NamedTuple outputs, #183207 FSDP2 with ignore_params does not work with foreach optimizers",5267a8dd220d22eb292deffa3a7c21e37e105724827e0407e80361b4d109e1a9
issue-collapse:2f01fc7c5de41aa2,module: bfloat16 needs triage queue (14 issues),duplicate/related candidate,needs triage,97,medium,title_component_similarity,14,0,0,0,"19106, 48251, 52741, 67797, 75171, 86532, 86594, 86595, 97876, 159238, 173492, 183812, 183815, 189298",,,,"module: bfloat16, module: convolution, module: cpp, module: cpu, module: crash, module: functorch, module: intel, module: mkldnn, module: multithreading, module: performance, module: regression, module: scatter & gather ops, module: windows, oncall: jit","bot-triaged, good first issue, module: bfloat16, module: convolution, module: cpp, module: cpu, module: crash, module: functorch, module: intel, module: mkldnn, module: multithreading, module: performance, module: regression, module: scatter & gather ops, module: windows, needs reproduction, oncall: jit, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19106 Performance issue with torch.jit.trace(), slow prediction in C++ (CPU), #48251 torch.eye(d) is slow and hogs cpu for d >= 182, #52741 torch.jit.script is swallowing a rfloordiv warning, #67797 Conv2d kernel performance regression on CPU since PyTorch 1.9, #75171 torch.jit.load fails when path contains non-ascii characters, #86532 Conv2d will crash by using `jit.trace`",16658998adf1f7ae0935af89c38d6029538799e566c475925ba58748684bae2c
issue-collapse:eeb484596cf25136,module: cpu needs triage queue (5 issues),duplicate/related candidate,needs triage,97,medium,title_component_similarity,5,0,0,0,"141668, 150011, 152108, 177508, 183401",,,,"module: cpu, module: inductor, module: regression, module: windows, oncall: cpu inductor, oncall: pt2","bot-triaged, module: cpu, module: inductor, module: regression, module: windows, needs reproduction, oncall: cpu inductor, oncall: pt2, topic: performance, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#141668 [inductor][cpu] AMP CPP wrapper performance regression in 2024-11-25 nightly release, #150011 [Windows][CPU] 51 UT of inductor/test_torchinductor_opinfo.py failed with CPP wrapper, #152108 [inductor][cpu] AMP static shape default wrapper AOTInductor performance regression in 2025_04_20 nightly release, #177508 [inductor][cpu] torch_multimodal_clip AMP multiple thread performance regression in 2026-03-16 nightly release, #183401 [inductor][cpu] resnet50_quantized_qat CPP wrapper torch._inductor.exc.InductorError: CppCompileError: C++ compile error in 2026-05-11 nightly release",22134297d8885f4d49fc76556def36d3f7fd12297225e623f9f19ed334fac7db
issue-collapse:fac821cdbb52e7ec,module: crash needs triage queue (4 issues),duplicate/related candidate,needs triage,97,medium,title_component_similarity,4,0,0,0,"116351, 177476, 177825, 182653",,,,"module: crash, oncall: quantization","bot-triaged, module: crash, oncall: quantization",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,3,Represents 4 open issues; raw issue rows remain available for audit.,"#116351 `torch.quantize_per_channel`: FPE, #177476 Crash in torch.ao.nn.quantized.functional.interpolate when input tensor has a 0-sized dimension, #177825 Internal Assert Failure in `torch.ao.nn.quantized.GroupNorm` with incorrectly sized bias tensor, #182653 CPU FPE crash in `torch.nn.quantized.functional.conv2d` (qnnpack)",4ed78bac40c4372ff271a9a20dfcf5c625884528e533507b2a75dff90a459ee6
issue-collapse:0a26b0ddec80f91f,oncall: jit needs triage queue (2 issues),duplicate/related candidate,needs triage,97,medium,title_component_similarity,2,0,0,0,"34116, 37015",,,,oncall: jit,"high priority, oncall: jit, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#34116 autograd with TorchScript does not match finite differences, #37015 Converting to Torch Script: cpp_module does not match nn_module",008bcd26f59c64fbbf13955a4aee7c71bb765ed8a35507bfa8ad903adabf3aa9
issue-collapse:17f807f47bf8aa79,module: serialization needs triage queue (2 issues),duplicate/related candidate,needs triage,97,medium,title_component_similarity,2,0,0,0,"46166, 82306",,,,"module: serialization, oncall: distributed","bot-triaged, high priority, module: serialization, oncall: distributed, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#46166 Error with DistributedDataParallel with specific model, #82306 when distribute training load pretrain model error",2654c6c7595171dbb3f07fa7915c6c376e7fd6d911aedf4c308fcd78cdd0e2ad
issue-collapse:98b6a37c6178e0c5,module: cpu high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,97,medium,title_component_similarity,2,0,0,0,"99509, 115640",,,,"module: cpu, module: crash, module: intel, oncall: quantization","high priority, module: cpu, module: crash, module: intel, oncall: quantization, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#99509 'Illegal instruction (core dumped)' for gpt-j bf16 generation task using greedy search, #115640 Quantisation results in a ""Illegal instruction (core dumped)""",8990605c0cdc59640f5eb19b11ac6802f0ab77cd4f8f7a8974fe3504ad097bc3
issue-collapse:98a7f332d7d7085e,module: autograd needs triage queue (7 issues),duplicate/related candidate,needs triage,94,medium,title_component_similarity,7,0,0,0,"73294, 143580, 146907, 158701, 182283, 185715, 186453",,,,"module: autograd, module: crash, module: cudacachingallocator, module: ddp, module: dynamic shapes, module: rpc, module: third_party, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: pt2","actionable, bot-triaged, high priority, module: CUDACachingAllocator, module: autograd, module: crash, module: ddp, module: dynamic shapes, module: rpc, module: third_party, needs reproduction, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: pt2, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#73294 Internal assert failed at rref_context.cpp, #143580 RuntimeError: expect_autograd_hooks_ INTERNAL ASSERT FAILED at ""../torch/csrc/distributed/c10d/reducer.cpp"", #146907 INTERNAL ASSERT FAILED at ""/pytorch/torch/csrc/distributed/c10d/reducer.cpp"":2134, please report a bug to PyTorch., #158701 RuntimeError: optimized_plan_ INTERNAL ASSERT FAILED at ""/opt/conda/envs/bld/conda-bld/libtorch_1741644046984/work/torch/csrc/jit/runtime/profiling_graph_executor_impl.cpp"":725, please report a bug to PyTorch., #182283 INTERNAL ASSERT FAILED at /pytorch/c10/core/Scalar.cpp:42, #185715 INTERNAL ASSERT FAILED: NYI SymInt equality in c10/core/Scalar.h when using torch.compile(dynamic=True) and torch.autograd.grad with torch.pow",1633923d1f7da1ef6f2ace499d4228b3183627e33e3c88942f153e6559ea8b89
issue-collapse:daefb64b15a0d14d,module: amp (automated mixed precision) has linked PR queue (3 issues),duplicate/related candidate,has linked PR,94,medium,title_component_similarity,3,0,0,0,"180740, 188261, 188912",,,,"module: amp (automated mixed precision), module: correctness (silent), module: cudnn, module: dynamic shapes, module: half, module: inductor, module: norms and normalization, module: reductions, module: regression, module: xpu, oncall: pt2","bot-triaged, module: amp (automated mixed precision), module: correctness (silent), module: cudnn, module: dynamic shapes, module: half, module: inductor, module: norms and normalization, module: reductions, module: regression, module: xpu, oncall: pt2, repro:agent:h100:repro_fail:skipped, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,2,Represents 3 open issues; raw issue rows remain available for audit.,"#180740 [Inductor] functorch_dp_cifar10 accuracy regression in dynamic_inductor_torchbench, #188261 [inductor][XPU] PR #184366 causes AMP fp16 training accuracy regression on XPU, #188912 vit_base_patch16_siglip_256 amp training accuracy regression on H100 (2.13.0)",6df09ad8530bb064d727df361d36cb5ca9f7ef29923ce2bab326a68b62fb93b1
issue-collapse:c0b843fec6917277,module: crash high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,94,medium,title_component_similarity,2,0,0,0,"130708, 145064",,,,"module: crash, module: empty tensor, module: error checking, module: macos, module: numpy","actionable, module: crash, module: empty tensor, module: error checking, module: macos, module: numpy, needs reproduction, topic: fuzzer, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130708 Converting a numpy array of size larger than 32,768 to a tensor causes a segmentation fault, #145064 `_pdist_forward` causes segmentation fault for 3D tensor with last dimension of size 0",8721eb81d14046a33fd5f22c45325c680904c6cd67dbce69bf4c434b5df09b2d
issue-collapse:c18604ba270e8c7a,module: complex high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,94,medium,title_component_similarity,2,0,0,0,"185405, 188069",,,,"module: complex, module: cpu, module: crash, module: inductor, oncall: pt2","bot-triaged, matrix multiplication, module: complex, module: cpu, module: crash, module: inductor, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#185405 [Inductor] AssertionError: len(index) == len(stride) in ir.py when using torch.compile(mode=""max-autotune"") on CPU BMM/GEMM, #188069 torch.compile(mode=""max-autotune"") crashes with too many positional arguments during bmm autotuning when template inputs are deduplicated",1a64ba3d8d95ba5d98402574dcca5ccea51f9a8f93ffcb4f38c031f0fb877f76
issue-collapse:381ada236f6ffedc,module: power needs triage queue (2 issues),duplicate/related candidate,needs triage,89,medium,title_component_similarity,2,0,0,0,"171870, 181638",,,,"module: power, oncall: cpu inductor, oncall: pt2","bot-triaged, module: POWER, oncall: cpu inductor, oncall: pt2",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#171870 [s390x/ZVECTOR] C++ compilation failure in argmax/argmin operations with torch.compile inductor backend, #181638 torch.compile(backend='inductor') hangs indefinitely on LayerNorm → flatten → cumprod",1f36fb28510063885d3a54d133907f1721c496ab63e57db4765d7a9c7cbb049e
issue-collapse:9c0d771938ec4982,module: binaries high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,88,medium,title_component_similarity,2,0,0,0,"66033, 170039",,,,"module: binaries, module: convolution, module: crash, module: macos, module: multithreading","module: binaries, module: convolution, module: crash, module: macos, module: multithreading, needs design, needs reproduction, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#66033 Crashed with `terminating with uncaught exception of type std::__1::system_error: condition_variable wait failed: Invalid argument`, #170039 PyTorch crashes with std::__1::system_error: mutex lock failed: Invalid argument",98159dae7663ab5ba75d23db65eed31781e9e03bac133fbc881ca5ed6188140e
issue-collapse:6e70166f9b5aab36,module: aotdispatch high-priority/blocker queue (6 issues),duplicate/related candidate,high-priority/blocker,82,medium,title_component_similarity,6,0,0,0,"38910, 173313, 174468, 174704, 183890, 187781",,,,"module: aotdispatch, module: compile ux, module: compile-time, module: crash, module: cuda, module: cudnn, module: dynamic shapes, module: dynamo, module: fsdp, module: memory usage, module: nn, module: pt2-dispatcher, module: regression, oncall: distributed, oncall: pt2","PT2-Bug-Bash, bot-triaged, feature, high priority, module: aotdispatch, module: compile ux, module: compile-time, module: crash, module: cuda, module: cudnn, module: dynamic shapes, module: dynamo, module: fsdp, module: memory usage, module: nn, module: pt2-dispatcher, module: regression, needs reproduction, needs research, oncall: distributed, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#38910 Memory Leak with Docker GPU, #173313 Bad perf for torch.compile for merge_attention with transposed inputs, #174468 Memory leak by torch.compile, #174704 Recompile time with torch.compile 2.10, #183890 torch.compile hits RecursionError on nn.Module with custom __getattr__, #187781 [RFC] Topology-aware nn.Module primitives for memory optimization in torch.compile",e078deabe30ba9811aaa345f1440ef73550880fa88062857c2cac8e9664fcee7
issue-collapse:27ab2559d1bc816a,module: arm high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,82,medium,title_component_similarity,2,0,0,0,"178538, 188255",,,,"module: arm, module: binaries, module: crash, module: windows","bot-mislabeled, bot-triaged, high priority, module: arm, module: binaries, module: crash, module: windows, needs reproduction, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#178538 [BUG] [CRASH] Torch cause notebook crash on any exception (import torch), #188255 Crash while import torch, in some special virtualization situation.",faaebc6f197ecd46c7588d3496bf73ac9ef8a428ddf3f3d9d5894c995e743c80
issue-collapse:346edad8fa213d4c,module: binaries high-priority/blocker queue (6 issues),duplicate/related candidate,high-priority/blocker,77,medium,title_component_similarity,6,0,0,0,"37449, 87209, 119547, 178085, 188958, 188959",,,,"module: binaries, module: convolution, module: crash, module: cuda, module: cudnn, module: ddp, module: dependency bug, module: edge cases, module: int overflow, module: memory usage, module: norms and normalization, module: regression, oncall: distributed, oncall: distributed parallelisms","bot-triaged, high priority, module: binaries, module: convolution, module: crash, module: cuda, module: cudnn, module: ddp, module: dependency bug, module: edge cases, module: int overflow, module: memory usage, module: norms and normalization, module: regression, needs reproduction, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#37449 RuntimeError: CUDA error: an illegal memory access was encountered with channels_last, #87209 Reproducible ""CUDA error: an illegal memory access was encountered"", #119547 CUDA Memory error while having a lot of free memory, #178085 DDP mode: CUDA error: an illegal memory access was encountered, #188958 [CUDA] illegal memory read on batch_norm_backward_reduce_channels_last_kernel, #188959 [CUDA] illegal memory read on col2im_batched_kernel",54c64d40c2ee541281356844de802cab371498d6ae9ed2b2e1459c53091b4cb4
issue-collapse:ea7fe8185ea74b55,module: arm needs triage queue (3 issues),duplicate/related candidate,needs triage,76,medium,title_component_similarity,3,0,0,0,"170877, 177247, 177249",,,,"module: arm, module: regression, oncall: jit","bot-triaged, module: arm, module: regression, oncall: jit, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#170877 AArch64 unit test failure test_transpose_mxn_16_16_bf16_fp16 on Neoverse-V1, #177247 AArch64 Unit Test Failure - test_jit_autocast, #177249 AArch64 Unit Test Failure - test_ops_jit.py test_variant_consistency_jit_linalg_lstsq_cpu_complex64",0c2e04247cb53cdf1685319d76c3f663be962e80c038cf9b18bdf3b704e8d957
issue-collapse:b19f7decc7eb4ae2,module: correctness (silent) high-priority/blocker queue (6 issues),duplicate/related candidate,high-priority/blocker,70,medium,title_component_similarity,6,0,0,0,"187444, 187445, 187446, 187447, 187448, 187449",,,,"module: correctness (silent), module: ddp, module: determinism, module: inductor, module: nccl, module: pt2 accuracy, module: pt2 optimizer, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, high priority, module: correctness (silent), module: ddp, module: determinism, module: inductor, module: nccl, module: pt2 accuracy, module: pt2 optimizer, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#187444 `torch.compile` vs eager DDP divergence with ResNet50 + AdamW lr=1.0 on two-node NCCL, #187445 `torch.compile` vs eager DDP loss/grad/logits divergence with ResNet34 + SGD lr=0.1, batch=2, TF32 allowed, and 1MB buckets, #187446 `torch.compile` vs eager DDP grad/parameter divergence with ResNet50 + Adam lr=0.05, batch=1, sqrt LR scaling and step scheduler, #187447 `torch.compile` vs eager DDP parameter-fingerprint divergence with ResNet50 + Adam lr=0.1, grad accumulation, grad clipping, and 1MB buckets, #187448 `torch.compile` vs eager DDP grad/logits divergence with ResNet50 + AdamW lr=0.01, warmup cosine schedule, grad accumulation, and DDP bucket views, #187449 `torch.compile` vs eager DDP logits-sum divergence with ResNet50 + AdamW lr=0.01 on two-node NCCL",b632cdaa45fe4ab29a0805af595639956e8037002fb227fb667a0f5e4257b93b
issue-collapse:598ad5aaf6445460,module: autograd high-priority/blocker queue (5 issues),duplicate/related candidate,high-priority/blocker,70,medium,title_component_similarity,5,0,0,0,"80946, 153937, 155693, 172798, 181024",,,,"module: autograd, module: cuda, module: dlpack, module: nn, module: regression, module: tests, module: typing, oncall: pt2","bot-triaged, high priority, llm-amenable, module: autograd, module: cuda, module: dlpack, module: nn, module: regression, module: tests, module: typing, needs reproduction, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,4,Represents 5 open issues; raw issue rows remain available for audit.,"#80946 torch.nn.functional.linear fails for multi-dimensional bias from torch 1.12, #153937 Dynamic compilation fails with torch 2.7, #155693 torch typing: register_buffer -> device, #172798 `backward` does not respect `with torch.device(device)`, #181024 [CUDA] `test_dlpack_tensor_on_different_device` fails with `torch.device` instead of `torch.cuda.device`",72e9f6d73de18bac5b055dc133548966e198fa2155a2dc01954715bfc4651a18
issue-collapse:02fe6e7e9e545fc1,module: abi high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,70,medium,title_component_similarity,2,0,0,0,"149658, 187320",,,,"module: abi, module: binaries, module: build, module: cpp-extensions, module: cuda, module: regression","bot-triaged, enhancement, has workaround, high priority, module: abi, module: binaries, module: build, module: cpp-extensions, module: cuda, module: regression, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#149658 unexpected kwarg 'no_python_abi_suffix' when installing stable version of pytorch with sam2, #187320 Stable ABI isn't stable when PyTorch changes the default CUDA version",d6001f0b50043f3f2f542b9e482d19675e14e775008629955f376b2e6ae8625b
issue-collapse:db6bca2aff60eff7,module: dynamo needs triage queue (4 issues),duplicate/related candidate,needs triage,65,medium,title_component_similarity,4,0,0,0,"144663, 166572, 167623, 175520",,,,"module: dynamo, oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, export-triaged, feature, module: dynamo, oncall: export, oncall: pt2",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#144663 Support loading and executing a ExportedProgram from torch.export in C++ environment, #166572 [Regression] torch.export.ExportedProgram.run_decompositions Starts to Break in Torch 2.8 with Mutable Buffer, #167623 torch.export: support external weights (.pt2 format), #175520 Question about torch._dynamo.export and nn.GRU support",fefb6b174bb361a8e30934e5fa613e351a65b2d399b727dac6f07ac399bdce1a
issue-collapse:5e680c38ede56635,module: c10d needs triage queue (3 issues),duplicate/related candidate,needs triage,65,medium,title_component_similarity,3,0,0,0,"74809, 144359, 163193",,,,"module: c10d, module: correctness (silent), module: dtensor, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms","bot-triaged, module: c10d, module: correctness (silent), module: dtensor, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,2,Represents 3 open issues; raw issue rows remain available for audit.,"#74809 Incorrect results for `torch.distributed.gather` for tensor created from permuted NumPy array, #144359 Incorrect Results with Tensor Parallelism, #163193 Incorrect results when adding partial DTensor with a scalar",d7629157adfa055fc5b7b8f218ce842ab3a058d7e2d3db00a4b63d687a024f1a
issue-collapse:15b2bd916e68de80,oncall: distributed needs triage queue (2 issues),duplicate/related candidate,needs triage,65,medium,title_component_similarity,2,0,0,0,"130345, 179238",,,,"oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130345 Provide shutdown() method for ProcessGroupGloo distributed backend, #179238 Shutdown race in Gloo backend",1502ade3cc5b3d81fda552207d36cf611a8bc7312823e39095097360d89f6591
issue-collapse:fcad92983847dc99,module: nn high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,65,medium,title_component_similarity,2,0,0,0,"86120, 175293",,,,"module: nn, oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, high priority, module: nn, oncall: export, oncall: pt2, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#86120 TransformerEncoder src_key_padding_mask does not work in eval(), #175293 TL-parse does not work for export.",bb15c8db8d669e6b744038b26d78ba6da898ca554ea79880792f431575f2f1a4
issue-collapse:4156673460481095,module: ddp needs triage queue (2 issues),duplicate/related candidate,needs triage,65,medium,title_component_similarity,2,0,0,0,"130113, 170045",,,,"module: ddp, module: nccl, oncall: distributed, oncall: distributed infra","bot-triaged, module: ddp, module: nccl, oncall: distributed, oncall: distributed infra, ptd-bot-triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#130113 ncclUnhandledCudaError: Call to CUDA function failed., #170045 ncclUnhandledCudaError: Call to CUDA function failed. Last error: Cuda failure 999 'unknown error'",acd75ed4f49c770aa0cefacb8162b8e03f3518741ce8503fc239da6b86977466
issue-collapse:477d3ffd19fb2f85,module: dataloader needs triage queue (2 issues),duplicate/related candidate,needs triage,65,medium,title_component_similarity,2,0,0,0,"87995, 167975",,,,"module: dataloader, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms","bot-triaged, module: dataloader, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#87995 multi-node distributed training rank0 hang at dataloader after a few epochs, #167975 Multi-node distributed training hangs indefinitely after NCCL initialization completes, specifically blocking at `verify_param_shape_across_processes` in `torch.distributed.utils`.",00ff6aa7b94b9142f631a2294c5a6994e2b0ece7a564e1e4f867149aa3c4c880
issue-collapse:769e050eca69564c,module: crash high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,65,medium,title_component_similarity,2,0,0,0,"39388, 143574",,,,"module: crash, module: nccl, oncall: distributed, oncall: distributed infra","bot-triaged, module: crash, module: nccl, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#39388 RuntimeError: NCCL error in: /pytorch/torch/lib/c10d/ProcessGroupNCCL.cpp:514, unhandled system error, NCCL version 2.4.8, #143574 Torch elastic restart fails with torch-2.6.0 nightly build: NCCL unhandled system error",deea9d04775d42c72801774b17ac08cff8add8cec642b84646a4660bd63a0626
issue-collapse:847fc14aa8961798,module: arm high-priority/blocker queue (2 issues),duplicate/related candidate,high-priority/blocker,64,medium,title_component_similarity,2,0,0,0,"40425, 149201",,,,"module: arm, module: binaries, module: crash, module: cuda, module: dependency bug, module: macos, module: openmp","module: arm, module: binaries, module: crash, module: cuda, module: dependency bug, module: macos, module: openmp, needs reproduction, triaged",,"Use one comparison pass to decide whether these issue reports share a close path, answer, reproduction, or implementation.",Compare the related issue set before writing separate maintainer responses.,1,Represents 2 open issues; raw issue rows remain available for audit.,"#40425 Segment Fault after use cusolverDnDestroy() with torch1.5, #149201 Segment fault when used with FAISS on arm",8f1b989ac30b603607a253da2079f7ef31dee3560341774708ec5235e8e53165
issue-collapse:1ef6c58bd6bac717,module: crash needs triage,batch action queue,needs triage,139,medium,bucket_component_queue,7,0,0,0,"73885, 122019, 153260, 170003, 186535, 189126, 189150",,,,"module: crash, module: cuda, module: cuda graphs, module: memory usage, module: multi-gpu, module: nccl, module: windows, oncall: distributed, oncall: distributed infra, oncall: export, oncall: jit, oncall: pt2, oncall: quantization","bot-triaged, high priority, module: crash, module: cuda, module: cuda graphs, module: memory usage, module: multi-gpu, module: nccl, module: windows, oncall: distributed, oncall: distributed infra, oncall: export, oncall: jit, oncall: pt2, oncall: quantization, ptd-bot-triaged, triage review, triaged",,"Review this module: crash queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#73885 operation not supported crash when initializing RPC tensorpipe, #122019 `F.max_pool3d`: Segfault due to incorrect `dilation` check in kernel, #153260 Export shouldn't run TS under the hood., #170003 batch_isend_irecv with nccl causes illegal memory access depending on P2P ordering, #186535 Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator, #189126 Report a bug, as requested by the error message",a05a65656e0b174c8b4ad8950c27c9d2f1f342012fb394407dd36bf21730747b
issue-collapse:ad84acd62ef6aee2,module: convolution needs triage,batch action queue,needs triage,139,medium,bucket_component_queue,4,0,0,0,"61268, 154354, 188323, 189239",,,,"module: convolution, module: cpu, module: crash, module: cuda, module: dynamic shapes, module: inductor, module: ios, module: macos, module: memory format, module: performance, module: regression, oncall: cpu inductor, oncall: mobile, oncall: pt2, oncall: quantization","bot-triaged, module: convolution, module: cpu, module: crash, module: cuda, module: dynamic shapes, module: inductor, module: ios, module: macos, module: memory format, module: performance, module: regression, needs research, oncall: cpu inductor, oncall: mobile, oncall: pt2, oncall: quantization, triage review, triaged",,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#61268 Unknown type name '__torch__.torch.classes.metal.Conv2dOpContext', #154354 integer convolution, #188323 [Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative, #189239 [Inductor]Unable to apply layout optimization on convolution backward causes training performance regression.",8f7d87cdb8fdcd122226b4c4140124f503c5d3c00883c787428dd45688a8bf1d
issue-collapse:2f270b93e718f73f,module: arm needs triage,batch action queue,needs triage,127,medium,bucket_component_queue,10,0,0,0,"48452, 88617, 125734, 144770, 146483, 146914, 146915, 148902, 157668, 173907",,,,"module: arm, module: cuda, module: ios, module: macos, module: nccl, module: nnpack, module: performance, module: tests, oncall: cpu inductor, oncall: distributed, oncall: distributed infra, oncall: jit, oncall: mobile, oncall: pt2, oncall: quantization","bot-triaged, high priority, module: arm, module: cuda, module: ios, module: macos, module: nccl, module: nnpack, module: performance, module: tests, needs reproduction, oncall: cpu inductor, oncall: distributed, oncall: distributed infra, oncall: jit, oncall: mobile, oncall: pt2, oncall: quantization, ptd-bot-triaged, triage review, triaged",,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#48452 After updating to Xcode 12 and LibTorch to 1.7.0. Facing issue when running unit test., #88617 The libtorch tests Simplify.{SimplifySymbolicMinMax,SimplifyNestedMax,SimplifyNestedMin} fail on Apple Silicon, #125734 ☂️ `torch.compile` generates slower code for LLMs than eager on ARM platform (M1/AARCH64), #144770 [ARM] multiple test failures in TestQuantizedConv on Aarch64, #146483 [ARM] Unit test failure on Neoverse-N1 - CPUReproTests.test_lstm_packed, #146914 TestSelectAlgorithmCPU.test_int8_woq_mm fails on Aarch64",4493c16c732581bc8d8af4ffca2b8ceb8dfbbcad10241ead92032676cff346dd
issue-collapse:5275a92a6bde7e5d,module: cpu needs triage,batch action queue,needs triage,127,medium,bucket_component_queue,5,0,0,0,"116240, 122813, 153145, 181621, 187912",,,,"module: cpu, module: crash, module: flex attention, module: higher order operators, module: inductor, module: intel, module: linear algebra, module: mkldnn, module: multithreading, module: performance, module: pt2-dispatcher, module: scatter & gather ops, oncall: cpu inductor, oncall: pt2, oncall: quantization","bot-mislabeled, bot-triaged, enhancement, high priority, intel, module: cpu, module: crash, module: flex attention, module: higher order operators, module: inductor, module: intel, module: linear algebra, module: mkldnn, module: multithreading, module: performance, module: pt2-dispatcher, module: scatter & gather ops, needs reproduction, oncall: cpu inductor, oncall: pt2, oncall: quantization, triage review, triaged",,"Review this module: cpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#116240 [inductor][cpu] RuntimeError: quantized_resize_cpu_ does not have a deterministic implementation, #122813 cpu performance for int4mm kernels, #153145 Segmentation faults with CPU FlexAttention, #181621 torch.compile (Inductor) generates invalid C++ kernel: tmp8 used outside its declared scope when torch.gather index is a self-referential expression, #187912 [CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared",402372143f1eed6cfaf6f3ab3ddf0668483dc39846bcfbc73a581657f374946f
issue-collapse:7ad5740bb8ad57d2,module: ddp needs triage  batch 1,batch action queue,needs triage,115,medium,bucket_component_queue,14,0,0,0,"47163, 60844, 65813, 68917, 69308, 69591, 71855, 74115, 76449, 85671, 106690, 111138, 112152, 124666",,,,"module: ddp, module: fsdp, module: memory format, module: pytree, module: regression, oncall: distributed, oncall: distributed parallelisms","Stale, better-engineering, bot-triaged, enhancement, high priority, module: ddp, module: fsdp, module: memory format, module: pytree, module: regression, oncall: distributed, oncall: distributed parallelisms, pt_distributed_rampup, ptd-bot-triaged, triage review, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47163 Grad strides do not match bucket view strides., #60844 DDP fails if you have multiple forward passes and a single backwards pass with `find_unused_parameters=True`, #65813 [RFC] Low-level speed optimizations for PowerSGD, #68917 Bugs related to NCCL on RTX 6000, code freezes with no output when using DistributedDataParallel, #69308 Enable DDP checkpointing tests for Gloo backend, #69591 Make DDP + Zero have the same communication volume as regular DDP",dd3b967b895203cd1019196f9433b893afd6dbf39b2bf8b91db54bc04e156c04
issue-collapse:1dff1dfb437c9628,module: c10d needs triage  batch 1,batch action queue,needs triage,115,medium,bucket_component_queue,14,0,0,0,"62780, 71303, 73542, 80349, 81684, 83289, 83824, 85234, 106164, 106633, 130251, 137007, 143593, 170707",,,,"module: c10d, module: crash, module: ddp, module: distributed_tool, module: docs, module: elastic, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms","better-engineering, bot-triaged, enhancement, feature, high priority, module: c10d, module: crash, module: ddp, module: distributed_tool, module: docs, module: elastic, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#62780 Seeing unit test failures for ProcessGroupShareTensorTest, #71303 [RFC] Cross-Process Performance Analysis: Straggler Detection, #73542 Investigate why tensor shapes are not populated when printing WorkNCCL for broadcast, #80349 Distributed Store `get` doesn't work well with `add`, #81684 Message exchange failure when perform alltoallv (cpus), #83289 Improvements to ProcessGroupGloo monitored_barrier",045639e181fa9c1929e29d8eefb36fa5d8a95aa6135fa7ac58b2c260ba8e1aa8
issue-collapse:487f2ee5938d532c,module: dtensor needs triage,batch action queue,needs triage,115,medium,bucket_component_queue,13,0,0,0,"149447, 161147, 162731, 173988, 173990, 174276, 174732, 175469, 177059, 179502, 180088, 182031, 185167",,,,"module: dtensor, module: nn, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, enhancement, feature, high priority, module: dtensor, module: nn, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: export, oncall: pt2, ptd-bot-triaged, triage review, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#149447 DTensor slicing on sharded dimension leads to replication, #161147 RuntimeError in DTensor evenly shard view op, #162731 DTensor cached op propagation results can be mutated when propagating other ops, #173988 [DTensor] Implement decomposed sharding propagation for SDPA operations, #173990 [DTensor] Propagate 'grad_dtype' attribute in 'to_local' and 'from_local' operations, #174276 [DTensor] Avoid forced redistribution in layer_norm strategy",ff9b8a6c6e7c035320460d0810d869e4d480f081afee6a0131c55d13787718b8
issue-collapse:4395972e2c95e564,module: aotinductor needs triage,batch action queue,needs triage,115,medium,bucket_component_queue,10,0,0,0,"140805, 145394, 145610, 148557, 151746, 152612, 154094, 159269, 165929, 166939",,,,"module: aotinductor, module: arm, module: crash, module: mps, module: user triton, oncall: cpu inductor, oncall: export, oncall: pt2","enhancement, export-triage-review, export-triaged, module: aotinductor, module: arm, module: crash, module: mps, module: user triton, oncall: cpu inductor, oncall: export, oncall: pt2",,"Review this module: aotinductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#140805 [AOTI] Test case `test_aot_inductor.py::test_multi_device_cpu` is Running GPU AOT Inductor in CPU container runner., #145394 [EXPORT AOTI] `aoti_compile_and_package` custom_ops dependecies, #145610 mmap fails on 64k page aarch64 systems for AOTI model loading, #148557 AOTI is OOM-ing when eager doesn't, #151746 [AotInductor][Export][Triton] how to export custom triton kernels when use torch.export.export, #152612 Enable AOTI for Metal inductor",eaaca26eff45f11d92a6cc7980c1b85d8664d1eac9fbe136c3d8632c64bc68ea
issue-collapse:c747c2aae74444ba,module: binaries needs triage,batch action queue,needs triage,115,medium,bucket_component_queue,7,0,0,0,"110259, 121597, 139745, 159265, 174488, 175045, 189194",,,,"module: binaries, module: ci, module: cuda, module: inductor, module: regression, module: rocm, module: vectorization, oncall: mobile, oncall: package/deploy, oncall: profiler, oncall: pt2, oncall: releng","bot-triaged, high priority, module: binaries, module: ci, module: cuda, module: inductor, module: regression, module: rocm, module: vectorization, needs reproduction, oncall: mobile, oncall: package/deploy, oncall: profiler, oncall: pt2, oncall: releng, triage review, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#110259 Cannot avoid kineto_LIBRARY-NOTFOUND error when using pre-built pytorch, #121597 LibTorch-Lite 2.1.0 Libs missing x86_64, #139745 cmake libtorch_jni.so, bug: 'fp16/psimd.h' file not found, #159265 Unpin dependency version when possible, #174488 [cd][testing] Add almalinux/9-base test to CD pipeline, #175045 Consolidate or retire pytorch/almalinux-builder Docker images after MAGMA deprecation",8041d281f19b1eb83f55e6b945db899a09526a28f78aaa45ce0774149650c3dd
issue-collapse:bbcb762764d1213d,module: inductor needs triage,batch action queue,needs triage,115,medium,bucket_component_queue,5,0,0,0,"135689, 135690, 136618, 150526, 173921",,,,"module: inductor, oncall: cpu inductor, oncall: pt2","PT2-Bug-Bash, bot-triaged, high priority, module: inductor, oncall: cpu inductor, oncall: pt2",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#135689 [inductor][cpu]GPTNeoForSequenceClassification AMP single/multiple thread static/dynamic shape default/cpp accuracy failure, #135690 [inductor][cpu] hf_BigBird float32 multiple threads performance regression in 2024-09-08 nightly release, #136618 [inductor][cpu]pyhpc_isoneutral_mixing singel thread AMP cpp wrapper performance regression in 2024-09-22 nightly release, #150526 Consider context when tuning kernels for max-autotune to more accurately reflect the performance of real workloads, #173921 [Inductor] Significant numerical divergence (4.7% relative error) in Conv2d on CPU with torch.compile",1e470fa2732a9f5be49fe54d27c6e7780bf06e45e42b2a1cc3941946b7130140
issue-collapse:411d1dfdad2bff5a,module: ci needs triage,batch action queue,needs triage,115,medium,bucket_component_queue,4,0,0,0,"153517, 161179, 162878, 169069",,,,"module: ci, module: cuda, module: nccl, module: regression, oncall: distributed, oncall: distributed infra, oncall: releng","better-engineering, bot-triaged, module: ci, module: cuda, module: nccl, module: regression, oncall: distributed, oncall: distributed infra, oncall: releng, ptd-bot-triaged",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#153517 [CI][CUDA][Distributed] test_non_blocking_with_eager_init timeout, #161179 Consolidate magma-cuda build for Linux and Windows, #162878 CUDA 13.0 Segmentation Fault in test/test_numba_integration.py::TestNumbaIntegration::test_array_adaptor, #169069 Triton Pin update 3.6 related - All periodic linux-jammy-cuda12.4-py3.10-gcc11 are in broken state. Some linux-jammy-cuda12.8-py3.10-gcc11-debug are broken as well",b8504113ea949bf92912f8dbf9b000a3ee0113fe967c32f33d5e347b9ae156ff
issue-collapse:875a8b8087aeb893,oncall: jit needs triage  batch 1,batch action queue,needs triage,97,medium,bucket_component_queue,14,0,0,0,"35998, 40867, 41289, 41695, 42317, 43548, 47218, 51856, 67541, 81085, 173643, 183520, 185441, 189309",,,,oncall: jit,"NNC, OSS contribution wanted, bot-triaged, days, good first issue, high priority, oncall: jit, triage review, triaged, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35998 JITed GRU too slow, #40867 [JIT] Infer type of argument foo = None to be Optional[Tensor], #41289 List[Any] support in TorchScript, #41695 Improve overload resolution order rules for `invokeOperatorFromPython`, #42317 Instance norm annotation is incorrect, #43548 JIT tracing incorrectly records some slice bounds",fca8119f7b38bfcf1cadcb530f1d6e70e242535dced942a90f39c6f7cdc97441
issue-collapse:1bbf42d061e311d8,oncall: export needs triage  batch 1,batch action queue,needs triage,97,medium,bucket_component_queue,14,0,0,0,"113572, 130861, 131941, 144683, 146152, 151654, 151726, 153705, 172930, 176148, 177510, 178618, 179176, 184647",,,,"oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, export-triage-review, export-triaged, good first issue, high priority, oncall: export, oncall: pt2, repro:agent:h100:repro_success:reproduced, triage review, triaged",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#113572 Deepcopying an exported model changes numerics for MobileNetV2, #130861 torch.fx.Tracer.record_stack_traces is broken in torch 2.4.0, #131941 [export] Schematize nn_module_stack serialization, #144683 torch.export treats two of the same parameters as the same node, #146152 [torch.export] Cannot export TorchVision fasterrcnn_mobilenet_v3_large_fpn, #151654 [sparse][exported-graph] Wrong Exported Graph for the Sparse Tensor",8210be4a18da487430470abadd1552e261d41c27712d48bebaac46158ae9258b
issue-collapse:97fc48a4b0920cef,oncall: distributed needs triage  batch 1,batch action queue,needs triage,97,medium,bucket_component_queue,14,0,0,0,"47999, 54524, 73602, 80151, 162466, 164929, 177730, 178473, 181061, 182102, 182915, 186938, 187138, 188193",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra","better-engineering, bot-mislabeled, bot-triaged, feature, high priority, needs design, needs reproduction, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47999 Torch DDP seems not prefer IPv6 over IPv4?, #54524 Store created by dist.new_group doesn't appear to respect timeout, #73602 Report DDP efficiency metric to users and guide them for setting up DDP correctly, #80151 activation checkpointing with non_reentrant implementation memory leaks, #162466 Launch one kernel take long time., #164929 get_optimizer_state_dict modifies optimizer state",b487f09443a58c3a9010799a37b92add57ecf885e679bab9c6347d1b7e562d4c
issue-collapse:3afb3a1cfe65aff9,module: bootcamp needs triage  batch 1,batch action queue,needs triage,97,medium,bucket_component_queue,14,0,0,0,"5489, 45549, 47565, 51673, 55577, 61344, 61966, 63917, 64060, 66094, 68014, 71261, 71277, 84318",,,,"module: bootcamp, module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: jit, oncall: profiler","NNC, OSS contribution wanted, better-engineering, days, good first issue, high priority, jit-backlog, module: bootcamp, module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: jit, oncall: profiler, pt_distributed_rampup, triage review, triaged",,"Review this module: bootcamp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5489 torch.jit.trace(network, data) fails if data is an OrderedDict, #45549 Can't initialize NCCL/GLOO process group if default process group is MPI, #47565 Keyword-only function not allowed with TorchScript, #51673 [JIT/Futures] Future callbacks/then() APIs should throw the correct exception, #55577 Source location range issue for dictionary, #61344 Get TensorType's device in python",a9a2b9dc17e4802d883ddba8eb700ff40bb104e78f3f6d1fcfe31ab4d24c5023
issue-collapse:d962d7f9ab077e6b,module: cuda needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,9,0,0,0,"45745, 53451, 65985, 72859, 93134, 132009, 167304, 169970, 181865",,,,"module: cuda, module: cuda graphs, module: nn, module: rpc, module: serialization, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: profiler, oncall: quantization","PT2-Bug-Bash, bot-triaged, high priority, module: cuda, module: cuda graphs, module: nn, module: rpc, module: serialization, needs research, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: profiler, oncall: quantization, ptd-bot-triaged, triage review, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#45745 Native CUDA-Aware MPI support for MVAPICH2 and other MPI libraries, #53451 Model loaded from full model state checkpoint does not run on multiple GPUs (device mismatch), #65985 torch.cuda.power_usage (ESG 🌱), #72859 Profiler crashes with ProfilerActivity.CUDA on AWS p4d.24xlarge with A100-SXM4-40GB, #93134 Failed to Open libnvrtc-builtins.so.11.7, #132009 2:4 semi structured sparsity matmul not supported on hopper",feb855d97f401dadb4573f39c651d28aeb61f3c7f146deff35c2c1d6f33a7a6a
issue-collapse:e035f60d67bad18b,module: tests needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,8,0,0,0,"56314, 58507, 64119, 64518, 66894, 67646, 83606, 92175",,,,"module: tests, oncall: distributed, oncall: fx, oncall: jit, oncall: profiler","NNC, better-engineering, high priority, module: tests, oncall: distributed, oncall: fx, oncall: jit, oncall: profiler, triage review, triaged",,"Review this module: tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#56314 test_variant_consistency_jit fails for torch.tensordot with dtype float32 with error INTERNAL ASSERT FAILED, #58507 OpInfo JIT tests do not work with Tensor kwarg arguments, #64119 Cleanup tests in rpc_test.py, #64518 TensorPipeDistAutogradTest is frequently failing, #66894 test_nccl_barrier_timeout_new_group_non_member fails intermittently, #67646 test_jit_fuser_legacy and test_jit_fuser_te are failing with SIGIOT",37a77765ea28fd240b90edba40beae3406c5880b895e356fef450a87983d43c3
issue-collapse:3b093908205cea63,module: fx needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,8,0,0,0,"76181, 103947, 119454, 120437, 149186, 149497, 163312, 175493",,,,"module: fx, module: tests, oncall: export, oncall: fx, oncall: jit, oncall: pt2","bot-mislabeled, bot-triaged, export-triaged, feature, fx, high priority, module: fx, module: tests, oncall: export, oncall: fx, oncall: jit, oncall: pt2, triage review",,"Review this module: fx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#76181 Allow `torch.fx` tracing on TorchScript models, #103947 [torch.fx] Deserialization Error - TypeError: ones() received an invalid combination of arguments - got (tuple, device=Attribute), #119454 Python 3.12 CPU test_fx.py::TestFunctionalTracing::test_nn_functional_adaptive_max_pool is failing, #120437 fx::GraphModule::to_folder doesn't seem to handle _PyTreeCodeGen correctly., #149186 torch.fx.symbolic_trace failed on deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B, #149497 symbolic_trace failed on deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",f46189fe44c6bc250d48960fa30d28aa9069dc300cf031846acc0d563229c435
issue-collapse:e0e51f100a081180,module: fsdp needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,7,0,0,0,"91165, 140183, 143536, 169448, 173680, 182988, 183702",,,,"module: fsdp, module: performance, module: rocm, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms","bot-triaged, enhancement, high priority, module: fsdp, module: performance, module: rocm, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Review this module: fsdp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#91165 [FSDP] FSDP with CPU offload consumes `1.65X` more GPU memory when training models with most of the params frozen, #140183 FSDP Hybrid Shard worse loss than than Full Shard, #143536 FSDP learning hangs when the program tries to save the model, #169448 FSDP2 model saving broken unless ignoring frozen params, #173680 Large CPU overhead with FSDP2, #182988 ROCm MI300 (gfx942): native segfault in mixed-precision BatchNorm2d backward",b62b296ddea9d7c6935c35be8d78dd4863ff46eb94d4fa783374a3e86f1134e2
issue-collapse:46e2f7f764ad619f,oncall: fx needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,6,0,0,0,"88574, 91000, 92812, 93002, 98422, 115818",,,,oncall: fx,"high priority, oncall: fx, triage review, triaged",,"Review this oncall: fx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#88574 Add support for `torch.Generator` in the FX IR, #91000 node.stack_trace does not handle escaping correctly, #92812 torch.fx fails to trace through ""+"" op between torch.Size and torch.fx.proxy.Proxy, #93002 Replace pattern fails on incompatible function arguments, #98422 [FX] Symbolic trace over `torch.Tensor.${fn}` APIs, #115818 Fx graph recompilation along with partitioner produces different node order even though graph operations are same",2991cd05baac2d34eaeb1dbc3a8544ece6db0fd157eff2e0e74d06edb1118f35
issue-collapse:2cbb418e9017b223,module: cpp needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,6,0,0,0,"18476, 68049, 73352, 84055, 88301, 89484",,,,"module: cpp, module: crash, module: docs, oncall: jit","module: cpp, module: crash, module: docs, oncall: jit",,"Review this module: cpp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#18476 [CPP] Allow binding config structs into the Python front end, #68049 Request to revise the Pytorch tutorial., #73352 Segfault on unloading a model, #84055 scripted fasterRCNN model cannot be loaded with libtorch c++ API, #88301 1.12.1 incompatible with c++ built for 1.12.0 and vice versa, #89484 kind_.is_prim() INTERNAL ASSERT FAILED at ""../torch/csrc/jit/ir/ir.cpp"":1098",a813991b8b332a8e08c37d11dca0aa99f5d9c38cb31b91f99274118691eccd6b
issue-collapse:fb5957fda6b4292d,module: build needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,6,0,0,0,"7036, 16602, 79003, 96185, 156915, 172348",,,,"module: build, module: crash, module: nccl, module: vulkan, oncall: distributed, oncall: distributed infra, oncall: mobile","bot-triaged, caffe2, module: build, module: crash, module: nccl, module: vulkan, oncall: distributed, oncall: distributed infra, oncall: mobile, ptd-bot-triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#7036 [Caffe2] compilation linker error `libtbb.so.2: undefined reference to std::__exception_ptr`, #16602 Error while installing pytorch, #79003 Libtorch C++ mobile build linking error, #96185 `ld: error: unknown argument '-force_load'` when linking libtorch on Android, #156915 Unable to compile with Vulkan support, #172348 Build failure in NCCLSymmetricMemory.cu: undefined identifier `offset_`",47155860e3f2838c2c9cc14b7e95ca7c96d40240f237078726376f4ffdbd70ea
issue-collapse:80c2c2986fb5f52d,module: dynamo needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,5,0,0,0,"112844, 123499, 145088, 151016, 156558",,,,"module: dynamo, oncall: export, oncall: jit, oncall: pt2","export-triage-review, export-triaged, module: dynamo, oncall: export, oncall: jit, oncall: pt2, pt2-pass-rate-regression",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#112844 torch.export does not support torchaudio.transforms.Spectrogram, #123499 Support Exceptions in Pytorch Export, #145088 [torchbench] torch._dynamo.exc.Unsupported: Graph break due to unsupported builtin None.morphologyEx, #151016 [Dynamo][Export] Untraceable issues when exporting the Stable Diffusion 3.5 model, #156558 Using collections. namedtuple in the forward method of the model resulted in compilation failure",ac661adae50f4d472555f5764609a7ef4aeee078cbe7fe2a55106f4ceffc4c2c
issue-collapse:7b81376692379608,module: regression needs triage,batch action queue,needs triage,97,medium,bucket_component_queue,3,0,0,0,"81115, 100253, 153941",,,,"module: regression, module: vision, oncall: jit, oncall: package/deploy, oncall: profiler","high priority, imported, module: regression, module: vision, oncall: jit, oncall: package/deploy, oncall: profiler",,"Review this module: regression queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#81115 torch.package can not be used to serialize `resnet18` from TorchVision-0.12, #100253 profiler.export_stacks doesn't return stack trace unless experimental_config is provided, #153941 TorchScript Fails Type Inference When Using .item()-Derived Scalar in Tensor Construction",b36918f39969adc66241e4df2fdfed875367c83b46022800c320f3845bee5d97
issue-collapse:d2e623c0b61a2203,module: inductor has linked PR  batch 1,batch action queue,has linked PR,94,medium,bucket_component_queue,14,0,0,0,"141595, 146306, 147075, 179272, 180823, 183899, 185533, 185885, 188213, 188217, 188225, 188226, 188578, 188635",,,,"module: inductor, module: interpolation, module: linear algebra, module: performance, module: pooling, module: regression, module: scatter & gather ops, oncall: pt2","bot-triaged, enhancement, inference mode, module: inductor, module: interpolation, module: linear algebra, module: performance, module: pooling, module: regression, module: scatter & gather ops, oncall: pt2, repro:agent:h100:repro_success:reproduced, topic: fuzzer, triaged",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#141595 FX graph cache does not do provenance tracking for inductor guards, #146306 torch.compile failed on single-node multigpu setting., #147075 [inductor] Performance Degradation and Hang in `torch.diff`, #179272 `torch.compile` returns different output stride/contiguity from eager for upsample_nearest3d on non-contiguous input, #180823 [Inductor] RuntimeError: weakref to UntypedStorage when compiling transpose(-2, -1) under inference_mode, #183899 torch.compile fails on addmm with beta=0 and non-broadcastable bias shape",ecaac068b8e0f112b25157c7b0f2f15895640fa7eb6180542eb33ade02c5db9d
issue-collapse:89638851d24b29ae,module: dynamo has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,14,0,0,0,"117584, 134950, 135011, 147212, 149989, 160757, 161765, 165385, 168307, 182706, 185886, 187593, 188536, 188796",,,,"module: dynamo, module: graph breaks, module: guards, module: library, module: nestedtensor, module: performance, module: pt2 optimizer, module: pt2-dispatcher, module: regression, oncall: distributed, oncall: pt2","bot-triaged, dynamo-autograd-function, dynamo-must-fix, dynamo-nn-modules, dynamo-triage-dec2025, dynamo-triage-june2024, enhancement, good first issue, high priority, module: dynamo, module: graph breaks, module: guards, module: library, module: nestedtensor, module: performance, module: pt2 optimizer, module: pt2-dispatcher, module: regression, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, recompilations, tensor subclass, triaged",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#117584 [pt2d] module register_pre_forward_hook and register_forward_hook triggered graph break when it's root module, #134950 Bert model training is not running on cpu (X86/ARM) with torch.compile() mode, generating value error., #135011 dynamo (re)compilation issues: shape (1,1), nn.Parameter, mark_dynamic, #147212 dict_tag optimization leads to wrong results with relational guards, #149989 [dynamo][higher order ops] Make support_aliasing and support_input_mutation default to False, #160757 Graph break when torch.compiling an autograd.Function",cfc42249c11458d445f106531f988401f07842d2b1ad2f530306d0c7396b063f
issue-collapse:73c7535c14c7a23f,module: docs has linked PR  batch 1,batch action queue,has linked PR,94,medium,bucket_component_queue,14,0,0,0,"32864, 112586, 129319, 129970, 131280, 134964, 170869, 172324, 173318, 180119, 180467, 180649, 188245, 188398",,,,"module: docs, module: dynamic shapes, module: edge cases, module: error checking, module: linear algebra, module: openreg, module: python frontend, module: regression, module: rnn, oncall: pt2","actionable, bot-mislabeled, bot-triaged, high priority, medium, module: docs, module: dynamic shapes, module: edge cases, module: error checking, module: linear algebra, module: openreg, module: python frontend, module: regression, module: rnn, oncall: pt2, release notes: python_frontend, small, topic: not user facing, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32864 [docs] Missing docs for torch.__version__ and torch.version, #112586 Fix docstring errors in _guards.py, _ops.py, _jit_internal.py, functional.py, _tensor_str.py, library.py, #129319 `int` type for `dims` of `tile()` without `dims=` works with a tensor against the doc, #129970 Setting the 0D tensor of only one element to `fill_value` of `full()`/`full_like()` works against error message, #131280 The shape of weights in lstm seems to be wrong., #134964 Add the pages of `torch.e` and `torch.pi` to PyTorch doc",bcca2770b13c22c5122b1eeac5678c82043d0ada1f869cb287b9bce6bf4d3768
issue-collapse:e42231b0cea97949,module: cuda has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,10,0,0,0,"85960, 103584, 153825, 159594, 165650, 170955, 173049, 174625, 181805, 183806",,,,"module: cuda, module: cuda graphs, module: cudacachingallocator, module: docs, module: error checking, module: linear algebra, module: magma, module: nans and infs, module: nn, module: performance, module: python frontend, module: regression, module: sparse","actionable, bot-triaged, enhancement, high priority, module: CUDACachingAllocator, module: NaNs and Infs, module: cuda, module: cuda graphs, module: docs, module: error checking, module: linear algebra, module: magma, module: nn, module: performance, module: python frontend, module: regression, module: sparse, topic: performance, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#85960 Discrepancy in output shape for batch_norm inference mode between CUDA and CPU, #103584 torch.cuda.mem_get_info to return 0 if CUDA context isn't initialized, #153825 'torch.sparse.to_sparse_semi_structured' significantly worsens performance on H100 GPUs, #159594 During CUDA graph capture, CUDACachingAllocator could potentially trigger cudaFree(), #165650 `torch.remainder` with integer divisor 0 should raise, but returns large values in CUDA, #170955 `grouped_mm` does grouping on CPU on 5060 Ti (Windows)",516030c0ba9b178859b34dcd1915f1f4776489bb238810825102be171a9bb540
issue-collapse:e1f5c38455cf18c8,module: crash has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,7,0,0,0,"27614, 184032, 184876, 185482, 187399, 188033, 188231",,,,"module: crash, module: cuda, module: edge cases, module: error checking, module: functorch, module: inductor, module: internals, module: linear algebra, module: mps, module: pybind, module: serialization, oncall: pt2","actionable, bot-triaged, high priority, module: crash, module: cuda, module: edge cases, module: error checking, module: functorch, module: inductor, module: internals, module: linear algebra, module: mps, module: pybind, module: serialization, oncall: pt2, security, topic: fuzzer, triaged",,"Review this module: crash queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#27614 scatter_add allows index tensor that doesn't match input size in forward pass but fails on backward pass, #184032 torch.UntypedStorage.<from_file/torch.load(weights_only=True)/_new_shared_cuda>.new() segfaults via null Allocator dereference at c10/core/StorageImpl.h:84, #184876 SIGSEGV when assigning to nn.Parameter.data inside a torch.func transform, #185482 [inductor]`torch.cond` Branch with sliced view and scatter_add crashes with non-tensor input, #187399 [MPS] torch.baddbmm — INTERNAL ASSERT for size-0 batch dimensions, #188033 [BUG] THPStream_richcompare casts arbitrary objects to THPStream* without type check",9b4e6e390bf3e3134316afde253adaf988b318a2bc8739af33f72f55bf3e2b97
issue-collapse:64a4802d966e1cae,module: pt2-dispatcher has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,6,0,0,0,"133443, 133592, 153044, 158561, 169150, 188358",,,,"module: pt2-dispatcher, module: torchbind, module: user triton, module: vllm, oncall: pt2","PT2-Bug-Bash, actionable, good first issue, high priority, module: pt2-dispatcher, module: torchbind, module: user triton, module: vllm, months, oncall: pt2, triaged, vllm-compile",,"Review this module: pt2-dispatcher queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#133443 [BE] Deduplicate auto_functionalized and triton_kernel_wrapper_functional, #133592 [torchbind x compile] Can't register a torchbind operator that mutates a tensor, #153044 Unexpected float32 overflow for amp training with torch.compile, #158561 `torch.compile` In-place operation raises Autograd error even when gradients are not required, #169150 [debug mode] do not hash collectives, #188358 [Dynamo] Eager and torch.compile behavior mismatch in F.batch_norm with multidimensional bias",32ce050e99ddcee8437e0f740b5e2212f0923fe9283ff8addad010509f41f1aa
issue-collapse:d77850ca1605a921,module: cpu has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,6,0,0,0,"127163, 140000, 146508, 170370, 173946, 181510",,,,"module: cpu, module: crash, module: cuda, module: docs, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: half, module: mps, module: nn, module: performance, module: python frontend, module: rnn, module: sanitizers, module: vectorization","actionable, bot-triaged, low priority, module: cpu, module: crash, module: cuda, module: docs, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: half, module: mps, module: nn, module: performance, module: python frontend, module: rnn, module: sanitizers, module: vectorization, topic: fuzzer, triaged",,"Review this module: cpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#127163 BUG: `torch.pow(-0.0)` wrongly returns `-0.0` (should be +0.0), #140000 torch.argmax slow on CPU, #146508 Something very wrong with float16 CPU implementation, #170370 [MPS][CUDA] EmbeddingBag on CUDA/MPS does not check offsets[0] like CPU, #173946 Segmentation fault in aten.lstm.data results if hx shapes are invalid (when running on CPU), #181510 torch.nn.utils.weight_norm triggers a UBSan null pointer error for nn.Linear(0, 1) on CPU",0abd4ac95e183b5012924491da36e87da90d2ffc3cae4b847215f1496e720c6c
issue-collapse:b22e3d3a50a84d7f,module: mps has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,5,0,0,0,"160739, 179259, 181650, 188198, 188728",,,,"module: mps, module: performance, module: regression, module: tests","bot-triaged, module: mps, module: performance, module: regression, module: tests, needs reproduction, triaged",,"Review this module: mps queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#160739 ConvTranspose 3D with BF16 or FP16 types is not supported on MPS, #179259 test_torch.py does not run any tests on MPS despite number of `skipIfMPS` decorators, #181650 MPS: `add_dense_scalar_cast_float` shader fails with ""Read-only bytes bound to shader argument with write access enabled"" assertion during pyannote.audio inference (M4 Pro, torch 2.11), #188198 [MPS] cat op dropped performance in torch==2.10 release, #188728 [MPS] F.linear does not respect the PYTORCH_MPS_PREFER_METAL flag",61b6c8da3cd8d4b7661298b56d6d2dd594456fed0f839acba91c2eedcf52a6ae
issue-collapse:e252054d583b7647,module: linear algebra has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,4,0,0,0,"101249, 185650, 186148, 187567",,,,"module: linear algebra, module: macos, module: mps, module: performance, module: viewing and reshaping","bot-triaged, enhancement, has workaround, high priority, low priority, module: linear algebra, module: macos, module: mps, module: performance, module: viewing and reshaping, triaged",,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#101249 `einsum` is about 40x slower on CUDA than manually multiplying and summing, #185650 [Bug] torch.kron fails with RuntimeError on non-contiguous inputs, #186148 `torch.matmul` misses `mm` fold for `[S, 1, H] @ [H, K]` when the size-1 batch dim has a non-contiguous stride, #187567 [MPS] torch.linalg.qr is 25-250× slower than CPU on tall-skinny matrices",96b9e281b06dba041ded43441cd32e6c9067caab623ed1657fbd4c16194cccd9
issue-collapse:d2cdf61b0ccc8ffe,module: cpp has linked PR,batch action queue,has linked PR,94,medium,bucket_component_queue,3,0,0,0,"150857, 157776, 186233",,,,"module: cpp, module: crash, module: edge cases, module: serialization, module: tensor creation","actionable, bot-triaged, enhancement, module: cpp, module: crash, module: edge cases, module: serialization, module: tensor creation, triaged",,"Review this module: cpp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#150857 No way to save/load model weights between PyTorch and LibTorch (C++) in a compatible way, #157776 Support for _Float16/C++23 std::float16_t, #186233 C++ API: at::range crashes due to division-by-zero in RangeFactories.cpp",d4e37e1685688028b00216739db41feba3ce874a16ef3d319c089a0e1ced1bb2
issue-collapse:3163a2d8cae07b56,oncall: quantization needs triage  batch 1,batch action queue,needs triage,89,medium,bucket_component_queue,14,0,0,0,"128114, 147818, 159505, 165829, 170169, 171675, 172082, 173075, 179109, 179111, 180995, 181181, 186363, 187048",,,,oncall: quantization,"bot-triaged, oncall: quantization, release-feature-request, topic: docs",,"Review this oncall: quantization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#128114 Import Error: cannot import name 'XNNPACKQuantizer' from 'torch.ao.quantization.quantizer', #147818 Bug in `torch.ao.nn.quantized.Sigmoid` Parameter Restoration after `state_dict ` Loading, #159505 Cannot pass through None for example_inputs in prepare_fx, #165829 Floating Point Exception (Core Dumped) with Dynamic Quantization on nn.Linear with Output Dimension 0, #170169 questions about default_weight_fake_quant config with pytorch quantizations, #171675 PyTorch quantization `QConfig` triggers INTERNAL ASSERT FAILED when printing config with invalid observer `dtype`",fafb95ff039d0dbb6137ecd95b0941d205c316dc969c62083162cff22b0e0916
issue-collapse:1699bc2efd5ea6db,oncall: profiler needs triage  batch 1,batch action queue,needs triage,89,medium,bucket_component_queue,14,0,0,0,"121471, 153614, 153949, 166121, 174273, 174555, 175790, 176288, 178087, 181238, 183842, 186026, 188481, 188490",,,,oncall: profiler,"OSS contribution wanted, bot-triaged, oncall: profiler",,"Review this oncall: profiler queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#121471 [profiler]Not passed shapes for lists of tensors to kineto_result, #153614 Add opentelemetry traces, #153949 Profiler does not capture steps on Blackwell, #166121 [profile] profile_memory doesn't work if _ExperimentalConfig(profile_all_threads = True) is set, #174273 using profiler on jvp function with record_shapes=True causes memory leaks., #174555 Profiler chrome traces express record_function and profiler step incorrectly when using with_stack=True",f0625714e42b08e0e9bc752737c1e9403152fdbc9401d179da033412bd3f3851
issue-collapse:dff3d8ce57e10e77,module: cuda high-priority/blocker  batch 1,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,14,0,0,0,"22259, 47844, 49877, 76191, 91754, 101529, 105872, 107363, 116928, 119000, 119261, 128819, 155607, 179214",,,,"module: cuda, module: cudnn, module: dependency bug, module: linear algebra, module: memory usage, module: multithreading, module: nestedtensor, module: nn, module: numerical-stability, module: nvfuser, module: performance, module: reductions, module: regression, module: tf32","bot-mislabeled, enhancement, high priority, llm-amenable, module: cuda, module: cudnn, module: dependency bug, module: linear algebra, module: memory usage, module: multithreading, module: nestedtensor, module: nn, module: numerical-stability, module: nvfuser, module: performance, module: reductions, module: regression, module: tf32, needs reproduction, quansight-nack, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22259 using multi thread lead to gpu stuck with GPU-util 100%, #47844 Incorrect output loss value under specific CUDA version, #49877 The same tensor requires more memory on RTX3090, #76191 [Feature request] Exclusive prefix sum, `torch.cumsum(input, dim=0, exclusive=True)`, #91754 CUDA error: initialization error, #101529 Inconsistent performance degradation of 3x3 convolution (torch 2.0.1+cu118)",0acff747fb4149d4cbb7587081ec34a25802ec139e639b3114fcad2ac4b0bc97
issue-collapse:f8055b9d6be4e154,module: crash high-priority/blocker  batch 1,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,14,0,0,0,"94765, 113586, 142048, 157704, 175614, 176426, 177714, 178038, 178892, 184880, 185494, 185839, 187343, 188218",,,,"module: crash, module: cublas, module: cuda, module: cudacachingallocator, module: dynamic shapes, module: dynamo, module: inductor, module: linear algebra, module: macos, module: memory usage, module: mps, module: nn, module: pickle, module: pooling, module: regression, module: rocm, module: serialization, module: windows, module: xpu, oncall: pt2","PT2-Bug-Bash, bot-mislabeled, bot-triaged, has workaround, high priority, module: CUDACachingAllocator, module: crash, module: cublas, module: cuda, module: dynamic shapes, module: dynamo, module: inductor, module: linear algebra, module: macos, module: memory usage, module: mps, module: nn, module: pickle, module: pooling, module: regression, module: rocm, module: serialization, module: windows, module: xpu, needs reproduction, oncall: pt2, triaged, upstream triton",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#94765 MPS internal error in `torch.gather` when last dimension is a singleton dimension, #113586 mps bug: failed assertion `[MPSNDArrayDescriptor sliceDimension:withSubrange:] error: subRange.start (8192) is not less than length of dimension[1] (512)', #142048 Device check missing in torch.linalg.solve_triangular leading to hard crash, #157704 Seg faults on macos / OSX, #175614 Segfault during device_map=""auto"" weight materialization on RTX 5090 D v2 (GB202-240, sm_120), #176426 Triton kernels with 2 or more tl.load() calls segfault at runtime on sm_120 (NVIDIA RTX PRO 6000 Blackwell). The kernels compile without errors but produce invalid code that crashes on execution.",8f4044dfe3a207126813cb5b6449905ccb359a879bd790d05d1f10fad3d66f30
issue-collapse:7ea7080a2a9864cb,module: binaries high-priority/blocker  batch 1,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,14,0,0,0,"41661, 88882, 108532, 109014, 120995, 159454, 161358, 161865, 163675, 165448, 168997, 178841, 182949, 186516",,,,"module: binaries, module: build, module: ci, module: cpp-extensions, module: crash, module: cuda, module: dependency bug, module: docker, module: macos, module: mps, module: openmp, module: python version, module: regression, module: rocm, oncall: releng","bot-mislabeled, bot-triaged, has workaround, high priority, module: binaries, module: build, module: ci, module: cpp-extensions, module: crash, module: cuda, module: dependency bug, module: docker, module: macos, module: mps, module: openmp, module: python version, module: regression, module: rocm, needs reproduction, oncall: releng, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41661 libgcc_s.so.1 must be installed for pthread_cancel to work, #88882 Can't import torch --> OSError related to libcublasLt.so.11, #108532 Breaking incompatibility with Cuda 12.2, pytorch stable, torchvision, #109014 Cannot install torchmetrics - ERROR 403, #120995 AttributeError: module 'mpmath' has no attribute 'rational', #159454 libtorch_cpu.so link libm.so issue:GLIBC_2.29 not found",840c29a40bed2ac621ead314ce484fe40bf7d988add7512f2f33a1560e3538a9
issue-collapse:3defa04014a6b894,module: autograd high-priority/blocker,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,12,0,0,0,"28980, 38231, 53796, 60835, 65366, 70108, 73349, 76543, 81532, 120622, 177561, 181534",,,,"module: autograd, module: bootcamp, module: ci, module: codegen, module: correctness (silent), module: cpp, module: crash, module: cuda, module: double backwards, module: fft, module: internals, module: loss, module: multithreading, module: nn, module: norms and normalization, module: performance, module: regression","actionable, bot-triaged, high priority, module: autograd, module: bootcamp, module: ci, module: codegen, module: correctness (silent), module: cpp, module: crash, module: cuda, module: double backwards, module: fft, module: internals, module: loss, module: multithreading, module: nn, module: norms and normalization, module: performance, module: regression, needs reproduction, shadow review, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#28980 Unsafe use of `at::parallel_for` in current codebase, #38231 Python autograd engine threads never terminate in Python 3.5-3.8, #53796 foreach ops in autograd, #60835 Pytorch V1.1 runs OK, but Pytorch v1.5 to 1.9 run wrong, #65366 Exception thrown in final autograd callback (queue_callback) not caught if not on CPU thread: terminate called after throwing an instance of 'python_error', #70108 Derivative for _ctc_loss_backward",f7171c79fa0d22ffe2ed10ad67b6a6cd2bf006b8d1d3e1381a54745ed96b417a
issue-collapse:6cbb3f2ef3c3d48b,module: cpu high-priority/blocker,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,11,0,0,0,"42784, 73174, 73185, 73186, 113037, 125247, 134198, 152979, 171606, 175436, 177906",,,,"module: cpu, module: crash, module: determinism, module: edge cases, module: error checking, module: intel, module: macos, module: mkl, module: multithreading, module: norms and normalization, module: openmp, module: performance, module: pooling, module: regression, module: third_party, module: windows","bot-triaged, high priority, matrix multiplication, module: cpu, module: crash, module: determinism, module: edge cases, module: error checking, module: intel, module: macos, module: mkl, module: multithreading, module: norms and normalization, module: openmp, module: performance, module: pooling, module: regression, module: third_party, module: windows, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#42784 libtorch 1.5 crashes when used on macs when using torch::max without AVX support, #73174 `LayerNorm` triggers INTERNAL ASSERT, #73185 Segmentation fault in fractional_max_pool2d, #73186 Segmentation fault in fractional_max_pool3d, #113037 `TensorIteratorBase::is_scalar` return `false` for empty numpy tensors, but true for empty Torch ones, #125247 RuntimeError: could not create a primitive descriptor for a matmul primitive",151e180ef8b39017fe562ac349e68d948500d05192046af14d9563f9ea69d8b5
issue-collapse:c9c753e38986a3aa,module: multiprocessing high-priority/blocker,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,4,0,0,0,"53625, 60626, 67247, 180458",,,,"module: multiprocessing, module: regression, module: serialization, module: tests, module: windows, oncall: distributed","bot-triaged, high priority, module: multiprocessing, module: regression, module: serialization, module: tests, module: windows, needs reproduction, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#53625 torch.arange() issue, #60626 Impossible to raise the limit on number of shared memory tensors, #67247 pytorch_linux_xenial_py3_6_gcc5_4_test may timeout during test_multiprocessing_spawn, #180458 [Bug] ValueError: I/O operation on closed file during torch.save() on Windows (PyTorch Nightly + CUDA 13.2)",2de0314f99b756452841130fe8f565694b1ff9a36c9143bac5d201372506c080
issue-collapse:3fe282ced83d65dc,module: performance high-priority/blocker,batch action queue,high-priority/blocker,82,medium,bucket_component_queue,3,0,0,0,"160296, 169823, 172537",,,,"module: performance, module: regression, oncall: cpu inductor","high priority, module: performance, module: regression, needs reproduction, oncall: cpu inductor, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#160296 [inductor][cpu] Performance regression in 2025-08-02 nightly release, #169823 Abnormal CPU usage with ComfyUI, #172537 [Regression] 20x slower vision model forward pass in PyTorch 2.9.1 vs 2.8.0",5ad9904680218dcd6860d0eec8c82b8de82ecc595f9d7ed8920a3630d94cd6c0
issue-collapse:0df92aff5f61058d,oncall: distributed needs triage  batch 2,batch action queue,needs triage,77,medium,bucket_component_queue,14,0,0,0,"76039, 124007, 130515, 166301, 181086, 184147, 186591, 187432, 188266, 188304, 188567, 188833, 189228, 189308",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: distributed parallelisms","bot-triaged, enhancement, feature, needs reproduction, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76039 [RFC] NPU device for PyTorch, #124007 gpu memory normal but one of gpus GPU-Util is 0 so cant train, #130515 [Bug Report] inconsistent global src and local rank in dist.nn.functional.broadcast, #166301 Improve Docs: Reproducibility and Randomness in dist training with multi-process data loading and explicit RNG seed, #181086 [Symmetric Memory] symm buffer init and rendezvous cannot be assigned to a specific CUDA stream, #184147 [AMD] MoRI support for MoE EP Combine/Dispatch torch.distributed.TokenSwitch Interface & AMD Pollara Support",55ba615e44718435ee74642a64f86d1ba28ea8cc38af7fc8f105c88367f34854
issue-collapse:055eb7e697681887,module: python version needs triage,batch action queue,needs triage,77,medium,bucket_component_queue,2,0,0,0,"124190, 172254",,,,"module: python version, oncall: jit, oncall: profiler","module: python version, needs reproduction, oncall: jit, oncall: profiler",,"Review this module: python version queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#124190 Re-enable skipped 3.12 dynamo-wrapped profiler tests, #172254 Python 3.14 test error: `AttributeError: 'MobileNetV2' object has no attribute '__annotations__'.`",1c4b68bb3fbcf48b4a347fcbacd175f9a07340a4c29dd42ca5f9d7f721f2fee5
issue-collapse:6fa8c83dc6739501,module: ci has linked PR,batch action queue,has linked PR,76,medium,bucket_component_queue,3,0,0,0,"150779, 172321, 175994",,,,"module: ci, module: openreg, module: regression, module: testing, module: tests","actionable, better-engineering, bot-triaged, high priority, module: ci, module: openreg, module: regression, module: testing, module: tests, triaged",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#150779 Decorator `skipIfXpu` disables tests when used on class, #172321 [OpenReg] Extend test_autograd.py coverage, #175994 CI: Dead CentOS code in docker setup scripts",f805e2f3812ce07eba96204fb5fadfbe9b327f1e7cbc1299e68f9d91f48bf7c7
issue-collapse:26b8fad16d05da38,module: 64-bit has linked PR,batch action queue,has linked PR,76,medium,bucket_component_queue,2,0,0,0,"167086, 170798",,,,"module: 64-bit, module: correctness (silent), module: cuda, module: edge cases, module: nn","actionable, high priority, module: 64-bit, module: correctness (silent), module: cuda, module: edge cases, module: nn, triaged",,"Review this module: 64-bit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#167086 Unchecked uint32 overflow in CUDA scan (cumsum/cumprod/cummin/etc), #170798 nn.functional.pad silently changes the padding value if input dtype is int64 and padding value is larger than 2^53",db3f4d111dc668362c9a4c490c8319d28705835ebb92abd545fa0b985d67bed5
issue-collapse:83dfe543f13e94d5,module: ci high-priority/blocker  batch 1,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,14,0,0,0,"54418, 64794, 81127, 82894, 83393, 88842, 123219, 137098, 148699, 158728, 161295, 164562, 172907, 187951",,,,"module: ci, module: cpu, module: cuda, module: inductor, module: intel, module: linear algebra, module: numba, module: python frontend, module: regression, module: testing, module: tests, module: typing, module: vllm, oncall: pt2","PT2-Bug-Bash, actionable, bot-triaged, high priority, module: ci, module: cpu, module: cuda, module: inductor, module: intel, module: linear algebra, module: numba, module: python frontend, module: regression, module: testing, module: tests, module: typing, module: vllm, oncall: pt2, triaged, unstable",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54418 Some Numba tests are failing, #64794 linalg.lstsq out variant fails internal assert because it uses non-inplace view op for some inputs, #81127 Docker updates cause subsequent builds to fail, #82894 linalg and lu tests fail when run in parallel on linux cuda, #83393 Test public bindings in CI gives weird output on error, #88842 test/test_ops.py is segfaulting on master build with DEBUG assets",dbab11fadc1e19469f9d9d13fc5af554784ec90a9f47ca06e5f59f1618ed3ee8
issue-collapse:9a2e6a8aa1d0c5eb,module: build high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,14,0,0,0,"35597, 38122, 51332, 80012, 94544, 98503, 129304, 135637, 146551, 153733, 156613, 162884, 164957, 178939",,,,"module: build, module: cpp, module: crash, module: cuda, module: doc infra, module: docs, module: functorch, module: linear algebra, module: magma, module: mkl, module: mps, module: power, module: regression, module: third_party","actionable, bot-mislabeled, bot-triaged, high priority, module: POWER, module: build, module: cpp, module: crash, module: cuda, module: doc infra, module: docs, module: functorch, module: linear algebra, module: magma, module: mkl, module: mps, module: regression, module: third_party, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35597 libtorch_global_deps.so not found., #38122 Pybind11 cpp extensions broken with pytorch v1.5.0, #51332 Cant compile from source 1.8.0a0 because 'magma_v2.h' file not found and undefined references to `magma_*', #80012 static builds are broken by MKL_DNN, #94544 Abort Caused by Virtual Function, #98503 Power VSX vectorization support disabled",677bf199788d90e60569861465e5700e1e138d1d88240b582fd88161207c75b3
issue-collapse:78b47f8c88326688,module: correctness (silent) high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,10,0,0,0,"58005, 65815, 73792, 107668, 132086, 154312, 163539, 165123, 178880, 187319",,,,"module: correctness (silent), module: crash, module: cuda, module: cudnn, module: custom-operators, module: distributions, module: fx, module: inductor, module: linear algebra, module: macos, module: memory usage, module: mps, module: multithreading, module: nans and infs, module: numerical-stability, module: pt2-dispatcher, module: reductions, module: regression, oncall: distributed, oncall: pt2","bot-triaged, has workaround, high priority, module: NaNs and Infs, module: correctness (silent), module: crash, module: cuda, module: cudnn, module: custom-operators, module: distributions, module: fx, module: inductor, module: linear algebra, module: macos, module: memory usage, module: mps, module: multithreading, module: numerical-stability, module: pt2-dispatcher, module: reductions, module: regression, oncall: distributed, oncall: pt2, repro:agent:h100:repro_success:reproduced, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#58005 torch.distributed.nn.all_reduce incorrectly scales the gradient, #65815 Overflow error in torch.linalg uncaught, #73792 Bug: torch.distributions.mixture_same_distribution._pad_mixture_dimension, #107668 torch.dot gives wrong result on Macos, #132086 Result of MPS memory operation can be corrupted if memory pressure is high, #154312 torch.logdet produces incorrect results for singular matrices on CUDA vs CPU",c2801d84649090510b3a553c1360b2fcafa0f001c4653277bec6a1b73db2b3a7
issue-collapse:c0a8e28eff48a50f,module: arm high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,10,0,0,0,"78924, 136299, 141083, 146857, 151680, 170787, 176285, 178726, 186952, 188336",,,,"module: arm, module: build, module: ci, module: correctness (silent), module: crash, module: cuda, module: dataloader, module: deadlock, module: inductor, module: macos, module: memory usage, module: performance, module: regression, module: xnnpack, oncall: cpu inductor, oncall: pt2","arm priority, bot-mislabeled, bot-triaged, high priority, module: arm, module: build, module: ci, module: correctness (silent), module: crash, module: cuda, module: dataloader, module: deadlock, module: inductor, module: macos, module: memory usage, module: performance, module: regression, module: xnnpack, oncall: cpu inductor, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#78924 If large enough tensor is being cloned, parallel dataloading hangs on M1 Mac, #136299 The results of baddbmm on CPU seem to have issues under certain conditions, #141083 ARM build failed with recent XNNPACK update: third_party/XNNPACK/src/reference/unary-elementwise.cc:125:14: error: invalid ‘static_cast’ from type ‘xnn_bfloat16’ to type ‘_Float16’, #146857 Segmentation Fault nn/test_convolution.py test_ConvTranspose2d_output_size_downsample_upsample on Aarch64, #151680 undefined reference to `at::native::blas_impl::fp16_gemv_notrans(int, int, float, c10::Half const*, int, c10::Half const*, int, float, c10::Half*, int)', #170787 [ARM] Unit test regression AArch64 - int4mm kernels require bf16 hw support",26801578617b65a3ce452a08250fe97a25c1b0a13be75fe36e25fc4a1552ece5
issue-collapse:66900b34cd72ebd9,module: regression high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,9,0,0,0,"45136, 68920, 73488, 86704, 114529, 150187, 152292, 152747, 175928",,,,"module: regression, module: rocm, module: serialization, module: sparse, module: tests, module: typing, module: vllm, module: xnnpack, oncall: pt2, oncall: releng","bot-triaged, high priority, llm-amenable, module: regression, module: rocm, module: serialization, module: sparse, module: tests, module: typing, module: vllm, module: xnnpack, needs reproduction, oncall: pt2, oncall: releng, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#45136 Multiple torch.load in one file, #68920 Breakes with -OO flag in script, #73488 After XNNPack update `TestXNNPACKSerDes.test_linear` started to fail, #86704 Performance tests mnist_hogwild-cpu_memory CPU memory increase by 30%, #114529 Bug in element-wise multiplication of `torch.sparse_csr_tensor`s on GPU - 0's in result considered significant - PyTorch 2.1.1, #150187 at::BlasBackend::Ck does not handle all ROCm BLAS gpus",df20c12441ee468cb7fb0a95bf40e1250f4de81e5c2d95ee9ddba567705b7a76
issue-collapse:14461f1dd78ab518,module: cpp high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,7,0,0,0,"52663, 53895, 71391, 83577, 106700, 140272, 186212",,,,"module: cpp, module: cpu, module: crash, module: cuda, module: nn, module: norms and normalization, module: python frontend, module: regression, module: static linking","bot-triaged, high priority, module: cpp, module: cpu, module: crash, module: cuda, module: nn, module: norms and normalization, module: python frontend, module: regression, module: static linking, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#52663 code linking to libtorch cannot use thrust/cub functions, #53895 terminate called after throwing an instance of 'c10::Error' what(): isTuple() INTERNAL ASSERT FAILED at ""/home/wenda/libtorch/include/ATen/core/ivalue_inl.h"":927, please report a bug to PyTorch. Expected Tuple but got GenericList, #71391 Calling .backward() inside of an LBFGS closure function throws an exception in Libtorch v1.6.0+, #83577 libtorch malloc cause coredump, #106700 C++ API `torch::nn::MultiheadAttention` Crashes by division by zero, #140272 There is probably a destructor ordering accident going on with registerPythonModule",4d3eb89f43679248e5377403bb1868a220b3208b8f27c47e01969c3b7377f591
issue-collapse:ec12cf870ca98b0d,module: convolution high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,5,0,0,0,"48400, 73179, 142209, 143481, 184027",,,,"module: convolution, module: cpu, module: crash, module: dependency bug, module: edge cases, module: empty tensor, module: error checking, module: inductor, module: intel, module: mkldnn, module: nn, module: nnpack, oncall: pt2","bot-triaged, high priority, module: convolution, module: cpu, module: crash, module: dependency bug, module: edge cases, module: empty tensor, module: error checking, module: inductor, module: intel, module: mkldnn, module: nn, module: nnpack, oncall: pt2, repro:agent:h100:repro_fail:not_reproduced, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#48400 conv_transpose3d returns different result when the input and kernel are mkldnn tensors, #73179 Floating point exception in _nnpack_spatial_convolution, #142209 Segmentation fault (core dumped) in `convolution`, #143481 Segmentation fault (core dumped) in `conv3d`, #184027 [inductor] `torch.compile` crashes on convolutions when weight is a plain tensor — stride prediction mismatch",9f5867afce34241d6b7279885220e4a8930938e658eec1a2a40cfb43dc147d4c
issue-collapse:c6f897f780bcb556,module: inductor high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,4,0,0,0,"181362, 185346, 188127, 188835",,,,"module: inductor, module: memory usage, module: numerical-stability, module: performance, module: pt2 accuracy, module: regression, module: rocm, oncall: pt2","bot-triaged, high priority, module: inductor, module: memory usage, module: numerical-stability, module: performance, module: pt2 accuracy, module: regression, module: rocm, oncall: pt2, triaged, upstream triton",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#181362 DISABLED deit_tiny_patch16_224.fb_in1k fail_accuracy on ROCm dynamic_inductor_timm_training, #185346 [Inductor] Performance regression of phlippe_densenet model due to non-fused Cat kernels from new fusion mode, #188127 torch.compile causes persistent RSS growth during repeated inference in PyTorch 2.9 with ROCM, #188835 DistillGPT2 fails accuracy on A100 in inductor smoke test",e1fe708116b620c42fd2fd23d6ffc419a52d3313bce0fe42b10a29acb9278d64
issue-collapse:7cfbc6849ba5bca6,module: cpp-extensions high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,4,0,0,0,"63749, 89006, 143780, 163501",,,,"module: cpp-extensions, module: crash, module: dispatch, module: dynamic shapes, module: pybind, module: regression","PT2-Bug-Bash, has workaround, high priority, llm-amenable, module: cpp-extensions, module: crash, module: dispatch, module: dynamic shapes, module: pybind, module: regression, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#63749 Segmentation fault when using C++/pybind11 module without also importing torch, #89006 Add smoke-tests for CPP extensions compilations, #143780 Looking for valid compiling option for extension based on torch-2.1.0+cpu.cxx11.abi, #163501 Improve Registration Error message with const SymIntArrayRef&in PyTorch 2.5+",95c80f38ee731254a9e2e0f4f377fc6d8591949b3fdbca725ad2a438108f01c4
issue-collapse:495c36db525c89a9,module: deadlock high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,3,0,0,0,"48382, 153050, 186174",,,,"module: deadlock, module: docs, module: error checking, module: macos, module: multiprocessing, module: regression, module: rocm, module: sdpa","bot-triaged, has workaround, high priority, module: deadlock, module: docs, module: error checking, module: macos, module: multiprocessing, module: regression, module: rocm, module: sdpa, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#48382 hangs indefinitely at os.waitpid(), #153050 Process never ends when sending tensors through multiprocessing queues in Python 3.12+ with filesystem strategy, #186174 ROCm: Flash/mem-efficient SDPA hangs during batched autoregressive generation with left-padded sequences",2480b4f761e7ba2663e5af291ad77a1b8af99796eeffe892541ecb03f179fc58
issue-collapse:91d3b7b2e7878f5e,module: complex high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,3,0,0,0,"46155, 146286, 188864",,,,"module: complex, module: correctness (silent), module: crash, module: cuda, module: error checking, module: fft, module: mps","bot-triaged, complex_autograd, high priority, module: complex, module: correctness (silent), module: crash, module: cuda, module: error checking, module: fft, module: mps, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#46155 Complex backward returns NaN values, #146286 Non-blocking GPU to CPU copy of complex numbers with the different conj status produces wrong results, #188864 [MPS] torch.fft.hfft aborts on 5D complex64 input with dim=0",b36170368df37bbe1ad47cef3413b44f9ab52f584b0643849868dc8dc38e3a6a
issue-collapse:4c75188ef435b084,module: activation checkpointing high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,3,0,0,0,"77284, 82293, 139989",,,,"module: activation checkpointing, module: dynamo, module: fsdp, module: higher order operators, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","PT2-Bug-Bash, bot-triaged, dynamo-triage-jan2025, enhancement, high priority, module: activation checkpointing, module: dynamo, module: fsdp, module: higher order operators, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#77284 non-rentrant checkpointing uses same memory as non-checkpointed code, #82293 pytorch's checkpoint_wrapper does not save memory while fairscale's checkpoint_wrapper saves huge memory, #139989 Support AC with graph break",7c25da69a0dc32e57bb6cddd143ef2e7e5986af3077649679936a721eb2ff81a
issue-collapse:dfdefdd5a620caa1,oncall: pt2 high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"161419, 188332",,,,oncall: pt2,"high priority, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#161419 mobilenetv3_large_100 aot_eager accuracy failure, #188332 Inductor: allgather fusion creates rank-specific cache that crashes when loaded by other ranks",4af780109dc2e59ed76cdd12cc2d2b186a88e708d167caf7c9413674ee4cc193
issue-collapse:d0bbe8a52e6c5bd9,module: nccl high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"18300, 152302",,,,"module: nccl, module: regression, oncall: distributed","high priority, module: nccl, module: regression, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#18300 NCCL backend fails when calling broadcast from different threads, #152302 NCCL out of memory error after updating to PyTorch 2.7",e5c65b38b85d777904431593a0a01bb73ddd63b9215587144327af14629f0712
issue-collapse:d1f5cb53592bd868,module: bfloat16 high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"155393, 184431",,,,"module: bfloat16, module: correctness (silent), module: crash, module: edge cases, module: numerical-stability, module: regression, module: sdpa, module: vllm","bot-triaged, high priority, module: bfloat16, module: correctness (silent), module: crash, module: edge cases, module: numerical-stability, module: regression, module: sdpa, module: vllm, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#155393 crash in torch.histc, #184431 [vllm] [2.12 regression][multimodal] Qwen2-Audio text-then-audio_embeds: prompt_embeds vs raw-text outputs diverge under --enforce-eager",f80ef389df1deaa13cb79a210a9891759b2dec273a343def78bff198f974f495
issue-collapse:9d2ab9ad91575eed,module: aotdispatch high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"179804, 181869",,,,"module: aotdispatch, module: autograd, module: inductor, module: pickle, module: pt2-dispatcher, module: regression, module: vllm, oncall: pt2","bot-triaged, high priority, module: aotdispatch, module: autograd, module: inductor, module: pickle, module: pt2-dispatcher, module: regression, module: vllm, oncall: pt2, pt2: numerics-discrepancy, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#179804 torch.compile + aot_eager: backward ignores inner autocast(enabled=False) vs eager, #181869 [vllm] [torch nightly] AOTAutogradCache.save: _pickle.PicklingError on triton launcher (DFlashDraftModel speculative-decode path)",bfb3690b3c936e71250a0bdb1aa646a65a84e2368fb14d3bb34d5327b11dea99
issue-collapse:27bd42a9bd363c27,module: amp (automated mixed precision) high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"47056, 188955",,,,"module: amp (automated mixed precision), module: inductor, module: memory usage, module: performance, module: regression, oncall: pt2","bot-triaged, module: amp (automated mixed precision), module: inductor, module: memory usage, module: performance, module: regression, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#47056 Training slows down and memory usage increases when upgrading from PyTorch 1.6 to 1.7, #188955 rms_norm autocast fp32 policy (2.12) materializes standalone cast kernels under torch.compile when dtypes already match. Peak-memory and throughput regression for bf16 models",4e7a72eb216dafd33592bb6b4667a5284838a0be8f3180b2327661ab45dd6b2e
issue-collapse:86dc240770a6053e,module: 64-bit high-priority/blocker,batch action queue,high-priority/blocker,70,medium,bucket_component_queue,2,0,0,0,"155293, 176784",,,,"module: 64-bit, module: crash, module: cuda, module: edge cases, module: mps, module: nn, module: random","bot-triaged, module: 64-bit, module: crash, module: cuda, module: edge cases, module: mps, module: nn, module: random, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#155293 [MPS] 5D 4Mln rng attempts fail with internal error, #176784 [CUDA] AdaptiveAvgPool2d: START_IND macro causes int32 overflow → OOB read",d55682950266b20eee37eee11acf886b29aba8ca0bfeca85ca1940acac57acc4
issue-collapse:d95e0091481d9d18,oncall: visualization needs triage  batch 1,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"39250, 51445, 65449, 65588, 68927, 68940, 72906, 75673, 80863, 83589, 90374, 93350, 174375, 184078",,,,oncall: visualization,"Stale, bot-triaged, enhancement, oncall: visualization",,"Review this oncall: visualization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39250 SummaryWriter `add_hparams` should support adding new hyperparameters, #51445 Saving embedding vectors as TSV for Tensorboard is not efficient, #65449 Make SummaryWriter.add_image supports PIL image directly, #65588 Make torch.utils.tensorboard.add_scalar accepts float global_step, #68927 Thank you for the diagnostic script., #68940 Breakpoint training in a Tensorboard log",d3d1a42d47a68f136657443d4e5d0200fbd1817a68ff4e8a913399abe4ca47b4
issue-collapse:2b55c570a9c147a5,oncall: package/deploy needs triage  batch 1,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"65305, 74964, 75467, 76682, 76683, 76685, 78172, 83143, 88735, 138341, 139297, 152532, 155321, 180307",,,,oncall: package/deploy,"bot-triaged, enhancement, imported, oncall: package/deploy",,"Review this oncall: package/deploy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#65305 Add support for HF NLP models in torch.package/deploy, #74964 Add Unit Tests for Torch.Package Subclasses [Waiting on changes to land], #75467 torch.package: log.info when loading packages w/ externed modules + documentation, #76682 [torch::deploy] Remove manager_ from AquireSession, #76683 [torch::deploy] remove reliance on manager_ for unload, #76685 [torch::deploy] move create_movable to interpreter_manager",5bb11314957ec2272f31a0423cd24e33ce80c7d1efc487fd41722dc517cb177b
issue-collapse:337129ee5d850364,oncall: mobile needs triage  batch 1,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"45398, 63476, 76807, 81552, 96639, 114102, 114283, 116795, 118069, 120976, 127013, 129633, 143565, 186670",,,,oncall: mobile,"bot-triaged, oncall: mobile",,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#45398 build_android.sh, #63476 Error loading Pytorch lite model in Android, #76807 There is a bug with latest stable torch version and the following Nightly versions related to `optimize_for_mobile`, #81552 Support Swift Package Manager (SPM) for iOS, #96639 Completely different output between .pt and .ptl, #114102 ""aten::prelu"" can not be used in the metal backend",3d9295341e47c9fb3b3ab6a7e008f94a86dd7c9c1f549f839f51cb2027f22e9f
issue-collapse:4e0b53ae3964d313,oncall: jit needs triage  batch 2,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"16592, 33692, 60930, 61210, 106846, 175708, 177239, 178044, 181231, 181473, 181480, 185440, 185442, 185443",,,,oncall: jit,"OSS contribution wanted, actionable, bot-triaged, good first issue, has workaround, needs reproduction, oncall: jit, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16592 [JIT] NVRTC unknown error, #33692 JIT tracing check fails with boolean tensor modifications, #60930 error for using faster rcnn on c++ & GPU, #61210 Jitted function cannot be pickled, #106846 untimeError: The following operation failed in the TorchScript interpreter. Traceback of TorchScript (most recent call last): RuntimeError: nvrtc: error: invalid value for --gpu-architecture (-arch), #175708 O(N * B) loop in _C._jit_pass_lint (potentially O(N^2))",419497de43d0debd83ee00b8a96b487241b20ddfd3977b232687d2fc82639af3
issue-collapse:90f8fae1c73cfe5f,oncall: distributed needs triage  batch 3,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"66046, 91535, 102186, 123983, 174669, 175145, 175724, 176276, 177371, 178191, 178756, 179770, 180550, 184010",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: distributed parallelisms","bot-triaged, enhancement, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triage review, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#66046 PicklingError when saving a ddp module with torch.save(), #91535 When dist.broadcast float32 to int64, it will silently generate wrong results, #102186 torch/distributed/_spmd/api.py should aot_module_export instead of make_fx directly, #123983 Segfault in TCPStore and FileStore compare_set(), #174669 `TypeError: cannot pickle code objects` with Python 3.13, #175145 [torch.disitributed] Persistent Deadlock when overlapping NCCL P2P and Collectives from separate Python Threads",7e5ca1df7c1bc3bef32faad33213e181d9bcf0ceef9de51632a21b46c0a05d55
issue-collapse:df00345cd4156791,oncall: cpu inductor needs triage  batch 1,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"134679, 137691, 147919, 148071, 148382, 149501, 151525, 151536, 152921, 160391, 162135, 167199, 183888, 188227",,,,"oncall: cpu inductor, oncall: pt2","bot-triaged, oncall: cpu inductor, oncall: pt2",,"Review this oncall: cpu inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#134679 [CPU] jx_nest_base float32 both inductor and eager performance regression in 2024-08-26 nightly release, #137691 [inductor][cpu]float32 and AMP performance regression in 2024-09-30 nightly release, #147919 [inductor][cpu]AOT inductor AMP static shape default wrapper occupied almost 3x disk than before, #148071 [inductor][cpu] maml performance regression in 2025-02-24 nightly release, #148382 [Inductor-CPU] Debug util request: fine-grained mechanism to disable out-of-template epilogues, #149501 Inductor produce significantly different inference results with the originl original model",2cca08a0d798f1b0e1056391579b5ecf94ed435f18b351aec242e7db0dfc91c6
issue-collapse:dbb2c840d2089e19,module: tensorboard needs triage  batch 1,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"30525, 36718, 39292, 40500, 42379, 45327, 46907, 64258, 75502, 90915, 101962, 125033, 144148, 173311",,,,"module: tensorboard, oncall: distributed, oncall: profiler, oncall: visualization","bot-triaged, module: tensorboard, needs reproduction, oncall: distributed, oncall: profiler, oncall: visualization, ptd-bot-triaged, triage review",,"Review this module: tensorboard queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30525 [Tensorboard] nothing appears in the projector tab, #36718 Support alpha channel in tensorboard.add_figure, #39292 Tensorboard，graph in pytorch 1.4 is more complicated than pytorch 1.1？, #40500 Not able to launch tensorboard using pytorch, #42379 TensorBoard Summaries Written with `SummaryWriter.add_scalars()` are not Purged, #45327 Tensorboard summary_iterator",ce5e3b76220c8bbd541e308e6c74f7f8186716097cc81131d1fdbf4067464970
issue-collapse:e326612a866cb620,module: c10d needs triage  batch 2,batch action queue,needs triage,65,medium,bucket_component_queue,14,0,0,0,"52607, 59612, 62007, 62283, 62300, 62314, 68088, 69594, 70054, 70155, 71049, 130407, 177959, 179848",,,,"module: c10d, oncall: distributed, oncall: distributed infra","Stale, bot-triaged, feature, module: c10d, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#52607 Enable timed barrier in c10d, #59612 Why Pytroch.distributed does not expose NCCL cuda stream?, #62007 Add Future return from default process group collective communications, #62283 c10d and discontiguous tensors with mismatch strides, #62300 ProcessGroupGloo CUDA Comm Synchronization Looks Wrong, #62314 dist.destroy_process_group() don't release master_port when there're at least 2 process groups",f4e9fc0190ad3cdf18ed19007df852ee65559d16e815be9771a82d14d6d71ede
issue-collapse:ad6bdf93dc48c5e3,module: docs needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,13,0,0,0,"6465, 37442, 46574, 52517, 54599, 67111, 67392, 74221, 87268, 92855, 95560, 130958, 139063",,,,"module: docs, oncall: distributed, oncall: distributed infra, oncall: jit, oncall: package/deploy, oncall: profiler, oncall: quantization","bot-triaged, caffe2, docs-hackathon, imported, module: docs, oncall: distributed, oncall: distributed infra, oncall: jit, oncall: package/deploy, oncall: profiler, oncall: quantization, ptd-bot-triaged, security, triage review, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#6465 [caffe2] ChannelShuffle example, #37442 Tensor.is_distributed not documented, #46574 Mismatch between the documentation and implementation of caffe2 PReLU operator, #52517 [doc] JIT-compatible c++ custom ops, #54599 Better documentation for `torch.jit.isinstance`, #67111 [doc] Can we update TorchScript union support?",64bf2b8b11ddb8874f516f1f5ead6c516a729e3dc43b7701fee3323e1d4945a9
issue-collapse:0e9b415418c0fe26,module: nccl needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,10,0,0,0,"52471, 68893, 70267, 86849, 109941, 114396, 129341, 137016, 172873, 178138",,,,"module: nccl, module: openmp, module: symm_mem, oncall: distributed, oncall: distributed infra","bot-triaged, feature, has workaround, module: nccl, module: openmp, module: symm_mem, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",,"Review this module: nccl queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#52471 all reduce hangs and does not throw exception when CUDA_VISIBLE_DEVICES is not set properly using the NCCL backend, #68893 NCCL Network is unreachable / Connection refused when initializing DDP, #70267 Inconsistent multi-node latency with NCCL and OpenMPI, #86849 `torch.distributed.all_reduce` allocates excess GPU memory when using NCCL backend, #109941 Need latest NCCL support to reduce GPU HBM consumption, #114396 is PGNCCL abortCommsFromMap working correctly?",fd97404340868a452e31c891c06f85ae64222bda7d48e41bc97a88d268d565f7
issue-collapse:6072d6ee18da4ebb,module: onnx needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,7,0,0,0,"47887, 74799, 120166, 133227, 134303, 141854, 172561",,,,"module: onnx, oncall: export, oncall: jit, oncall: pt2, oncall: quantization","module: onnx, needs reproduction, oncall: export, oncall: jit, oncall: pt2, oncall: quantization, onnx-needs-info, onnx-triaged",,"Review this module: onnx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#47887 ONNX export of a scripted submodule fails with ""Modules that are called during a trace must be registered as submodules of the thing being traced"", #74799 Irrelevant warning during ONNX export of torch.jit.ScriptFunction: Model has no forward function, #120166 QAT using pytorch-quantization cause accuracy loss after exporting to onnx, #133227 Issue Exporting PyTorch Model with pytorch3d.ops.ball_query to ONNX: Bug or Compatibility Problem?, #134303 [FakeTensor] Error: SymIntArrayRef expected to contain only concrete integers in UpSample2D block, #141854 RuntimeError: quantized::conv2d_relu (ONEDNN): data type of input should be QUint8.",63eeb814404fbc5bac0320f8801d18e85f328267902ab845b81256cf0639d939
issue-collapse:bdfb2139c81baa13,module: autograd needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,5,0,0,0,"49157, 57457, 73972, 74590, 166577",,,,"module: autograd, module: cuda, module: data parallel, module: dataloader, module: forward ad, module: hub, module: linear algebra, module: mta, module: multiprocessing, module: nn, module: optimizer, module: pipelining, module: rpc, module: sparse, module: tests, module: viewing and reshaping, oncall: distributed, oncall: distributed parallelisms, oncall: jit, oncall: quantization, oncall: r2p","actionable, bot-triaged, module: autograd, module: cuda, module: data parallel, module: dataloader, module: forward ad, module: hub, module: linear algebra, module: mta, module: multiprocessing, module: nn, module: optimizer, module: pipelining, module: rpc, module: sparse, module: tests, module: viewing and reshaping, needs reproduction, oncall: distributed, oncall: distributed parallelisms, oncall: jit, oncall: quantization, oncall: r2p, onnx-triaged, ptd-bot-triaged, triage review, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#49157 softmax fail, #57457 Valgrind Shows Potential Memory Leaks, #73972 Forward_AD and Torchscript Functions results in Nones or wrong values., #74590 [CTA] Let's Stamp Out Flaky Tests!, #166577 Pipelining - cannot calculate dWeight if a custom torch.autograd.Function is used with a schedule that performs separate dWeight and dInput backward passes",80d67fbf8df06996b55efde9b5e78a2ee3513f66a231b203dc703faada837ffa
issue-collapse:e6d6aec9c393cd4f,module: rpc needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,4,0,0,0,"51616, 66624, 86962, 106711",,,,"module: rpc, oncall: distributed, oncall: distributed infra","better-engineering, bot-triaged, module: rpc, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review",,"Review this module: rpc queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#51616 [RPC] rref.get_type() should have default return type of future, #66624 Separate libtorch and non-libtorch specific files under `torch/csrc/distributed`, #86962 PyTorch RPC crashed when using IB, #106711 'CUDA out of memory' when using a GPU services for reinforcement learning in Torch rpc tutorial",f08253ce034520c905677620f2ba58c4c595ab435fbe7c587c59864cc752f744
issue-collapse:e7b4fc0f2a8cfb59,module: elastic needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,4,0,0,0,"63832, 100310, 103221, 181429",,,,"module: elastic, oncall: distributed, oncall: distributed infra","bot-triaged, feature, module: elastic, oncall: distributed, oncall: distributed infra, ptd-bot-triaged",,"Review this module: elastic queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#63832 [RFC] Add `torch.distributed.run` as a console script in pytorch's setup.py, #100310 Using ddp training with different machine, #103221 [help] did torch.distributed.launch can be applied on k8s cluster with pytorch-operator, #181429 Make torchrun options ( --nnodes, --nproc-per-node, etc.) same a mpirun/mpiexec",15db9b2821a6eca2e23197d9a68a5cd2096281f002cd9825f608579134867894
issue-collapse:766024b2ecfe4d8b,module: symm_mem needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,3,0,0,0,"167966, 172617, 176439",,,,"module: symm_mem, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, feature, module: symm_mem, oncall: distributed, oncall: distributed infra, ptd-bot-triaged",,"Review this module: symm_mem queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#167966 Symmtric_memory: torch.ops.symm_mem.all_to_all_vdev_2d restricts the number of experts per rank (ne) to less than 16, #172617 Symmetric utils missing, #176439 [wip][RFC] Symmetric Memory pool explicit lending semantics",15bfd0f0f3897cded3ffc6cfa6f4209339e233c613be9bfd927275f5fc834d64
issue-collapse:5936bc43a4dd24d4,module: rocm needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"177309, 182373",,,,"module: rocm, oncall: distributed, oncall: distributed infra, oncall: profiler","bot-triaged, module: rocm, oncall: distributed, oncall: distributed infra, oncall: profiler, ptd-bot-triaged",,"Review this module: rocm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#177309 [ROCm][NCCL watchdog] Cross-thread stream-capture mode restrictions in hipEventQuery/hipEventSynchronize cause false watchdog failures, #182373 [ROCm] torch.profiler drops GPU activity in spawn child if parent calls torch.cuda.init() (regression in 2.12, 2.10/2.11 pass)",d1ce4cfc8b6b5cc585b9b12fc23ec76c57063af4045dab983aef8f1c28d70b3e
issue-collapse:787e28a00b1d619a,module: privateuse1 needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"166205, 169357",,,,"module: privateuse1, oncall: profiler","feature, module: PrivateUse1, oncall: profiler",,"Review this module: privateuse1 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#166205 Enable Graph Capture & Profiler Integration for PrivateUse1 Backends, #169357 prepareProfiler() skips collecting info of ""privateuseone""",fb5bdb8a23bd35b4991a82c8b6e84a7ca4ce9d98bc4dc19e43589ff24b33cf53
issue-collapse:dfb51ee43727db91,module: linear algebra needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"52655, 188553",,,,"module: linear algebra, module: performance, module: rocm, oncall: jit","bot-triaged, feature, module: linear algebra, module: performance, module: rocm, oncall: jit, triage review, triaged",,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#52655 Optimize `einsum` in TorchScript profile guided optimization, #188553 test_cholesky_solve_batched_many_batches_cuda_float64 fails on ROCm gfx942 (Tensor-likes are not close)",e409377fd9dc564ee7bbc659e5268c0136f260f1deb0194395fde377af01e321
issue-collapse:a852403064b567d7,module: devicemesh needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"172218, 173815",,,,"module: devicemesh, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, feature, module: DeviceMesh, oncall: distributed, oncall: distributed infra, ptd-bot-triaged",,"Review this module: devicemesh queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#172218 `_expand_group` can get DeviceMesh ranks more directly, #173815 [RFC] Abort a DeviceMesh",20959db8152b94e7bd798e5fcbf1db49274b92389619dd73a19975d9ce8b374d
issue-collapse:5afbe788755c68b3,module: deploy needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"72482, 76686",,,,"module: deploy, oncall: package/deploy","imported, module: deploy, oncall: package/deploy",,"Review this module: deploy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#72482 Improve torch::deploy documentation, #76686 [torch::deploy] Remove `manager_` from the constructor and deconstructor of `InterpreterSession`",69fad7b8bb2f9968e2890ff7104778d99e1345f798763e7095e430dfa2731a85
issue-collapse:7ed75d2c4e381799,module: cpp-extensions needs triage,batch action queue,needs triage,65,medium,bucket_component_queue,2,0,0,0,"46505, 189245",,,,"module: cpp-extensions, module: deadlock, module: torchbind, oncall: jit","bot-triaged, internals, module: cpp-extensions, module: deadlock, module: torchbind, oncall: jit, triage review",,"Review this module: cpp-extensions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#46505 Cannot script when a static method or cross module function calls custom op, #189245 [cpp_extension] Stale .lock file from FileBaton causes permanent deadlock when compilation process is killed",cdb39f266264526de621d4fc19918e53e09d4a7ae4a42e3828287ba3d118a2a1
issue-collapse:2ace50dac67c2849,module: crash high-priority/blocker  batch 3,batch action queue,high-priority/blocker,64,medium,bucket_component_queue,14,0,0,0,"29498, 31362, 81180, 128691, 141358, 153203, 154424, 154887, 157097, 158154, 161760, 171284, 172011, 184878",,,,"module: crash, module: edge cases, module: empty tensor, module: error checking, module: int overflow, module: intel, module: linear algebra, module: mkl, module: mps, module: multithreading, module: python frontend, module: rnn, module: sanitizers, module: sparse, module: third_party","bot-mislabeled, bot-triaged, low priority, module: crash, module: edge cases, module: empty tensor, module: error checking, module: int overflow, module: intel, module: linear algebra, module: mkl, module: mps, module: multithreading, module: python frontend, module: rnn, module: sanitizers, module: sparse, module: third_party, needs reproduction, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#29498 ""malloc(): memory corruption (fast)"", action=3) at malloc.c, #31362 sklearn and pytorch incompatibility issue, #81180 MPS: Placeholder tensor is empty!, #128691 Segmentation fault (core dumped) in `torch._remove_batch_dim`, #141358 PyTorch crashes with `mkl_lapack_dgetrf` if `from_numpy and set_num_threads` are used, #153203 FPE Crash in torch.fmod with large negative integer modulo value",cab371228b538d28905124a0835f6d79363f97fe87fef637b4f6cd40447138cd
issue-collapse:64cd8f0c81840c5f,module: crash high-priority/blocker  batch 2,batch action queue,high-priority/blocker,64,medium,bucket_component_queue,14,0,0,0,"26120, 35792, 48555, 54578, 62308, 62331, 87446, 89714, 91533, 99200, 121235, 127275, 149941, 156723",,,,"module: crash, module: cuda, module: dependency bug, module: linear algebra, module: macos, module: numpy, module: vision, module: vulkan","high priority, module: crash, module: cuda, module: dependency bug, module: linear algebra, module: macos, module: numpy, module: vision, module: vulkan, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#26120 Process fails with assertion error in magma-cuda100, #35792 Python 3 build PyTorch got core dump, #48555 segment fault for Image model training by four GPUs, #54578 cublasSgemmStridedBatched failure when calling grad of grad, #62308 Error while running FINETUNING TORCHVISION MODELS, #62331 broken pipe error",106d0528e95d1c5bef7d0166ed6d31cb281008c0718a59f3e639bf8da6d7ef72
issue-collapse:d353e0c3a14d3a17,module: dataloader high-priority/blocker,batch action queue,high-priority/blocker,64,medium,bucket_component_queue,3,0,0,0,"45198, 93846, 122935",,,,"module: dataloader, module: memory usage, module: multithreading, module: openmp, module: regression","high priority, module: dataloader, module: memory usage, module: multithreading, module: openmp, module: regression, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#45198 DataLoader with cv2 and some numpy/cv2 import order causes workers to not work, #93846 large number of temporary files generated when using dataloader with num_workers>0, #122935 Pytorch 2.1 / 2.2 uses 4x more memory during data loading than Pytorch 2.4 or 1.13",ffe1fc366781bd0bfda513ae17ee5a13b0fad7d0b09fc27b207a4e75af03479e
issue-collapse:236fc21519dbac36,module: serialization high-priority/blocker,batch action queue,high-priority/blocker,64,medium,bucket_component_queue,2,0,0,0,"20497, 140230",,,,module: serialization,"high priority, module: serialization, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#20497 Statically make `__setstate__` set all attributes/parameters, #140230 Inconsistent Model Loading with `weights_only=True` for Custom Models",fe78d1f3cbe37b13ff9761646bd2ff9de68c20dcaa4d447060c9cad2698bb85b
issue-collapse:234906ef1593e81c,module: data parallel high-priority/blocker,batch action queue,high-priority/blocker,64,medium,bucket_component_queue,2,0,0,0,"36871, 40457",,,,"module: data parallel, module: determinism, module: regression, oncall: distributed","high priority, module: data parallel, module: determinism, module: regression, needs reproduction, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#36871 Stochasticity for DistributedDataParallel on CPU but not on GPU, #40457 DataParallel with Torch 1.5",97d20d9a3a85584f6941c19580d40de72aabd13a76d633742a1b02a175e8fb4c
issue-collapse:86d177f95ef1161c,module: dtensor has linked PR,batch action queue,has linked PR,62,medium,bucket_component_queue,4,0,0,0,"108744, 170367, 180346, 188182",,,,"module: dtensor, module: performance, oncall: distributed, oncall: distributed parallelisms","actionable, bot-triaged, enhancement, good first issue, module: dtensor, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#108744 switch more test cases to use MultithreadTestCase, #170367 DTensor redistributes on foreach_X operators are inefficient, #180346 [DTensor] clip_grad_norm_ fails when per-parameter norms come from different DeviceMesh objects, #188182 torch.fft has no sharding strategy",e678377ed0326529c3f258d7e138f8542927603bedd016dacb93db71292ca765
issue-collapse:6ca8aaccde99f053,oncall: profiler needs triage  batch 2,batch action queue,needs triage,59,medium,bucket_component_queue,14,0,0,0,"126703, 151969, 152617, 153901, 157728, 159256, 162481, 164145, 166311, 167605, 169666, 170319, 170774, 170867",,,,oncall: profiler,"feature, needs reproduction, oncall: profiler",,"Review this oncall: profiler queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#126703 PyTorch's Execution Trace Observer Producing Invalid JSON, #151969 Unrecognized in PyTorch traces w/ Cuda 12.6, #152617 Pytorch Profiler crashes while using it with Pytorch Lightning module, #153901 torch.profiler support for tf32, #157728 fill_ overflows on uint64 in range [2**63, 2**64) when profiler is engaged, #159256 can kineto profile only one thread with corresponding gpu tasks among multithreads",e2d73617f197393c0d3511a184685c82e42e8cd40ad7f4cc8a9c69c774fc7521
issue-collapse:898a42811c870851,oncall: jit needs triage  batch 3,batch action queue,needs triage,59,medium,bucket_component_queue,14,0,0,0,"15342, 49460, 65607, 160392, 161854, 161935, 163945, 164208, 165400, 165900, 167412, 168077, 170680, 171377",,,,oncall: jit,"low priority, needs reproduction, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15342 [JIT] Batch op. batch_sum return a mask which is on cpu, #49460 Libtorch: Segmentation fault when running torch::jit::load, #65607 torch.dtype is int in JIT instead of torch.dtype, #160392 bug in libtorch optimize_for_inference, #161854 torch.ScriptObjectProperty.name.deleter occurs Segmentation fault (core dumped), #161935 torch.jit.trace_module.register_parameter :cannot assign 'torch.cuda.FloatTensor' object to parameter 'new_param' (torch.nn.Parameter or None required)",2bff66cf0be026f3ef0961b6259f3fcdbb396a5f5b9db7097e78cd4433596c57
issue-collapse:9f3d262f057d0d79,module: convolution has linked PR,batch action queue,has linked PR,56,medium,bucket_component_queue,5,0,0,0,"169688, 177237, 178125, 178128, 188338",,,,"module: convolution, module: correctness (silent), module: cuda, module: cuda graphs, module: dynamo, module: error checking, module: meta tensors, module: rocm, oncall: pt2","actionable, bot-triaged, enhancement, module: convolution, module: correctness (silent), module: cuda, module: cuda graphs, module: dynamo, module: error checking, module: meta tensors, module: rocm, needs reproduction, oncall: pt2, topic: fuzzer, triaged",,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#169688 [CPU] Stride 3 bfloat16 convolutions appear incorrect, #177237 torch.conv1d incorrectly accepts invalid weight shape in meta backend while raising correct error in eager execution, #178125 conv_transpose1d meta implementation allows invalid output_padding that real CUDA execution rejects, #178128 conv_transpose3d meta execution produces output for invalid bias shape instead of matching eager validation, #188338 [ROCm]: CUDAGraph safe Convolutional Kernel e.g. `F.conv_transpose1d`",ae485bcf43807757cf12e525543f50f95e14c51b0ddc0852e9205cc43b758bbf
issue-collapse:ec8aac0f04820890,module: crash high-priority/blocker  batch 9,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"74547, 79014, 82813, 83070, 84181, 84990, 85214, 85215, 85217, 88036, 89718, 90535, 90676, 91590",,,,"module: crash, module: cublas, module: cuda, module: cudnn, module: ddp, module: distributions, module: edge cases, module: error checking, module: functorch, module: fx, module: half, module: linear algebra, module: mkldnn, module: named tensor, module: norms and normalization, module: regression, module: tensorboard, module: third_party","fx, high priority, module: crash, module: cublas, module: cuda, module: cudnn, module: ddp, module: distributions, module: edge cases, module: error checking, module: functorch, module: fx, module: half, module: linear algebra, module: mkldnn, module: named tensor, module: norms and normalization, module: regression, module: tensorboard, module: third_party, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#74547 importing open3d before pytorch causes matmul to produce a segfault, #79014 RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR, #82813 functionalize and make_fx are not composable resulting in segfault and cuda error, #83070 make_fx + aot_autograd segfaults, #84181 Session of Google Colab crashes when `torch.utils::SummaryWriter` is called after importing `torchaudio`, #84990 Segmentation fault in `torch.futures.collect_all`",688676b4467e710a0d5db49b9a4336008b08bf07843a3a7839505490fbf984c5
issue-collapse:5b36e6f2bb99197e,module: crash high-priority/blocker  batch 8,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"50037, 94125, 94669, 94696, 94698, 96277, 98136, 106362, 107429, 115416, 117254, 117431, 126385, 127102",,,,"module: crash, module: cuda, module: custom-operators, module: edge cases, module: int overflow, module: loss, module: macos, module: nn, module: opcheck, module: openmp, module: single threaded, module: sparse, oncall: quantization","enhancement, high priority, low priority, module: crash, module: cuda, module: custom-operators, module: edge cases, module: int overflow, module: loss, module: macos, module: nn, module: opcheck, module: openmp, module: single threaded, module: sparse, oncall: quantization, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50037 Torch quantized_lstm_cell op out-of-bounds access, #94125 A Floating Point Exception can be trigerred in torch._C._nn.slow_conv3d, #94669 A segment fault can be triggered in torch.svd, #94696 A segment fault can be triggered in torch.max_pool1d_with_indices, #94698 A segment fault can be triggered in torch.avg_pool1d, #96277 A Segment Fault can be triggered in torch.adaptive_max_pool1d with an edge case",e82c9a3132d5c8b0f76abd32751f899e00e78ad58de552897a6883e82cb7efce
issue-collapse:acb7e9a8cb9745e8,module: crash high-priority/blocker  batch 7,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"128503, 128690, 128692, 128695, 133334, 136217, 136704, 136709, 136714, 136719, 138425, 139987, 140945, 141481",,,,"module: crash, module: edge cases, module: empty tensor, module: error checking, module: floatx (formerly float8), module: intel, module: linear algebra, module: macos, module: nn, module: openmp, module: power, module: python frontend, module: sleef","high priority, module: POWER, module: crash, module: edge cases, module: empty tensor, module: error checking, module: floatx (formerly float8), module: intel, module: linear algebra, module: macos, module: nn, module: openmp, module: python frontend, module: sleef, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#128503 crash@sleef_tryVXE2 () while trying to run torch.compile() BERT model, #128690 `torch._transform_bias_rescale_qkv`:FPE, #128692 Segmentation fault (core dumped) in `torch._weight_int4pack_mm`, #128695 Segmentation fault (core dumped) in `torch._weight_norm_interface`, #133334 torch._scaled_mm row-wise hits CUDA invalid memory access when M % 256 != 0, #136217 Segmentaion fault on Apple M1 caused by libomp on v2.4.1",2c986bbb11390c7f6685780a45f6253f8837e7764c3980ba8588f4483d22aa26
issue-collapse:216b6b200cbf926b,module: crash high-priority/blocker  batch 6,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"96056, 103966, 104342, 119367, 133625, 141229, 141847, 142221, 142837, 142838, 143482, 143483, 143485, 143490",,,,"module: crash, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: mkldnn, module: mps, module: pooling, module: sdpa","module: crash, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: mkldnn, module: mps, module: pooling, module: sdpa, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#96056 MPS device throws error for `F.adaptive_avg_pool2d`, #103966 'MPS' Issue Running HuggingFace Transformer Pix2Struct Model, #104342 Segmentation error while using F.cross_entropy with mps(for code that works fine with device= ""cpu""), #119367 Updating a slice on CPU with a slice from MPS results in an error, #133625 Binary Cross Entropy crashes kernel on apple silicon with singleton dimensions, #141229 FPE in `torch._scaled_dot_product_attention_math`",dbc0adf8e591f2f2aab2e73c881b76da6d946fff23ad9c1d27ef602f16d32d4d
issue-collapse:042d43d5e2bce711,module: crash high-priority/blocker  batch 5,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"92776, 97109, 143480, 145065, 145071, 145259, 145267, 145604, 146787, 146789, 146791, 146793, 146794, 149820",,,,"module: crash, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: linear algebra, module: mps, module: nn, module: sparse","module: crash, module: edge cases, module: embedding, module: empty tensor, module: error checking, module: linear algebra, module: mps, module: nn, module: sparse, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#92776 segfault when running torch.lu_unpack, #97109 ""Adaptive pool MPS: input sizes must be divisible by output sizes"", I keep getting this error even when I try to adjust for size, #143480 Segmentation fault (core dumped) in `conv1d`, #145065 SIGFPE error when passing very large kernel_size to `avg_pool1d`, #145071 Illegal memory access and segmentation fault due to large `storage_offset` in `as_strided`, #145259 No Range Check for `storage_offset` in `as_strided` Function",be558b93f9944492e62dbabafd3412ba087e40fee916d0ef821b63497c86f239
issue-collapse:24edec2e763be408,module: crash high-priority/blocker  batch 4,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"19739, 147362, 150388, 152777, 152824, 152913, 153231, 154422, 154882, 155201, 155981, 157098, 159713, 161874",,,,"module: crash, module: edge cases, module: empty tensor, module: error checking, module: flaky-tests, module: lint, module: macos, module: mkldnn, module: mps, module: nn, module: openmp, module: pybind, module: rnn, module: third_party, oncall: quantization","actionable, low priority, module: crash, module: edge cases, module: empty tensor, module: error checking, module: flaky-tests, module: lint, module: macos, module: mkldnn, module: mps, module: nn, module: openmp, module: pybind, module: rnn, module: third_party, oncall: quantization, topic: fuzzer, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19739 Importing open3d after PyTorch causes free(): invalid pointer, #147362 Floating point exception (core dumped) in torch.nn.functional.conv3d, #150388 double free or corruption (out) in torch.as_strided_scatter, #152777 Segmentation fault (core dumped) in torch.nn.functional.alpha_dropout, #152824 `mypy` stage of `lintrunner -a` has intermittent but continuing crashes, #152913 Segmentation fault (core dumped) in torch.nn.functional.max_unpool2d",dc7075aaf4a97a923c283f6cc18e14bac0a07683aca292ad36d8d584f8261c02
issue-collapse:fed1e353f7dea367,module: crash high-priority/blocker  batch 10,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,14,0,0,0,"6998, 35151, 38623, 42452, 43005, 43484, 47161, 49520, 50038, 51134, 51732, 59645, 72182, 74422",,,,"module: crash, module: cuda, module: dependency bug, module: distributions, module: loss, module: meta tensors, module: nn, module: numpy, module: padding, module: pybind, module: rnn, module: third_party, module: vmap","ZeroTensor, better-engineering, module: crash, module: cuda, module: dependency bug, module: distributions, module: loss, module: meta tensors, module: nn, module: numpy, module: padding, module: pybind, module: rnn, module: third_party, module: vmap, todo, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6998 [pytorch] Not handling python reload properly, #35151 Segfault when using misaligned data pointer (from joblib), #38623 Providing CUDA tensor to model on CPU causes a crash, #42452 `device-side assert triggered` - when probability > 1.0, #43005 `torch.distributions.Categorical` crashes with illegal instruction, #43484 pybind11_object_dealloc error",114a8b3d6227e38a3306db1d985ec9aeb0838b05fbd2fe22ac1be2b7c12738ba
issue-collapse:05f49805f92555e4,module: binaries high-priority/blocker  batch 2,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,4,0,0,0,"35418, 36152, 47116, 60628",,,,"module: binaries, module: build, module: crash, oncall: releng","high priority, module: binaries, module: build, module: crash, oncall: releng, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#35418 PyTorch / libtorch executables fail when built against libcuda stub library, #36152 Seg Fault: import vaex with torch, #47116 trying to Installing pytorch on python3.9 via pip results in an non-descript error, #60628 pip install nightly torch torchaudio and torchvision together will install 0.3.0 version of torchvsion on Ubuntu",48f3f6fc5ab01b4a37e34e3e756b8623573860bd24f4ef694d670fde70eecc42
issue-collapse:10eb704509769311,oncall: jit high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"36040, 42487, 61223",,,,oncall: jit,"high priority, oncall: jit, triaged, weeks",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#36040 [JIT] Huge delay (1274s vs 0.031s) when running scripted model, #42487 Support recursive data type in TorchScript, #61223 optimize_for_inference does not handle MKLDNN for BatchNorm1d",f41c460b1d2388d66ac84b46d2b24cffe0d18287101fd0efee07772066088838
issue-collapse:fc7bd0756db1c8a7,module: mps high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"122394, 141836, 143123",,,,"module: mps, module: performance, module: regression","module: mps, module: performance, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#122394 [MPS] Severe Performance Degradation with Deterministic Algorithm on macOS 14.4, #141836 StrideAPI caused regression in channels-last logic, #143123 MPS incompatibility: Calls into the C++ engine to run the backward pass",fc2ff48dd6be2e9854bb3fceb094655d77cce2ba02f7d8fefad139b4a32cac0d
issue-collapse:4e0982e0e778771b,module: memory usage high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"59196, 84051, 95432",,,,"module: memory usage, module: multiprocessing, module: nn","actionable, high priority, module: memory usage, module: multiprocessing, module: nn, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#59196 torch.zeros memory leak : cpu, #84051 model.load_state_dict won't work in Child process if a sufficiently large tensor was padded in the Parent (even if empty padded), #95432 Regression bug in `torch.nn.ReLU6` and `torch.nn.Hardtanh` that `inplace=True` doesn't work in PyTorch 1.10.0~1.13.1",2c7a5df7e534646c79f6fc6ec6b15b5f6f087b7f68bf34285d6b3a5baf1d2f68
issue-collapse:ae038580327d1db5,module: functorch high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"92033, 121711, 142032",,,,"module: functorch, module: regression","high priority, module: functorch, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#92033 Unable to use vmap atop torch.distribution functionality, #121711 torch.func.grad hangs on nested differentiation, #142032 Python 3.12 ""from functorch.einops import rearrange"" fails with ""RuntimeError: First class dim doesn't work with python 3.12""",1b1e1d12ed631da07d5ff7b1536f9d8d57212a8befc0d702d62ad076c36fcef4
issue-collapse:9ea16dcf4003d639,module: crash high-priority/blocker  batch 11,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"5353, 16953, 25518",,,,"module: crash, module: cuda, module: loss, module: memory usage, module: power","has workaround, module: POWER, module: crash, module: cuda, module: loss, module: memory usage, todo, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#5353 Bugs: Score Function approach in REINFORCE for PONG, #16953 Seg fault with test_rnn_retain_variables on ppc64le, #25518 Assign torch.cuda.FloatTensor to List tensor",e1a87c5da192bc9263413e4e5d92b679c54d60e20426b3e44458d3451f9203eb
issue-collapse:1f5e5f4f39ec0685,module: ci high-priority/blocker  batch 2,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,3,0,0,0,"49392, 51863, 72189",,,,"module: ci, module: crash, module: cuda","ZeroTensor, high priority, module: ci, module: crash, module: cuda, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#49392 Uninitialized variable was not detected in ASAN CI config, #51863 TestVectorizedMemoryAccess.CopyKernel starting to fail after driver to 460.39, #72189 test_fn_fwgrad_bwgrad_[trapezoid|trapz]_cuda_complex128 causes CUDA memory exception",3977cb0227e77743bf2a0f0698178bb5c30eab08b9c642b7a9eed6f9ef77d901
issue-collapse:8f07f287e3b931c2,module: rpc high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"46491, 83243",,,,"module: rpc, oncall: distributed","high priority, module: rpc, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#46491 cannot call rpc.init_rpc twice within a single process, #83243 RPC: wait method of Future object return 0 sometimes in rpc framework",68e01e8ba61d8533fee9eb5fa9b9864612fdd2bff4bbf2111bc40793a2cd1e25
issue-collapse:bda5610ad61de95e,module: numpy high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"59237, 60081",,,,module: numpy,"high priority, module: numpy, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#59237 Multiplication of `torch.tensor` with `np.array` does the operation with numpy., #60081 [Bug] numpy is no longer a required dependency",fec2be6e630c2848d352b7a9a2a9f0d9ccb2cb3fb30758f1e7cec310d85df474
issue-collapse:7dfbdeb1157b1280,module: nn high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"80488, 82068",,,,"module: nn, module: regression","high priority, module: nn, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#80488 Negative values still produced by torch.nn.functional.kl_div, #82068 upsample_bilinear2d() received an invalid combination of arguments",afa6e29d8276cb6988156a7011eac10aa383f13f5ec1e0f4cb60c1350ce9c206
issue-collapse:314f73ab72c22f44,module: infra high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"135885, 136374",,,,"module: infra, module: macos","high priority, module: infra, module: macos, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#135885 ☂️ 150+ MacOS tests were marked flaky recently, #136374 Runbook for GCP A100 runners should be updated",27e6c154e8ab8e41997f70769053721c38c689d43fc35d904b4f5f06dcfb1494
issue-collapse:c28c0200a0ab5220,module: docs high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"57178, 67237",,,,"module: docs, module: pooling","high priority, module: docs, module: pooling, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#57178 MaxPool2D Returns Wrong Shape With Ceil_Mode, #67237 [docs] Google finds docs pages that give 404",e0d10a9dfac5161654792b604bb714f0ab3b8217ce033925fe8ecded1361b6a2
issue-collapse:4733bceadf7d845d,module: cudnn high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"50879, 137347",,,,"module: cudnn, module: regression, module: rnn, module: sdpa","high priority, module: cudnn, module: regression, module: rnn, module: sdpa, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#50879 Unexpected slow dropout in stacked RNN/LSTM/GRU, #137347 cuDNN dBias error starting on 10/2 nightly",6ed64b122568db944370d78316e4d5c68b264657cc47292f7f7e1e47124fab25
issue-collapse:404e8eb2c93eba45,module: advanced indexing high-priority/blocker,batch action queue,high-priority/blocker,52,medium,bucket_component_queue,2,0,0,0,"107699, 139242",,,,module: advanced indexing,"actionable, high priority, module: advanced indexing, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#107699 index.Tensor_out & index_put.out errors or segfaults with indices list containing only null tensors, #139242 Tensor.index_reduce produces incorrect result",bd891819879a5d209694a4b05d2e06a9ad61e2dd9b5145ad1951466eb2716873
issue-collapse:0c81782a692cf1b0,oncall: quantization needs triage  batch 3,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"115985, 118165, 121230, 122183, 122303, 123824, 125004, 125564, 126140, 126266, 126974, 127076, 131055, 132029",,,,oncall: quantization,oncall: quantization,,"Review this oncall: quantization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#115985 Faster Pytorch dequantize() + matmul for quantized models, #118165 [Quantization-aware training] Can not quantize nn.MultiheadAttention module, #121230 INTERNAL ASSERT FAILED in torch.pow, #122183 [FX Graph Mode Quantization] Prepare model fails with Linear-BN when `affine=False`, #122303 How to exclude some modules from quantization?, #123824 Precision problem in `quantize_fx.convert_fx` and question about the `example_inputs` for `quantize_fx.prepare_fx`.",eeaf415afdf6a8b17b0a54dc21720a8fdb34a5d46bfa58834eeb2c8d02c37227
issue-collapse:9da51ea3f0d67c21,oncall: quantization needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"133552, 133557, 134220, 136094, 136828, 138820, 139464, 143289, 147112, 148171, 149500, 152813, 153358, 154323",,,,oncall: quantization,"enhancement, oncall: quantization",,"Review this oncall: quantization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#133552 fx qat symmetric config zero_point not zero, #133557 torch quantize error, #134220 [Operator Benchmark] [CPU]OP: weight_norm_sparsifier_step functionality failed with parts of long config shapes, #136094 Quantized model is way slower than regular model., #136828 false INTERNAL ASSERT FAILED in `torch.empty`/`torch.ones`, #138820 RuntimeError: HalfTensor is not supported",a99696c44e8a7c3124aafe80f97d7f9816ca3ddd35e79ca36c686dec77363bdf
issue-collapse:ad80806ec740e2ba,oncall: profiler needs triage  batch 3,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"69782, 124674, 125272, 127215, 128292, 132738, 133444, 135003, 138193, 140684, 144613, 144778, 147604, 147722",,,,oncall: profiler,"OSS contribution wanted, oncall: profiler, triage review",,"Review this oncall: profiler queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#69782 Torch Profiler does not count FLOPs for backward pass, #124674 Improving format of communication metadata in PyTorch Execution Trace., #125272 Profiler Kineto Config: Warning: Failed to recover relationship between all profiler and kineto events: 51361 vs. 0 reassociated. (function reassociate), #127215 No way to config profiling scope in torch.autograd.profile, #128292 put profiling results from different ranks into one webpage, #132738 Profiler getProfilerConfig interface can not support Ascend npu platform",c060de220aa77a06fe86db379cb21a70529f9d54c505683f75f0ca52ade949c8
issue-collapse:674130ddf4cfca20,oncall: mobile needs triage  batch 5,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"36182, 37016, 38312, 39240, 39724, 42039, 42308, 43488, 44630, 46979, 47177, 47342, 48410, 48703",,,,oncall: mobile,oncall: mobile,,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36182 ## 🐛 Bug: QNNPACK tests failing on master on Nexus 6, #37016 Guidance on implementing a new backend, #38312 Status of support for training on mobile, #39240 [mobile] Running on armeabi-v7a is inconsistent with arm64-v8a, #39724 pickle_save on mobile (no longer works with 1.5.0 release), #42039 Segmentation fault (core dumped) when running optimize_for_mobile",b633fb9c3e6d7219667d61298cd53412da11b6b5d0103fa83b1f623745e13633
issue-collapse:5868d2ba73b897d3,oncall: mobile needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"32901, 34682, 44469, 46995, 52144, 53264, 55779, 55782, 56007, 57887, 57893, 61765, 65662, 67599",,,,oncall: mobile,"enhancement, mobile_perf, oncall: mobile",,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32901 QNNPACK linear doesn't preserve dimensions, #34682 Pytorch not compatible with react native android, #44469 Error in the converting to pytorch mobile.., #46995 max_pool2d always fails on native Android app, #52144 Concatenation of Java tensors (Android for mobile), #53264 How to load trained . torch from conversion to .mlmodel",c02ce5e74586c12e01a478e2eeffad4a4857982adbdadc1272af2641e0f0a0c4
issue-collapse:0659b8a8d76ccfd4,oncall: mobile needs triage  batch 3,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"49782, 50943, 65490, 67603, 68536, 68984, 69347, 69354, 70159, 72252, 74272, 75308, 75864, 76029",,,,oncall: mobile,"Stale, mobile_perf, oncall: mobile",,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#49782 TorchScript model's output different on Mobile and PC, #50943 Exporting opnames fails when using custom classes, #65490 optimize_for_mobile() assert fail: missing op for prepacked::conv2d_clamp_prepack, #67603 The minimum supported version of pytorch when using toch-mobile？, #68536 Exporting a model with `_save_for_lite_interpreter()` fails with Segmentation fault, #68984 speed_benchmark_torch crashes when trying to run a model using Vulkan backend on Android",381fff28ece77ebd9261ee706b34c3a149af71e8bd8840251831b591b1d5e1b1
issue-collapse:aa98188e5f8359a5,oncall: mobile needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"50714, 68987, 69609, 77840, 80142, 80321, 81651, 82312, 83161, 83250, 86572, 91387, 109579, 110379",,,,oncall: mobile,"Stale, oncall: mobile",,"Review this oncall: mobile queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50714 [PyTorch Mobile] Android speed benchmark binary crashes when using vulkan, #68987 Operator is not supported in mobile module., #69609 [mobile_optimizer] RuntimeError: We don't have an op for metal_prepack::conv2d_prepack, #77840 Legacy model format is not supported on mobile, #80142 CPUProfilingAllocator greedy allocation plan generation failed, #80321 java.lang.ExceptionInInitializerError at org.pytorch.NativePeer.initHybrid(Native Method)",7acea550bfad9426f2dcdb67da12d733ffce219ee16343fba06af35c9b463846
issue-collapse:3c92dadbd1c3e246,oncall: jit needs triage  batch 9,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"69250, 70450, 70544, 134922, 134923, 134924, 134937, 134971, 136654, 136829, 137561, 138973, 139556, 140101",,,,oncall: jit,"Stale, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#69250 TorchScript incomptaible with non-persistent buffers, #70450 torch.jit.optimized_execution is not mentioned anywhere in the docs, #70544 Tensor loses `bool` method during scripting, #134922 [torch.jit.script] An INTERNAL ASSERT error will be raised when permuting the inputs of a jit scripted graph with mismatched indices., #134923 A crash would be raised when using `._StackString` with empty lists, #134924 `TorchScriptTesting._PickleTester` caused a segmentation fault with empty input lists in a jit traced function",60b3a1bc3c1a2f80398e60fb0029b4edee80030b993fe73b08e92e7a4c1236c1
issue-collapse:f6befa88a66682c3,oncall: jit needs triage  batch 8,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"76026, 81855, 81912, 92041, 111142, 121581, 132240, 137202, 140468, 140539, 140702, 141855, 143831, 143854",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76026 Jit torchscript for prediction is missing 'forward' when using forward hooks, #81855 Enable jit error when using FSDP, #81912 Unknown builtin op: aten::broadcast_shapes, #92041 jit testing fails on 3.11 debug build, #111142 'torch._C.Node' object has no attribute 'cs', #121581 The results of PyTorch and TorchScript models are inconsistent.",91b35470b9ca7d85de30045771d69676fc4fa9b40481ef2b871517c553b3b366
issue-collapse:61edb8fe5f77cbac,oncall: jit needs triage  batch 7,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"75960, 76774, 112372, 114112, 143857, 145995, 146757, 146758, 146759, 146760, 146773, 146775, 146778, 147055",,,,"oncall: jit, oncall: quantization","feature, oncall: jit, oncall: quantization",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#75960 jit/_trace.py"", line 71, in _unique_state_dict filtered_dict[k] = v.detach() AttributeError: 'torch.dtype' object has no attribute 'detach', #76774 Support Positional-only Arguments in JIT, #112372 [Tracer] RuntimeError: _Map_base::at when tracing using autograd.Function, #114112 [JIT] - torch.script - 'Optional[Tensor]' object has no attribute or method 'size', #143857 `@torch.jit.script` causes `pytest-cov` to miss function body, #145995 torch.jit.trace does not move tensors to GPU.",b6eb81c145054461cd06ffb0e9f0fd3157bda09165d2a38926fdb2b125d097e9
issue-collapse:dc4770a3e7b85fac,oncall: jit needs triage  batch 6,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"36061, 82550, 107568, 147444, 147739, 147844, 148447, 148543, 149223, 149251, 149391, 149619, 149736, 149800",,,,oncall: jit,"enhancement, oncall: jit, triage review, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36061 [JIT] support list of nn.Module in torchscript, #82550 `torch.cat` can break `torch.jit.ScriptModule` when in inference mode, #107568 RuntimeError: Unsupported value kind: Tensor while torch.jit.script nn.Module, #147444 NaN values originating from the normalize_weight_jit function during the backward pass, #147739 torchvision export model error:: torchvision.models.detection.retinanet_resnet50_fpn_v2, #147844 Python warnings are printed multiple times",23b8a90cc846ae25c03ca29ee261583e255b40229f66d1f9baf073093f8473bf
issue-collapse:d7c4105fa77b01f4,oncall: jit needs triage  batch 51,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"7904, 9595, 14270, 15116, 18312, 28191, 28267, 29125, 29128, 33810, 33993, 34851, 35526, 35606",,,,oncall: jit,"low priority, oncall: jit, triage review, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7904 TracedModules don't support parameter sharing between modules, #9595 [JIT] Unify python -> SugaredValue construction + fix inlining of graphs with Tuple-typed inputs, #14270 Pytorch C++ API with cuda : Expected object of backend CPU but got backend CUDA for sequence element 1 in sequence argument at position #1 'tensors', #15116 torch.save does not work if nn.Module has partial JIT., #18312 [jit] Trace perform weirdly on BatchNorm, #28191 [JIT] Allow user to provide aliasing information on input tensors and model parameters",f95189e91ebd0838a5fbe623b6acfa85d74b8fcb68639a2cc7e5f7dc32e1949a
issue-collapse:05d4249c63ed75ad,oncall: jit needs triage  batch 50,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"33281, 36211, 36603, 36926, 38744, 39309, 39539, 39569, 41141, 41418, 41433, 42081, 42145, 42170",,,,oncall: jit,"days, enhancement, oncall: jit, triage review, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33281 Uninitialised value was created by a stack allocation, reported by valgrind, #36211 [JIT] support self.named_buffers and self.named_parameters in TorchScript, #36603 TorchScript Support for Named Tensors, #36926 JIT string ops and other miscellaneous ops probably shouldn't be in aten namespace, #38744 [jit] `jit.annotate` and multiple compilation of functions with different types, #39309 [JIT] nn.ModuleList loses None objects inside it after scripting",9cd98e714ed6d94e828832f1461abc0b8a4bbdb85dd50263496e22d00c785818
issue-collapse:48145f86fd89eb84,oncall: jit needs triage  batch 5,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"94974, 150342, 150577, 150607, 151165, 151233, 151428, 151468, 153224, 153722, 153931, 154101, 154407, 154423",,,,oncall: jit,"oncall: jit, security",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#94974 Slow inference of torchscript model in C++, #150342 Torch trace doesn't respect @torch.jit.ignore on torch.nn.Module forward method, #150577 [TorchScript] Enum scripting failures in python 3.11+, #150607 `torch.jit.script` does not respect `torch.set_default_dtype`, #151165 Bug on running TorchScript on H100, #151233 Code Injection via `torch.jit.annotations.parse_type_line()`",dc983654315292737c288e706aa6ebfb698a66f848b0c85d46bb267ba77a94e0
issue-collapse:ebb854d34f728832,oncall: jit needs triage  batch 49,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"17126, 41291, 41683, 42947, 42948, 43119, 43134, 43378, 44446, 44503, 44608, 44800, 44992, 45012",,,,oncall: jit,"days, jit-backlog, months, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17126 Support callables in scripted functions, #41291 Return list of tuples from custom operator, #41683 'torch::jit::script::ErrorReport' from 'torch::jit::load('modelpath') in Visual Studio 2017, #42947 [jit] std::exception_ptr can't be torchbind/convert to ivalue, #42948 [jit] TorchBind std::chrono::milliseconds, #43119 Tracing can leak CUDA memory when using tensor constants",9aa5ed3f395df4c6d473ca3338ce853ed9c19427e804ecdd37cda9cc95764b64
issue-collapse:d5113e861c47ea23,oncall: jit needs triage  batch 48,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"40696, 46368, 46654, 47010, 47335, 47682, 47883, 48089, 48090, 48094, 48307, 48984, 49253, 49377",,,,oncall: jit,"days, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40696 Trace model with floor, Inconsistent performance, #46368 [JIT] Inconsistent behavior of torchScript parser in python and c++, #46654 Converting object detector model to TorchScript, #47010 [JIT] Tracing turns None Device Input into a Constant, #47335 [jit] support length inference for list comprehension, #47682 Common dunder methods fail when called directly",34690d68bc1b6ea3593673705014d30c0416a0668f5e5d49c92fc3452ddfecdc
issue-collapse:ea6ff41b2f10339d,oncall: jit needs triage  batch 47,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"41046, 41415, 41513, 42673, 43261, 48191, 48408, 48869, 49569, 49585, 49650, 50127, 50772, 50834",,,,oncall: jit,"TSRootCause:DefaultTypes, TSRootCause:InvalidCustomClass, TSRootCause:PoorIRVisibility, TSRootCause:UnsupportedConstructs, TSUsability, days, oncall: jit, triage review, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41046 Human-readable names and operations for TorchScript model graphs, #41415 [jit] `set` builtin support, #41513 Python object None check not supported, #42673 Unexpected behavior of ""to"" method inside a torch.jit.script decorated function, #43261 OSError: could not find class definition when export torchscript module, #48191 [JIT] Contained Module Attributes should be recursively compiled",bb0e855873303da469a700952be7172416292c16a3ff649bd41239dcc5767cf6
issue-collapse:34f34d223c3ecbb3,oncall: jit needs triage  batch 46,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"39690, 43873, 45199, 49935, 51379, 51409, 51808, 52147, 52599, 52720, 52768, 52772, 52777, 52795",,,,oncall: jit,"NNC, TSRootCause:DefaultTypes, TSRootCause:InvalidCustomClass, TSRootCause:PyTorchParityGap, TSUsability, days, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39690 [Quantization] Output tensor type is lost after serializing and loading back a quantized model, #43873 PackedSequence objects that are created within jit.script are not PackedSequence, #45199 JIT doesn't support rfloordiv with scalars correctly, #49935 [FX] intermediate types of empty lists/dicts not preserved during torch.fx tracing, #51379 TorchScript class compilation failure with pybind11-2.6.2, #51409 C++ load model error",b1a7bcd90ee8cb35d5251e590776b2ac29a37115c196d354772b002ad7ff9198
issue-collapse:035e6404e1fd55cd,oncall: jit needs triage  batch 45,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"41512, 47692, 50971, 52589, 52821, 53140, 53414, 53753, 53824, 53862, 53867, 53877, 53878, 53921",,,,oncall: jit,"NNC, TSRootCause:BetterEngineering, TSUsability, days, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41512 Tensor.new_tensor is not supported, #47692 pyre not able to infer the type of torch.jit.script, #50971 Failing to trace Huggingface models using Trace API from Torchscript, #52589 [JIT-autodiff] batch_norm isn't differentiable, #52821 [nnc] Compute expressions with > 4 axes is pretty tedious, #53140 [JIT] Document __jit_ignored_attributes__",fc4addcc1a639fbd52641dca40f23a7d738647d884e1fbfc242f8563dacda3f2
issue-collapse:dcf86e6cf551b29a,oncall: jit needs triage  batch 44,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"13164, 35215, 45524, 50933, 51055, 54146, 54348, 54365, 54435, 54598, 54625, 55072, 55194, 55541",,,,oncall: jit,"TSRootCause:TypeRefinement, TSUsability, days, oncall: jit, triage review, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13164 Tracing custom ops, #35215 Get hold of Backward graph from a C++ module., #45524 [JIT] JIT Autograd is saving more results than are necessary for bprop like the result of RELU when it is an intermediate, #50933 [JIT] Type refinement failure in if-elif-else structure, #51055 Native code assert or hang on process exit with a torch.jit.script model on multi-GPU, #54146 Conversion of Pytorch model to Torchscript for model LaneATT(Lane Detection)",093097f3cce0c4cfda7b688dca887960c30c356eb1c803789069d42970a9b7d8
issue-collapse:dc2eb2731f9d5b97,oncall: jit needs triage  batch 43,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"43249, 43765, 49261, 49614, 49820, 51465, 51942, 52309, 53344, 53904, 56096, 56159, 56571, 57035",,,,oncall: jit,"TSRootCause:TypeChecking, TSUsability, days, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43249 [JIT] Cannot iterate over nn.ModuleList in JITted code when accessed through one level of indirection, #43765 JIT fails sanity checks during tracing torch.rand_like, #49261 [TorchScipt] Custom Objects C++ API, #49614 Improve error on missing/problematic Module member type annotation, #49820 [JIT] Don't support math.exp and math.log in the Script, #51465 torch::jit::Module cannot be forward declared and used into a `unique_ptr`",8ca2e003c40eb504a102b7abe4fdad27221c5eaca26c86b2395c33a79f66815c
issue-collapse:08b3e953740a4fdf,oncall: jit needs triage  batch 42,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"42267, 48066, 53472, 55999, 57313, 57592, 57639, 57717, 57739, 57793, 58516, 58804, 58836, 58857",,,,"oncall: jit, oncall: mobile","days, oncall: jit, oncall: mobile",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42267 assertRaisesRegxWithHighlight isn't triggered correctly, #48066 [JIT] Saving and Loading JIT Model Does Not Preserve __dir__, #53472 After upgrade to 1.8.0 memory leak in libTorch after using torch::jit::load(), #55999 [JIT][ProfilingExecutor] A reliable and consistent API/protocol to query & propagate profiling data, #57313 [jit] pickling custom classes may lead to invalid pickle files, #57592 `jit.trace` fails to capture single-element tensor passed into op requiring scalar-type argument",1c555562a2aaab03b961d75e7f9c63d2affb23c32b9f703ed100d9dece3b1225
issue-collapse:b1d4fa63aeb0ecbc,oncall: jit needs triage  batch 41,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"50734, 56969, 57894, 58907, 59114, 59153, 59174, 59528, 59652, 59806, 60359, 60732, 60847, 62556",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50734 [TorchScript] Document TorchScript IR, #56969 [JIT / NNC] Incomplete Ops Coverage, #57894 JIT ""optimizations"" take way too long, #58907 JIT fails when calling an exported method inside another one, if kwargs with defaults are not passed, #59114 [JIT] OpInfo / Autodiff tests dont seem to actually invoke symbolic backwards, #59153 JIT compiling a ParameterList raises an unnecessary warning",228f0c97aef17249c1b93b9e9cfbbbf96ff03ad15a3184413924538d394a6a85
issue-collapse:99b9c9f57aa90cd7,oncall: jit needs triage  batch 40,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"8473, 42592, 53744, 54593, 54937, 55742, 56660, 57559, 62154, 62394, 62543, 62603, 62854, 63124",,,,oncall: jit,"TSRootCause:TypeRefinement, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8473 RuntimeError: /pytorch/torch/csrc/jit/tracer.h:117: getTracingState: Assertion `var_state == state` failed., #42592 Lower performance when forwarding JIT model with libTorch 1.6 compared with 1.5, #53744 TorchScript divison by zero fails to error out, #54593 Multiple refinement with `isinstance` doesn't work, #54937 [JIT] export fast-transformer to TorchScript, #55742 [JIT] TorchScript to represent typed tensor annotations in IR graph",672597db7d22af64d125f7452c3871a0d837ada9dd9a1a0d4ceec12b45300304
issue-collapse:cc5c9eaf4a1b9063,oncall: jit needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"55282, 111480, 130309, 154478, 154502, 154607, 155003, 155004, 155007, 155009, 156557, 157103, 157603, 158227",,,,oncall: jit,"NNC, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55282 Poor error message when trying to jit a function instead of a module (RuntimeError: Cannot insert a Tensor that requires grad as a constant.), #111480 torch.jit.script persistently changes default from utf-8 to ascii, #130309 【Pytorch mobile android】torch jit forward fail,Calling torch.geqrf on a CPU tensor requires compiling PyTorch with LAPACK. Please use PyTorch built with LAPACK support., #154478 `__jit_ignored_attributes__` is not respected in `torch.jit.trace_module`, #154502 `typing.get_type_hints` fails on TorchScript model after loading with `torch.jit.load`, #154607 Internal Assertion Failure with torch.cuda.Stream in @script_method on CPU-only Systems",061513dc0d6bd83e6f1d370493fb9132220ac50d11ba6ae61952ce0de12108d5
issue-collapse:e634502202d28c12,oncall: jit needs triage  batch 39,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"34713, 53421, 55749, 60143, 62325, 62775, 63405, 63648, 64096, 64247, 64260, 64334, 64487, 64628",,,,oncall: jit,"oncall: jit, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#34713 torch.jit.script report error when using index to subscript nn.ModuleList, #53421 [bug][JIT] Error when tracing model with partial hooks in PyTorch v1.8.0, #55749 Fail to introduce torch::jit::Module as a parameter of a cusomized operator., #60143 torch.jit.save gives error - RuntimeError: Could not export Python function call 'NMSop'. Remove calls to Python functions before export. Did you forget to add @script or @script_method annotation? If this is a nn.ModuleList, add it to __constants__, #62325 [jit]][script] handling of 'void' returns, #62775 [TorchScript] Support for registering builtin operators with control flow",81f90ee77dd5cf2b3856d75331b12f3637f7f20f5cf3fbf1c65a5d0ba6161db8
issue-collapse:a2f6db08e62834e3,oncall: jit needs triage  batch 38,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"18520, 49643, 64490, 64775, 64896, 64918, 64957, 65098, 65144, 65297, 65397, 65398, 65611, 65630",,,,oncall: jit,"oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18520 Jit fail with TracingCheckError with tracing model with layers created after init., #49643 Tracing doesn't work with spectral normalization, #64490 Load torchscript model with custom operators (GNN) on Windows, #64775 [JIT] Canonicalize aten::rsub, #64896 pytorch1.9.0 can not use torch.jit.load() to load model which save in pytorch1.3, #64918 TorchScript error loading module with float literal default argument",2b493f4dce82e9e182d38fd5860d96fbe3ad3971fdbedfe37247c6d1a1378f49
issue-collapse:d8ff1f17b3855b07,oncall: jit needs triage  batch 37,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"51154, 51892, 51941, 57929, 65776, 65824, 65915, 66065, 66150, 66651, 66791, 66841, 67099, 67160",,,,oncall: jit,"oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51154 [JIT] Rough edges found while exploring expressions supported by TorchScript, #51892 [cuda jit extentions] global shared destination conflicts with virtual envs, #51941 RuntimeError: torch.dtype passed as int from JIT to ignored function, #57929 [JIT] provide a pass for alias reduction, #65776 `torch.jit.save` fails on Windows, #65824 Failing to compile /torch/csrc/jit/frontend/ir_emitter.cpp with pointer is null warning",286bb1172ff3853a73ae90ec7fd8a4766ab610fb1db19f8e2b9a9769cebd1e1f
issue-collapse:952ad07002946d80,oncall: jit needs triage  batch 36,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"54821, 58456, 65814, 66816, 67420, 67458, 67473, 67474, 67542, 67659, 67838, 67889, 68064, 68065",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54821 Decouple TorchScript compiler from operator library, #58456 TSAN issues in torch::jit::ProfilingRecord, #65814 INTERNAL ASSERT FAILED when reusing JIT compiled module, #66816 Jit fuser half precision support, #67420 Avoid warnings when jitting pytorch modules, #67458 OpInfo JIT tests fail for torch.nonzero",c9ec0f82b44ddc025e76d38481ccb2aa824648b71a1d9944af22bd1d26ff5d19
issue-collapse:2ab9e6b2d7de0943,oncall: jit needs triage  batch 35,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"14455, 17932, 41416, 41492, 43237, 52214, 52271, 56595, 63395, 69320, 69608, 69938, 70573, 70627",,,,oncall: jit,"TSRootCause:UnsupportedConstructs, TSUsability, better-engineering, days, months, oncall: jit, triage review, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14455 [JIT] Tracing a script function/module where not all args are Tensors, #17932 JIT torch.ones_like with dtype starts failing on master, #41416 [jit] support for generators and `yield`, #41492 Make torch.iinfo/torch.finfo torchscriptable, #43237 [jit] remove shape/device specialization from tracing after ONNX done with its own shape inference, #52214 `hasattr` check not supported on objects other than `self`",be5b4df209b6c4a3622e826e74dc6b610fe3a9605acd54c03328abc3de663edb
issue-collapse:b7a781d6dabc157b,oncall: jit needs triage  batch 34,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"54612, 59905, 65643, 65983, 68759, 68800, 68968, 69084, 71151, 71165, 71286, 71288, 71784, 72029",,,,oncall: jit,"NNC, Stale, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54612 Exceptions thrown in JIT interpreter always are translated into RuntimeError, #59905 [JIT] batch operators in training/inference, #65643 [JIT] Start Using and Optimizing on Symbolic Shape Inference in Freezing, #65983 traced EmbeddingBag cannot be used with different batch size, functional.embedding_bag is not de, #68759 `test_variant_consistency_jit` errors for `torch.nn.functional.prelu`, #68800 [1.10 regression][jit][cuda fuse] cat codegen wrong var name",4a193bf59124009af69eb2ec623dd93f76e35d4f470b78d8551e3e8e9e075d5e
issue-collapse:d69836f82aafcb1f,oncall: jit needs triage  batch 33,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"55211, 69435, 69814, 69830, 70102, 70223, 70240, 71912, 72338, 72617, 72746, 73072, 73419, 73571",,,,oncall: jit,"Stale, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55211 [TorchScript] Can we support type refinement by Python type annotation?, #69435 torch.is_tensor(obj) doesn't work with JIT, despite the fact that isinstance(obj, Tensor) already works with JIT, #69814 JIT / TorchScript docs missing any mention of typing.cast, #69830 Conversion error from pytorch model to libtorch model, #70102 torch jit script segm fault, #70223 jit: Confusing behavior w/ torch.autograd.grad, iterative loop, and printing?",b0697570e9611771a18a744d83c91d6f10e014fea9a571cb55261e20405b3ad0
issue-collapse:2b2020269a317092,oncall: jit needs triage  batch 32,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"41417, 58744, 69434, 71929, 72516, 74030, 74095, 74589, 75005, 75038, 75217, 75309, 75654, 75655",,,,oncall: jit,"TSRootCause:DynamicBehaviors, TSRootCause:UnsupportedConstructs, TSUsability, months, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41417 [jit] support class polymorphism, #58744 Using pytorch 1.1 model on pytorch 1.8 environment, #69434 JIT RuntimeError: 'Union[Tensor, List[float], List[int]]' object is not subscriptable, #71929 [JIT] Results are different when saving in TorchScript Format through train/eval mode, #72516 test_del (jit.test_builtins.TestBuiltins) fails due to highlight assertions, #74030 [JIT] tuple of float is accepted in List[float] but not in Union[float, List[float]]",8e60b2342f254f015955d6d1f41b5e538e9c4c6666748fb7791414a8bde07d5a
issue-collapse:930fda2c4487d210,oncall: jit needs triage  batch 31,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"72220, 73608, 75280, 75794, 75956, 76503, 76672, 77027, 77354, 77837, 78133, 78486, 78559, 78805",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72220 [JIT] Assert that the autodiff implementation of backward() returns the correct number of values, #73608 `torch.jit.load` fails when function parameters use non-ASCII characters, #75280 [JIT] log_extract.py improvements, #75794 Different result with JIT, #75956 [JIT] [Autocast] JIT Autocast Pass operations' list should be extendable and consistent with imperative path, #76503 [JIT] make IRAttributeError extend jit::Error",65697e87217ef6411671051e95adef9026a9c5f3d5632763feee0bd2a0b87cb4
issue-collapse:6dcc2a2fac6d4983,oncall: jit needs triage  batch 30,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"47162, 67142, 70511, 75935, 78812, 78920, 79008, 79275, 79375, 79684, 79751, 80829, 80939, 81100",,,,oncall: jit,"Stale, feature, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47162 Custon C++ classes are not traceable, #67142 enumerate(..., start=idx) is not working correctly while scripting, #70511 Strange behavior of torch.jit.trace when moving parameters across device, #75935 torch.jit.trace error when custom autograd function used in the model, #78812 TorchScript inference get intermediate result?, #78920 Strange tracing result with torchscript",12b6fabe001eb93c705691872d947aecce97e714406f767efa2fbce4c6688a36
issue-collapse:58c92df8f4aac78e,oncall: jit needs triage  batch 29,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"30365, 46460, 62313, 69931, 81245, 81460, 82084, 82324, 82451, 82634, 82635, 82762, 82872, 83045",,,,oncall: jit,"Stale, oncall: jit, triage review, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30365 TorchScript Performance: 150x gap between TorchScript and Native Python, #46460 C++ API: Training and inference of torchscript modules on multiple GPU, #62313 On Triton inference server, the TorchScript .pt file, exported by 1.9.0, serves unsuccessfully, #69931 [JIT] torch.exp roughly 20 times slower in TorchScript vs. PyTorch, #81245 Libtorch cannot load TrochScript Module correctly, when a network contains conv2d(inchannels=64, outchannels=128, kernelsize=1) ., #81460 [JIT] Request Constant Propagation to keep fake_quantize_per_tensor_affine and fake_quantize_per_channel_affine on the graph",fc7540169eae337ccc3766d49294e46b413f4a230aec1419a13de4fc1e9146f5
issue-collapse:b483c194d58d65de,oncall: jit needs triage  batch 28,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"46375, 76013, 81635, 83349, 83585, 84049, 84135, 84187, 84189, 84588, 84615, 84628, 84646, 84673",,,,oncall: jit,"oncall: jit, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#46375 JIT: error in LSTM with `flatten_parameters`, #76013 test_jit.py TestWarn.test_warn and friends doesn't work under pytest, #81635 Windows Debug binaries crash on forward: assert fail on IListRefIterator destructor, #83349 checkpoint function is not jit compatible, #83585 Segfault when profiling with_stack=True on model with jit.optimize_for_inference, #84049 TorchScript unsupport tuple unpacking as function inputs.",203edf62bb3b129a0a94c2e1417c40f78cb3a851d3990a98aecb637d8bf81e38
issue-collapse:f0d9f4a964561fdd,oncall: jit needs triage  batch 27,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"44951, 84681, 84685, 85072, 85342, 85505, 85514, 85568, 85570, 85573, 85585, 85588, 85804, 85805",,,,oncall: jit,"NNC, days, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#44951 copy.deepcopy not working properly for jit.TopLevelTracedModule, #84681 JIT will affect the gradient computation of forward mode, #84685 [TensorExpr] applying `rfactor` for a Mul Reducer with init value different than 1 results in wrong results, #85072 Segmentation fault in `torch.jit.wait`, #85342 JIT fuser issues with {ceil,floor,round,trunc}(int8), #85505 The reload model has different (and strange) forward computation from original model with `LSTMCell`",067ff2baf27b2ce1b709de7bb229398e7e2e71941f14aeff8991a7100f92d491
issue-collapse:341bae2ea33fa004,oncall: jit needs triage  batch 26,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"47336, 81337, 85706, 85813, 85893, 86110, 86112, 86116, 86124, 86276, 86287, 86351, 86455, 86553",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47336 [jit] Properly support slicing of ModuleList, #81337 JIT trace takes forever on a simple method, #85706 RuntimeError: [enforce fail at CPUAllocator.cpp:68] . DefaultCPUAllocator: can't allocate memory: you tried to allocate 4860032 bytes. Error code 12 (Cannot allocate memory), #85813 TorchScript error for `Enum` inside a module, #85893 TorchScript does not recognize mix-in types with `Enum`, #86110 JIT return a tensor with different datatype from the tensor w/o gradient and normal function",b5fcb307aa8d685b1b9490e855ed1ed51d535c2195f0f511bf0781301bbfffed
issue-collapse:58572084399e6e5e,oncall: jit needs triage  batch 25,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"51376, 77167, 85513, 86552, 86598, 86816, 86818, 87289, 87390, 87487, 87781, 87787, 87792, 88491",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51376 Huggingface model not being torch scriptable, #77167 nn.functional.pad accepts bool values but raises internal assert when converted to JIT, #85513 Measure impact of JIT decompositions, reconsider the design, #86552 `max_unpool` and `max_pool` will trigger INTERNAL ASSERT FAIL in JIT, #86598 `squeeze_` fails with JIT but succeeds without it, #86816 JIT model returns different value on cpu with uniform-initialized input",d21b92d3269b81ce8470de7081d8c71bf99f71dfb37e97d60631217ed5af1a1b
issue-collapse:3ad559495bae3a44,oncall: jit needs triage  batch 24,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"21064, 37883, 58109, 69006, 84304, 88765, 89065, 89076, 89483, 90181, 90301, 90305, 90369, 90466",,,,oncall: jit,"TSRootCause:PoorIRVisibility, TSUsability, enhancement, oncall: jit, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21064 [JIT] Expose subgraph execution for intermediate output extraction, #37883 Support @property decorator in TorchScript, #58109 torch.jit.trace memory leak, #69006 [JIT] allow `Optional[torch.Generator]` be tracable in Python, #84304 op for aten::bitwise_and during torch.jit.trace, #88765 torch.jit.trace() - AttributeError: 'NoneType' object has no attribute '__module__",4dc584f5da4155803b68570b8fc443ffe4571e7084f24b1cd00be2777b3e41f9
issue-collapse:348d81a291011126,oncall: jit needs triage  batch 23,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"86662, 91098, 91229, 91358, 91360, 91396, 91504, 91562, 91686, 92230, 92375, 92548, 92558, 92563",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#86662 libtorch throws `required keyword attribute 'profiled_view_size' has the wrong type` on Linux, #91098 [TorchScript] Failed to Forward Correct Number of Arguments to Different Functions, #91229 JIT mishandles torch.round() in PyTorch 1.10.1, #91358 [JIT] .backward() not supported by JIT trace, #91360 Offer `get_buffer` and `get_submodule` in `ScriptModule`?, #91396 [JIT] Zero-channel conv2d cannot be applied with `optimize_for_inference`",e7ce99bbb37cca4987d12d67c51350d5682c1dcca135a7df6c45553eec4288e8
issue-collapse:48f5c10766201ea8,oncall: jit needs triage  batch 22,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"36910, 72498, 84340, 92654, 92674, 92740, 92742, 92866, 92888, 93255, 93826, 95194, 95497, 96098",,,,oncall: jit,"oncall: jit, small, triage review, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36910 jit.script leads to RuntimeError: attribute lookup is not defined on python value of type in some cases, #72498 torch.jit.script + torch.split + onnxruntime incompatibility, #84340 Floordiv is deprecated., #92654 Traced model output differs on C++ and Python, #92674 [JIT] `Linear` + `BatchNorm2d` Trigger Inconsistency between Eager Mode and JIT, #92740 [JIT] Applying `conv2d` over Constants Leads to Exception",6902419b60015279035624670158e60764f5edc998aa840d6cfa05db02f2db7a
issue-collapse:d8e9f48c1970959b,oncall: jit needs triage  batch 21,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"68921, 73142, 79875, 80007, 95789, 95858, 95960, 96420, 96735, 97343, 97913, 98064, 98210, 98888",,,,"oncall: jit, oncall: mobile","Stale, oncall: jit, oncall: mobile, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68921 [Bug] forward function error occurs when the scriptmodule (made in Python) including the hook function is loaded from C++(Libtorch1.8.2), #73142 Support passing Python code as a string to torch.jit.script, #79875 A/libc: Fatal signal 6 (SIGABRT), code -6 in tid 11742 (objectdetection), #80007 when forward use **kwargs，how to construct the example_ Inputs parameter in jit.trace?, #95789 drastic speed regression of torch.jit.load starting with the 20230301 nightly, #95858 [JIT] Support string type annotations in NamedTuples",a5a53e14152adfce10c00f7a60dd64d891fcfade95b40e470a7f4d22f3701a1f
issue-collapse:ae63165a23581b30,oncall: jit needs triage  batch 20,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"33491, 51969, 58789, 62221, 70058, 99821, 100636, 100730, 101055, 101855, 102653, 103250, 103788, 104041",,,,oncall: jit,"Stale, oncall: jit, triage review, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33491 jit.trace checker fails for LSTM, #51969 Support backward hooks in JIT, #58789 Type annotations for torch.jit.* decorators, #62221 Torchscript Optimize for Mobile RuntimeError: Can't deepcopy IValue with tag: Enum, #70058 Unknown builtin op: spconv::get_indice_pairs, #99821 Tracer cannot infer type of Seq2SeqLMOutput",7c2bca0e017714df306bbad4ab1798b3165d5722eb29ba593f360d2dd47ba514
issue-collapse:8704820a59f465bd,oncall: jit needs triage  batch 19,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"43520, 48082, 104450, 104458, 104712, 104732, 104906, 105196, 105900, 105917, 105918, 105997, 106435, 106699",,,,oncall: jit,"months, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43520 [jit] cannot use functools.partial with TorchScript, #48082 Microsoft C++ exception: torch::jit::ErrorReport at memory location 0x0000001C0C3BBB40., #104450 Numpy/scipy module works fine with Torch modules, but not TorchScript. How to torchscript a numpy/scipy module?, #104458 [Feature Request] Add a new overload of torch::jit::load to restore traced shape and type, #104712 torch.jit slicing error (styleganv2), #104732 torch.jit.trace says ""Arguments for call are invalid"" on torch.ops.aten.sub(3, x, alpha=3)",84358933aa683b93441ab41d72577cda1fafdc7c970ce909ee89bcca0d69bb55
issue-collapse:361c6d3e6e46fe17,oncall: jit needs triage  batch 18,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"40561, 41869, 60937, 67146, 93913, 105821, 107771, 108226, 108332, 108665, 108670, 108865, 109113, 109699",,,,oncall: jit,"days, oncall: jit, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40561 Cannot manually assign a tensor to .grad from TorchScript, #41869 TorchScript pack_padded_sequence and pad_packed_sequence run time error, #60937 [jit] About traced script module using in c++, #67146 `torch.jit.is_scripting()` not set when scripting a Module, #93913 [BUG] jit.trace not working for torchvision ViT models, #105821 JIT input aliasing does not support aten::fill_",7af1caa33ce3679d86fb8975b94b7b7966e04bb09c3537b064315d029420f599
issue-collapse:ccaae339cc4c9437,oncall: jit needs triage  batch 17,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"42885, 66930, 68568, 76295, 89114, 89185, 97047, 98695, 103070, 109488, 111135, 111792, 111863, 112425",,,,"oncall: jit, oncall: mobile","NNC, TSRootCause:DynamicBehaviors, TSRootCause:ModuleInheritance, TSUsability, actionable, oncall: jit, oncall: mobile, triage review",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42885 Unable to call `super` method with TorchScript, #66930 JIT vs. eager mismatches for jit.traced `int8` to `int32` casting, #68568 About ModuleDict indexing when save model based on ""torch.jit.script()"" method, #76295 TorchScript: jit.script fails when using 'tolist' with int32 tensors., #89114 jit.script() fails to resolve/cast Optional[Tensor] fields of sub-modules or base classes of the object being scripted, #89185 [feature request] Add ability to preserve traced shape during torch.jit.save and torch.jit.load",4971dea1fea07c52658bf53d325f7ae3c3be2fbadcee4af81f9721c8a8c0da1a
issue-collapse:0cef11c6671e7b47,oncall: jit needs triage  batch 16,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"41909, 55603, 62962, 75788, 90885, 114543, 115622, 115639, 115733, 115736, 115914, 116010, 116498, 116612",,,,oncall: jit,"NNC, oncall: jit",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41909 Dropout in TorchScript under torch.no_grad() produces incoherent results, #55603 Faster Inference time for first .pt model prediction, #62962 [JIT] NVRTC unknown error, #75788 `torch.jit.script` Script functions do return `requires_grad = False` if `torch.no_grad()` has been used, #90885 TorchScript with complex abs doesn't work in backward, #114543 [torch.jit.script] Torchscript produces incorrect result when argmax result is used in indexing",17f0892b7dea12b25cddb992a46f4b09206d15b7850856b1ff776f9da4688f14
issue-collapse:a35728c302a5844f,oncall: jit needs triage  batch 15,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"48325, 49538, 54140, 82417, 113945, 117541, 118086, 118189, 118297, 118300, 118302, 118303, 118400, 118401",,,,oncall: jit,"days, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#48325 [JIT] Constant Propagation Shouldn't Run on Values that Escape the Graph, #49538 jit tracer doesn't work with unflatten layer, #54140 [Torchscript] Error detected in torch::jit::(anonymous namespace)::DifferentiableGraphBackward on custom RNN model, #82417 [JIT] _unsafe_view returns alias when size(input) = size argument, #113945 torch.jit.trace has incorrect execution for += operation during compilation, #117541 Jit model running on cuda:1 uses memory on cuda:0",3d7ad45a93e1662a46db1d629cd5cc7db4e36f0658dfb8d143a4474a06113c36
issue-collapse:325e9df1545f1a2f,oncall: jit needs triage  batch 14,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"14867, 43388, 52363, 58845, 66017, 69425, 70051, 75556, 87145, 89386, 119192, 120485, 122020, 123066",,,,"oncall: jit, oncall: mobile","Stale, oncall: jit, oncall: mobile",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14867 [C++/Pytorch] Get input shapes, #43388 Abort message: ‘terminating with uncaught exception of type c10::Error: _ivalue_ INTERNAL ASSERT FAILED at ../torch/csrc/jit/api/object.cpp:19, please report a bug to PyTorch. (_ivalue at ../torch/csrc/jit/api/object.cpp:19), #52363 RuntimeError: Can't redefine method: forward on class: __torch__.torch.nn.quantized.modules.conv.Conv1d (of Python compilation unit at: 0x5571a223a770), #58845 Use jit in flask, #66017 Torchscript `jit.script` breaks with OSError if dataclass used to pass results during model execution, #69425 Integrate Libtorch into Unreal Engine 4: _ivalue_INTERNAL ASSERT FAILED",52ffd3b7d58a20a353a45000760cb453a784590b392499e1d71fc8b1ee2e656a
issue-collapse:4ece20c1512f9d36,oncall: jit needs triage  batch 13,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"59530, 66623, 67983, 76881, 80861, 123961, 124458, 125003, 125293, 125749, 126167, 126277, 126843, 127679",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#59530 Batch size is hardcoded using torch.jit.trace with LSTMCell, #66623 Constant folding in symbolic shape inference fails: expected scalar type Long but found Float, #67983 torch.jit.export does not working, #76881 RecursionError when running torch.jit.script inside JitTestCase, #80861 jit.freeze throws RuntimeError: stack_out && stack_out->size() == 1 INTERNAL ASSERT FAILED at ""../torch/csrc/jit/passes/frozen_conv_folding.cpp"":281, #123961 JIT reports RuntimeError when decorating forward with torch.jit.script",b903b01610c88ce9413eb9383c2bc00fa9159717c123e227686498901d747342
issue-collapse:e081dfee77c09802,oncall: jit needs triage  batch 12,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"42207, 48387, 76486, 89064, 104405, 127283, 128019, 128125, 128261, 128491, 128496, 128585, 129282, 129347",,,,"oncall: jit, oncall: mobile","months, oncall: jit, oncall: mobile",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42207 torchscript does not support unicode strings, #48387 Error occurred when trying to call ""torch::jit::load""method for loading "".pt"" file on IOS, #76486 [Tracer] RuntimeError: _Map_base::at when tracing fake quantization, #89064 UserWarning: The TorchScript type system doesn't support instance-level annotations on empty non-base types in `__init__`., #104405 Detailed error: Tensor-likes are not close! When use torch.jit.trace, #127283 [BUG][JIT] `torch.jit.script` is not compatible with `DeprecationWarning` and `FutureWarning`",100e970dac4442e30577c859735c2fb06a048a0fd1d9d733ab0b6117c0a0b3ae
issue-collapse:f7556008439d79ed,oncall: jit needs triage  batch 11,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"51328, 67834, 123168, 130726, 130735, 131022, 132315, 132385, 133537, 133675, 133678, 133795, 133840, 133847",,,,"oncall: jit, oncall: mobile, oncall: quantization","actionable, oncall: jit, oncall: mobile, oncall: quantization, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51328 Support torch::jit::script::Module converting to torch::nn::Module., #67834 Model tracing not working, #123168 TorchScript: Union cannot be used as a tuple, #130726 TorchDispatchMode fails on jit.trace, #130735 Problem with combination of jacref, vjp, tanh in torchscript, #131022 Code bug in call_prepare_scriptable_func_impl function.",5308f137df1e038154d2de64c4c270f65470fbbd3d5a59993386353f971a27f9
issue-collapse:5c03ced14eae3e41,oncall: jit needs triage  batch 10,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"40870, 58298, 82099, 93943, 130229, 134559, 134562, 134849, 134898, 134903, 134918, 135580, 136127, 136256",,,,oncall: jit,"TSRootCause:UnsupportedConstructs, TSUsability, oncall: jit, weeks",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40870 [JIT] cannot statically infer the expected size of a list in this context for positional arguments to Tensor.view(), #58298 [JIT] Can't handle creating new objects in compiled methods: ""Class does not have an __init__ function defined"", #82099 UserWarning: operator() sees varying value in profiling, #93943 `torch.jit.trace` memory usage increase although forward is constant, and gets much slower than forward with model depth increase, #130229 RuntimeError: 0 INTERNAL ASSERT FAILED at ""../torch/csrc/jit/ir/alias_analysis.cpp"":615, please report a bug to PyTorch. We don't have an op for aten::full but it isn't a special case. Argument types: int[], bool, int, NoneType, Device, bool,, #134559 [JIT][foreach_add] Arguments for call are not valid",e41eda084e1019358a84e6caadc6a9e43742f5c19c847109c029f642089d2725
issue-collapse:c0038266542e16cb,oncall: export needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"121191, 122934, 123345, 125007, 126291, 126770, 127357, 127643, 128385, 130555, 130820, 132459, 133255, 139470",,,,oncall: export,"better-engineering, export-triage-review, export-triaged, oncall: export, onnx-needs-info",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#121191 Export-based GraphModules fail with deepcopy, #122934 Cannot export DETR model with torch.export, #123345 torch.export does not support Sparse CSR tensors, #125007 Support exporting the compute graph JiT compiled by inductor?, #126291 Pytorch -> ExecuTorch conversion fails with BatchNorm2d layer, #126770 _convert_input_to_fake doesn't work when there are no inputs",ae88d4cea4806847d4058a24531ccbd3949e51e1b7e4002803cb6d78425ac975
issue-collapse:080feea074695cc3,oncall: export needs triage  batch 3,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"127571, 141116, 141893, 141996, 143379, 144331, 145456, 146315, 146576, 147187, 152283, 152465, 154853, 155550",,,,"oncall: export, oncall: pt2","export-triaged, oncall: export, oncall: pt2",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#127571 [export] Cannot mutate tensors with frozen storage, #141116 How to fuse batchnorm to conv2d in the graph exported by torch.export, #141893 Forward from pytorch/executorch: export error with Macbook pro M4, #141996 [export] Move pytree serialization into export serializer, #143379 Running decompositions over torch.export program adds one more output to the graph result, #144331 [Export] UserWarning: Attempted to insert a get_attr Node with no underlying reference in the owning GraphModule",3b73340c588c29a8dc041773001eba65fde2fe3a98faf0dd19dcbcb90ddfe65e
issue-collapse:93f77fdc26e3edf4,oncall: export needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"153429, 155862, 156882, 157625, 159316, 163791, 164083, 165395, 167730, 169009, 170548, 170673, 170783, 171516",,,,"oncall: export, oncall: pt2","export-triaged, feature, oncall: export, oncall: pt2",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#153429 [Export] Non-strict mode can't handle conditionals on tensor subclass types, #155862 Export Huggingface models with StaticCache, #156882 Export always give a value range with max length - 1, #157625 Is there some official method to extract the featuremap of each node in pt2 graph like the function torchvision.models.feature_extraction.create_feature_extractor(), #159316 Support for non-scalar valued Loss Tensor in `_export_forward_backward()`, #163791 torch.export fails on Phi-3.5 model",37412164a80d61ef736faa83dc7728a7dad3333471c83d3347155beb571b22c1
issue-collapse:b146907fc3e3f344,oncall: distributed needs triage  batch 5,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"35372, 51574, 51931, 51932, 54912, 56111, 58982, 59441, 59823, 63455, 64130, 64150, 64799, 65052",,,,oncall: distributed,"better-engineering, has workaround, oncall: distributed, pipeline parallelism",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35372 How to support single-process-multiple-devices in DistributedDataParallel other than CUDA device, #51574 [Pipe] conversion to Sequential: various complex scenarios, #51931 [Pipe] Tied layers, #51932 [Pipe] Memory-Efficient Model Construction, #54912 [autocast] DataParallel in a single process - possibly outdated docs, #56111 ProcessGroupGlooTest and other distributed tests don't capture stdout/stderr in test UI view",26284f6cbc8896ad90772b19fe24f6da9d1958597a5bb75262c01d02904fd249
issue-collapse:f7c4754f57f62af0,oncall: distributed needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"60518, 62381, 62474, 64419, 65907, 65918, 66547, 67972, 68256, 68785, 69522, 69563, 70037, 70099",,,,"oncall: distributed, oncall: r2p","Stale, better-engineering, oncall: distributed, oncall: r2p, pt_distributed_rampup",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#60518 different gpus to train, #62381 torch.distributed and subprocess do not work together?, #62474 Distributed: RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [2048]] is at version 4; expected version 3 instead, #64419 Windows Torch.distributed Multi-GPU training with Gloo backend not working, #65907 Inplace error of BatchNorm layer in DistributedDataParallel module, #65918 C++ extensions don't correctly set USE_C10D_FOO preprocessor macros",ad08dec561a18970b19ba7b67c5b7ef606995bfbb989bf0a3d2bf2eb4bd367ff
issue-collapse:ba9f968c594bb247,module: android needs triage  batch 1,batch action queue,needs triage,47,medium,bucket_component_queue,14,0,0,0,"56703, 57271, 59690, 60172, 61709, 61951, 62165, 62676, 62821, 64333, 65058, 65734, 66545, 68330",,,,"module: android, module: arm, module: fft, module: ios, module: memory usage, module: vulkan, oncall: jit, oncall: mobile","enhancement, mobile_perf, module: android, module: arm, module: fft, module: ios, module: memory usage, module: vulkan, oncall: jit, oncall: mobile",,"Review this module: android queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#56703 [Pytorch Mobile] Error running build_pytorch_android.sh, #57271 RuntimeError after optimize_for_mobile, #59690 Inference ran on new thread leak memory on Android, #60172 [Android] Upgrading to 1.9.0 causes NNAPI model loading to fail, #61709 [pytorch android] use Vulkan backend crash, #61951 Vulkan: error -5 when trying to use it with Nvidia card",b0a6477b16cc5d8d67fd285b2f2f02a8d5e068d79b4bfec147b4e1179f626a9b
issue-collapse:4c3355dfae735efb,oncall: jit needs triage  batch 52,batch action queue,needs triage,47,medium,bucket_component_queue,13,0,0,0,"6007, 6010, 8453, 8556, 11532, 13170, 13477, 14332, 14484, 15043, 16263, 18376, 18975",,,,oncall: jit,oncall: jit,,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#6007 Clean up the extra copies that occur in the execution engine and other places, #6010 Develop a strategy for writing leak tests for pytorch., #8453 [JIT] Generalize checkTrace() to also compare gradients of parameters, not just inputs, #8556 [JIT] Interleaved C++-Python execution loses inner Python stacks, #11532 [JIT][tracer] Slicing shape is specialized to tensor rank, #13170 Allow traced modules to return dictionaries",aa21bc6c541782f0c45a2495b40251e0b92fcd78afd9096f163711a27728b4c3
issue-collapse:8a540612fb26e1dc,oncall: quantization needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,12,0,0,0,"116250, 116251, 116252, 116253, 116865, 116992, 117102, 117229, 117239, 117920, 120705, 122224",,,,oncall: quantization,oncall: quantization,,"Review this oncall: quantization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#116250 `torch.Tensor.rot90`: Heap-buffer-overflow, #116251 `torch.Tensor.flip`: Heap-buffer-overflow, #116252 `torch.Tensor.fliplr`: Heap-buffer-overflow, #116253 `torch.Tensor.flipud`: Heap-buffer-overflow, #116865 Optimizing Per-Channel Quantization for Improved Inference Performance, #116992 torch.fake_quantize_per_tensor_affine with inf input provides wrong output value",56667e017f1f10f370164fa043481d66ab8b3a71e287af1163e82592be4f6ccf
issue-collapse:8dcbc44b9d2a76b9,oncall: profiler needs triage  batch 4,batch action queue,needs triage,47,medium,bucket_component_queue,9,0,0,0,"69506, 70951, 74605, 76309, 82951, 83769, 91311, 91903, 98124",,,,oncall: profiler,"Stale, oncall: profiler",,"Review this oncall: profiler queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#69506 [performance] a profiler util to show a rough break-down of the types of ops used by a model, #70951 A diagnostics mode to report when constraints aren't being met for optimal performance, #74605 torch.profiler.profile does't work well for CPU model when not using torch.profiler.schedule, #76309 More understandable name column of the table of the profiling result., #82951 torch.profiler's FLOPs measure only counts operations involving '+' and '*' ., #83769 [TorchTidy] Check if `set_to_none` would change optimizer semantics.",12e490b9fd5304ebc87c73d54ca41b4c038a4174ba1eca83eaec72ca63146d96
issue-collapse:d8437721bd1b7a98,oncall: visualization needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,8,0,0,0,"30863, 31603, 43323, 43328, 47265, 49823, 56010, 57163",,,,oncall: visualization,oncall: visualization,,"Review this oncall: visualization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#30863 [TensorBoard] Installed tensorboard 1.14.0, TestTensorBoardWriter.test_writer failed., #31603 Typo in `torch.utils.tensorboard.add_image`, #43323 Set SummaryWriter step globally, #43328 torch.utils.tensorboard.SummaryWriter.add_histogram() execution time explodes with each epoch., #47265 [Feature Request] tensorboard add_graph support for multiple method, #49823 Why trace method used insted of script method in graph creation?",cc53cd5c7fd1bf4edeca46944137bd5589eda859f4f4accf7af927286d9b8a1f
issue-collapse:181cb301ddf32d56,module: bootcamp needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,8,0,0,0,"37387, 41419, 43409, 48045, 60062, 63171, 63247, 63403",,,,"module: bootcamp, module: ddp, module: tensorboard, oncall: distributed, oncall: jit, oncall: visualization","NNC, caffe2, days, feature, module: bootcamp, module: ddp, module: tensorboard, oncall: distributed, oncall: jit, oncall: visualization, pt_distributed_rampup",,"Review this module: bootcamp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#37387 The pytorch's graph is lack of common names for nodes, #41419 [jit] Support NamedTuple in tracing, #43409 Switch C10_EXPORT_CAFFE2_OP_TO_C10 to new operator registration API, #48045 [NNC] Improve ""UNSUPPORTED DTYPE"" error messages, #60062 [c10d] Make `broadcast_object_list` accept a device parameter, #63171 [BE] Clean up DDP Single-Program Multi-Device Vestige in reducer.cpp",bb586319f9f12ec583029f77c44ea9e45d33b7f42de4a9640f9973929effd012
issue-collapse:5346e77190aa6bba,oncall: export needs triage  batch 5,batch action queue,needs triage,47,medium,bucket_component_queue,7,0,0,0,"115055, 115488, 116544, 118410, 119484, 124507, 124796",,,,"oncall: export, oncall: pt2","export-triage-review, export-triaged, oncall: export, oncall: pt2, onnx-needs-info",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#115055 Better test suite for torch.export.export, #115488 torch.export.export is overriding source_fn inside higher order op submodules, #116544 [export] namedtuple inputs can't be retraced, #118410 [export] Error loudly when tracing over a module not in the module hierarchy, #119484 Follow-up #119356: Preserve node names for torch.export.export, #124507 torch.export of ResNet with dynamic height fails due to constraint violation",cd5c433a188cce4f30dba57d6dac08a7901dd75433c8bf2647abef2ed1bced59
issue-collapse:6f936c64b4c74a15,oncall: cpu inductor needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,7,0,0,0,"124697, 135569, 140689, 143886, 144039, 146824, 146974",,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2",,"Review this oncall: cpu inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#124697 [INDUCTOR] [CPU] [GPT-FAST-MOE] large perf regression with coordinate_descent_tuning disabled, #135569 [inductor][cpu]AOT inductor performance regression and accuracy failure in 2024-09-04 nightly release, #140689 [inductor][cpu]opacus_cifar10 and functorch_dp_cifar10 AMP performance regression in 2024-11-11 nightly release, #143886 [RFC] Identifying dynamic int8 symmetric vs asymmetric quantization of activation/input in Inductor-CPU, #144039 [Mac/M1] torch.compile() -- expm1 returns an inaccurate result compared to the interpreted version, #146824 Inductor-CPU might load (and store) fewer elements than the vector-width",60aada32900d4621fb685e25530c3fe4e92de5caeddec3829a41644ade27514d
issue-collapse:f601f234896efaae,oncall: package/deploy needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,6,0,0,0,"60089, 60189, 64462, 72659, 74909, 74956",,,,oncall: package/deploy,"imported, oncall: package/deploy",,"Review this oncall: package/deploy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#60089 torch.package: TypeError: 'NoneType' object is not iterable, #60189 [package] Tutorial not accessible, #64462 torch.package: re-export docs suggestions don't work + other sharp edges, #72659 Add a unit test which uses an extension module + ordered importer, #74909 torch.package fails to import if dataclass and __future__.annotations present, #74956 Allow `torch.package` to accept `*.pyc` files",49aa5ddfff8895304dd31f7148dc893fe66465ee89ec01ad27885424d7b6de0e
issue-collapse:f511de5245fc80f3,module: ios needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,6,0,0,0,"33285, 59551, 66768, 72330, 82216, 85149",,,,"module: ios, module: macos, oncall: mobile","module: ios, module: macos, oncall: mobile, shadow review",,"Review this module: ios queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#33285 iOS libtorch superpoint model bug, #59551 error when trying to call ""torch::jit::load""method, use Metal backend in PyTorch Mobile, #66768 Error: `copy_to_metal_ is implemented only for float dtype`, #72330 Add and Mul torch tensors on Metal (IOS), #82216 iOS TestApp from mobile performance recipes tutorial doesn't build on macOS, #85149 Pytorch on iOS (iPhone X & XR) throwing can't allocate memory exception. Ref Logs:",f00875f2582125bcb4053567d632d6cf583b3b155375b05134710997482e5d89
issue-collapse:d077a61ee1133e72,module: ddp needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,6,0,0,0,"48478, 62403, 62542, 63445, 68297, 68862",,,,"module: ddp, module: nccl, oncall: distributed","better-engineering, module: ddp, module: nccl, oncall: distributed, pt_distributed_rampup",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#48478 Callbacks of Futures shouldn't wait inline on another future, #62403 DDP, #62542 3 Times memory cost when loading the model to torch.nn.parallel.DistributedDataParallel, #63445 getNcclVersion() in NCCLUtils.cpp handles ""2.10.3"" incorrectly, #68297 [DDP] Debug mode should ensure reduction is finished in backward pass, #68862 find_unused_parameters of DDP",be3f6c973434df34f311646ccdf63c7d8c014fd93fc807b7f1336638dd9c4a26
issue-collapse:ec468fcd63ebd5cb,module: android needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,6,0,0,0,"18380, 40906, 50531, 52970, 59865, 64488",,,,"module: android, module: docs, module: serialization, oncall: mobile","caffe2, module: android, module: docs, module: serialization, oncall: mobile",,"Review this module: android queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#18380 C++ inference (CPU-only) stalls in Android, and crashes on Mac/Linux, #40906 will you support SRU on mobile?, #50531 PyTorch mobile perf recipes, fails when building android, #52970 Wrong link in readme for the android nightly version, #59865 Custom build of c++ libtorch for Android program or dynamic so lib does not reduce program or lib size, #64488 It' so strange!!!",e2d30868537aa2ca9bccfbbf09cc54ebacebe52e7bcf39209eb767c3e1219c61
issue-collapse:bc32943ce1c0e619,oncall: java needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,5,0,0,0,"33631, 49536, 78483, 83551, 89868",,,,"oncall: java, oncall: mobile","oncall: java, oncall: mobile",,"Review this oncall: java queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#33631 bitmapToFloat32Tensor() 1 channel Tensor [feature] [mobile], #49536 Pytorch mobile Vulkan API bug during call to ReLU, #78483 is the issue resolved? windows not pytorch_jni in path, #83551 OpenJDK libtorch_cpu.so stack guard warning, #89868 No pytorch_jni.dll file in libtorch 1.13.0 lib folder",e7f12f806e2ae955364bcd89a60efce8d09dae1f026cb13ee15cd4180b1650c0
issue-collapse:7afd4be1da6382b0,module: vulkan needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,4,0,0,0,"56006, 82065, 84831, 116662",,,,"module: vulkan, oncall: jit, oncall: mobile","module: vulkan, oncall: jit, oncall: mobile",,"Review this module: vulkan queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#56006 Unable to get a Vulkan tensor using `to('vulkan')`, #82065 optimize_for_mobile vulkan_prepack::conv2d_clamp_prepack, #84831 INTERNAL ASSERT FAILED for _jit_pass_vulkan_optimize_for_mobile (Google Colab), #116662 test/test_vulkan.py fails with vulkan_prepack::conv2d_clamp_prepack not found",27ea9b6ff26c96f167e9f40b7c00a8d4eb9813326cfc4449a5fdd21a9ed74fb2
issue-collapse:c422c81edb5383d2,module: amp (automated mixed precision) needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,4,0,0,0,"50231, 59257, 69616, 104959",,,,"module: amp (automated mixed precision), module: cuda, module: data parallel, oncall: distributed, oncall: jit","Stale, has workaround, module: amp (automated mixed precision), module: cuda, module: data parallel, oncall: distributed, oncall: jit",,"Review this module: amp (automated mixed precision) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#50231 Tracing with autocast failed with error: ""Cannot insert a Tensor that requires grad as a constant"", #59257 Optimizer step stuck during distributed training, #69616 Multi-GPU training stuck when using grad scaler, #104959 Saving a LightningModule torch.jit.ScriptModule is incompatible with torch.amp.autocast",ccbf134946db9afbb782a62ea28923483f817cdecdf40b0dc06455f06df27544
issue-collapse:7ae4913617c66793,oncall: distributed needs triage  batch 6,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"8818, 48710, 48928",,,,oncall: distributed,oncall: distributed,,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#8818 MPI causing job to hang --- unresponsive to external (termination) signals, #48710 RuntimeError: CUDNN_STATUS_BAD_PARAM when testing environment, #48928 When multiple GPUs run multiple processes, it is found that any process not running in GPU 0 will have some more memory (such as 200m) in GPU 0. What is the cause of this?（多个GPU跑多进程时候，发现只要不在0号GPU跑的进程都会在0号GPU多出一些内存(如200M)，请问这是什么情况导致的？）",137fe0c31e095a8f5d1e538ce9fcc346f921932c003b6b722ac677cf7a747ec0
issue-collapse:7cbd62e7b670a9ba,module: nvfuser needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"75903, 80606, 94908",,,,"module: nvfuser, oncall: jit","NNC, module: nvfuser, oncall: jit",,"Review this module: nvfuser queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#75903 torch.jit.script'd function very slow on first invocation on latest nightly, #80606 [jit.script] jit.script give uncertain results using torch.half, #94908 ASSERT(initialized()) Debug Error after JIT fusion on Windows",6d04a88108ac18d034b767f091bf6e6f0eb8766d82584cd04efc9f82936e8df0
issue-collapse:eaad8f579abbe8e8,module: mkldnn needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"56024, 61267, 86667",,,,"module: mkldnn, oncall: distributed, oncall: jit","module: mkldnn, oncall: distributed, oncall: jit, triage review",,"Review this module: mkldnn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#56024 Support for mkldnn + ddp, #61267 Incorrect trace with MKLDNN (adding scalar), #86667 Adding a linear layer leads to failure of `optimize_for_mobile`",e9221b79ebc8e5dfdf292ebf98368f4909301eb7d5fa85fd23c5aa6a0c3a758f
issue-collapse:30ad9e2397acf9a1,module: dynamic shapes needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"126127, 132930, 153179",,,,"module: dynamic shapes, module: error checking, oncall: export, oncall: pt2","export-triaged, module: dynamic shapes, module: error checking, oncall: export, oncall: pt2",,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#126127 torch.export.export() fails with dynamic shapes when more than one shape is dynamic, #132930 Treating SymNode compute more fungibly for the purpose of export flattener (or not?), #153179 export: No error raised on data-dependent assertion.",d8eec5031530cf229de80cb359b0a0cb3ec9acf7070d08a2585c4721ab847490
issue-collapse:c3c918fafc1cc02c,module: dataloader needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"62885, 64986, 69352",,,,"module: dataloader, oncall: distributed","module: dataloader, oncall: distributed",,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#62885 Type of first constructor parameter of ElasticDistributedSampler and DistributedSampler, #64986 RandomSampler / DistributedSampler does not seem really random, #69352 I want to know how to read the LMDB file once when using DDP",a1f3b6c3c015b081263908bb783e42038e8ab04e4afc7ce9bcdd79d06a0c1bc8
issue-collapse:b84dd43b1f1739be,module: correctness (silent) needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,3,0,0,0,"86597, 86616, 153315",,,,"module: correctness (silent), oncall: jit","module: correctness (silent), oncall: jit",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#86597 JIT returns different values for `cos + frac` on cpu, #86616 JIT returns different values for a model on cuda and returns a strange error message on cpu, #153315 [JIT] Compilation-induced discrepancy in F.instance_norm when passing input as running stats",bd63018d04ace133c200fe63ba0d908afec9c67ad94527d04445da41262dc4c2
issue-collapse:afe3cae890f50502,module: windows needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"146127, 161815",,,,"module: windows, oncall: jit, oncall: releng","better-engineering, module: windows, oncall: jit, oncall: releng",,"Review this module: windows queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#146127 When I try to run rt-detr model in C++ libtorch i face the given error, #161815 Windows AMI release issues and improvements",03117af9f2cb0757534b67637258b3709b09733582b8f0d16fc215d5ee7ea08e
issue-collapse:2b94d12b532fb9ee,module: vision needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"82598, 91738",,,,"module: vision, oncall: fx, oncall: jit","module: vision, oncall: fx, oncall: jit, triage review, triaged",,"Review this module: vision queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#82598 Deep copy models with `create_feature_extractor` produces different parameters, #91738 tracing torchvision detection model results in an error",8e46b6bb4a63a9f5c3d6dce560f3a03b4c31967332d0c9dad9b49c2f91b6cbc3
issue-collapse:86ea587471b0b711,module: tensorboard needs triage  batch 2,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"35095, 37740",,,,"module: tensorboard, oncall: visualization","module: tensorboard, oncall: visualization",,"Review this module: tensorboard queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#35095 tensorboard add_graph 's ""operator_export_type"", #37740 torch Summary writer does not display torchvision.io.read_video output",4567cdb9626642bce22d9a15efd766bbde39a5cf2682e411f9ee3c9a1e339a8f
issue-collapse:190dcc69394b2d43,module: serialization needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"38273, 57716",,,,"module: serialization, module: vision, oncall: jit","module: serialization, module: vision, oncall: jit, triage review",,"Review this module: serialization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#38273 Adding model metadata in TorchScript model file, #57716 Unknown builtin op: torchvision::nms in LibTorch",72db9fde6b822d173b50deb8008310e03226e42f7479521360a6f0491af920d9
issue-collapse:8c551a08cf8c51f7,module: python frontend needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"106764, 142170",,,,"module: python frontend, oncall: jit, oncall: package/deploy","feature, module: python frontend, oncall: jit, oncall: package/deploy, triage review",,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#106764 [Feature request] Add new API Tensor.device_as, #142170 [3.13] Fix failing torch.package dynamo-wrapped test",c9c88a06cd54c35d5c2f74677619a3dbbbd78e61d3bb2a4822d9ea9477bc8230
issue-collapse:3f8b4d8b09f9d3f0,module: power needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"41242, 110332",,,,"module: power, oncall: jit","module: POWER, oncall: jit",,"Review this module: power queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#41242 test_backward_deadlock fails with ""Directory not empty"", #110332 `torch.jit.load()` might unresponsive in IBM s390x when loading some certain torchscript saved by x86 machine.",68e6f1704e3a1c1fcf31434d9de8c5df30cede42bc6cfdf3b7a4b896f74f03f0
issue-collapse:cdc7381d7e57c28e,module: multithreading needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"52047, 64795",,,,"module: multithreading, module: tests, oncall: distributed, oncall: jit, oncall: profiler","days, module: multithreading, module: tests, oncall: distributed, oncall: jit, oncall: profiler",,"Review this module: multithreading queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#52047 Detect OpenMP Loop and this application may hang warnings, #64795 profiler crashes the program when trying to stop on different threads without context manager",c0f50b6e8d8722436c0103c1f10ac421dfb8f6f83356084b4eadffafebd71ec5
issue-collapse:0f60384ec8d8f942,module: lazy needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"63610, 64005",,,,"module: lazy, oncall: jit","module: lazy, oncall: jit",,"Review this module: lazy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#63610 Enable some amount of CI tests on lazy_tensor_staging branch, #64005 [LTC] Find a way to convert at::Generator into torch::jit::NamedValue",b145c3a0c7c013f54edd4fe92f9a87a9c56b245325fb1db1b9c6efc359754d2b
issue-collapse:6ea5cabe95500135,module: flaky-tests needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"158331, 164534",,,,"module: flaky-tests, module: unknown, oncall: cpu inductor, oncall: jit","module: flaky-tests, module: unknown, oncall: cpu inductor, oncall: jit, skipped",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#158331 DISABLED test_conv2d_unary_cpu_cpp_wrapper (__main__.TestCppWrapper), #164534 DISABLED RecordDebugHandles.ScopedCallbacks (__main__.test_jit)",4d395e4245f4711445de7bdcabfd6b911c540c80137f0fc665bb95d4efc1952f
issue-collapse:e98f1e941753f951,module: deadlock needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"67241, 145743",,,,"module: deadlock, module: nccl, oncall: distributed, oncall: jit","module: deadlock, module: nccl, oncall: distributed, oncall: jit",,"Review this module: deadlock queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#67241 init_process_group hangs for multi-node, Pytorch > v1.3.1 and file init_method, #145743 TorchScript run_method fails from 2.5.0 onward on Ubuntu",d6c0ac14e1b3d20c5dc178e942cb32e5cb5c25746c4d511e6012bbcec98ed300
issue-collapse:7f3c920c2647af65,module: custom-operators needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"33110, 51994",,,,"module: custom-operators, oncall: jit","actionable, feature, module: custom-operators, oncall: jit, weeks",,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#33110 InlineAutodiffSubgraphs in JIT inlines non-differentiable custom groups unexpectedly., #51994 Allow Custom Classes to register a handler for .to operations",7819077e395c08fbe9487aee13bd0d17cfccbbd35f93b5727687079526e4b434
issue-collapse:bd4b6640d5b616c0,module: complex needs triage,batch action queue,needs triage,47,medium,bucket_component_queue,2,0,0,0,"154324, 156814",,,,"module: complex, oncall: jit","module: complex, oncall: jit",,"Review this module: complex queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#154324 torch.jit.script gives false results for autograd if complex data types are involved, #156814 torchscript: RuntimeError Unsupported dtype complex",b18b177d07bbb041a68c399a6a769d5b447437e904ee7ad95f6f9e743fd655e0
issue-collapse:d7773349157a27d6,module: autograd has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,13,0,0,0,"67887, 80025, 91898, 94336, 144580, 160016, 162350, 174017, 175324, 184572, 184783, 185331, 186556",,,,"module: autograd, module: correctness (silent), module: derivatives, module: distributions, module: docs, module: edge cases, module: forward ad, module: mta, module: multiprocessing, module: multithreading, module: pipelining, module: scatter & gather ops, module: typing, module: windows, oncall: distributed, oncall: distributed parallelisms","actionable, better-engineering, bot-triaged, enhancement, module: autograd, module: correctness (silent), module: derivatives, module: distributions, module: docs, module: edge cases, module: forward ad, module: mta, module: multiprocessing, module: multithreading, module: pipelining, module: scatter & gather ops, module: typing, module: windows, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#67887 Add more explanation on multithreaded graph building of Autograd, #80025 `torch.special.gammainc` backward pass with respect to the first argument, #91898 Lazily start worker threads in the autograd engine, #94336 `scatter` fails the gradient computation in reverse mode for `src` when `index` is empty, #144580 `torch._foreach_mul` does not support autograd, #160016 Requesting improved torch.auto_grad.detect_anomaly() for NaN detection",32a99913fc1e3b629644cf33c86cc4ff51338c69b00c740b4c5aef754472377f
issue-collapse:39fc022c31bc85da,module: error checking has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,11,0,0,0,"126673, 134841, 137468, 152178, 169236, 175368, 176244, 181576, 184280, 185006, 185481",,,,"module: error checking, module: inductor, module: lrscheduler, module: mps, module: nn, module: norms and normalization, module: optimizer, module: python array api, module: python frontend, oncall: jit, oncall: pt2, oncall: quantization","actionable, bot-triaged, enhancement, low priority, module: LrScheduler, module: error checking, module: inductor, module: mps, module: nn, module: norms and normalization, module: optimizer, module: python array api, module: python frontend, oncall: jit, oncall: pt2, oncall: quantization, topic: fuzzer, triaged",,"Review this module: error checking queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#126673 torch.nn.AdaptiveAvgPool2d lacks checking of input dimension, #134841 LPPool2d lacks a check for the validity of norm_type, #137468 overflow in `torch.quantize_per_tensor`, #152178 Raise an Error when File Not Found in `torch.jit.load()`, #169236 [MPS] nn.ConvTranspose{1,2,3}d ignore validation check for output_padding < stride (Silent Failure), #175368 `F.embedding_bag` segfaults when intermediate offsets exceed indices length",6358dbb34f69d537cbe10f919f8409430f331fd7bd5cbfb7abf94f2ec5343587
issue-collapse:74b738551d61d7c7,module: nn has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,10,0,0,0,"73009, 104860, 109652, 110865, 128712, 176381, 184037, 185022, 185023, 188166",,,,"module: nn, module: typing","actionable, bot-triaged, module: nn, module: typing, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#73009 Improve loading for nn.modules.lazy.LazyModuleMixin, #104860 torch.nn.Conv2d/Conv1d's padding mode circular cannot accept 3-dim input, #109652 InstanceNorm does not catch dim mismatch, #110865 BCEWithLogitsLoss: Check if labels / targets are within zero and one, #128712 Add Swiglu activation function, #176381 Output difference anomaly under equivalent migration in RReLU operator",2d8e31092f07b602e13f415e3051cd02d8a26adc097e6baa84172b669bc9d263
issue-collapse:3288b6b971217d9e,module: correctness (silent) has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,9,0,0,0,"179608, 181806, 184031, 185246, 185248, 185464, 185696, 187295, 187296",,,,"module: correctness (silent), module: cuda, module: decompositions, module: dynamo, module: edge cases, module: error checking, module: half, module: inductor, module: loss, module: mps, module: nans and infs, module: nn, module: numpy, module: pooling, module: random, module: testing, module: trigonometric functions, oncall: pt2","actionable, bot-mislabeled, bot-triaged, module: NaNs and Infs, module: correctness (silent), module: cuda, module: decompositions, module: dynamo, module: edge cases, module: error checking, module: half, module: inductor, module: loss, module: mps, module: nn, module: numpy, module: pooling, module: random, module: testing, module: trigonometric functions, oncall: pt2, topic: fuzzer, triaged",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#179608 [MPS] avg_pool1d produces negative values from non-negative input when zeros follow large floats, #181806 CUDA torch.signbit returns false for negative float16 NaN, #184031 torch.nn.functional.gumbel_softmax returns NaN when tau=0 (division by zero), #185246 torch.compile silently accepts invalid Bernoulli probabilities that eager mode rejects, #185248 `torch.compile` silently accepts negative standard deviations in `torch.normal` that eager mode rejects, #185464 torch.compile produces wrong results for F.multilabel_margin_loss when target contains -1",9bca32c25b4751874ca0df2137177de8afcb60feba07e5cb17bacf0fe3d4933d
issue-collapse:df3c209f55d63d59,module: edge cases has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,7,0,0,0,"154262, 171249, 174668, 174984, 186901, 187294, 187299",,,,"module: edge cases, module: empty tensor, module: error checking, module: fft, module: half, module: intel, module: nans and infs, module: nn, module: padding, module: special","actionable, bot-triaged, enhancement, module: NaNs and Infs, module: edge cases, module: empty tensor, module: error checking, module: fft, module: half, module: intel, module: nn, module: padding, module: special, topic: fuzzer, triaged",,"Review this module: edge cases queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#154262 max_pool2d padding assert incorrect with dilation, #171249 [Bug] softplus returns inf for large beta while other frameworks return finite value, #174668 torch.histc is not consistent, #174984 `torch.fft.rfft` errors out with MKL FFT error when given an empty tensor, #186901 Inconsistent behavior of torch.sign on NaN input values, #187294 torch.special.zeta returns NaN for q=+Inf instead of 0",cd71fc8b59eea2f9b2e060d3ecbb037067933fe7bbb23cacab9acca5564f15bb
issue-collapse:6bcaaef4f7f27945,module: inductor has linked PR  batch 2,batch action queue,has linked PR,44,medium,bucket_component_queue,6,0,0,0,"105560, 133462, 139077, 180957, 182430, 182928",,,,"module: inductor, module: performance, oncall: pt2","bot-triaged, enhancement, feature, module: inductor, module: performance, oncall: pt2, small, triaged",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#105560 aten._trilinear, #133462 PT2 inference slowdown after updating Huggingface pin on CI, #139077 Add support for binding tensor kwargs in max-autotune templates, #180957 Missing CSE in TorchInductor, #182430 Inductor generates redundant upcast_to_fp32, #182928 [Inductor] Filter extremely bad triton configs by device information for triton heuristics",bcd2e6bb854378ccb0c36785da58ea8b76ed62c45bd24c55d90fabbfc8561a8e
issue-collapse:ed0f377bdc9b4d85,module: python frontend has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,5,0,0,0,"149097, 152750, 161863, 170298, 188088",,,,"module: python frontend, module: testing","actionable, bot-triaged, enhancement, function request, has workaround, module: python frontend, module: testing, triaged",,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#149097 Aten arange behavior when dtype is int64 and step size is greater than range, #152750 Error on padding 0-sized tensors, #161863 torch.testing.assert_close doesn't accept tensor-likes, #170298 `resize_` updates tensor shape metadata even when storage resize fails, creating corrupted ""Zombie"" tensors, #188088 torch.clamp rejects mixed Tensor and Number min/max bounds although Tensor+Tensor and Number+Number both work",29906a366f0e687ff07166fc20aad49d189fa043f19d7c8aa5e5181073834359
issue-collapse:44770641c109af38,module: backend has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,5,0,0,0,"77170, 77176, 177968, 188370, 188882",,,,"module: backend, module: codegen, module: meta tensors, module: mps, module: privateuse1, module: tests","bot-triaged, enhancement, feature, module: PrivateUse1, module: backend, module: codegen, module: meta tensors, module: mps, module: tests, triaged",,"Review this module: backend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#77170 Move the MPSGuardImpl to inherit from NoOpDeviceGuardImpl, #77176 Improve the overall design of MPSGraphCache, #177968 Add custom precision or tolerance support for out of tree backends, #188370 [Meta][PrivateUse1]: Data-dependent ATen ops need eager shape-oracle support for PU1 backends., #188882 Empty autograd section in backend YAML is rejected",aa341f4a3380a8dd85c7748380a2e82d8571a9aa894a8425e872f25793e09459
issue-collapse:4a128e1771f3f644,module: typing has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,4,0,0,0,"146249, 159259, 162373, 166905",,,,module: typing,"actionable, better-engineering, feature, module: typing, triaged",,"Review this module: typing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#146249 Replace `*args : Any` with `typing_extensions.TypeVarTuple`, #159259 Use aten/src/ATen/native/native_functions.yaml to automatically generate tests for type hints, #162373 Pylint E1102: torch.fft.rfft/istft/irfft and more is not callable, #166905 Type Annotation for `named_modules()`",f7fb1d039b9231231aa2f6cf431cfaa6ad9d2ee71caf533f9e8b6764451e71c7
issue-collapse:6e20dd2276870ba5,module: optimizer has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,4,0,0,0,"116203, 119300, 131765, 142378",,,,"module: optimizer, module: typing","actionable, better-engineering, low priority, module: optimizer, module: typing, triaged",,"Review this module: optimizer queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#116203 Refactor precision handling in test/test_optim.py, #119300 Optimizer step pre/post hook tests should try harder to ascertain the hook happens before/after the step call, #131765 [typing] Add missing `__all__` to modules., #142378 Support int step for nonfused optimizer implementations",51818fe6c84ec2804b5e4fbf7997a0e7ea38495b94adec0b90fb12733666f499
issue-collapse:548cb48f10b05bbb,module: build has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,4,0,0,0,"89489, 174682, 177265, 186544",,,,"module: build, module: cpp, module: cpp-extensions, module: debug-build, module: lint, module: pybind, module: spin, module: windows","actionable, bot-triaged, module: build, module: cpp, module: cpp-extensions, module: debug-build, module: lint, module: pybind, module: spin, module: windows, needs research, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#89489 build: cmake: need to uniformize installation of libraries in CMAKE_INSTALL_LIBDIR (not lib), #174682 spin lint doesn't regenerate cuda headers, #177265 Compiling CppExtensions in parallel causes race condition, #186544 [Windows] Debug build fails: LNK1104 cannot open file 'python310_d.lib'",4c018036b949eec858c4910d72c326ed98603507724aafa1de13954838077392
issue-collapse:fa584772871cff08,general has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,4,0,0,0,"125388, 130641, 164478, 185581",,,,,"actionable, topic: docs, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#125388 torch.multinomial raises no error when sampling from zero weight and replacement=False., #130641 `dims` of `flip()` without `list` or `tuple` works with a tensor agaist the doc, #164478 Out of date references in CONTRIBUTING.md and Wiki, #185581 `torch.compile` fails at FakeTensor validation for ConvTranspose2d",9532341d58c4a8495e033d9cac5c73f7f4ca007bb052a8161e11a96b3f36e681
issue-collapse:61ac53ce8db3bcc3,module: loss has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"123339, 126642, 185647",,,,"module: loss, module: memory usage, module: optimizer","actionable, module: loss, module: memory usage, module: optimizer, triaged",,"Review this module: loss queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#123339 Inconsistency: BCELoss asserts target in [0, 1], BCEWithLogitsLoss does not, #126642 [optim] add missing 'maximize' parameter to LBFGS, NAdam and RAdam optimizers, #185647 Unexpected storage from MSE loss output",d17d020882d281db186f055b9980b59693273a59b1e92ff24dd91f901a3848da
issue-collapse:e176e167a639a7ac,module: fx has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"171981, 185640, 188538",,,,"module: fx, module: proxytensor, module: pt2-dispatcher, oncall: pt2","bot-mislabeled, bot-triaged, module: ProxyTensor, module: fx, module: pt2-dispatcher, oncall: pt2, triaged",,"Review this module: fx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#171981 torch.fx.Tracer.trace fails on wrapped forward methods with arbitrary arguments. ValueError: code: co_varnames is too small, #185640 `make_fx` generates broken input codegen for a function with a `(tuple, dict)` positional signature, #188538 torch.fx _format_target emits invalid or semantically incorrect Python for valid module names",094e14cd97f06b1b7ea09140ef8db6c2a5d44b8499f596dea09e46a4873393a7
issue-collapse:29bab6d222704d52,module: functorch has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"154284, 175521, 175522",,,,"module: functorch, module: higher order operators, module: pt2-dispatcher, module: pytree, module: vmap, oncall: pt2","bot-triaged, feature, module: functorch, module: higher order operators, module: pt2-dispatcher, module: pytree, module: vmap, oncall: pt2, triaged",,"Review this module: functorch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#154284 Using `compile` on `hessian(hessian)`, #175521 torch.vmap working with non-tensor leaves in output pytrees, #175522 while_loop vmap compatibility",b6c8fc671f98874b88cb28c6bf10ee9ae0b14a88a2f3e1cef975fe1bd0cfb4d9
issue-collapse:fe4e3dc9b6ea52a4,module: flex attention has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"133254, 145208, 150321",,,,"module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: pt2","module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: pt2, repro:agent:h100:repro_fail:not_reproduced, triaged",,"Review this module: flex attention queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#133254 Shared memory out of resource when using flex attention, #145208 Some FlexAttention learned bias bugs/limitations, #150321 Masking out Loads from mask_modded out regions",1e0b3625220975784261c566f77ad4a4baa956ad1d5f05920917b9a56edfaf10
issue-collapse:0d83ce534aa2b72b,module: dynamic shapes has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"112044, 133655, 187547",,,,"module: dynamic shapes, module: dynamo, module: fx, module: rnn, oncall: pt2","bot-triaged, enhancement, module: dynamic shapes, module: dynamo, module: fx, module: rnn, oncall: pt2, triaged",,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#112044 DDE in pack_padded_sequence/pad_packed_sequence decomposition., #133655 PythonMod vs Mod spookiness, #187547 Dynamo graph-breaks on raw unbacked SymInt BlockMask lengths in FlexAttention",89c5b2968b977b1b96334d9dd119940ed3a20915640dc2d5ffe182cff25fb898
issue-collapse:410e26cb26e8d362,module: custom-operators has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"177109, 185046, 186140",,,,"module: custom-operators, module: library, module: performance, module: pt2-dispatcher, module: vllm, oncall: pt2","actionable, bot-triaged, enhancement, module: custom-operators, module: library, module: performance, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged",,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#177109 [RFC] Make torch.library.custom_op faster in the common case, #185046 [Bug] `torch.library._register_fake` crashes when any unrelated `LazyLoader` module in `sys.modules` fails to import, #186140 fix qualname for OpOverload/OpOverloadPacket",f869c43de52637ac47f76b717b87d30dd1cdf350cc16a33018e572d51dc704da
issue-collapse:916a2d3e5189b0f6,module: complex has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"184282, 187107, 187233",,,,"module: complex, module: cudnn, module: error checking, module: linear algebra, module: mps, module: nn, module: pooling","bot-triaged, enhancement, module: complex, module: cudnn, module: error checking, module: linear algebra, module: mps, module: nn, module: pooling, topic: fuzzer, triaged",,"Review this module: complex queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#184282 LazyBatchNorm2d complex64 input has inconsistent CPU and CUDA errors, #187107 [MPS] linalg.lu_factor not implemented, causing RunTimeErrors on torch.linalg.det and torch.linalg.solve, #187233 torch.nn.functional.max_unpool3d — internal MPS RuntimeError for complex64 input",bb9ee6e6ae2ca4c39363ec25427896ea40fc791e4a9599888fe56ee3e4c0f795
issue-collapse:708411268a66343d,module: binaries has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"164883, 167818, 181565",,,,"module: binaries, module: cpp, module: privateuse1","actionable, bot-triaged, enhancement, module: PrivateUse1, module: binaries, module: cpp, topic: fuzzer, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#164883 Wheels include protobuf headers (and headers for other projects), #167818 undefined symbol for `at::meta::_index_put_impl_` when running or compiling executable on my own torch-related project., #181565 [C++ API][Packaging] nlohmann/json.hpp is exposed again by public headers in PyTorch 2.10 wheels",da438b83641ae0b1a742c62d8a580f54665662a07b76aa93396cac12cf965ff9
issue-collapse:eeec3c14509cb3c2,module: arm has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"125438, 168972, 177250",,,,"module: arm, module: autograd, module: inductor, module: linear algebra, module: nn, module: tests, oncall: pt2","bot-triaged, module: arm, module: autograd, module: inductor, module: linear algebra, module: nn, module: tests, oncall: pt2, triaged",,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#125438 Fix broken linalg unittests on ARM platform, #168972 Aarch64 unit test failures - test/test_cpu_repro.py CPUReproTests.test_float32_to_uint8, #177250 AArch64 Unit Test Failure - TestNN test_upsampling_bfloat16",2d7625e2d82250edff5435f82bec9f11788eb86f911ed11cdd4cb9de316e2a4a
issue-collapse:4877463d19cec488,module: aotdispatch has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"132896, 185296, 188390",,,,"module: aotdispatch, module: dtensor, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, module: aotdispatch, module: dtensor, module: pt2-dispatcher, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, tensor subclass, triaged",,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#132896 AOTAutograd.process_inputs doesn't work on nested subclasses, #185296 [AOTAutograd] [DTensor] subclass output SymInt arity mismatch causes partitioner assertion, #188390 [DTensor][Compile][AOTAutograd cache] DTensor stable hashing generates identical AOTAutograd hash keys across ranks, causing false cache hits.",02cf24e78b2a4ed5605dd5463bc0e3fb4ee4779084177de589a7afe0ac922b75
issue-collapse:6b4217cda9a4e748,module: __torch_function__ has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,3,0,0,0,"147995, 167223, 182318",,,,"module: __torch_function__, module: activation checkpointing, module: autograd, module: correctness (silent), module: docs, module: dynamo, module: graph breaks, oncall: pt2","actionable, bot-triaged, has workaround, module: __torch_function__, module: activation checkpointing, module: autograd, module: correctness (silent), module: docs, module: dynamo, module: graph breaks, oncall: pt2, triaged",,"Review this module: __torch_function__ queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#147995 Checkpoint doesn't work with torch_function if torch_function change tensor metadata, #167223 [feature request] Re-enable `__torch_function__` dispatch when invoking autograd hooks for Tensor subclasses, #182318 Dynamo codegen leaks torch function enabled state via _EnableTorchFunction destructor",369aac5e2b9fc1d11c15458241770066d75473649ea1dd5fe3b98ced210fca5e
issue-collapse:adea41497d45f6fb,module: symm_mem has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"169099, 185216",,,,"module: symm_mem, oncall: distributed","module: symm_mem, oncall: distributed, triaged",,"Review this module: symm_mem queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#169099 [RFC] Implement StoreExchange.broadcast, #185216 `torch.ops.symm_mem.one_shot_all_reduce` causes `CUDA_ERROR_ILLEGAL_ADDRESS`",9ddb8a73a4e67e53e7371c96e2914a7efaf4173cae3f4cdfccf9f37ae036f244
issue-collapse:9f02483a37c3821e,module: privateuse1 has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"170737, 172538",,,,module: privateuse1,"module: PrivateUse1, triaged",,"Review this module: privateuse1 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#170737 Expect to open the is_privateuseone branch in the LSTM interface of PyTorch's RNN.cpp to develop corresponding LSTM functionality on NPU., #172538 Support native structured kernel and out_as_primary: true for PrivateUse1",ea98fc34274d2540b61f4057d46dec215a8ba75fe71c59c3129265455fbd951c
issue-collapse:529db4ae48195a90,module: numpy has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"41245, 116465",,,,module: numpy,"actionable, module: numpy, triaged",,"Review this module: numpy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#41245 torch.sign is divergent from numpy.sign on NaN, #116465 torch.cartesian_prod inconsitent return shape for one input tensor",a814758031cca91f36c7a9288076d8af97ebeaee9c755d69be378c1570d82936
issue-collapse:477a958c6a394853,module: decompositions has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"137546, 185543",,,,"module: decompositions, module: derivatives, module: inductor, module: linear algebra, oncall: pt2","bot-triaged, module: decompositions, module: derivatives, module: inductor, module: linear algebra, oncall: pt2, triaged",,"Review this module: decompositions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#137546 `torch.linspace` doesn't agree with decomposition., #185543 [inductor] Gradient mismatch between eager and inductor for torch.quantile on tied values",0d1d968139c40a5cc86e7364470d58cabd5a50fd5582564b86762901711be3b6
issue-collapse:8dcfba57d8007332,module: bfloat16 has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"175489, 187089",,,,"module: bfloat16, module: cpu, module: half, module: nn, module: tensor creation","bot-triaged, enhancement, module: bfloat16, module: cpu, module: half, module: nn, module: tensor creation, triaged",,"Review this module: bfloat16 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#175489 CPU `F.interpolate(..., antialias=True)` raises NotImplementedError for bfloat16/float16 (bilinear/bicubic), #187089 Support BFloat16 in ATen CPU tensor creation (tensor_cpu)",e4fb667dd25c183064c0e0940b484a195c8d550b99e1d4a1cbad98e361958374
issue-collapse:bd48bce12bd518b4,module: aotinductor has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"151027, 186176",,,,"module: aotinductor, module: inductor, module: serialization, oncall: export, oncall: pt2","bot-triaged, has workaround, module: aotinductor, module: inductor, module: serialization, oncall: export, oncall: pt2, triaged",,"Review this module: aotinductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#151027 Add size/strides/alignment assertions for cpp_wrapper, #186176 AOTInductor serializes entire underlying storage for each constant tensor, causing severe bloat for shared-storage weights",8acf8096e9ad14d2215202bedc412b62c654e34f181706cb98c16e507f7ee9dc
issue-collapse:9f65656e40d89880,module: amp (automated mixed precision) has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"133924, 184890",,,,"module: amp (automated mixed precision), module: aotdispatch, module: autograd, module: functorch, module: norms and normalization, module: pt2-dispatcher, module: vmap, module: xla, oncall: pt2","bot-triaged, has workaround, module: amp (automated mixed precision), module: aotdispatch, module: autograd, module: functorch, module: norms and normalization, module: pt2-dispatcher, module: vmap, module: xla, oncall: pt2, triaged",,"Review this module: amp (automated mixed precision) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#133924 AOTAutograd not tracing AMP correctly when using PyTorch/XLA., #184890 vmap of autograd.grad fails on the output of model run on AMP",ae505c4e7c84fe67edca35b6c7497b47777900230c5487a6bbc44a2cc46aa343
issue-collapse:d3e100004a438c7f,module: activation checkpointing has linked PR,batch action queue,has linked PR,44,medium,bucket_component_queue,2,0,0,0,"171355, 186572",,,,"module: activation checkpointing, module: correctness (silent), module: inductor, module: reductions, oncall: pt2","bot-triaged, has workaround, module: activation checkpointing, module: correctness (silent), module: inductor, module: reductions, oncall: pt2, triaged",,"Review this module: activation checkpointing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#171355 torch.utils.checkpoint.CheckpointError with qwen3 MoE, #186572 [Inductor] Checkpointed reductions can choose different configs between forward and recompute",b38a6a2aa703ff565ce34e3b14a30eaeb712c739829006e57a22f1a44ddfaa54
issue-collapse:5ed9388002b69ff8,module: nn needs design decision  batch 1,batch action queue,needs design decision,38,medium,bucket_component_queue,14,0,0,0,"44945, 80168, 109691, 122680, 142320, 148843, 153035, 154906, 155368, 155878, 158043, 166826, 175862, 177537",,,,"module: nn, module: python frontend, module: sdpa, oncall: distributed","bot-mislabeled, bot-triaged, enhancement, feature, function request, has workaround, module: nn, module: python frontend, module: sdpa, needs design, needs research, oncall: distributed, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#44945 Fuse softmax and masking in MultiheadAttention, #80168 GEGLU activation, #109691 Extends the functionality of `nn.BatchNorm1d`., #122680 Please consider adding MIG (MI-rror with G-radient modification) to torch.nn, #142320 Proposal for Official Cross-Platform Support for 3D Gaussian Splatting, #148843 make `to_empty` a no-op if parameter/buffer already on `device`",de9f3479b9b57eb2babb69c8f9890c81460dc73ee33db8b23ad6ae15a3cf2d94
issue-collapse:1c9137ce5dda1b2e,module: autograd needs design decision  batch 1,batch action queue,needs design decision,38,medium,bucket_component_queue,14,0,0,0,"4132, 106360, 138118, 140378, 144159, 144228, 149133, 150698, 154767, 155322, 155983, 158709, 161129, 177623",,,,"module: autograd, module: docs, module: edge cases, module: memory usage, module: nans and infs, module: optimizer, module: privateuse1, module: python frontend, oncall: distributed","bot-triaged, enhancement, feature, has workaround, module: NaNs and Infs, module: PrivateUse1, module: autograd, module: docs, module: edge cases, module: memory usage, module: optimizer, module: python frontend, needs design, needs research, oncall: distributed, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4132 x.grad should be 0 but get NaN after x/0, #106360 Improving save_on_cpu's performance by overlapping memory transfers with compute, #138118 [autograd] do not clear all variables when retain_graph=False, #140378 Differentiable quadratic programing solver, #144159 Allow generic python data structure input for torch.autograd.Function, #144228 grad_fn function disobeys broadcast rules",64abf45f6c0d94436143fc8d6fc5a6624bfcc4b6e6576a9a3f624535b6a99b06
issue-collapse:8b9c12e2182960de,module: cuda needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,11,0,0,0,"44777, 45295, 81333, 109052, 115993, 124296, 131881, 137680, 141678, 152875, 156424",,,,"module: cuda, module: cudacachingallocator, module: memory usage, module: nn, module: reductions, module: third_party, module: windows","actionable, enhancement, feature, has workaround, low priority, module: CUDACachingAllocator, module: cuda, module: memory usage, module: nn, module: reductions, module: third_party, module: windows, needs design, needs research, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#44777 Hybrid Memory, #45295 Adaptive usage of memory during training inference, #81333 Reductions on tensors larger than GPU memory, #109052 Allow reductions to write into pinned memory, #115993 [FR] Add a way to reserve memory that survives `torch.cuda.empty_cache()`, #124296 Request for modifying native allocator to use nvidia unified memory `cudaMallocManage`",7a9085ba4067f320b25085354d09454932f7cd68735dfc4c9c0c5dbc320b040a
issue-collapse:97269c972de60946,module: python frontend needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,7,0,0,0,"77171, 79842, 106784, 130428, 141994, 145739, 149118",,,,module: python frontend,"enhancement, feature, module: python frontend, needs design, needs research, topic: new features, triaged",,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#77171 Allow users to express fused matmul/bias/relu, #79842 Automatically use CUDA, #106784 [ux] Suppot torch.tensor(set([1,2,3])), #130428 Extend the device context manager to support existing tensors as well, #141994 Add torch.set_default_int_dtype() / extend set_default_dtype() to allow setting default signed integer dtype, #145739 No broadcasting by default.",fca50c0fdadbfa91f4b44f9e2bf998dce7c223630bcdbf729c2684565b61a07d
issue-collapse:a4a0a59556b2ee4a,module: numpy needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,7,0,0,0,"41803, 45051, 48830, 61901, 71479, 105053, 125422",,,,"module: numpy, module: python frontend, module: reductions, module: sorting and selection, module: ux, module: viewing and reshaping","enhancement, feature, function request, module: numpy, module: python frontend, module: reductions, module: sorting and selection, module: ux, module: viewing and reshaping, needs design, needs research, triaged",,"Review this module: numpy queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#41803 out-variant for tensor.bitwise_and (exists for torch.bitwise_and) + bitwise_friends, #45051 Custom Datatypes in Tensors, #48830 Please add ""dim"" feature for function ""torch.masked_select"", #61901 Reducing over empty dimensions for reductions without identity, #71479 [numpy compat] torch.stack and torch.tensor doesn't support nested list+tensors (NumPy does support) - at least document the difference in the error message, #105053 Multiple dimensions support for `torch.max`",eb4213715bf84b9870ad5b6341c5f2a0ab28050a8c64aa961c6b1494f46d3287
issue-collapse:ea6e58fc633e25ce,module: loss needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,5,0,0,0,"50712, 88325, 95374, 97788, 188216",,,,"module: loss, module: nn","bot-triaged, feature, function request, module: loss, module: nn, needs design, needs research, triaged",,"Review this module: loss queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#50712 ignore_index for nn_mse_loss, #88325 Group losses in a common namespace, #95374 Add oscillating activation functions to PyTorch., #97788 Add SSIM as Loss Function, #188216 Feature request: built-in loss extrapolator",143aed881aa80df23d19cdc6409a51f2340314362a447a7da2075461bc33d02b
issue-collapse:9a6f2ea366c32a87,module: dynamo needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,4,0,0,0,"99873, 135816, 141452, 170224",,,,"module: dynamo, module: vllm, oncall: pt2","actionable, enhancement, feature, module: dynamo, module: vllm, needs design, no-scrub, oncall: pt2, triaged, vllm-compile",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#99873 Dynamo config patching in our code is brittle, #135816 Can we remove @torch._dynamo_disable from torch.utils.checkpoint.checkpoint?, #141452 [Dynamo] Keep weak reference of parameters instead of accessing them by attributes, #170224 Way to ask Dynamo to hoist some warmup logic to the beginning of the compile region",46932ecd021258d179d7a1592a00d5d758872b0b9055a8fd1d727ff1e17393fb
issue-collapse:c569c218081837ba,module: binaries needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,4,0,0,0,"72712, 148345, 150742, 180215",,,,"module: binaries, module: ci, module: privateuse1, oncall: releng","bot-mislabeled, bot-triaged, enhancement, feature, has workaround, module: PrivateUse1, module: binaries, module: ci, needs design, oncall: releng, release notes: releng, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#72712 Better support for pypip packages implementing torch cuda extentions, #148345 Installation of `pytorch==2.6.0+cu124` doesn't install `triton` and `nvidia` libraries, #150742 Install pytorch from pypi using local CUDA build, #180215 [RFC] Best practices for backend installations requiring +cpu / non-PyPI wheels",ebde909ffaf13e9fd25d48955d6ad995610279ce23fc0d2d07f456cf4ec7a85a
issue-collapse:251da3d3e548af73,module: custom-operators needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,3,0,0,0,"113022, 139444, 149559",,,,"module: custom-operators, module: dispatch, module: docs, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: dispatch, module: docs, module: pt2-dispatcher, needs design, oncall: pt2, triaged",,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#113022 Operator with only `Tensor[][]` args unsupported by dispatcher, #139444 registering op through function does not work, #149559 Improve handling for custom ops with no returns",c75c05b46780050828cbddb97ad264dd94ff0674fb23b42dbf301824a49a9ac2
issue-collapse:a1e283b10bf66d56,module: activation checkpointing needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,3,0,0,0,"113737, 129867, 136433",,,,"module: activation checkpointing, module: aotdispatch, module: autograd, module: multithreading, module: pt2-dispatcher, oncall: distributed, oncall: pt2","module: activation checkpointing, module: aotdispatch, module: autograd, module: multithreading, module: pt2-dispatcher, needs design, needs research, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, triaged",,"Review this module: activation checkpointing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#113737 torch.compile + SAC: mutations in backward are not preserved, #129867 `saved_tensors_hooks`/`_pop_saved_tensors_default_hooks` API extension, #136433 SAC doesn't support nesting with different recompute plans",7829b29f42f0feb78f9e020fad0a84ee9451315fb7775f2897676cca861cfd2c
issue-collapse:8024ecddc3c0a388,module: functionalization needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"140835, 152323",,,,"module: functionalization, module: inductor, module: pt2-dispatcher, oncall: pt2","module: functionalization, module: inductor, module: pt2-dispatcher, needs design, oncall: pt2, triaged",,"Review this module: functionalization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#140835 auto-grad graph replicate split_with_sizes(lengths) X times where X = len(lengths) effecting compile time, #152323 compile generates inefficient code when mutating small slice of a graph input",ed2ff081f5c02f688198c7c871b215e5961d3bc0782ca4fa4736ac61cc294dab
issue-collapse:42c6af97c279e9da,module: flex attention needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"141627, 176837",,,,"module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","bot-triaged, feature, module: flex attention, module: higher order operators, module: pt2-dispatcher, needs design, oncall: pt2, triaged",,"Review this module: flex attention queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#141627 Support reductions in FlexAttention's score_mod/mask_mod, #176837 FlexAttention: extend AuxRequest to return min_scores and key indices for row-wise max/min",15d6dfc4d675ce7075366abdb50d4286b332e5f9f40c724cb547378680837b40
issue-collapse:e7c443ad673c05fc,module: correctness (silent) needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"65392, 178586",,,,"module: correctness (silent), module: cuda, module: cuda graphs, module: sparse, oncall: pt2","bot-triaged, module: correctness (silent), module: cuda, module: cuda graphs, module: sparse, needs design, oncall: pt2, triaged",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#65392 torch.sparse.sum result has wrong dtype when reducing over all dimensions, #178586 `make_graphed_callables` silently drops in-place mutations on user-provided input tensors",7789b5255e353efdee84f2fe575e09cd5181a15ce6aa6fabd74231963f4e3c0b
issue-collapse:1207d1abef80f3b3,module: convolution needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"74363, 99149",,,,"module: convolution, module: nn, module: norms and normalization","module: convolution, module: nn, module: norms and normalization, needs research, triaged",,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#74363 Efficient very large (e.g., 31x31) depth-wise convolution, #99149 Spectral Normalization can not be applied to Conv{1,2,3}d",f495a620da92d7e5372322d261e7e260de81c3c4aad6dbc28e3007d8d5df6508
issue-collapse:47d25952adbc8aae,module: build needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"115937, 158346",,,,"module: build, module: cpp-extensions","enhancement, module: build, module: cpp-extensions, needs design, needs research, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#115937 Building extensions with CMake, #158346 Compatibility with C++20 `std::format`",f1bd159b91448367dbe8de195bbae981669b6ae485d9d86e570015202c552662
issue-collapse:fcbefcaf1b0a396f,module: aotdispatch needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"110485, 151142",,,,"module: aotdispatch, module: dynamo, module: pt2-dispatcher, oncall: export, oncall: pt2","module: aotdispatch, module: dynamo, module: pt2-dispatcher, needs design, needs research, oncall: export, oncall: pt2, triaged",,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#110485 [export] `torch.tensor(0)` should not get burned in as a constant, #151142 AOT Dispatcher converts a single `detach` call to multiple `aten.detach`",78863902dd135b703d46ebfc8227da2ef1559d08678751039a648461fae5e436
issue-collapse:a812df627b0a5a3f,module: __torch_dispatch__ needs design decision,batch action queue,needs design decision,38,medium,bucket_component_queue,2,0,0,0,"79205, 132651",,,,"module: __torch_dispatch__, module: aotdispatch, module: compile-time, module: optimizer, module: pt2-dispatcher, oncall: pt2","module: __torch_dispatch__, module: aotdispatch, module: compile-time, module: optimizer, module: pt2-dispatcher, needs design, needs research, oncall: pt2, tensor subclass, triaged",,"Review this module: __torch_dispatch__ queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#79205 out-of-place functional optimizers: functional optimizers may not be composite compliant, #132651 fix unnecessary re-trace in AOTDispatcher in subclass code path",c34887b29296a97a739825d74505b96a0b791d6e5b51ede25c1ad3a918e13d35
issue-collapse:ed0aeca2b24291ea,module: cuda needs reproduction  batch 1,batch action queue,needs reproduction,32,medium,bucket_component_queue,14,0,0,0,"133548, 153172, 156010, 161104, 162957, 164342, 170676, 171499, 173315, 173476, 174010, 177020, 177028, 188411",,,,"module: cuda, module: cuda graphs, module: cudacachingallocator, module: cudnn, module: edge cases, module: embedding, module: half, module: linear algebra, module: memory usage, module: nans and infs, module: numerical-stability, module: performance, module: python frontend, module: python version, module: reductions, module: rnn, module: windows","bot-mislabeled, bot-triaged, module: CUDACachingAllocator, module: NaNs and Infs, module: cuda, module: cuda graphs, module: cudnn, module: edge cases, module: embedding, module: half, module: linear algebra, module: memory usage, module: numerical-stability, module: performance, module: python frontend, module: python version, module: reductions, module: rnn, module: windows, needs reproduction, topic: fuzzer, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#133548 Floating point exception on H20, #153172 triggered internal assert in matmul, #156010 `torch.prod` or `torch.special.entr` triggers `CUDA driver error: invalid argument` on GPU unless kernel cache is cleared, #161104 Eager CUDAGraph + stream performance, #162957 torch.linalg.eigh uses a large amount of memory in pytorch 2.8.0, #164342 Official support for sm_120 (RTX 50-series / Blackwell) in stable PyTorch builds",df02a9639c138133efdcab24e9fadfece09a4cec51b898af56c9fd578ce11b82
issue-collapse:7173b8842f55b438,module: build needs reproduction  batch 1,batch action queue,needs reproduction,32,medium,bucket_component_queue,14,0,0,0,"50779, 110883, 111419, 113150, 120081, 128290, 133808, 134331, 143338, 151201, 152592, 153574, 158986, 182760",,,,"module: build, module: cpp, module: cuda, module: doc infra, module: docs, module: macos","module: build, module: cpp, module: cuda, module: doc infra, module: docs, module: macos, needs reproduction, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50779 pytorch is built without _GLIBCXX_USE_CXX11_ABI and can cause std::regex crashes (probably), #110883 MAX_JOBS ignored when compiling pytorch from source, #111419 nonnull error, #113150 Aborted (core dumped) after Run Pytorch2.0.0, which I compiled by myself, #120081 undefined reference to `__atomic_exchange_1', #128290 [apple M2]libc++abi: terminating due to uncaught exception of type c10::Error: Could not find schema for aten::randn",419eea9157b87a6526fd2c0142374b5a0e2a410243c61e2e6b0ad3d2391fb71b
issue-collapse:07e88d061fd26188,module: binaries needs reproduction  batch 1,batch action queue,needs reproduction,32,medium,bucket_component_queue,14,0,0,0,"71545, 104259, 144071, 148169, 152783, 159207, 160261, 165798, 167833, 168362, 168897, 171092, 172628, 187824",,,,"module: binaries, module: cpp, module: cuda, module: cudnn, module: third_party","has workaround, module: binaries, module: cpp, module: cuda, module: cudnn, module: third_party, needs reproduction, triaged, upstream triton",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#71545 pip installation SSLError [SSL: CERTIFICATE_VERIFY_FAILED], #104259 ImportError: libcudnn.so.8: cannot open shared object file: No such file or directory, #144071 torch-nightly doesn't support tesla v100, #148169 PyTorch's nightly version no longer includes the CU118, CU124, and CU121 versions, #152783 undefined symbol: __nvJitLinkCreate_12_8, version libnvJitLink.so.12, #159207 Add official support for CUDA sm_120 (RTX 5090 / Blackwell architecture)",f775e6394d7dec999c4d51cabae518d4b7279b3daabdbb7b09beea22a4599057
issue-collapse:2d4749b639517a7b,general needs reproduction  batch 1,batch action queue,needs reproduction,32,medium,bucket_component_queue,14,0,0,0,"97635, 110944, 116882, 117032, 117962, 118263, 135693, 137260, 145827, 153564, 157709, 159224, 178049, 178145",,,,,"bot-triaged, needs reproduction, release notes: python_frontend, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#97635 JAX + PyTorch produces `OMP: Error #13: Assertion failure at kmp_affinity.cpp(532)`, #110944 gdb core dump when enable DEBUG mode to compile cpu torch in centos!!!, #116882 RuntimeError: CUDA out of memory., #117032 losing precision in torch.sum, #117962 runtime error: unknown parameter, #118263 Getting error while loading LLAMA_2_7B",ec37652cd6008bf37c54ff8fcc0da5629e7bdd3435b0dacef60989707c46512f
issue-collapse:ba5ffc037394fc90,oncall: distributed needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,12,0,0,0,"16654, 29891, 43859, 47832, 97872, 121606, 121944, 130530, 137524, 139579, 143906, 165959",,,,"oncall: distributed, oncall: distributed infra","bot-triaged, needs reproduction, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#16654 pytorch was blocked at loss.backward, #29891 Get wrong precision when multi nodes run in docker, #43859 Multi-machine multi-gpu training won't start on CNGrid, #47832 Error during distributed training, #97872 nn.linear not support bfloat16, #121606 RuntimeError: Rank 2 successfully reached monitoredBarrier, but received errors while waiting for send/recv from rank 0. Please check rank 0 logs for faulty rank.",4aa3f962b88cc4d1d03e9b5982dcfa399c31efeb097ecf3e4314c4c00e7f1efc
issue-collapse:c6f33dc8afc8f3b9,module: arm needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,10,0,0,0,"55663, 64156, 71704, 72591, 78831, 104284, 125754, 137078, 157237, 173389",,,,"module: arm, module: build, module: convolution, module: correctness (silent), module: macos, module: nans and infs, module: nn, module: nnpack, module: performance","module: NaNs and Infs, module: arm, module: build, module: convolution, module: correctness (silent), module: macos, module: nn, module: nnpack, module: performance, needs reproduction, triaged",,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#55663 torch.nn.functional.log_softmax on Arm CPU gives partly NaN results, #64156 Get zero when using torch.matmul and torch.dot with 1-D tensor in torch-1.9.0-cp39-none-macosx_11_0_arm64.whl, #71704 M1 Pro Apple Silicon chip support., #72591 Can't forward pass conv2d with kernel_size=1, and padding=1, #78831 Mac M1 Build Failure on DEBUG=1, #104284 `F.conv1d` and `F.conv2d` propagate `nan`'s incorrectly when minibatch > 15",dcc0e952c3645b0cf2fef8fea198ce3e2e058714da4b408feefef4b43bf0f9e8
issue-collapse:82e7628e15bde58e,module: cpp needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,9,0,0,0,"19029, 27694, 30507, 59628, 82546, 114572, 122707, 133860, 177738",,,,"module: cpp, module: cuda graphs, module: docs, module: nn, oncall: jit","bot-triaged, module: cpp, module: cuda graphs, module: docs, module: nn, needs reproduction, oncall: jit, triaged",,"Review this module: cpp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#19029 C++ custom module not thread safe, #27694 Deployment training model at C + + end, #30507 LibTorch, Error in 'xxx': free(): invalid pointer, #59628 what(): result type Float can't be cast to the desired output type Long, #82546 Libtorch C++ torch::stack error, #114572 In version 2.1, libtorch needs to be woken up every time it is called after the model is initialized, which means that every time the model is called, it is very slow to predict the first picture.",f01fb9e46637b12805bad77d28eb3c4a40bcad3137b089ea0c80e8f352eb599a
issue-collapse:8342ce1d86f7ce3e,module: correctness (silent) needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,9,0,0,0,"126862, 137001, 137964, 151667, 154302, 157398, 163510, 169082, 181901",,,,"module: correctness (silent), module: cpp-extensions, module: cuda, module: dependency bug, module: functorch, module: inductor, module: macos, module: mps, module: nans and infs, oncall: pt2","NVGEMM, bot-triaged, has workaround, module: NaNs and Infs, module: correctness (silent), module: cpp-extensions, module: cuda, module: dependency bug, module: functorch, module: inductor, module: macos, module: mps, needs reproduction, oncall: pt2, triaged",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#126862 worse results by using the MPS backend, compared to the CPU, #137001 BCE loss mps device, #137964 Network training silently fails with MPS backend, #151667 [MPS] MultiheadAttention with masks and dropout produces NaNs, #154302 Inconsistent output of `torch.func.jvp` calculation, #157398 Cannot copy data from one gpu to another using torch",9cd942b3c597ddf0fefba9594521b9a27376c2df6624c345bfc10cd284e12774
issue-collapse:a9aa68fb61c16d35,module: serialization needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,8,0,0,0,"37066, 48580, 64143, 76643, 146579, 150998, 160846, 179904",,,,module: serialization,"bot-triaged, has workaround, module: serialization, needs reproduction, triaged",,"Review this module: serialization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#37066 Saved model behaves differently on same data, #48580 torch.save() fails when attempting to save to mounted drives, #64143 AttributeError: module 'torchvision.models' has no attribute 'load_model', #76643 Pytorch can't process special unicode, #146579 Pytorch model (.pth format) not loading in KServe environment, #150998 Unsupported operand 118",275e9900dc1774f305b5bb02610852863bc95f817cff7ac2af8f45ce8783639a
issue-collapse:7e9b9b295b272235,module: optimizer needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,8,0,0,0,"3863, 52989, 68144, 83901, 108493, 111508, 149508, 173066",,,,"module: optimizer, module: performance","awaiting response (this tag is deprecated), module: optimizer, module: performance, needs reproduction, triaged",,"Review this module: optimizer queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#3863 Considerable slowdown in Adam.step after a number of epochs with multiple losses, #52989 lr_scheduler _triangular2_scale_fn calculation is overflowed, #68144 Unstable buggy calculation., #83901 pytorch 1.12.1 Adam Optimizer Malfunction!!!, #108493 RWKV + Adam exp_avg_sq will change from positive to negative after loss.backward(), #111508 LBFGS accuracy difference between CPU and GPU",759ae27e54bb6ddc8b07f1dccb13042f37bc2aca41ca02410fe09a7f8df4a929
issue-collapse:31cf6ddf49de4fb9,module: convolution needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,8,0,0,0,"47990, 120905, 133093, 143651, 169342, 170016, 170138, 175637",,,,"module: convolution, module: correctness (silent), module: cuda, module: error checking, module: mkldnn, module: mps, module: nn, module: performance","bot-triaged, module: convolution, module: correctness (silent), module: cuda, module: error checking, module: mkldnn, module: mps, module: nn, module: performance, needs reproduction, security, triaged",,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#47990 Is nn.Conv2d equivalent with Unfold + Matrix Multiplication + Fold ?, #120905 F.conv2d double precision error, #133093 Inconsistent results of Conv2d in pytorch between CPU and CUDA, #143651 `MemoryDenyWriteExecute` in systemd service causes `RuntimeError: could not create a primitive`, #169342 [MPS] Batch inference produces incorrect results when chunk() views are passed through conv operations, #170016 Conv2d produces numerically inconsistent results compared to other DL frameworks (TensorFlow / Paddle / MindSpore)",69008420acd6aa8413aeee12fdca540a04fe8f074680a75b5562e213e2f6effd
issue-collapse:49c3fe05d9d75dae,module: dataloader needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,7,0,0,0,"2539, 28950, 54716, 55035, 61660, 91245, 165532",,,,"module: dataloader, module: nn, module: numpy","module: dataloader, module: nn, module: numpy, needs reproduction, triaged",,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#2539 DataLoader converts cuda FloatTensor into cpu DoubleTensor when shape is (n,), #28950 ConnectionResetError when using dataLoader with pin_memory=True, #54716 Segmentation fault in PyTorch dataloader, #55035 DataLoader performance drop on 4-channel images?, #61660 Pytorch model load failure in Gunicorn with Gevent workers, #91245 Segmentation faults in DataLoader (in latest torch version).",5d889c9c96962f9b98e94a37d4627a4feac4fc7c2b97ca140a563d76c6eaa968
issue-collapse:27e26aeed339ecc7,module: onnx needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,4,0,0,0,"86684, 87550, 149937, 177952",,,,module: onnx,"OSS contribution wanted, bot-triaged, module: onnx, needs reproduction, onnx-triaged, triaged",,"Review this module: onnx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#86684 [ONNX] AssertionError: A mismatch between the number of arguments (5) and their descriptors (4) was found at symbolic function 'scatter', #87550 [onnx] export repeat_interleave TypeError: z_(): incompatible function arguments, #149937 torch.inverse can't export onnx opset16, #177952 Fix onnx tests running on GPU/CUDA",23ba4b1220f2746911cf218ddcfc31a9494d9c45b619a7d485b6f1389238423a
issue-collapse:039cd1f9b7a64ada,module: ddp needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,4,0,0,0,"50803, 88251, 98542, 114861",,,,"module: ddp, module: deadlock, module: elastic, oncall: distributed, oncall: distributed parallelisms, oncall: r2p","bot-triaged, module: ddp, module: deadlock, module: elastic, needs reproduction, oncall: distributed, oncall: distributed parallelisms, oncall: r2p, ptd-bot-triaged, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#50803 pytorch DDP hangs at .backward() call, #88251 Weird random SIGTERM occurance, #98542 Training runs 50% slower when using 2 GPUs comparing to 1, #114861 Torchrun threads with different computing speed.",d48e04ab214013c251389053473deee5abf9477c2abc70c77faae36679a34769
issue-collapse:7e4d2de52a8c99d8,module: cublas needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,4,0,0,0,"52515, 70519, 107276, 130253",,,,"module: cublas, module: cuda","module: cublas, module: cuda, needs reproduction, triaged",,"Review this module: cublas queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#52515 torch.bmm incorrect with pytorch 1.7.1 and cuda 11, #70519 RuntimeError: cublas runtime error, #107276 CUBLAS_STATUS_NOT_SUPPORTED, #130253 RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling `cublasGemmEx( handle, opa, opb, m, n, k, &falpha, a, CUDA_R_16F, lda, b, CUDA_R_16F, ldb, &fbeta, c, CUDA_R_16F, ldc, CUDA_R_32F, CUBLAS_GEMM_DEFAULT_TENSOR_OP)`",d7a8e948357b3c03b9bd7d37fbf059e87d05329d984894603edf6100e0497ed8
issue-collapse:f414b25dbdcdc7e7,oncall: pt2 needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,3,0,0,0,"159196, 159669, 172547",,,,oncall: pt2,"needs reproduction, oncall: pt2, triaged",,"Review this oncall: pt2 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#159196 torch.compile: device of tensor switches after the next statement is run. no repro, #159669 Torch Compile. CudaGraphs Memory leak, #172547 Cold start models with compile-cache have significant overhead",c2f96c39403cdfcc53684334718da6aef9ba00c8fe9b67d94c1eb4fc59353dfe
issue-collapse:7e9d3de83068d657,module: cpu needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,3,0,0,0,"75949, 143800, 173922",,,,"module: cpu, module: cuda, module: memory usage, module: optimizer, module: performance, module: rnn","bot-triaged, module: cpu, module: cuda, module: memory usage, module: optimizer, module: performance, module: rnn, needs reproduction, triaged",,"Review this module: cpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#75949 Potential memory leak in Adam optimizer in AMD chips (CPU), #143800 The tensor-based computation of exponentiation and logarithmic operations is much slower than using NumPy, #173922 [Numerical Accuracy] 100% relative error in LSTM output between CUDA and CPU backends in eval mode",85062c396b20792c4e13d99e3091cfa34e72f868167c0ae66471e6e7d2ae3581
issue-collapse:be3245233589ae4f,module: cpp-extensions needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,3,0,0,0,"151310, 155344, 166633",,,,"module: cpp-extensions, module: cuda, module: rocm, module: windows","actionable, module: cpp-extensions, module: cuda, module: rocm, module: windows, needs reproduction, triaged",,"Review this module: cpp-extensions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#151310 cpp_extension.load 'sources' does not support Chinese paths, #155344 PyTorch CPP Extensions fail when same kernel is compiled more than once on ROCm servers, #166633 Command '['ninja', '-v']' returned non-zero exit status 255.",c2bd3a2161d27a8e2f0432e3652d5cc826b98c87a13038df14c6d47164a94cf5
issue-collapse:068cee43c76a2257,module: higher order operators needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,2,0,0,0,"144803, 145451",,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",,"Review this module: higher order operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#144803 Something is fishy with discard_graph_changes, #145451 [Dynamo]while_loop raise an exception",f1a71fb145c2303931f310d303a04db6e9b8105e4914a23d2d0fb724dd45e34d
issue-collapse:6fd0db649bbb9322,module: error checking needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,2,0,0,0,"111158, 167414",,,,"module: error checking, module: rocm","module: error checking, module: rocm, needs reproduction, triaged",,"Review this module: error checking queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#111158 RuntimeError in run_streaming_llama.py When Using Accelerate with Streaming LLMa Model on A4500 GPU, #167414 C10_CUDA_KERNEL_LAUNCH_CHECK uses getLastError without clearing the error.",9e573f2b2f0af34b07e8ee92ab97cc73f11f558c80dc4be212e65b32ab482cdc
issue-collapse:e3f976b0485955e7,module: dynamic shapes needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,2,0,0,0,"143494, 175057",,,,"module: dynamic shapes, module: fx, module: inductor, oncall: pt2","module: dynamic shapes, module: fx, module: inductor, needs reproduction, oncall: pt2, triaged",,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#143494 sympy.C.ConstantInteger has no method name, #175057 torch.compile / Inductor generates invalid C++ (undeclared zuf0/zuf1) during compile (CppCompileError) on RTX 3090",dea1cf4845d7c9890d066fa5c108488b288ef555fb43ae6573bb8059f9e8830f
issue-collapse:943f926d75ecb1e0,module: ci needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,2,0,0,0,"170278, 172217",,,,"module: ci, module: macos, module: third_party","module: ci, module: macos, module: third_party, needs reproduction, triaged",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#170278 macOS Parallel Builds Fail with jobs > 4, #172217 test_numba_integration consistently fails with SIGSEGV on CUDA-enabled environments (Ubuntu 22.04 & RHEL 9)",478cdd17679f2d1cb631a96237ea5565bf684cb133da07729e8f4b99260f9681
issue-collapse:4f6e4d3468b62e76,module: aotdispatch needs reproduction,batch action queue,needs reproduction,32,medium,bucket_component_queue,2,0,0,0,"146994, 148811",,,,"module: aotdispatch, module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: aotdispatch, module: custom-operators, module: pt2-dispatcher, needs reproduction, oncall: pt2, tensor subclass, triaged",,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#146994 subclasses + HOPs fail with `Attempting to use FunctionalTensor on its own`, #148811 Custom Triton operator not registered when loading AOT-compiled .so in C++ environment",306655cdd746d3657f95aaa9b76dae6a44939b6198efaf04d72be88dcd9aba35
issue-collapse:cc2bb71d307cb812,oncall: distributed needs maintainer decision  batch 7,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"76760, 78638, 79388, 85949, 88626, 89197, 90245, 90509, 99614, 137505, 145726, 152442, 152710, 159635",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: pt2, oncall: r2p","PT2-Bug-Bash, bot-triaged, feature, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: pt2, oncall: r2p, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76760 hanging process with init_process_group(backend='mpi') cannot be killed, #78638 torch.distributed.init_process_group(backend=""nccl"") NCCL version error, #79388 Init connect timeout when use torch.distributed.run, #85949 [Distributed] Loading distributed checkpoint with FSDP fails with varying key errors (pos.embedding, shared.weight), #88626 Cloud-based rendezvous backend / distributed store?, #89197 Collective operations do not work with `torch.BoolTensor`s on `gloo` and raise `Invalid scalar type`",9c4cd7906ec1fa75ba2e36869221880e6a1c25465dbc28adea7b874fe1ed3431
issue-collapse:f6125ca9348b4692,oncall: distributed needs maintainer decision  batch 6,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"103626, 107177, 111131, 111646, 111943, 113114, 113376, 117159, 117755, 119758, 119856, 119890, 120516, 121543",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra","better-engineering, bot-triaged, enhancement, ezyang's list, feature, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#103626 2D model checkpointing hangs on a ViT model, #107177 Timeout during NCCL initialization due to store, #111131 ""Invalid Scalar type"" when using bf16 allreduce with Gloo backend, #111646 torchrun: elastic training not restarted on missing keep-alive heartbeat/scale-down event, #111943 Implement scatter and broadcast in functional collectives, #113114 Allow to construct distributed Work object from Future",a2f9a76f5d833ee0e291c8598e248b6d3ca2f2b42bc1b33338e8e0a81dc860c5
issue-collapse:215eaaab7ea39414,oncall: distributed needs maintainer decision  batch 5,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"121605, 123815, 124101, 124204, 124317, 131737, 136122, 136248, 136425, 136866, 139280, 139739, 144239, 144851",,,,"oncall: distributed, oncall: distributed infra, oncall: r2p","bot-triaged, enhancement, feature, oncall: distributed, oncall: distributed infra, oncall: r2p, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#121605 Make the data type of local_used_map_ in the Reducer class configurable, #123815 my_role_info = role_infos group_rank index error, #124101 [C10D] Standardize torch distributed launcher interface, #124204 RPC + DDP VRAM overflow to rank 0 GPU., #124317 Many tests under test/distributed/elastic not running in OSS CI, #131737 Intra-Node one-shot/two-shot low latency kernels for fp32",208430efbe3531bcd327c5018c0230d38abe9b290b8d78500704356d577a8b8e
issue-collapse:9713b431d18ecaa0,oncall: distributed needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"144856, 145170, 145376, 146565, 146813, 152427, 153350, 153369, 153822, 154102, 154116, 154327, 154345, 154370",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra","bot-triaged, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#144856 Connection Limitation in PyTorch Distributed (Vanilla) with c10d Rendezvous Backend, #145170 CUDA initialization error with vLLM 0.5.4 and PyTorch 2.4.0+cu121, #145376 distributed.new_group with backend GLOO hangs when distributed.split_group was called before, #146565 use register_full_backward_hook with c10d in torch.compile and raise error, #146813 Oneshot AllReduce not being triggered when there's nested intra- and inter-node process groups, #152427 Add switch to disable truncation to long list print",aa44161662bb1a2e611c006fc88610785ca4f22fab2fe542bfc402a5b6601bca
issue-collapse:ef1d4441bf1ddb6f,oncall: distributed needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"154563, 154621, 155467, 155664, 155892, 155916, 156002, 156627, 156942, 157318, 157781, 158016, 159563, 159686",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra","bot-triaged, feature, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#154563 Distributed Breakpoint doesn't exit safely, #154621 functional all_gather does not work with scalar tensor and fail with torch.compile, #155467 torch.distributed TCP init method binds socket to all interfaces in all cases, #155664 [Misc] skip_if decorators aborts the test process in distributed/test_composability.py, #155892 [RFC] `DeviceGroup`: a mixin of `ProcessGroup` and `Backend`, #155916 [`torch.distributed`] Watchdog monitor error handling should match watchdog error handling",f44878c1a0cf3582eeaea381acdf29c1f06bc9ae5a1928b3fd3ffa0ce323557a
issue-collapse:f8dacefe7e4e60d3,oncall: distributed needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"160006, 160731, 160950, 162032, 162237, 163546, 163678, 163913, 165653, 167288, 167305, 167832, 169275, 169841",,,,"oncall: distributed, oncall: distributed infra","bot-triaged, feature, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#160006 Support NUMA Binding for Callable entrypoints to `elastic_launch`, #160731 Cannot initialise process group on CPU device with GLOO backend, #160950 fx.node._sort_key dones't work right with torch.distributed._functional_collectives, #162032 torch thread yield after launch nccl kernel, #162237 Does it support new CPU group after passing `device_id` to `init_process_group`, #163546 Better nccl timeout handling",21f0190486ee91953054d20a39216669e149958fa8072b15645f3747c445bee2
issue-collapse:53f95b428632ab00,oncall: distributed needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"72034, 74666, 74824, 75657, 75936, 76387, 76395, 76527, 170290, 173608, 173789, 187996, 188317, 188542",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: r2p","bot-triaged, feature, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: r2p, ptd-bot-triaged, sharded_tensor, triaged",,"Review this oncall: distributed queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72034 Bug about distributed launch, #74666 Re-initializing torch.distributed process_group hangs with destroy_process_group, #74824 C10d Elastic Training master_addr ERROR, #75657 1.11.0 distribution train different with 1.8.1, #75936 AllGather with backward support async_op=True, #76387 [checkpoint] Make prepare_sharded_tensor_read and prepare_sharded_tensor_write public",c3f3943385db54dbe04a0736329c865db2f2130a045c477ef96a115586413280
issue-collapse:33f9625f603d5dcd,module: nn needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"74746, 105632, 112827, 113708, 133168, 141827, 157345, 166166, 174045, 175372, 176628, 178072, 181663, 189236",,,,"module: nn, module: norms and normalization, module: padding, module: performance, module: sdpa, module: typing, module: ux, oncall: distributed","bot-triaged, enhancement, feature, function request, module: nn, module: norms and normalization, module: padding, module: performance, module: sdpa, module: typing, module: ux, oncall: distributed, topic: docs, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#74746 Change the type hint for nn.Module.__call__ to be friendly to overrides., #105632 torch.nn.TransformerDecoderLayer lacks parameter validation check, #112827 Multi Scale Deformable Attention Support, #113708 `log_softmax` could be `2**124` to `2**1021` times more accurate on small outputs, #133168 The more detailed GELU formula should be added to the doc of `GELU()`:, #141827 Compiled and masked MHA running with DDP and `no_grad()` leads to Graph break due to unsupported builtin torch._C._distributed_c10d.PyCapsule._broadcast_coalesced",10d9c8648cdaa0da64598bcf5d1275488a9498d946690a85c227910c4407cc6d
issue-collapse:3865d2a41e6c37b4,module: inductor needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"121122, 132865, 145401, 150296, 159413, 161842, 163932, 173793, 174066, 174792, 175538, 176507, 179520, 181076",,,,"module: inductor, module: nans and infs, module: numerical-stability, module: performance, module: pt2-dispatcher, module: reinplacing, module: tests, oncall: pt2","bot-triaged, enhancement, feature, module: NaNs and Infs, module: inductor, module: numerical-stability, module: performance, module: pt2-dispatcher, module: reinplacing, module: tests, months, oncall: pt2, pt2: numerics-discrepancy, release-feature-request, repro:agent:h100:repro_fail:not_reproduced, repro:agent:h100:repro_success:reproduced, security, topic: fuzzer, triaged",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#121122 Evaluate defaulting `TORCHINDUCTOR_CACHE_DIR` to `~/.cache` on Linux systems, #132865 aotautograd CSE interferes with Inductor reinplacing, #145401 torch.compile has different numerics for var_mean, #150296 [RFC] zentorch Integration, #159413 Inductor Assertion Error with extremely large input, #161842 Exhaustive Autotuning causes `cudaErrorIllegalAddress`",84bc864747a79e6220b917461d2d7c9546fd5777624ecbe2c626267dc0acf02c
issue-collapse:289144e3dd21b551,module: inductor needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"106006, 121135, 130039, 138980, 161609, 165280, 165501, 165785, 170562, 180168, 181006, 184394, 184903, 187241",,,,"module: inductor, module: python version, module: typing, module: vllm, module: xpu, oncall: pt2","bot-triaged, dynamo-functorch, feature, module: inductor, module: python version, module: typing, module: vllm, module: xpu, oncall: pt2, repro:agent:h100:repro_fail:not_reproduced, repro:agent:h100:repro_success:reproduced, topic: performance, triaged, vllm-compile",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#106006 [torch.compile] assertion sometimes ignored with inductor backend, #121135 [torch.compile] `randperm_index_add_pattern` doesn't check the shape mismatch between self and source tensor, #130039 `min_cut_rematerialization_partition` is not optimal memory usage., #138980 compile_fx inplace modifly the input graph, #161609 Bucketize lowering crash when using a sliced computed tensor as boundaries, #165280 [inductor] avoid triton kernel when creating zero tensor",b3f675a941790b3bb625746760d0db4d9b292b8b5111182f5d54d2160f32727b
issue-collapse:d07b52450b19e0e3,module: inductor needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"156818, 168995, 180948, 180974, 182868, 185534, 187019, 188238, 188239, 188392, 188393, 188687, 188740, 188840",,,,"module: inductor, module: linear algebra, module: nn, module: performance, module: random, module: reductions, module: tests, module: xpu, oncall: pt2","bot-triaged, enhancement, feature, has workaround, module: inductor, module: linear algebra, module: nn, module: performance, module: random, module: reductions, module: tests, module: xpu, oncall: pt2, skipped, triaged",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#156818 Tighter invariants for the inductor graph, #168995 MXFP8 + torch.compile error: NotImplementedError: make_reindexer NYI on DtypeView, #180948 DISABLED test_emulate_precision_casts_preserves_explicit_precision_cast_float16 (__main__.CudaReproTests), #180974 [RFC][inductor] Support pre/post kernel execution hooks in compiled graphs, #182868 [RFC] [Inductor] L2 cache aware welford/two-pass selction, #185534 Sub-optimal heuristics in MixOrderReduction prevent fusion",fae8f56a2fc1d1d431bfab90eefc90906934f33b7a6436764815b5bb1e104248
issue-collapse:b34334806977756e,module: fsdp needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"136950, 154657, 156784, 157027, 160068, 160169, 160272, 160340, 161153, 162160, 162257, 165933, 167036, 168143",,,,"module: fsdp, module: performance, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: fsdp, module: performance, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: fsdp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#136950 Incomplete Parameter Gathering on Rank 0 with FSDP Model Saving, #154657 [FSDP2] offer public API to share communication context aross fsdp roots, #156784 [FSDP2] allow different dtypes for the model params with gradients, #157027 [RFC] Remove the FSDP data copy from compute stream critical path, #160068 FSDP2 not compatible with transformers >= 4.54.0 GenericForTokenClassification, #160169 When using DP + TP, DP only parameters diverge across TP ranks if using operations with non-deterministic implementations",070282013264824abb2ce29521a8f7e9821d94876ac4f66b74d2a43a1391cd43
issue-collapse:a88e9e4b87c8d261,module: flaky-tests needs maintainer decision  batch 6,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"100074, 145189, 158443, 159151, 180437, 181224, 181318, 182174, 182259, 186868, 186873, 186910, 186911, 186912",,,,"module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: nn, module: rocm, module: unknown, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: pt2, oncall: r2p","bot-triaged, module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: nn, module: rocm, module: unknown, oncall: distributed, oncall: distributed checkpointing, oncall: distributed infra, oncall: pt2, oncall: r2p, ptd-bot-triaged, rocm-skipped-tests, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#100074 DISABLED test_checkpointing_resets_persistent_refs (__main__.CudaGraphTreeTests), #145189 DISABLED test_slice_scatter_reinplace_cuda (__main__.GPUTests), #158443 DISABLED test_gradients_synchronizations (__main__.TensorPipeTensorPipeCudaDistAutogradTest), #159151 DISABLED test_fake_crossref_backward_no_amp_nn_functional_bilinear_cuda_float32 (__main__.TestFakeTensorCUDA), #180437 DISABLED test_find_learning_rate_ensembling_FeatureAlphaDropout_mechanism_functional_call_cpu (__main__.TestExamplesCorrectnessCPU), #181224 DISABLED test_cross_stream_stride_copy (__main__.TestUserStreamCompile)",143e575803eb039f5fa6de694452a307febba665286905a9781f3790a3a035f0
issue-collapse:460667af922273b6,module: flaky-tests needs maintainer decision  batch 5,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"179811, 179874, 180322, 181322, 185452, 186871, 186999, 187156, 187202, 187203, 187424, 187645, 188123, 188144",,,,"module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: meta tensors, module: rocm, oncall: distributed checkpointing, oncall: export, oncall: jit, oncall: profiler, oncall: pt2","module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: meta tensors, module: rocm, oncall: distributed checkpointing, oncall: export, oncall: jit, oncall: profiler, oncall: pt2, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#179811 DISABLED test_hessian (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #179874 DISABLED test_lazy_init_with_view (__main__.LazyFuncionalizationTest), #180322 DISABLED test_vjp_vjp_cpu (__main__.TestComposabilityCPU), #181322 DISABLED test_trace_return_dataclass_nested (__main__.TestFX), #185452 DISABLED test_torchvision_models_convnext_large (__main__.TestVisionTracing), #186871 DISABLED test_addmm_relu_tunableop_rocm_cuda_float32 (__main__.TestLinalgCudaOnlyCUDA)",1352c438020f155c9e41697b0c827d6f68b70cd6646d0ff47cbc138685ce6687
issue-collapse:4b6e8b0240931323,module: flaky-tests needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"149080, 179853, 179897, 179990, 180204, 180324, 180366, 180439, 182260, 186872, 186909, 187000, 187046, 187290",,,,"module: flaky-tests, module: inductor, module: linear algebra, module: nn, module: rocm, oncall: export, oncall: profiler, oncall: pt2","module: flaky-tests, module: inductor, module: linear algebra, module: nn, module: rocm, oncall: export, oncall: profiler, oncall: pt2, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#149080 DISABLED test_aoti_debug_printer_codegen_cuda (__main__.AOTInductorTestABICompatibleGpu), #179853 DISABLED test_hessian_argnums (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #179897 DISABLED test_jacfwd (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #179990 DISABLED test_jacfwd_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #180204 DISABLED test_jacrev_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #180324 DISABLED test_jvp_two_tensors_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)",5c987febdad2fe367deca754d8b68d2e82f78e378888de7cfa23664b3e881d4c
issue-collapse:1ee349d1c7602756,module: flaky-tests needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"145344, 153236, 180414, 180490, 180542, 181629, 187176, 187378, 187425, 187658, 188627, 188628, 188668, 188767",,,,"module: flaky-tests, module: fsdp, module: fx, module: inductor, module: rocm, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, module: flaky-tests, module: fsdp, module: fx, module: inductor, module: rocm, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#145344 DISABLED test_cat (__main__.NumBytesMetricTests), #153236 DISABLED test_clip_grad_norm_2d (__main__.TestClipGradNormWorldSize4), #180414 DISABLED test_torchvision_models_efficientnet_b2 (__main__.TestVisionTracing), #180490 DISABLED test_vmap_two_inputs (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #180542 DISABLED test_vmap_two_inputs_tuple_in_dims (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #181629 DISABLED test_torchvision_models_efficientnet_b5 (__main__.TestVisionTracing)",99d5814d59b0f364ea86e026ca0f644d2670d84dbecb0bd6ede7d541f361d5c4
issue-collapse:f09c7f19a93ca902,module: flaky-tests needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"124152, 176966, 180337, 187718, 187738, 188041, 188895, 188897, 188899, 188983, 188985, 189015, 189066, 189067",,,,"module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: macos, module: mps, module: rocm, oncall: pt2","module: flaky-tests, module: functorch, module: fx, module: inductor, module: linear algebra, module: macos, module: mps, module: rocm, oncall: pt2, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#124152 DISABLED test_torchvision_models_efficientnet_v2_m (__main__.TestVisionTracing), #176966 DISABLED test_trace_return_namedtuple (__main__.TestFX), #180337 DISABLED test_vmap_grad_cpu (__main__.TestComposabilityCPU), #187718 DISABLED test_grad_two_tensor_all_grad_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #187738 DISABLED test_grad_two_tensor_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #188041 DISABLED test_jacfwd_randomness (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd)",6aa34b6a550c9250c961448c9ed7aa4b90ca670177407109462a968bdee29450
issue-collapse:936df391bbfc2954,module: flaky-tests needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"152632, 179742, 179799, 179879, 180194, 180195, 180213, 187289, 188798, 189151, 189221, 189223, 189244, 189255",,,,"module: flaky-tests, module: functorch, module: fx, module: inductor, module: macos, module: mps, module: rocm, oncall: distributed, oncall: distributed checkpointing, oncall: pt2","module: flaky-tests, module: functorch, module: fx, module: inductor, module: macos, module: mps, module: rocm, oncall: distributed, oncall: distributed checkpointing, oncall: pt2, skipped, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#152632 DISABLED test_torchvision_models_efficientnet_v2_l (__main__.TestVisionTracing), #179742 DISABLED test_grad_freevar_python_scalar (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #179799 DISABLED test_grad_has_aux (__main__.FuncTorchHigherOrderOpTestsWithCompiledAutograd), #179879 DISABLED test_vmap_vmap_cpu (__main__.TestComposabilityCPU), #180194 DISABLED test_vjp_grad_cpu (__main__.TestComposabilityCPU), #180195 DISABLED test_vjp_vmap_cpu (__main__.TestComposabilityCPU)",5da6af8cf38f5a5c94a47a1d5b265175979bf7585a5a7ab88aa48c0e919d11e8
issue-collapse:29f6f121942eb88f,module: faketensor needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"82616, 84925, 91655, 100985, 102187, 104194, 107582, 113589, 119407, 119589, 120105, 122806, 159150, 185684",,,,"module: faketensor, module: flaky-tests, module: inductor, module: logging, module: meta tensors, module: primtorch, module: pt2-dispatcher, module: serialization, module: sparse, module: unknown, module: windows, oncall: pt2","better-engineering, bot-triaged, feature, module: fakeTensor, module: flaky-tests, module: inductor, module: logging, module: meta tensors, module: primTorch, module: pt2-dispatcher, module: serialization, module: sparse, module: unknown, module: windows, oncall: pt2, rocm-skipped-tests, skipped, triaged",,"Review this module: faketensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#82616 FakeTensor Support For Pickling, #84925 `is_pinned()` support in PrimTorch and FakeTensor., #91655 FakeTensors not moving between device properly on Module.cuda(), #100985 native_batch_norm has different size results on ""CPU"" vs ""META"" device, #102187 Extend fake fast path to more situations, #104194 View ops on fake tensors can dispatch `detach`es to backend kernels",d0e59bcca8315750039e50a3504746f82204ca9386b63b337c00f169f786f436
issue-collapse:c250a79f571a31ab,module: edge cases needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"87131, 87172, 87556, 89431, 106665, 138464, 173525, 173885, 176384, 176385, 176412, 176415, 182560, 188382",,,,"module: edge cases, module: error checking, module: inductor, module: linear algebra, module: mps, module: nans and infs, module: nn, module: numerical-stability, module: numpy, module: python frontend, module: special, module: vectorization, module: windows, oncall: cpu inductor","actionable, bot-mislabeled, bot-triaged, module: NaNs and Infs, module: edge cases, module: error checking, module: inductor, module: linear algebra, module: mps, module: nn, module: numerical-stability, module: numpy, module: python frontend, module: special, module: vectorization, module: windows, oncall: cpu inductor, triaged",,"Review this module: edge cases queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#87131 torch.clamp does not clamp out of -0 from 0 when ran on the CPU, #87172 Unstable results in sin/arcsin/arccos calls, #87556 The behavior of cast `NaN` is different on cpu and cuda, #89431 The problem caused by the parameter dim of torch.norm, #106665 Inconsistency between CPU and GPU for `Linear()` layer with input size 0, #138464 `torch.special.zeta` ignores `nan` input when `other=-inf`.",e91fc2f73e9c0c3a6cf043ae53469bb04bc41ba8482260b803250644f6583f09
issue-collapse:8c25959853400272,module: dynamo needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"107631, 112344, 113643, 118211, 123252, 136149, 136495, 146129, 155632, 158519, 170558, 173452, 175269, 176692",,,,"module: dynamo, module: faketensor, module: graph breaks, module: pt2-dispatcher, oncall: distributed, oncall: pt2","PT2-Bug-Bash, better-engineering, dynamo-triage-dec2025, enhancement, feature, google-tpu, module: dynamo, module: fakeTensor, module: graph breaks, module: pt2-dispatcher, oncall: distributed, oncall: pt2, tensor subclass, triaged",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#107631 torch.fx.Interpreter modules don't get compiled, #112344 [test/dynamo] BE: cleanup `test_misc.py`, #113643 Distinguish immutable/mutable fake tensor mode, #118211 Dynamo codegen for mapping outputs back to user variables should contain safety checks (no fake tensors, no variable trackers), #123252 FakeTensor support of pin_memory, #136149 Dynamo: warning when we incorrectly inline autograd.Function.backward due to subclass gradients",929043207c944340ea4872be27ab012b4dc12baaac2546dffda0c53588f05843
issue-collapse:f5d4f69fccd86c96,module: dynamo needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"128121, 137749, 144805, 170361, 175869, 178511, 179643, 180839, 182684, 186864, 186865, 187073, 188077, 188234",,,,"module: dynamo, module: flaky-tests, module: fsdp, module: fx, module: guards, module: higher order operators, module: performance, module: pt2-dispatcher, module: sdpa, module: sparse, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","better-engineering, bot-triaged, dynamo-variable-tracker, enhancement, feature, has workaround, module: dynamo, module: flaky-tests, module: fsdp, module: fx, module: guards, module: higher order operators, module: performance, module: pt2-dispatcher, module: sdpa, module: sparse, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, skipped, topic: build, triaged",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#128121 2nd compile of deepcopy(model) fails on multiple ubuntu-pc (fatal error: Python.h: file not found), #137749 Graph break on wait() of Torchrec's Awaitable instance, #144805 ""GenericHOPVariable"" / abstract out Dynamo support for HOPs, #170361 [Dynamo] Use VariableBuilder/SourcelessBuilder consistently, #175869 [dynamo] Tracking issue for itertools polyfills, #178511 [torch.compile] allow_in_graph called inside compiled function causes Unsupported error",69f0a38820a84437e5739295f09af55e2c0f33183c1a3f45149d44261cbbe112
issue-collapse:d57e585ee6fbe488,module: dynamic shapes needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"90900, 118521, 121818, 125604, 137560, 140463, 151997, 152944, 165506, 181108, 181525, 181653, 188151, 188871",,,,"module: dynamic shapes, module: dynamo, module: flaky-tests, module: flex attention, module: functorch, module: fx.passes, module: higher order operators, module: inductor, module: performance, module: pooling, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, better-engineering, bot-triaged, dynamo-functorch, enhancement, feature, module: dynamic shapes, module: dynamo, module: flaky-tests, module: flex attention, module: functorch, module: fx.passes, module: higher order operators, module: inductor, module: performance, module: pooling, module: pt2-dispatcher, oncall: pt2, skipped, triaged",,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90900 Check that SymPy semantics match Python semantics, #118521 Improve constant propagation in the dynamic shape case, #121818 torch._check but trust me, #125604 Compiling GPT2 with tensorrt backend and dynamic shapes results in guard failures., #137560 vmap specializing on dynamic shapes with index_put_, #140463 Ban all operations on sympy expressions in Inductor",392e0a74b5aae1bc2dce8035032b612a65fa97487b8eaaf4c3e849f4b2253953
issue-collapse:81e687998ce99d6c,module: dtensor needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"118094, 119930, 120787, 124019, 127772, 146045, 155562, 167540, 168099, 177663, 178091, 178616, 179448, 179858",,,,"module: dtensor, module: dynamo, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, llm-amenable, module: dtensor, module: dynamo, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, pt2d-triage-nov2024, ptd-bot-triaged, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#118094 DTensorConverter::successful returns False on no Tensor args causing test failures, #119930 [dtensor] ops coverage tracker, #120787 Tensor Parallel Inference Performance Issue with Llama-7b-chat-hf Model caused by straggler problem, #124019 [FSDP+TP] RuntimeError: 'weight' must be 2-D, #127772 Run existing eager DTensor tests under torch.compile, #146045 Redundant collectives are deduplicated but stick around as dead code",b9b450adfe21c7a6298b14b80ea9b5f3c3e145eb3f123bc98877411b3ecc7924
issue-collapse:5a2e274498500b15,module: dtensor needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"109477, 116683, 135178, 139681, 140227, 146601, 153615, 156842, 161091, 161368, 162290, 163043, 166930, 167106",,,,"module: dtensor, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: dtensor, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#109477 [DTensor] optimizer step performance is still too bad, #116683 DTensor tensor subwrapper not be handle with apex, #135178 [DTensor] loss_parallel not worked with shifted logits and labels, #139681 Unexpected type cast from DTensor to Tensor when applying SequenceParallel on RMSNorm., #140227 `Torch.distributed.tensor.parallel.Style.Rowwiseparallel` when send tensor, the use of the underlying operator is `aten.to.dtype_layout`, lack of distributed strategy. But after adding a policy, `AsyncCollectiveTensor` no placements error, #146601 Enabling efficient model-level redistribution between FSPD-TP",d0b7836f9312537b29e57214cc656d1e42c85ed579142634996f45cd5e1a4fed
issue-collapse:4f18cd3465a2e612,module: dtensor needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"120158, 143748, 143985, 145218, 146884, 155657, 156523, 156533, 159959, 160911, 160968, 161218, 161974, 162215",,,,"module: dtensor, module: performance, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: dtensor, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, topic: bug fixes, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#120158 The forward accuracy cannot be aligned when using DTensor and ColwiseParallel, RowwiseParallel for TP training on the Llama2, #143748 [DTensor]`Linear` fails on 3D DTensor with `batch size > 1` and `Replicate` input redistributed from `Shard`, #143985 [DTensor] Allow multiple dimensions to be sharded together (as if flattened), #145218 getting different results when adding `torch.Tensor` or python number to a DTensor - Is that expected?, #146884 [DTensor]Operator aten.mT.default/aten.mv.default/aten.dot.default does not have a sharding strategy registered, #155657 [DTensor] DTensor is not well supported on older versions of GPUs, such as A10",c767b2f8fab26665de9eddbf2cf590699706856f9182dad458fd629de65c79ed
issue-collapse:253029a1aea047fc,module: dtensor needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"163692, 167074, 167090, 167219, 167657, 167917, 169320, 169814, 172485, 172491, 173123, 186030, 187020, 187561",,,,"module: dtensor, module: flaky-tests, oncall: distributed, oncall: distributed parallelisms","bot-triaged, feature, module: dtensor, module: flaky-tests, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, skipped, triaged",,"Review this module: dtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#163692 [feature][dtensor] support nn.Module inputs containing DTensor parameters in local_map, #167074 DTensor with `torch.compile` fails with `aten.as_strided.default` error, #167090 DTensor `F.conv2d` fails with `bias=None` - AssertionError in sharding propagation, #167219 Are there limitations to dtensor's registration strategy?, #167657 make_fx trick for tracing DTensor compute doesn't work with redistribute, #167917 Operator aten.detach_.default does not have a sharding strategy registered",da6ac7fc894ce06c5b65ac52b60dcf79c7837e45e6f21b51aa5f3a9c29664678
issue-collapse:18c597365c3677b5,module: docs needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"83494, 103841, 113669, 121587, 144779, 157395, 158775, 161847, 176065, 177957, 182490, 183036, 186496, 188389",,,,"module: docs, module: fsdp, module: linear algebra, module: optimizer, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: pt2","PT2-Bug-Bash, bot-triaged, enhancement, module: docs, module: fsdp, module: linear algebra, module: optimizer, no-scrub, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: jit, oncall: pt2, ptd-bot-triaged, rst-to-md, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#83494 `torch.pinverse` produces wrong output!, #103841 [question] [docs] Short/mid/long-term status of TorchScript / JIT / torch.jit.trace / FX / symbolic tracing and its replacement by Dynamo, #113669 document functional_collectives, #121587 Add clearer docs for `torch.distributed.<OP>` vs. `torch.distributed.nn.<OP>, #144779 Adding Infiniband to RDZV Backend for optimal torch run training, #157395 [FSDP2] figure out the contract for mp_policy and tensor subclass extention",1d900830f469a5fd5dbf37090511ecbfce17868c647bd7007c8b1894fa15ffaf
issue-collapse:961b40eba159030d,module: ddp needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"86456, 88006, 89546, 89716, 90171, 90688, 91408, 95724, 99908, 99978, 100582, 102197, 102438, 103622",,,,"module: ddp, module: forward ad, module: functorch, module: performance, module: rpc, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: ddp, module: forward ad, module: functorch, module: performance, module: rpc, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, tensor subclass, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#86456 `SyncBatchNorm` doesn't work with subclass of `torch.Tensor`, #88006 C++ Extensions can't import c10d/reducer.hpp, #89546 Unable to launch CUDA Graph with DDP model, #89716 DDP hangs on forward pass of transformer, #90171 Option to let DistributedDataParallel know in advance unused parameters at each forward pass, #90688 In distributed get SIGTERM and run crash",9a469c3a8640b46eac976d532c25928eefd2f3a512532eda7fef73eff1641932
issue-collapse:9afe9b68812fd6c1,module: ddp needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"104164, 105723, 109103, 111834, 112029, 112277, 114765, 114857, 121181, 124680, 135511, 140107, 152042, 155993",,,,"module: ddp, module: distributed_tool, module: dtensor, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: profiler","bot-triaged, enhancement, feature, module: ddp, module: distributed_tool, module: dtensor, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, oncall: profiler, ptd-bot-triaged, tensor subclass, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#104164 DDP enhancement, #105723 Parameter ... has been marked as ready twice, #109103 DDP - ""No backend type associated with device type cpu"" with new Model Phi 1.5 despite everything loaded on GPUs, #111834 Numerical inaccuracies in ""ddp_apply_optim_in_backward"" unit tests for gloo backend, #112029 DDPWrapper, #112277 test_DistributedDataParallel fails with static_graph=True",73eeaf4ded01854bc5a17adcf65196c3135a9c03f1dca27c097e1b41ffe936b4
issue-collapse:e66e75e20ee06fc4,module: ddp needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"73081, 74162, 75031, 76389, 135358, 143647, 143721, 156130, 157862, 159836, 163722, 166523, 187276, 187279",,,,"module: ddp, module: docs, module: fsdp, module: mpi, oncall: distributed, oncall: distributed parallelisms","better-engineering, bot-triaged, enhancement, feature, module: ddp, module: docs, module: fsdp, module: mpi, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#73081 Additional out of the box DDP comm hooks, #74162 consume_prefix_in_state_dict_if_present can not remove prefix of _metadata., #75031 comm hook error in BWD pass, #76389 Error in DistributedDataParallel with 'CPU' device, #135358 UserWarning: Grad strides do not match bucket view strides. This may indicate grad was not created according to the gradient layout contract, or that the param's strides changed since DDP was constructed. This is not an error, but may impair performance. grad.sizes() = [1, 64, 64, 64], strides() = [1, 4096, 64, 1] bucket_view.sizes() = [1, 64, 64, 64], strides() = [262144, 4096, 64, 1], #143647 A very weird bug involving ddp",8b9a346291fa86997aa4d5d274c668ed3fe94596b5448d7ac9efe98fac00c24b
issue-collapse:06a118cc5a3074a4,module: custom-operators needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"127621, 129395, 139257, 142388, 146936, 152270, 155240, 158186, 160472, 180206, 180620, 187470, 187505, 187862",,,,"module: custom-operators, module: docs, module: faketensor, module: flaky-tests, module: fx, module: inductor, module: performance, module: pt2-dispatcher, module: python frontend, module: sdpa, oncall: pt2","agents-banned, bot-triaged, compile-cache, enhancement, feature, module: custom-operators, module: docs, module: fakeTensor, module: flaky-tests, module: fx, module: inductor, module: performance, module: pt2-dispatcher, module: python frontend, module: sdpa, oncall: pt2, skipped, topic: fuzzer, tracker, triaged",,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#127621 custom_op API: better type anntation for Tuple, #129395 [custom_op] Figure out what to do with random custom ops, #139257 Error compiling the torch.library.custom_op with input mutations with set_, #142388 User-defined way to hook into Inductor autotuning (+ custom op), #146936 flexible custom operators: custom operators that accept arbitrary input/output type, #152270 Question about that support of torch.compile for a custom CUDA operator?",4d261945a0bac9468beceb49f5b26ff2626825c12e7a7dcb65125f036173fd7b
issue-collapse:7a3d8d0bbb218b2e,module: cuda needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"34242, 53721, 76632, 154077, 170535, 171398, 173290, 173516, 173729, 173904, 174227, 174983, 175311, 175737",,,,"module: cuda, module: cuda graphs, module: cudacachingallocator, module: docs, module: edge cases, module: helion, module: inductor, module: linear algebra, module: memory usage, module: python frontend, module: rocm, module: serialization, module: sorting and selection, module: ux, module: vllm, oncall: pt2","PT2-Bug-Bash, actionable, better-engineering, bot-mislabeled, bot-triaged, enhancement, feature, module: CUDACachingAllocator, module: cuda, module: cuda graphs, module: docs, module: edge cases, module: helion, module: inductor, module: linear algebra, module: memory usage, module: python frontend, module: rocm, module: serialization, module: sorting and selection, module: ux, module: vllm, oncall: pt2, release-feature-request, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#34242 RuntimeError with torch.unique: radix_sort: failed on 2nd step: invalid argument, #53721 PyTorch on Embedded Hardware, #76632 RestrictPtrTraits in CUDA potentially has no effect., #154077 accept 0-d tensor for `fill_value` parameter of `full`, `full_like`, #170535 Graphsafe RNG clone_state, #171398 ISAAC inference",18f3134c835a9959d3c8751a2228769904c7fd991e7e9b403dcfaead9e1cfe1e
issue-collapse:242e588887ae5f22,module: cuda needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"111824, 127605, 147478, 150725, 153555, 169015, 175409, 177477, 177910, 178052, 178864, 179124, 181379, 182286",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: deadlock, module: determinism, module: edge cases, module: error checking, module: floatx (formerly float8), module: helion, module: inductor, module: interpolation, module: jiterator, module: loss, module: memory format, module: memory usage, module: nn, module: performance, module: sdpa, module: symm_mem, module: tests, module: type promotion, module: wsl, oncall: export, oncall: pt2","actionable, bot-mislabeled, bot-triaged, enhancement, feature, function request, has workaround, module: CUDACachingAllocator, module: cuda, module: cudnn, module: deadlock, module: determinism, module: edge cases, module: error checking, module: floatx (formerly float8), module: helion, module: inductor, module: interpolation, module: jiterator, module: loss, module: memory format, module: memory usage, module: nn, module: performance, module: sdpa, module: symm_mem, module: tests, module: type promotion, module: wsl, oncall: export, oncall: pt2, topic: performance, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#111824 GroupNorm & InstanceNorm does not handle channels_last correctly, #127605 RuntimeError: CUDA error: an illegal memory access was encountered with specific input shape and Conv1d, #147478 Blackwell Only Bugs, #150725 Continuous calls to nn.Linear in fp32 on the 5090D cause severe performance degradation, #153555 Updated Scaled_mm to support more scaling formats via CuBlas, #169015 [CI][CUDA][B200] Test_weight_norm_conv2d_cuda Failure",74c5b2f6fc7a11d4a499b39657c1d8ca1e1c330f46043950cf5db7821977e882
issue-collapse:a1ec9b7348c4ce32,module: cuda needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"104695, 120902, 131754, 136187, 136567, 153122, 157456, 164015, 167353, 175636, 177578, 181094, 183391, 186213",,,,"module: cuda, module: cudacachingallocator, module: edge cases, module: fft, module: flaky-tests, module: inductor, module: linear algebra, module: multiprocessing, module: performance, module: sparse, module: symm_mem, module: tests, module: type promotion, oncall: distributed, oncall: distributed infra, oncall: pt2","NVGEMM, bot-triaged, enhancement, feature, matrix multiplication, mempool-bug, module: CUDACachingAllocator, module: cuda, module: edge cases, module: fft, module: flaky-tests, module: inductor, module: linear algebra, module: multiprocessing, module: performance, module: sparse, module: symm_mem, module: tests, module: type promotion, oncall: distributed, oncall: distributed infra, oncall: pt2, ptd-bot-triaged, release-feature-request, skipped, topic: fuzzer, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#104695 DISABLED test_cuda_memory_leak_detection (__main__.TestCudaMultiGPU), #120902 RuntimeError: cuFFT error: CUFFT_INTERNAL_ERROR, #131754 Coredump for pytorch ddp when job finished, #136187 Unable to Specify CUDA Stream for Collective Operations Using with torch.cuda.stream() context, #136567 [RFC] Offload collectives to NVSwitch when possible, #153122 [CUDA] test_c10d_nccl test_extra_cuda_context failure due to _helper_test_extra_cuda_context_by_memory",a3b995ff5933ff1e6ad196dfb89de2f13500e539ae06f103cb00c6eaa38fa987
issue-collapse:4a459ac55a0182ec,module: cuda needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"68506, 104701, 130330, 143281, 147971, 153460, 159113, 159547, 179152, 179745, 179798, 182001, 187045, 188705",,,,"module: cuda, module: cudnn, module: flaky-tests, module: floatx (formerly float8), module: half, module: inductor, module: linear algebra, module: memory usage, module: nn, module: performance, module: tests, oncall: pt2","bot-triaged, enhancement, feature, matrix multiplication, module: cuda, module: cudnn, module: flaky-tests, module: floatx (formerly float8), module: half, module: inductor, module: linear algebra, module: memory usage, module: nn, module: performance, module: tests, oncall: pt2, skipped, topic: performance, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68506 cudnn_convolution_relu is 17x slower for 7x7 convolutions and channels last, #104701 System memory leak when using different input size of torch.nn.Conv3d, #130330 why not enable expandable_segments by default？, #143281 torch.linalg.qr is significantly slower on GPU compared to CPU and SVD for batched small matrices, #147971 Fp8 scaled-mm row-wise is substantially slower than tensor-wise, #153460 DISABLED test_mempool_ctx_multithread (__main__.TestMemPool)",e429aa92a095f4ebf4b4e60590d40d955b52f117c0893176557528aa842cf6d1
issue-collapse:4914d16c76174384,module: correctness (silent) needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"152028, 160740, 164563, 177654, 178482, 178614, 182960, 183607, 183768, 185587, 187180, 187227, 187715, 188074",,,,"module: correctness (silent), module: cuda, module: cuda graphs, module: determinism, module: dynamic shapes, module: dynamo, module: edge cases, module: embedding, module: functorch, module: half, module: inductor, module: int overflow, module: mps, module: nn, module: python frontend, module: rocm, module: sdpa, module: special, module: tensor creation, oncall: pt2","bot-mislabeled, bot-triaged, has workaround, module: correctness (silent), module: cuda, module: cuda graphs, module: determinism, module: dynamic shapes, module: dynamo, module: edge cases, module: embedding, module: functorch, module: half, module: inductor, module: int overflow, module: mps, module: nn, module: python frontend, module: rocm, module: sdpa, module: special, module: tensor creation, oncall: pt2, repro:agent:h100:repro_success:reproduced, topic: fuzzer, triaged",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#152028 [CUDA] MultiheadAttention with masks and dropout produces NaNs, #160740 MPS torch.argmax/torch.argmin fail for non-contiguous input, #164563 [Blackwell][Inductor] Numerical mismatches in test_max_autotune.py::TestMaxAutotune::test_non_contiguous_input_mm_plus_mm, #177654 `import functorch.dim` silently breaks `torch.Tensor.expand(size=...)`, #178482 torch.compile(dynamic=True) ignores incompatible `out` tensor for `torch.empty` and returns wrong shape, #178614 bag boundry is not calculated correctly in EmbeddingBag when setting include_last_offset = True",6c283bd9c8f6790b5f5d7068495f458aad4f80cf09a28d89520c95fe95c56ae2
issue-collapse:8db16f8ad3094f94,module: ci needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"146094, 167616, 167994, 175022, 177135, 177425, 177896, 177970, 178345, 182336, 189032, 189164, 189261, 189262",,,,"module: ci, module: devx, module: infra, module: lint, module: rocm, module: spin, module: tests, oncall: distributed, oncall: distributed infra","bot-mislabeled, bot-triaged, enhancement, feature, module: ci, module: devx, module: infra, module: lint, module: rocm, module: spin, module: tests, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged, unstable",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#146094 [CI][Distributed] @skip_if_lt_x_gpu(2) seems to be broken, #167616 UNSTABLE trunk-rocm-sandbox / linux-jammy-rocm-py3.10 / test, #167994 CI Not Detecting Failing Tests in test/distributed/elastic/*, #175022 [RFC] Cross-Repository CI Relay for PyTorch Out-of-Tree Backends, #177135 release notes commit script should respect user-specified labels, #177425 binary_populate_env.sh: PYTORCH_ROOT optional check is inconsistent with later required use",21acca6a751e29ebd36b259b0d16757bbe044897b05bb3637695fb6fb179bf19
issue-collapse:32c8455041707a6a,module: c10d needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"72025, 72099, 72346, 72366, 72435, 72835, 74323, 74602, 113281, 115861, 120741, 137010, 139129, 142356",,,,"module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms","better-engineering, bot-triaged, feature, module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72025 [c10d] destruction of Store objects, #72099 halt or exit function implementation, #72346 torch.distributed.new_group() should consolidate the type of rank list, #72366 Custom ProcessGroup Destructor Not Correctly Called in PT 1.10, #72435 ProcessGroupWrapper: Additional Improvements, #72835 torch.distributed hangs at barrier()",80c69f64566f46d9b86408030c91c8a4ff145129388b2ce0e1494af490215574
issue-collapse:1fe3a85b9f0fa957,module: build needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"14152, 103222, 144608, 155512, 167365, 167822, 181499, 181939, 184397, 185215, 185698, 186004, 186405, 188935",,,,"module: build, module: cpp, module: cpp-extensions, module: cuda, module: dependency bug, module: rocm, module: third_party, module: windows, module: xpu, oncall: pt2","bot-triaged, enhancement, has workaround, module: build, module: cpp, module: cpp-extensions, module: cuda, module: dependency bug, module: rocm, module: third_party, module: windows, module: xpu, oncall: pt2, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14152 Failed to run 'bash ../tools/build_pytorch_libs.sh --use-cuda --use-nnpack --use-mkldnn --use-qnnpack caffe2', #103222 compilation fails `error: invalid argument '-std=c++17' not allowed with 'C'`, #144608 Build breaks on FreeBSD on arm platforms: Unrecognized CMAKE_SYSTEM_NAME = FreeBSD, #155512 Unable to find caffe2 when building libtorch from the source and then trying to use it in a cmake project, #167365 PyTorch is no longer importable from `build` folder, #167822 [RFD] Composable Kernel (CK) moving to C++20 — request for feedback/concerns",17dbacf99b455f49997f8a12cc47ec9295efeb9f0918f9a181ac4fb219760617
issue-collapse:8d8d2d6edaa07690,module: binaries needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"71381, 135508, 150655, 152355, 154299, 167411, 177050, 178196, 178368, 182008, 182272, 185746, 186905, 187834",,,,"module: binaries, module: build, module: ci, module: cpp-extensions, module: cuda, module: cudnn, module: docker, module: docs, module: rocm, module: windows, oncall: pt2","bot-triaged, enhancement, feature, has workaround, module: binaries, module: build, module: ci, module: cpp-extensions, module: cuda, module: cudnn, module: docker, module: docs, module: rocm, module: windows, oncall: pt2, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#71381 Cannot import PyTorch in Alpine Docker Container, #135508 BUG: `torch.cuda.is_available()` returns `False` in certain torch, CUDA and driver version, #150655 Create better alerting for binary size validations and time it takes to build the binary, #152355 [source.wheel] Pin setuptools runtime dependency, #154299 ImportError: libcudnn.so.9: cannot open shared object file: No such file or directory, #167411 Torch nightly rocm6.4 pytorch-triton-rocm requirement mismatch",a53078d8f32ccd823c0a515344089107b329a7863d9c730dd24cac030ed05e09
issue-collapse:56b2be92a9461e6e,module: autograd needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"76185, 86735, 117662, 153016, 162638, 179652, 181723, 182381, 184466, 185639, 188205, 188375, 188693, 189093",,,,"module: autograd, module: bfloat16, module: complex, module: correctness (silent), module: cpu, module: cuda, module: cuda graphs, module: dtensor, module: dynamo, module: edge cases, module: faketensor, module: flaky-tests, module: functorch, module: higher order operators, module: nn, module: norms and normalization, module: optimizer, module: pipelining, module: pt2-dispatcher, module: tests, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","actionable, bot-mislabeled, bot-triaged, enhancement, feature, has workaround, module: autograd, module: bfloat16, module: complex, module: correctness (silent), module: cpu, module: cuda, module: cuda graphs, module: dtensor, module: dynamo, module: edge cases, module: fakeTensor, module: flaky-tests, module: functorch, module: higher order operators, module: nn, module: norms and normalization, module: optimizer, module: pipelining, module: pt2-dispatcher, module: tests, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, release-feature-request, skipped, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76185 Ensure custom Function are correct in double backward setting, #86735 DISABLED test_reentrant_parent_error_on_cpu_cuda (__main__.TestAutogradDeviceTypeCUDA), #117662 dynamo <> autograd.Function assumes that grad_outputs have the same subclass-ness as forward outputs, #153016 inconsistent grads between two types of `allgather`s, #162638 Gradient Clipping in Pipeline Parallelism Schedules, #179652 `torch.library.custom.register_autograd` incompatible with `jacrev`",945a13efc286ebc17e41a5697af11cd250c2654b8ce74e00d29d4ce26fddc206
issue-collapse:782fbccb35489206,module: aotdispatch needs maintainer decision  batch 1,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,14,0,0,0,"91469, 104053, 110700, 125977, 129964, 131974, 136272, 139152, 141349, 163716, 174970, 178883, 188222, 188232",,,,"module: aotdispatch, module: autograd, module: correctness (silent), module: cuda, module: decompositions, module: dynamo, module: edge cases, module: functionalization, module: fx, module: inductor, module: linear algebra, module: pt2-dispatcher, module: tests, oncall: pt2","PT2-Bug-Bash, actionable, better-engineering, bot-triaged, enhancement, ezyang's list, feature, internal ramp-up task, module: aotdispatch, module: autograd, module: correctness (silent), module: cuda, module: decompositions, module: dynamo, module: edge cases, module: functionalization, module: fx, module: inductor, module: linear algebra, module: pt2-dispatcher, module: tests, months, oncall: pt2, tracker, triaged",,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#91469 torch.compile with aotautograd does not support double backwards, #104053 Tracking issue for optimizer graph not being an inference graph, #110700 More informative variable names in AOTAutograd, #125977 [compile time] AOT Autograd is taking long time in tracing, #129964 run dynamo CI shard with aot_eager_decomp_partition backend to see fallout, #131974 Missing grad_fn information while torch.compile with customized gradient function",91377a258573b0f20366266ed6996d46521c0e09d1d55975ab90293943ffd1fe
issue-collapse:04cc131de7a4a567,module: build needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,13,0,0,0,"28648, 94792, 113149, 126804, 149809, 153546, 159444, 165070, 172807, 172886, 175193, 177014, 178988",,,,"module: build, module: cpp, module: cuda, module: docs, module: infra, module: lint, module: macos, module: mpi, module: risc-v, module: rocm, module: third_party","actionable, better-engineering, bot-mislabeled, bot-triaged, has workaround, module: build, module: cpp, module: cuda, module: docs, module: infra, module: lint, module: macos, module: mpi, module: risc-v, module: rocm, module: third_party, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#28648 Can't successfully install pytorch with python3.6 on my pi 4, #94792 Compiling libtorch from Source on Mac Beyond v1.11.0, #113149 Compilation error on loongarch64, #126804 Torch fails to build with cub from CCCL v2.3.2 or newer, due to -D__CUDA_NO_HALF_OPERATORS__, #149809 LoadHIP.cmake should find_package(composable_kernel), #153546 Make implicit packages (PEP420) explicit PyTorch",1d38de44a0b72320deb25ed6155a99dd6cac2bab90decc20375f7ec65a08ff4a
issue-collapse:36bc9647753a0ad1,module: pipelining needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,11,0,0,0,"110775, 116183, 124372, 136363, 155589, 159711, 160648, 162741, 167425, 167897, 171656",,,,"module: pipelining, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: pipelining, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: pipelining queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#110775 torch.distributed.pipeline skip module throws assert error that portal.grad is not None, #116183 Pipeline parallel doesn't seem to work with mixed precision, #124372 Pipeline Parallelism support for ZeroBubble, #136363 [pipelining] output shape validation / lazy inference, #155589 pipeline() fails when a sub-module uses ""no_grad()""; impacts RoPE implementation on HF models, #159711 Support dict as input/output for pipeline parallelism",309519d77614030b0fae9ee132c5d732d66cf4a1b3025def493b6e4c5f010e99
issue-collapse:e16d5b2a1f31e6bf,module: c10d needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,11,0,0,0,"76913, 76927, 77523, 92694, 119455, 124380, 124468, 151204, 162668, 178600, 178820",,,,"module: c10d, module: ddp, module: dynamo, module: elastic, module: nccl, oncall: distributed, oncall: distributed infra, oncall: pt2","PT2-Bug-Bash, bot-triaged, feature, module: c10d, module: ddp, module: dynamo, module: elastic, module: nccl, oncall: distributed, oncall: distributed infra, oncall: pt2, ptd-bot-triaged, triaged",,"Review this module: c10d queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#76913 Pytorch return TCPStore( RuntimeError: Connection reset by peer), #76927 Can't pickle model torch._C._distributed_c10d.ProcessGroupNCCL' object, #77523 DDP window TCP bug [socket.cpp:558] [c10d] The client socket has failed to, #92694 API to check for errors in c10d.ProcessGroupNCCL, #119455 Users confusing torch.distributed.nn.functional with functional collectives, #124380 Error when running DDP examples to train between nodes",77aefca9067514f5cd98a904962e8ab877a300cd14ce05b0ce10ea1753845eb6
issue-collapse:d528144a67237fea,module: cuda graphs needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,10,0,0,0,"131009, 155679, 158551, 167766, 170241, 176611, 177719, 180960, 188677, 188837",,,,"module: cuda graphs, module: ddp, module: dependency bug, module: error checking, module: higher order operators, module: inductor, module: nccl, module: performance, module: pt2-dispatcher, module: tests, module: vllm, oncall: distributed, oncall: pt2","PT2-Bug-Bash, actionable, bot-triaged, enhancement, has workaround, inference mode, module: cuda graphs, module: ddp, module: dependency bug, module: error checking, module: higher order operators, module: inductor, module: nccl, module: performance, module: pt2-dispatcher, module: tests, module: vllm, oncall: distributed, oncall: pt2, repro:agent:h100:repro_success:reproduced, skipped, topic: new features, triaged, vllm-compile",,"Review this module: cuda graphs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#131009 [cudagraph] an issue related to dynamic shapes for cudagraph backend with sd2.1, #155679 Torch compile CUDA graphs leads to a large number of CUDA streams, #158551 CUDAGraphs RuntimeError: Accessing Overwritten Tensor Output Despite Clone and cudagraph_mark_step_begin in PyTorch 2.3.0, #167766 Fix cudagraph flaky tests, #170241 Allow for multiple stream captures to a cuda graph cudagraph at once., #176611 [cudagraph][torch.compile] `cudagraph_or_error` can be silently ignored even when the cudagraph is not captured.",6a178ba3a792519f013d5f395d87d971b85a2cb36e1b8e052948dcba5b340474
issue-collapse:961e66741d58b33c,module: cpu needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,10,0,0,0,"136748, 140970, 170743, 174266, 176274, 177599, 180650, 183686, 186618, 188926",,,,"module: cpu, module: cuda, module: edge cases, module: empty tensor, module: error checking, module: half, module: inductor, module: linear algebra, module: memory usage, module: nn, module: norms and normalization, module: numerical-stability, module: performance, oncall: cpu inductor, oncall: pt2","actionable, bot-mislabeled, bot-triaged, enhancement, module: cpu, module: cuda, module: edge cases, module: empty tensor, module: error checking, module: half, module: inductor, module: linear algebra, module: memory usage, module: nn, module: norms and normalization, module: numerical-stability, module: performance, oncall: cpu inductor, oncall: pt2, topic: performance, triaged",,"Review this module: cpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#136748 Inconsistent behavior of cdist with half-precision inputs, #140970 `torch._inductor.cpu_vec_isa.pick_vec_isa` takes ~9 seconds to run, #170743 Reference count modifications led to performance drop, #174266 BatchNorm1d CPU outputs differ across devices/threads with identical inputs, #176274 PyTorch + Math libraries: redundant `copy_` calls when working with buffer views, #177599 arange CPU kernel has different precisions for scalar and vectorized executions",ffe030facbfc3bfa7402db1e3cb54ca8f55b04c9f271dc747efeb4771acfd8ef
issue-collapse:b5f1e1737b7a0000,module: arm needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,10,0,0,0,"154444, 162780, 168168, 169926, 170117, 171501, 174088, 174373, 179710, 186521",,,,"module: arm, module: binaries, module: build, module: ci, module: cuda, module: intel, module: openblas, module: tests, module: third_party, module: windows","bot-triaged, feature, module: arm, module: binaries, module: build, module: ci, module: cuda, module: intel, module: openblas, module: tests, module: third_party, module: windows, triaged",,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#154444 [RFC] Removing ideep git submodule dependency from PyTorch & add oneDNN git submodule, #162780 Prebuilt Wheels supporting T4G, #168168 [CD][ARM][CUDA] PYTORCH Released Docker Images Should Ideally Support Multi-ARCH But Currently It is only x86_64, #169926 Failed to build PyTorch with CUDA by using clang with undefined reference of const_data_ptr in TensorBase, #170117 PyTorch Nightly's Torch Compile Feature Does Not Work on THOR due to Triton 3.6.x regression., #171501 Several Windows-related GitHub Actions not running — are they intentionally disabled?",7829d4aa113781c4a389a2b88320f2eb5aaf1ea433f3061977920775602bf2b1
issue-collapse:d801f6fb1a1967a8,module: rocm needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"55937, 109204, 141958, 150168, 171568, 176188, 177548, 179951",,,,"module: rocm, module: sdpa, module: tests, module: third_party, module: windows, oncall: distributed","actionable, bot-triaged, in progress, module: rocm, module: sdpa, module: tests, module: third_party, module: windows, oncall: distributed, skipped, topic: performance, triaged",,"Review this module: rocm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#55937 test_stream_event_nogil: Is the test making a wrong assumption?, #109204 Pytorch ROCM windows builds, #141958 Support SDPA flash attention/ memory efficient attn on ROCm gfx908, #150168 [ROCm] PyTorch slow on TTS, #171568 [ROCm][Inductor] Flash/efficient attention meta kernel shape mismatch on CDNA (MI200/MI300), #176188 test_structured_trace.py's test_ddp_graphs and test_graph_breaks flaky on all platforms",eb675d7b7a643c6bae4ca641603a9c9216a9813480420a4ab4ad69ee801f1af6
issue-collapse:aa3e444026d024fd,module: performance needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"110281, 120308, 131739, 136226, 136380, 175252, 180915, 181488",,,,"module: performance, module: pipelining, module: rocm, module: serialization, module: tensoriterator, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: pt2","bot-triaged, enhancement, feature, module: TensorIterator, module: performance, module: pipelining, module: rocm, module: serialization, oncall: distributed, oncall: distributed infra, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triaged",,"Review this module: performance queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#110281 feat(Pipeline Parallelism): use mincut optimization for local communication optimization, #120308 ROCm: foreach is substantially slower than for loop MI250x, #131739 One Shot Low-Latency kernel is 2x higher than expected, #136226 [pipelining] try not to dry run module when creating PipelineStage?, #136380 [pipelining] Improve recv buffer management, #175252 Slow reading tensor meta-data",251698ebdff5be3667738464981ae13e4d160a8ced073e5221036a3e9032678f
issue-collapse:3449ece47c3e8e69,module: onnx needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"76344, 129279, 143954, 149605, 171759, 173369, 175738, 179709",,,,"module: onnx, oncall: export, oncall: pt2","OSS contribution wanted, PT2-Bug-Bash, bot-triaged, module: onnx, oncall: export, oncall: pt2, onnx-triaged, triaged",,"Review this module: onnx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#76344 [ONNX] Use topk to export max(dim,keepdim) to onnx, #129279 [ONNX] Create unit tests for the new export path by adapting all existing tests, #143954 No ONNX function found for <OpOverload(op='quantized_decomposed.dequantize_per_channel', overload='default')>, #149605 torch.export does not support nested tensor operations, #171759 [ONNX] Create a tutorial, #173369 Exporting ONNX model with captum heatmaps generation",65246fe0ec09fffba3786f8d4644c22e4e53487a1d5612b35cda22fb2e25d3e5
issue-collapse:6567ec1d3bbfe0de,module: higher order operators needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"122588, 133721, 141259, 146860, 146861, 147210, 150062, 169672",,,,"module: higher order operators, module: onnx, module: pt2-dispatcher, oncall: export, oncall: pt2","actionable, better-engineering, feature, module: higher order operators, module: onnx, module: pt2-dispatcher, oncall: export, oncall: pt2, triaged",,"Review this module: higher order operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#122588 [ONNX] Applying consecutive conv2d layers in a for loop, #133721 [HOP] Add BC tests for hop in export, #141259 Consider relaxing torch.cond's requirement that the outputs of each branch have matching requires_grad metadata, #146860 [BE] Put all the HOP tests in one location, #146861 [BE] the majority of HOP OpInfo tests should be in one place, #147210 Consider relaxing the in-place mutation restriction for torch.cond and torch.while_loop",a02c8dfe1a9c78440b09e1f033556cdbae0ae3c9e819b91be3f14e0a232a10ca
issue-collapse:055339978453b4a0,module: dataloader needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"34130, 45728, 72107, 173661, 176166, 179707, 179979, 186908",,,,"module: dataloader, module: ddp, module: error checking, module: flaky-tests, module: multiprocessing, module: performance, module: rocm, module: sparse, module: typing, module: windows, oncall: distributed, oncall: distributed parallelisms","better-engineering, bot-triaged, module: dataloader, module: ddp, module: error checking, module: flaky-tests, module: multiprocessing, module: performance, module: rocm, module: sparse, module: typing, module: windows, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, skipped, triaged",,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#34130 ExceptionWrapper cannot handle exceptions with more than one argument, #45728 Low shared memory, #72107 DistributedDataParallel creates too many threads, #173661 [Perfomance issue] Dataloader has unexpected overhead, #176166 DataLoader: collate_fn leads to high RAM usage, #179707 Improve static type hints for dataloader with known `collate_fn`",3568fb2852dc894ea378080242b016f87cef1c57c46772f0410a48fda273bdf8
issue-collapse:68ba73317f86fd92,module: ci needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"159520, 172427, 173950, 174220, 174930, 175829, 175899, 176687",,,,"module: ci, module: devx, module: infra, module: rocm, module: tests","bot-triaged, feature, module: ci, module: devx, module: infra, module: rocm, module: tests, triaged, unstable",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#159520 [RFC] Enable native Windows CI/CD on ROCm, #172427 [CI] Permission denied while trying to connect to the docker API, #173950 [CI] enabling CUDA 13.0 periodic test, #174220 Dr CI produces confusing issue attribution report, #174930 UNSTABLE inductor / inductor-test-cuda13 / test (inductor_torchbench), #175829 Why the same test is running twice in regular and slow shard",d6b777342cf6be21baded68082a91582046deda7267d5c07c70dc8fede32c50e
issue-collapse:3612eb84a7ce90b2,module: autograd needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,8,0,0,0,"52241, 89757, 131972, 171620, 172797, 174165, 174355, 177386",,,,"module: autograd, module: cuda, module: custom-operators, module: determinism, module: dispatch, module: docs, module: loss, module: nans and infs, module: nestedtensor, module: nn, module: pt2-dispatcher, oncall: pt2","actionable, bot-triaged, feature, has workaround, module: NaNs and Infs, module: autograd, module: cuda, module: custom-operators, module: determinism, module: dispatch, module: docs, module: loss, module: nestedtensor, module: nn, module: pt2-dispatcher, oncall: pt2, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#52241 CTCLoss gradient is incorrect, #89757 third-order gradient of torch.pow with tensor args and certain input returns NaN, #131972 backward of adaptive max pool (adaptive_max_pool2d_backward_cuda) doesn't have a deterministic implementation, #171620 Feedback about torch.autograd.graph.Node.next_functions, #172797 autograd.Function.apply yields surprising strides, #174165 Fallback to Composite Implicit Autograd for Purely Functional Composite Custom Ops",6a93433eb461b2fbbf26323936c4ab2d3b389e4aaff7d2913d9ab21c49c8e288
issue-collapse:12e29a0e9e887908,module: nccl needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"75334, 107174, 114884, 119196, 129862, 136390, 139581",,,,"module: nccl, module: performance, oncall: distributed, oncall: distributed infra","better-engineering, bot-triaged, enhancement, module: nccl, module: performance, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: nccl queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#75334 ProcessGroupNCCL is relying on UB to support bool data type, #107174 Surface NCCL and CUDA version incompatibility, #114884 Undefined reference to ncclCommSplit, #119196 Possible NCCL race condition `destroy_process_group` followed by `init_process_group`., #129862 The call to ncclCommSplit does not have the correct config parameters set, #136390 [c10d][nccl][cuda] Regression (unspecific cuda launch error) with test_c10d_nncl",282b323a111c8622f8c75cb535da5ffa7553646be2fc870ef6a6add14dcfc341
issue-collapse:ccbbec47ef4a537d,module: mps needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"96602, 99037, 123148, 150121, 160732, 177225, 187827",,,,"module: mps, module: performance, module: primtorch, module: rnn, module: tests, oncall: distributed, oncall: distributed infra, oncall: profiler","OSS contribution wanted, bot-triaged, feature, module: mps, module: performance, module: primTorch, module: rnn, module: tests, oncall: distributed, oncall: distributed infra, oncall: profiler, ptd-bot-triaged, triaged",,"Review this module: mps queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#96602 [MPS] softmax returns NaN attention probabilities for large tensors, in float16 and float32., #99037 AttributeError: type object 'torch._C._profiler.ProfilerActivity' has no attribute 'MPS', #123148 GRU is super slow on MPS backend, #150121 torch.compile on MPS progress tracker, #160732 GLOO backend should have MPS capability, #177225 mps test got Exception: mean(): could not infer output dtype.",e35da7c4b7df25cac88f79246c4a3e9d53087e65bda40ae49b9d774781206b18
issue-collapse:804175f78d0cbd44,module: linear algebra needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"9983, 75225, 76528, 151761, 176413, 188135, 188391",,,,"module: linear algebra, module: numpy, module: tests, module: tf32, module: windows, module: xpu","bot-triaged, function request, module: linear algebra, module: numpy, module: tests, module: tf32, module: windows, module: xpu, skipped, triaged",,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#9983 [feature request] Add matrix functions, #75225 One of the backends for `lu_factor` fails on windows., #76528 Discrepancy in einsum when done in batch vs non-batch, #151761 Inconsistent `sum`/`dot`/`norm` behavior, #176413 Output difference anomaly under equivalent migration in logdet operator, #188135 DISABLED test_compare_cpu_linalg_polar_xpu_float32 (__main__.TestCommonXPU)",561133bf3baf39a132e7ebf81ebec1493d8d6d12567a8e9550fcff85e5844736
issue-collapse:b372aa6f199b28df,module: fsdp needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"102402, 108381, 124260, 136408, 141237, 156966, 178749",,,,"module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: fsdp, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, release notes: distributed (composable), triaged",,"Review this module: fsdp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#102402 [Composable] Unified way to check if modules are managed by composable API, #108381 FSDP always puts parameters to fp32 when loading state_dict, even if state_dict has bf16 params, #124260 [FSDP] Unreleased reserved memory after FSDP wrap, #136408 [RFC] Composable parallel documentation, #141237 test_fsdp_tp_integration fails for non-power-of 2 GPUs, #156966 Perf drop when running with FSDP and torch.compile",8c086f17ee6a2b8a03d27f1712c10d160afa5fb56444f41d9604ae454ec246af
issue-collapse:9552690d387bc291,module: floatx (formerly float8) needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"130015, 148420, 153809, 156771, 174417, 187807, 188676",,,,"module: floatx (formerly float8), module: inductor, module: mps, module: tests, module: windows, module: xpu, oncall: pt2","bot-triaged, feature, module: floatx (formerly float8), module: inductor, module: mps, module: tests, module: windows, module: xpu, oncall: pt2, release-feature-request, skipped, triaged, upstream triton",,"Review this module: floatx (formerly float8) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#130015 Improve the Inductor generated kernel for the pattern of `output1 = pointwise(intput); output2 = transpose(output1)`, #148420 Float8_e4m3fn, #153809 Optimize rowwise scaled mm CUTLASS kernel for large values of K, #156771 Support fp16 and fp32 outputs for rowwise scaled_mm, #174417 [release 2.11] [triton] test/inductor/test_loop_ordering.py::LoopOrderingTest::test_fp8_cast_and_t, #187807 [PyTorch 2.13][API-Unstable] Intel XPU Enhancements",7fe4c86ec0adb6b1c6691af05a204a9a476f5d464edeada5e2f835833a946c9f
issue-collapse:a2542a81dc195321,module: elastic needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"75087, 76287, 80824, 121486, 124058, 143823, 160130",,,,"module: elastic, module: tests, oncall: distributed, oncall: distributed infra, oncall: r2p","bot-triaged, enhancement, feature, module: elastic, module: tests, oncall: distributed, oncall: distributed infra, oncall: r2p, ptd-bot-triaged, triaged",,"Review this module: elastic queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#75087 extend torch.distributed's `--tee` to log the nodename, #76287 torch.elastic fails to shutdown despite crashed processes, #80824 When running GPT trainning with megatron, the program quit due to torch.distributed.elastic.agent.server.api:Received 1 death signal, shutting down workers, #121486 Kubernetes API server as Rendezvous backend, #124058 [RFC] Support Torchrun for Out-of-the-tree Devices, #143823 Possible race condition found in TailLogTest.test_tail",8533bdcbc67578536da496f2c5abc84727272aefb0e9955f9e3584fac2bb1133
issue-collapse:ee49a019ee95d542,module: correctness (silent) needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"159155, 163327, 165987, 170837, 173638, 176091, 176625",,,,"module: correctness (silent), module: cpu, module: cuda, module: edge cases, module: error checking, module: inductor, module: linear algebra, module: memory format, module: mps, module: nans and infs, module: nestedtensor, module: openmp, module: sdpa, module: windows, oncall: pt2","actionable, bot-mislabeled, bot-triaged, module: NaNs and Infs, module: correctness (silent), module: cpu, module: cuda, module: edge cases, module: error checking, module: inductor, module: linear algebra, module: memory format, module: mps, module: nestedtensor, module: openmp, module: sdpa, module: windows, oncall: pt2, topic: fuzzer, triaged",,"Review this module: correctness (silent) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#159155 Audit all usages of constrain_to_fx_strides. Many of them might want constrain_to_fake_tensor, #163327 [MPS] scatter_add_ is incorrect on MPS with non-zero offset slices, #165987 `triu_/tril_` ignores memory overlaps which can lead to silent correctness issues, #170837 MPS backend - inconsistent results for batched inference on BERT/RoBERTa, #173638 `torch.linalg.slogdet` does not propagate `NaN` in CUDA, #176091 Windows CPU: SDPA FlashAttention gives wrong results with multiple intra-op threads (mixed OpenMP runtimes)",48c7a6bc1dcd972c158797cdedd6fd130f386b09eb1a48adcddf0841d1e62804
issue-collapse:088da685a573352f,module: compile ux needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"173621, 173642, 175229, 175484, 177830, 180824, 181851",,,,"module: compile ux, module: custom-operators, module: dynamo, module: flex attention, module: inductor, module: pt2-dispatcher, oncall: pt2","bot-triaged, enhancement, feature, function request, module: compile ux, module: custom-operators, module: dynamo, module: flex attention, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",,"Review this module: compile ux queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#173621 Feasibility of Persistent Caching for Dynamo Frontend, #173642 [Feature Request][torch.compile] Improve UX when autotuning repeated layer, #175229 Idiomatic way of matching against flex attention regions in SAC policy with wrap_inductor_compiled_regions, #175484 Unable to disable warning about tf32 precision, #177830 Why register_replacement fails to support fuzzy matching for optional operator parameters in search_fn/search_fn_pattern, #180824 `torch.compile fails with RecursionError on nn.Module with custom __getattr__, while eager execution succeeds`",ec4fe0cd65edfff2d324c8d07715f999d3b899d1714d1d673c6ceefc1c0c3501
issue-collapse:4ade73c4cf18a098,general needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,7,0,0,0,"7887, 127882, 163650, 177021, 185295, 187974, 189279",,,,,"bot-triaged, caffe2, release notes: python_frontend, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#7887 [Caffe2] convert ONNX to caffe2, #127882 tensor.size does not accept `dim=None` argument despite documentation because of some confusion with named dimensions, #163650 [DO NOT CLOSE] Autorevert actions shadow mode stream, #177021 linear attention / mamba2 OP, #185295 [Inductor][feature] Block FP8 quantization: fuse scale layout transformation with quant math, #187974 reserving issue for myself",9a575c49b6b4792fc5796d7f42c19da3777cf9df70afda88801d4521fa24dd75
issue-collapse:7a08b39351fc09d2,module: tests needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"78885, 99082, 160514, 169752, 177260, 177393",,,,"module: tests, module: typing, module: vllm, oncall: distributed, oncall: distributed infra","better-engineering, bot-triaged, enhancement, module: tests, module: typing, module: vllm, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#78885 [distributed_test.py] Improve `test_barrier`, #99082 Look into test coverage for `UntypedStorage`, #160514 BenchmarkingTest.test_benchmark_on_non_zero_device mock hit failed, #169752 if __name__ == ""__main__"": run_tests() cannot be found in the test files under distributed/elastic/rendezvous, #177260 Standardize device-agnostic test cases via instantiate_device_type_tests, #177393 [release 2.11][vllm] Language Models Test (Extended Generation) test_models[True-False-5-32-bigcode/starcoder2-3b]",5590d34448780363f74f79a9ad67914c0a407fcf7614f54246047773a3bd8116
issue-collapse:5788e300a75ef985,module: symm_mem needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"167684, 169035, 170479, 171495, 188001, 188005",,,,"module: symm_mem, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, module: symm_mem, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: symm_mem queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#167684 [RFC] Access to ncclDevComm in kernel, #169035 [Question] Why torch.ops.symm_mem.multimem_all_reduce_() don't support e4m3, e5m2, fp16?, #170479 nvshmem [GPU 0] Peer GPU 8 is not accessible, #171495 [internode] Symmetric mem rendezvous erro, #188001 [RFC] Unify rendezvous caching across SymmetricMemory backends, #188005 [RFC] Protect SymmetricMemory signal pads against cross-stream races",0701253d6fb5b63fba94af69437fd186551760118b1b5c4c34329ba06e8fd7d7
issue-collapse:f1a6db0bf552e352,module: openreg needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"169597, 172319, 172320, 172322, 172803, 177268",,,,"module: openreg, module: privateuse1, module: tests, oncall: distributed","bot-triaged, enhancement, module: PrivateUse1, module: openreg, module: tests, oncall: distributed, triaged",,"Review this module: openreg queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#169597 Standardize Testing in OpenReg, #172319 [OpenReg] Add OpInfo-based operator tests using op_db, #172320 [OpenReg] Extend device infrastructure tests (device, event, streams), #172322 [OpenReg] Extend test_rng.py and test_misc.py coverage, #172803 [OpenReg] PrivateUse1 backend missing distributed backend registration causes KeyError, #177268 [OpenReg] Add testing operator framework usage examples",0149d9aec46b86f5b8cce351818abe2589b005f2c6b1258a1415a421c4513f7c
issue-collapse:414eba3f3b701d08,module: nn needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"38035, 140458, 141288, 172338, 173160, 175032",,,,"module: nn, module: typing","OSS contribution wanted, bot-triaged, enhancement, feature, module: nn, module: typing, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#38035 Quantile Regression Loss, #140458 Making nn.Module generic, #141288 Enhancement Request: improve this error message, #172338 Fix: Support is_causal=True with need_weights=True in Multi-Head Attention, #173160 Feature Request: load_state_dict should optionally support partial tensor filling in non-strict mode, #175032 Dead code: reduction of post AI tech debt",d7f731f8648d6d134f80e9a47db85a397bc9fb6dde039ed99950aead7c2c4e77
issue-collapse:e20256afc644d521,module: distributions needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"133702, 136532, 145679, 171283, 172059, 179430",,,,"module: distributions, module: docs, module: numerical-stability, module: python frontend","bot-triaged, feature, function request, module: distributions, module: docs, module: numerical-stability, module: python frontend, topic: new features, triaged",,"Review this module: distributions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#133702 Add `beta`, `betainc`, etc. to `torch.special`, #136532 Wrong results when sampling from the beta distribution for small alpha=beta, #145679 Gamma cdf function, #171283 aten.uniform cannot be called with keyword argument from, #172059 Add Tweedie distribution to torch.distributions, #179430 Add Frechet/Inverse Weibull distribution",83b48df9c699fa6132895e134614dfa0b9c54c785240d63a180e62d5d963b948
issue-collapse:30631a1a3e977d18,module: convolution needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"130614, 177245, 177277, 178134, 178472, 178839",,,,"module: convolution, module: cpu, module: cuda, module: error checking, module: faketensor, module: inductor, module: memory format, module: memory usage, module: meta tensors, module: mkldnn, module: nn, module: pt2 accuracy, module: pt2-dispatcher, module: rocm, oncall: pt2","actionable, bot-triaged, has workaround, module: convolution, module: cpu, module: cuda, module: error checking, module: fakeTensor, module: inductor, module: memory format, module: memory usage, module: meta tensors, module: mkldnn, module: nn, module: pt2 accuracy, module: pt2-dispatcher, module: rocm, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",,"Review this module: convolution queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#130614 Significant Memory Increase with Conv2d using padding_mode 'replicate', 'reflect', or 'circular', #177245 AArch64 Unit Test Failures - Multiple failures in test/nn/test_convolution.py TestConvolutionNNDeviceTypeCPU for oneDNN, #177277 Conv3D does not preserve dim_order for some inputs, #178134 torch.compile (CUDA fp32) vs eager gives large max relative error for Conv2d, #178472 conv1d on meta device path skips groups validation and returns an invalid output shape, #178839 [ROCm] MIOpen Gemm solver receives workspace_size=0 via legacy Find1 API",b0f809e4c86f028d30ff189c5237c43072c2b078359301d869ba704841d9ff5d
issue-collapse:515dfcb7e045b03f,module: aotinductor needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"144459, 169182, 180811, 186143, 187907, 189189",,,,"module: aotinductor, module: build, module: complex, module: cpp-extensions, module: cuda, module: determinism, module: docs, module: dynamic shapes, module: inductor, module: performance, oncall: export, oncall: pt2","PT2-Bug-Bash, bot-triaged, has workaround, module: aotinductor, module: build, module: complex, module: cpp-extensions, module: cuda, module: determinism, module: docs, module: dynamic shapes, module: inductor, module: performance, oncall: export, oncall: pt2, repro:agent:h100:repro_fail:skipped, triaged",,"Review this module: aotinductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#144459 In AOTI Runtime, acquiring a write lock on model_exec_mutex_ may cause starvation., #169182 Kineto library not installed, #180811 [AOTInductor] max_autotune non-deterministically selects cuBLAS over fused Triton kernels for small GEMM, causing significant performance regression, #186143 AOTInductor Lowering Backward Compatibility, #187907 AOTInductor: reverse-mode autograd.grad through sqrt/reciprocal bakes a symbolic-shaped constant under a dynamic batch (strict export), #189189 Subtraction between two `torch.complex64` tensors fails on modules compiled with `aoti_compile_and_package`",0168bb42aa57ef11a7a28803757ea00dbd9a5ea8335a201af0cc6cbf0d784237
issue-collapse:447201ae583ca5c3,module: aotdispatch needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"103650, 104146, 109237, 130028, 136092, 172213",,,,"module: aotdispatch, module: inductor, module: pt2-dispatcher, oncall: distributed, oncall: pt2","feature, module: aotdispatch, module: inductor, module: pt2-dispatcher, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, tensor subclass, tracker, triaged",,"Review this module: aotdispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#103650 AOT autograd: Avoid dependency on strides for manual regeneration of outputs that are aliased to inputs, #104146 make_fx: torch.where scalar promotion burns in device, #109237 AOTAutograd should track view chains so it can replay them, instead of using as_strided., #130028 Support python attribute mutations of tensor subclass graph inputs from within __torch_dispatch__, #136092 User API For Donated Buffers, #172213 Eager buffer mutation after `torch.compile` + backward + `torch.export.export` hits `_map_assigned_buffer_to_proxy` `FakeTensor` assertion",7fbf99e6d09c4056a7c876d868fc24fbebd5a846a23f476f45b2a6f39f52995e
issue-collapse:b1f9cf8964ce3cf4,module: __torch_function__ needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,6,0,0,0,"54881, 115741, 186203, 186259, 188684, 188686",,,,"module: __torch_function__, module: bfloat16, module: complex, module: convolution, module: cuda, module: docs, module: dynamo, module: flaky-tests, module: pt2-dispatcher, module: tests, oncall: pt2","PT2-Bug-Bash, bot-triaged, feature, module: __torch_function__, module: bfloat16, module: complex, module: convolution, module: cuda, module: docs, module: dynamo, module: flaky-tests, module: pt2-dispatcher, module: tests, oncall: pt2, skipped, tracker, triaged",,"Review this module: __torch_function__ queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#54881 Often the documentation does not match the actual signature of the function, #115741 torch_function mode explodes with torch.compile, #186203 DISABLED test_redispatch_nn_functional_linear_cross_entropy_cuda_float16 (__main__.TestTorchFunctionRedispatchOpsCUDA), #186259 DISABLED test_redispatch_nn_functional_linear_cross_entropy_chunked_cuda_float16 (__main__.TestTorchFunctionRedispatchOpsCUDA), #188684 DISABLED test_redispatch_nn_functional_conv_transpose3d_cuda_bfloat16 (__main__.TestTorchFunctionRedispatchOpsCUDA), #188686 DISABLED test_redispatch_nn_functional_conv_transpose3d_cuda_complex32 (__main__.TestTorchFunctionRedispatchOpsCUDA)",8b2fc85da273816fe6d1a21418e39661f14754d3085458bc83f4aad37edc9434
issue-collapse:17ba7a403b3e0fd2,oncall: pt2 needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"125557, 143337, 162728, 172522, 177053",,,,oncall: pt2,"enhancement, oncall: pt2, topic: fuzzer, triaged, upstream triton",,"Review this oncall: pt2 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#125557 result of 2 ** s differs between eager mode and inductor + triton + cuda when in float32 denormal range, #143337 An idea to improve Welford implementation in Inductor, #162728 Significant Numerical Inconsistencies After torch.compile() for Local Attention Model, #172522 Support out-of-graph mutations with side streams, #177053 [inductor] realize_reads_threshold could be tuned differently with auto-chunking",b1b9d6f975e54d69987ad5ba763d856388ec8d340111a008a0a0848cdce00faf
issue-collapse:3d0aed210d1d22d8,module: xpu needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"134956, 142135, 143239, 173897, 185198",,,,"module: xpu, oncall: distributed","bot-triaged, has workaround, module: xpu, oncall: distributed, triaged",,"Review this module: xpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#134956 Pytorch XPU Windows build failed in cmake rerun loop due to the source code deep path, #142135 Device assert throws a runtime error in cuda backend and results in a crash in xpu backend, #143239 xpu: torch.nn.DataParallel fails on multi-XPU environment with ""module 'torch._C' has no attribute '_scatter'"", #173897 XPU dist.all_reduce is async: torch.xpu.synchronize does not wait for collective completion, #185198 [xpu] Intel GPU telemetry management APIs may not work on Arc A-Series and data center GPUs",29d0a65dcddb73ba2ab65c0b4e02f7c0748312f23f8b0dd4a92d05d087b13607
issue-collapse:7bae885e1ca831e7,module: pt2-dispatcher needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"129430, 145522, 160333, 172549, 174131",,,,"module: pt2-dispatcher, module: vllm, oncall: export, oncall: pt2","feature, internal ramp-up task, module: pt2-dispatcher, module: vllm, oncall: export, oncall: pt2, tensor subclass, triaged, vllm-compile",,"Review this module: pt2-dispatcher queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#129430 OperatorBase.py_impl should support with_keyset=True extra argument., #145522 use_const_ref_for_mutable_tensors doesn't work with out= overloads, #160333 Tensor subclasses don't work with torch.library.triton_op, #172549 AOTAutograd error when compiled graph has two DTensor outputs that alias each other, one differentiable one not, #174131 turn on autograd_cache_normalize_inputs in fbcode",fbabcfbf0e5976907e2f00711c569bdff73fa6e9b83e608d1f2b00908ce5d261
issue-collapse:b336a1f2c9c470e0,module: numerical-stability needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"56340, 108521, 126158, 132989, 139964",,,,module: numerical-stability,"matrix multiplication, module: numerical-stability, triaged",,"Review this module: numerical-stability queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#56340 log_softmax(x) != x - logsumexp(x), #108521 resutl of (torch.mm(a,b) does not match result of (a[:part,:], b), #126158 [torch.arange] Output mismatch with Numpy / TensorFlow, #132989 Different random seeds can cause differences in arithmetic precision, #139964 Report issue for torch.nn.Linear when forwarding a 3-dim tensor.",cd4a7f647a13b2bd3979f954367f8cd86629011d6287f8c8cc527a4babe2721e
issue-collapse:4c79579d5d44769f,module: inductor needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"125254, 146780, 165398, 168573, 174313",,,,"module: inductor, module: mps, module: rocm, module: vllm, oncall: pt2","bot-triaged, module: inductor, module: mps, module: rocm, module: vllm, oncall: pt2, rocm-skipped-tests, triaged, vllm-compile",,"Review this module: inductor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#125254 AssertionError: Device mps not supported, #146780 TypeError when using torch.compile with RegionViT under torch.inference_mode(), #165398 Skipping combo kernel for large pointwise nodes, #168573 TestModule: test/inductor/test_codecache.py::TestFxGraphCache, #174313 [release 2.11][triton] trunk / linux-jammy-rocm-py3.10 / test failures",4f8ba5567cb596b6880aaaca9bdca6624a2ed4d779b9477b87db09080d668f60
issue-collapse:5049151eeccfaba5,module: flex attention needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"141784, 153221, 159770, 173071, 188212",,,,"module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, module: rocm, oncall: pt2","bot-triaged, enhancement, feature, module: flex attention, module: higher order operators, module: inductor, module: pt2-dispatcher, module: rocm, oncall: pt2, triaged",,"Review this module: flex attention queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#141784 [`FlexAttention`] Request for the Support of Dropout, #153221 FlexAttention: Support for post_mod, #159770 extract statistics from attention weights in FlexAttention, #173071 Allow privateuse1 devices in FlexAttention device validation, #188212 [RFC] Add Gluon Backend To Inductor Backend",55885c89dd2033e1d314a5acd885fbf244c45376a3a4b82777e0fe335b56369a
issue-collapse:0362b78d88914ce5,module: determinism needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"88096, 166856, 172475, 174711, 175478",,,,"module: determinism, module: distributions, module: inductor, module: random, oncall: pt2","bot-triaged, feature, function request, module: determinism, module: distributions, module: inductor, module: random, oncall: pt2, triaged",,"Review this module: determinism queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#88096 Add nondeterministic alert to `torch.Tensor.scatter()`, #166856 CPU `normal_()` Non-Deterministic Behavior with Non-Contiguous Tensors, #172475 [Inductor] Track deterministic `alloc_id` assignment across cached compilation artifacts, #174711 add an alternative to compile time benchmarking in inductor for determinism that is reasonably performant, #175478 torch.distributions.Gamma().sample() should take a torch.Generator",861d5bebe5996d62fb34701794e7639b09135a62c469f28873ebb36a642a7eb6
issue-collapse:483d76900adb6537,module: data parallel needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"90052, 91629, 101402, 114437, 156392",,,,"module: data parallel, module: dataloader, module: ddp, module: functorch, module: performance, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, module: data parallel, module: dataloader, module: ddp, module: functorch, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: data parallel queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#90052 DDP overlapped optimizer: set grads to None enhancements, #91629 vmap + nn.SyncBatchNorm.convert_sync_batchnorm, #101402 DataParallel for nested modules, #114437 [Help Needed] [DataParallel/DataLoader/AMP] Severe slowdown on large dataset with fp16 precision and DataParallel wrapper, #156392 DataParallel gather NaN across multiple gpus",0f743b7d2563d589a389348ecb2e85b7feff3eeab5cb7feb683fe4641801ae99
issue-collapse:d98076db6b13a760,module: amp (automated mixed precision) needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"127176, 163688, 164455, 172383, 174763",,,,"module: amp (automated mixed precision), module: bfloat16, module: docs, module: functorch, module: inductor, module: linear algebra, module: mta, module: pipelining, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","actionable, bot-triaged, has workaround, module: amp (automated mixed precision), module: bfloat16, module: docs, module: functorch, module: inductor, module: linear algebra, module: mta, module: pipelining, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, triaged",,"Review this module: amp (automated mixed precision) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#127176 RuntimeError: ""_amp_foreach_non_finite_check_and_unscale_cuda"" not implemented for 'BFloat16', #163688 [torch.distributed.pipelining] Gradients are None in first training step with ScheduleGPipe, #164455 AMP + Torch Compile bug, #172383 torch.func.linearize produces NaNs with autocast + LayerNorm (jvp works), #174763 torch._grouped_mm is not autocast-compatible, causes dtype mismatch",1b39690124c5244260cd84a199579c3f8ff851a19c26d4ed267cbfae609464fd
issue-collapse:6a167e2f90c63878,module: accelerator needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,5,0,0,0,"143887, 156805, 177771, 177935, 186260",,,,"module: accelerator, module: cuda","bot-triaged, feature, function request, module: accelerator, module: cuda, triaged",,"Review this module: accelerator queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#143887 `torch.accelerator` cross-device utilities and properties, #156805 [RFC] Support the unified memory API GetAllocator and getDeviceAllocator for in-tree and out-of-tree backend, #177771 Accelerators: torch.Stream.record_event doesn't handle torch.cuda.Events even though torch.cuda.Event is a torch.Event subclass, #177935 [RFC] Add A Cross-Backend Device Name API: `torch.accelerator.current_device_name()`, #186260 [Accelerator] Maybe host memory should be included in torch.accelerator.mem_stats",bb97b71d57a15adcb19a691afb28157c4fdf8c3740589a03f5b790eae7487533
issue-collapse:be82ddaa7de75db9,module: sdpa needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"117974, 161757, 179382, 187770",,,,module: sdpa,"bot-triaged, function request, module: sdpa, release-feature-request, triaged",,"Review this module: sdpa queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#117974 RuntimeError: derivative for aten::_scaled_dot_product_efficient_attention_backward is not implemented, #161757 [RFC] Flash decoding optimization based on FlexAttention on X86 CPU, #179382 Support dropout in varlen_attn, #187770 torch.nn.attention.varlen: make backend selection configurable",4e8df44118955de8738ba11cc19dcd101f1d718a3fbb064f0c97cb85bc450a21
issue-collapse:5e9ec2c382f654d2,module: rpc needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"80774, 81622, 95487, 106851",,,,"module: rpc, module: tensorpipe, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, module: rpc, module: tensorpipe, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: rpc queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#80774 RPC: Make RRefProxy callable, #81622 [Distributed] test_dynamic_rpc_existing_rank_can_communicate_with_new_rank_cuda fails in caching allocator, #95487 Torch RPC on multiple nodes with GPU returns a EOF error, #106851 RPC all_gather doesn't work with dynamic world size (world_size=None)",c36bd986d22c54efd57f7938a1f19be818e18c3cfc32d79404c9921730faed62
issue-collapse:fbe5a7a7aa048601,module: docs needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"133511, 152213, 171866, 175479",,,,"module: docs, module: dynamo, module: minifier, module: rnn, module: vllm, oncall: pt2","PT2-Bug-Bash, actionable, bot-mislabeled, bot-triaged, module: docs, module: dynamo, module: minifier, module: rnn, module: vllm, oncall: pt2, triaged, vllm-compile",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#133511 Do dynamo/inductor support custom classes like TorchScript., #152213 Outdated install commands, #171866 Why are there three run_repro functions implemented in pytorch?, #175479 Feedback about LSTM",aea197d6ae5b7be5acf3e3e6d36d202061f06e31e54918eaafd5c8a0a0e20822
issue-collapse:252d2ab1990b822d,module: devx needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"167775, 177322, 186939, 188446",,,,"module: devx, module: windows, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, feature, module: devx, module: windows, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: devx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#167775 Graceful Ctrl+C handling from torchrun, #177322 Support TORCH_SHOW_CPP_STACKTRACES=1 on Windows, #186939 pitch: fast development mode where we use pre-build nightly .so files, #188446 Add issue de-duplication system",eed5d4f60d0f33446cbc096f92d4a2aafbbb4ca54136723b93b25a45725771b9
issue-collapse:17b89efd690988d1,module: deadlock needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"113228, 144634, 162872, 174359",,,,"module: deadlock, module: mps, module: multithreading, module: openreg, module: third_party","bot-triaged, module: deadlock, module: mps, module: multithreading, module: openreg, module: third_party, release notes: dataloader, triaged",,"Review this module: deadlock queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#113228 dataloader use lmdb stuck, #144634 [MPS] `torch.mps.synchronize` hangs on error, #162872 [MPS] dead lock when calling `Event.synchronize()` before `Event.elapsed_time()`, #174359 The OpenReg runtime is stuck",963e6655a3e4c8dbe230bba6d5526de52a572aa93e4de23151c0fed9f4feaf8b
issue-collapse:3991108aac374b6d,module: ddp needs maintainer decision  batch 4,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"80496, 80595, 81383, 82153",,,,"module: ddp, module: fsdp, module: optimizer, module: performance, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, feature, module: ddp, module: fsdp, module: optimizer, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: ddp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#80496 DDP find_unused_parameters=True does not work for Sparse gradients, #80595 Overlapping Optimizer.step() with DDP backward, #81383 Investigate ncclRedOpCreatePreMulSum operator for gradient reduction, #82153 DistributedDataParallel hangs when not using GPU 0",fef07a99ee75cc4a5a7160c8b927d87cb22ff786871670cdad616bff58dec1d6
issue-collapse:3119c95d4da5cafa,module: cpp needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"108794, 163974, 176153, 177520",,,,"module: cpp, module: cuda, module: doc infra, module: docs, module: memory usage, module: nn, module: windows","bot-mislabeled, bot-triaged, enhancement, module: cpp, module: cuda, module: doc infra, module: docs, module: memory usage, module: nn, module: windows, triaged",,"Review this module: cpp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#108794 [C++ Frontend] Simple Changes for Cleaner Options, #163974 C++ build dependencies, #176153 Libtorch - make call reset_parameters during init optional, #177520 Memory management, memory not returned after CPU -> GPU",a17c77a6c2f492ccd52dcfab5eea53f6dc1bade9385a84dd38213637b1b89ca9
issue-collapse:531ff8f82672e300,module: complex needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"99652, 139313, 146691, 184769",,,,"module: complex, module: ddp, module: linear algebra, module: mps, module: performance, oncall: distributed, oncall: distributed parallelisms","actionable, bot-triaged, enhancement, module: complex, module: ddp, module: linear algebra, module: mps, module: performance, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: complex queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#99652 DistributedDataParallel doesn't work with complex buffers, #139313 BFloat16 support for `torch.fft.rfft`, #146691 MPS support `torch.linalg.norm` on complex numbers, #184769 Optimize Metal Kernels with sincos",0194c4e02bfeb637353b581f4418a265d7d2f548cfdc0d1e675e52d3147e92f6
issue-collapse:40867ecc9dc1fb30,module: binaries needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"74282, 119461, 171194, 172351",,,,"module: binaries, module: ci, module: cuda, module: cudnn, module: docker, module: vision","module: binaries, module: ci, module: cuda, module: cudnn, module: docker, module: vision, topic: binaries, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#74282 RuntimeError: cuDNN error: CUDNN_STATUS_VERSION_MISMATCH for torchvision models, #119461 Promote pytorch nightly docker image, #171194 CUDA Smoke tests: Extend smoke tests to include profiler coverage, #172351 [RFC][Release-2.11] Dropping Volta support from CUDA-12.8",dd9b50a6e940da4a46e209d049558ea461a44b1309d1b39ba9cf940a68b26a9b
issue-collapse:9a7ca2c2978b9357,module: bfloat16 needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"90574, 170758, 177859, 181718",,,,"module: bfloat16, module: complex, module: cpu, module: cublas, module: cuda, module: edge cases, module: linear algebra, module: numpy, module: performance","bot-triaged, enhancement, matrix multiplication, module: bfloat16, module: complex, module: cpu, module: cublas, module: cuda, module: edge cases, module: linear algebra, module: numpy, module: performance, onnx-needs-info, triaged",,"Review this module: bfloat16 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#90574 Bfloat16 tensor .numpy() support, #170758 bf16 layernorm has wrong answer, #177859 Combining `bfloat16` with complex scalars raises `NotImplementedError`, #181718 bf16 matmul throughput drops on RTX 4090 for shapes where N%16==8",b733d36f071844f4134b35c54bb499945dfb3775feb26795127e655bfe177781
issue-collapse:0fa75cf889339355,module: benchmark needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"164593, 171376, 172495, 188851",,,,"module: benchmark, module: ci, module: dependency bug, module: rocm, oncall: pt2","bot-triaged, module: benchmark, module: ci, module: dependency bug, module: rocm, oncall: pt2, triaged",,"Review this module: benchmark queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#164593 [CI] Make sure B200 benchmark testing covers regression such as https://github.com/triton-lang/triton/issues/8328, #171376 PyTorch ThroughputBenchmark.add_input crashes when split_sizes is incompatible with input tensor shape, #172495 Add T5Gemma2 to TorchInductor benchmark HF suite, #188851 [CI][ROCm] DebertaV2ForMaskedLM fails to import from transformers on gfx950 (dynamic_inductor_huggingface eager_fail_to_run)",f373360ab952cd1bb57d2bf35ef779de01b34cdeb94ef333d9f74a17418a8269
issue-collapse:a288063f2ab3f8bc,module: activation checkpointing needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,4,0,0,0,"120945, 121594, 155547, 175306",,,,"module: activation checkpointing, module: compile ux, module: ddp, module: distributed_tool, module: fsdp, module: higher order operators, module: pt2-dispatcher, module: rocm, oncall: distributed, oncall: distributed parallelisms, oncall: pt2","bot-triaged, enhancement, module: activation checkpointing, module: compile ux, module: ddp, module: distributed_tool, module: fsdp, module: higher order operators, module: pt2-dispatcher, module: rocm, oncall: distributed, oncall: distributed parallelisms, oncall: pt2, ptd-bot-triaged, rocm-skipped-tests, triaged",,"Review this module: activation checkpointing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#120945 DISABLED distributed/_composable/fsdp/test_fully_shard_training.py::TestFullyShard1DTrainingCompose::test_train_parity_with_activation_checkpointing, #121594 [DDP] Gradient Synchronization Failure Induced by model.gradient_checkpointing_enable() in multi-task setting, #155547 [Misc] The keys of the sac_milp return dict were modified, but the test case was not updated, #175306 SAC should treat inductor_compiled_code and other HOPs specially (pull out a more unique name) to prevent mixups",6377a904dc341015b199bc03cbf295da81768d7490fa329b18637027cb7fd0eb
issue-collapse:f7fbf4ec9334dc85,module: sparse needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"97668, 171558, 181494",,,,module: sparse,"bot-triaged, feature, module: sparse, triaged",,"Review this module: sparse queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#97668 sparse_csr_tensor matmul wrong output in bfloat16, #171558 `torch.sparse.spdiags` produces incorrect results, #181494 Add support for sparse block-diagonal matrices",c9145f19f3123d3a6dcbc947d74d59b383c1b401f9db5fa6de4c3287a05cb2f5
issue-collapse:baf9b2460c2846ee,module: nans and infs needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"163997, 173796, 182156",,,,"module: nans and infs, module: nn, module: sdpa","bot-triaged, feature, module: NaNs and Infs, module: nn, module: sdpa, triaged",,"Review this module: nans and infs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#163997 `F.scaled_dot_product_attention` returns all-`NaN` rows during evaluation but not training, #173796 `torch.nn.functional.PoissonNLLLoss` returns `NaN` imaginary part on CPU (vs 0 on GPU) for overflowing complex inputs, #182156 Add torch.utils.nan_guard for forward-pass NaN/Inf debugging",bf7cccae343c596b20125606d3cbebd4bef65478424bdbf9a3986445f359e95d
issue-collapse:34d663378a7ef8bb,module: flaky-tests needs maintainer decision  batch 7,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"150877, 167272, 167297",,,,"module: flaky-tests, module: higher order operators, module: inductor, module: pt2-dispatcher, module: rocm, oncall: pt2","module: flaky-tests, module: higher order operators, module: inductor, module: pt2-dispatcher, module: rocm, oncall: pt2, rocm-skipped-tests, skipped, tracker, triaged",,"Review this module: flaky-tests queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#150877 DISABLED test_multiple_module (__main__.TestInvokeSubgraphExportStrict), #167272 DISABLED test_max_autotune_addmm_max_autotune_gemm_backends_ATen,Triton,CK_x_shape2 (__main__.TestCKBackend), #167297 DISABLED test_max_autotune_addmm_max_autotune_gemm_backends_ATen,Triton,CK_x_shape1 (__main__.TestCKBackend)",d203cf7c2897d18878e1324acbebaf78ff0c3e225b3489180adab448bbfd8b65
issue-collapse:a0d8de4d075cf58d,module: error checking needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"174602, 188087, 188360",,,,"module: error checking, module: nn, module: reductions","bot-triaged, enhancement, module: error checking, module: nn, module: reductions, triaged",,"Review this module: error checking queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#174602 [nondeterminism][inconsistency] ctc_loss sometimes return incorrect results, #188087 torch.nanmedian(..., dim=None) raises a named-dimension RuntimeError while omitting dim works, #188360 Inconsistent and unclear error messages for `offsets` validation when `input.dim() == 1` in `torch/nn/functional.py`",703b3ddd44786ab7f26263d6f51b1dea5de67d92ab907580f019ed525bc5590e
issue-collapse:664990fe462cfd3c,module: distributed_tool needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"108212, 126302, 162755",,,,"module: distributed_tool, module: tests, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, feature, module: distributed_tool, module: tests, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,"Review this module: distributed_tool queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#108212 Is there a standard procedure to check the consistency of environment across all nodes in PyTorch DDP training?, #126302 Make it easier to run distributed tests without run_test.py, #162755 Ideally, we should not use `require_exact_world_size(4)` if we only have 8-GPU runners",2619e922a005ce81d1212292293a79416c8fd38ff55766ff503f25ec1c350f77
issue-collapse:c7cd49ea4ba584a6,module: custom-operators needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"125030, 138355, 164696",,,,"module: custom-operators, module: inductor, module: pt2-dispatcher, module: reinplacing, oncall: fx, oncall: pt2","feature, function request, internal ramp-up task, module: custom-operators, module: inductor, module: pt2-dispatcher, module: reinplacing, oncall: fx, oncall: pt2, triaged",,"Review this module: custom-operators queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#125030 Allow operators with no returns to be ""out="" operations, #138355 tlparse's fx_graph_runnable should include custom op info, #164696 Support torch._inductor.config.inplace_buffers for custom_op whenever possible",5d946c30f7fa612d8f8d5a2c3acda35369cbb443611a23ed4dc8befb2ebdd249
issue-collapse:fc9dd530e5305245,module: cublas needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"172231, 174949, 181737",,,,"module: cublas, module: cuda, module: numerical-stability, module: vllm","bot-triaged, matrix multiplication, module: cublas, module: cuda, module: numerical-stability, module: vllm, triaged",,"Review this module: cublas queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#172231 [cuBLAS] setting `cublas/cublaslt` backend seems to be a no-op for `torch.addmm` when dispatched to cuBLASLt, #174949 [vllm] CUBLAS_STATUS_INVALID_VALUE in cublasGemmEx after upgrading to PyTorch 2.10, #181737 image diffusion model output changes for the worse after 530ff92655",8494ed5ce6ce3c81e625487423ffbd9acaa1876176983f6c7b11927d3ef16f8a
issue-collapse:04b61a70b2cc311a,module: cpp-extensions needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"127445, 158032, 166123",,,,"module: cpp-extensions, module: custom-operators, module: pt2-dispatcher, module: rocm, module: windows, oncall: pt2","feature, module: cpp-extensions, module: custom-operators, module: pt2-dispatcher, module: rocm, module: windows, oncall: pt2, triaged",,"Review this module: cpp-extensions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#127445 PyTorch version agnostic C++ extensions, #158032 extension-cpp claims ROCM support with no additional changes; we should test this., #166123 Can't compile torch-cublas-hgemm project in Windows on a torch newer than 2.8.0, I get complains involving compiled_autograd.h",e5b98faa36fd7df4bf0a636c906cc3fae4f4d7f273bb5b1e8d79fc79917e0fee
issue-collapse:dbe1053daeff7eac,module: abi needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"103359, 121837, 174507",,,,"module: abi, module: binaries, module: build, module: cpp","bot-triaged, feature, module: abi, module: binaries, module: build, module: cpp, triaged",,"Review this module: abi queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#103359 Libtorch compile error when defining D_GLIBCXX_DEBUG, #121837 Revive libtorch-shared-without-deps flavor, #174507 PyTorch Stable ABI support",f7f9cd13f4e94d8d412496fb75e2094204672ebb51b89c4e7d0411dea4cecbdd
issue-collapse:5c4c2f9e51a065d6,module: __torch_dispatch__ needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,3,0,0,0,"62982, 113089, 117115",,,,"module: __torch_dispatch__, module: bootcamp, module: docs, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, better-engineering, feature, module: __torch_dispatch__, module: bootcamp, module: docs, module: pt2-dispatcher, oncall: pt2, tensor subclass, triaged",,"Review this module: __torch_dispatch__ queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#62982 Don't use try-catch to handle overload resolution handling in torch.ops, #113089 Add docs for __tensor_flatten__ / __tensor_unflatten__, #117115 request: torch.compile support for tensor subclass subgraph boundaries",d955ca9255eff77624142116ab853bc2f09d03d9bbde049f7826a2e13f0b9756
issue-collapse:87c92d986835ff8e,oncall: releng needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"136275, 159516",,,,oncall: releng,"better-engineering, oncall: releng, triaged",,"Review this oncall: releng queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#136275 Please offer `<version>+cpu`-versioned packages for MacOS and linux ARM CPU-only builds, #159516 Validate PyPI Wheel Binary Size failing for >3months",3487460e26db4a84fb754fadab17f19ed182048e2b8be589595625cd3c64bb99
issue-collapse:e0d7af2daf65329a,oncall: export needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"186381, 186594",,,,"oncall: export, oncall: pt2","bot-triaged, oncall: export, oncall: pt2, triaged",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#186381 [export] No-op contiguous() elided during tracing bakes trace-time layout into the graph; downstream view becomes layout-specific, #186594 torch.ops.higher_order.associative_scan cannot be torch.export'ed",ece4e6fc78727db4f1c81fac43e13897c06eed383b71e040c59e24edc7b9667b
issue-collapse:6fcb0b986d9b5d9f,module: windows needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"157989, 167062",,,,"module: windows, module: xpu, oncall: pt2","module: windows, module: xpu, oncall: pt2, triaged",,"Review this module: windows queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#157989 [XPU] torch.xpu.mem_get_info() query failed or reports Out-Of-Memory even when there are enough free memory, #167062 How to use torch.compile on Windows GPU?",d91b6b97619a8828dffa1a739b69e26cd07d1665f8d8b0a7b4b643399b755659
issue-collapse:73cf360c075b3ba9,module: typing needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"165628, 181861",,,,module: typing,"enhancement, module: typing, triaged",,"Review this module: typing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#165628 [Meta] Type Hints in Pytorch, #181861 nn.Sequential type hints too general for nested lists",f6b58503deec72842f8659a62d43a246ab23054e863c3e18df22c5d6d8aedf05
issue-collapse:e0d13dde4c429c53,module: serialization needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"55777, 71389",,,,module: serialization,"enhancement, has workaround, module: serialization, triaged",,"Review this module: serialization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#55777 torch.load is very slow with gzip.open, #71389 Create a nested directory while saving objects using `torch.save`",2267b0b46b054f1a0ac6144b0feab8a71bb54f720e8706cfaa7c3972157da0fd
issue-collapse:7da26dad57779c73,module: scipy compatibility needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"85366, 179784",,,,"module: scipy compatibility, module: special","bot-triaged, enhancement, module: scipy compatibility, module: special, triaged",,"Review this module: scipy compatibility queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#85366 More windows for filtering and spectral analysis, #179784 torch.xlogy(0, 0) returns NaN, while in many numerical computing libraries (e.g., SciPy) and common machine learning practice, it is defined as 0.",2a7549464d4cf0b08d9114d867da1ecd3863b5774c2713553a920e17812d6152
issue-collapse:bd68d939ab4ffd1b,module: pytree needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"119328, 188648",,,,"module: pytree, module: serialization","bot-triaged, module: pytree, module: serialization, triaged",,"Review this module: pytree queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#119328 [pytree] Handling of `None` in torch.utils._pytree is inconsistent with JAX., #188648 treespec_dumps does not preserve tuple keys in dict/OrderedDict/defaultdict pytrees, so the round-trip fails with `unhashable type: 'list'`",8e7174a4259aa256bb03ca0d4a2757567126161f137816a3e06965eecaf88f0a
issue-collapse:933343c3b47782bc,module: python frontend needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"170419, 176078",,,,module: python frontend,"bot-triaged, inference mode, module: python frontend, triaged",,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#170419 Cannot set version_counter for inference tensor, #176078 custom_op with mutable optional arguments",f0bcc08a6ae233c756af06161c6c63f5404a5a5acfb61dbae038b287c7471ea5
issue-collapse:9a55fac356e14cb3,module: privateuse1 needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"158911, 188688",,,,"module: privateuse1, module: testing, module: tests, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, module: PrivateUse1, module: testing, module: tests, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, skipped, triaged",,"Review this module: privateuse1 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#158911 [RFC] Refactor test_c10d_nccl.py to Support Backend-Agnostic Testing for PyTorch, #188688 DISABLED test_backend_simple (__main__.PrivateUse1BackendTest)",3e844fab83dee985c00bd2adbe4c560c6f8625d58d842a0cf0ccc69131ccd2e6
issue-collapse:37a07f7416ebd9af,module: optimizer needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"50282, 179143",,,,module: optimizer,"enhancement, feature, module: optimizer, triaged",,"Review this module: optimizer queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#50282 Add MultiScheduler, #179143 [RFC] AdamTR: Adam variant for deterministic Token-Routed architectures",7c0f88932d53d48e074b50eb93f2a5b1788bb5689b8b2b3ca8959fdf4f9dccf3
issue-collapse:09687c31de3520e9,module: nestedtensor needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"172623, 187582",,,,"module: nestedtensor, module: typing","bot-triaged, module: nestedtensor, module: typing, triaged",,"Review this module: nestedtensor queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#172623 Nested Tensors should have their own type, #187582 torch.nested NJTs with identical layout fail on binary operation",7157966968777ab9541c0e27076f0b26dbbc78fb0c76fcaf2580f9f9cab0b01d
issue-collapse:fb67368d0eb1b836,module: mta needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"133367, 172273",,,,"module: mta, module: optimizer, module: performance","actionable, module: mta, module: optimizer, module: performance, triaged",,"Review this module: mta queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#133367 Adafactor foreach impl performance tracker, #172273 Numeric difference between div and foreach_div",981918381f2ef5b15f0284e70cd1166b88083ca7d05adef988de5cab1fcd1fc6
issue-collapse:99525eeb30435778,module: loss needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"25174, 134767",,,,"module: loss, module: nn","module: loss, module: nn, triaged",,"Review this module: loss queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#25174 BCEWithLogitsLoss expects wrong shape of weight (#classes instead of batch size), #134767 Setting a negative, `int` and `bool` value to `label_smoothing` argument of `nn.CrossEntropyLoss()` works against the error messages",a1e0a1fb20ef21a12348162c229f8a178974985c4a212d161faec49bf7d3b13d
issue-collapse:5d2ca95bd025ab90,module: functorch needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"151782, 175850",,,,"module: functorch, module: vmap","actionable, bot-triaged, function request, module: functorch, module: vmap, triaged",,"Review this module: functorch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#151782 JVP: Option to Disable Gradient Caching for Tangents, #175850 RFC: Vmap Extension Points for Custom Container Types",2a8db2924d6ce51b02cdf0db3e1bb6ec6372087530e69001022f80c477df0137
issue-collapse:908096fe62c536c4,module: functionalization needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"171375, 177372",,,,"module: functionalization, module: functorch, module: pt2-dispatcher, module: scatter & gather ops, oncall: pt2","bot-triaged, function request, has workaround, module: functionalization, module: functorch, module: pt2-dispatcher, module: scatter & gather ops, oncall: pt2, triaged",,"Review this module: functionalization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#171375 aten.lift triggers internal assert failure in Inductor functionalization pass, #177372 Add torch.narrow_scatter operator to complete the *_copy/*_scatter operator family.",0bac4a17f2f29f57c5bbf2cca35ff542b5251a3235fd6b1903865d75e10c6411
issue-collapse:14bb96b1b430d524,module: fft needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"171022, 188496",,,,"module: fft, module: onnx","bot-triaged, has workaround, module: fft, module: onnx, triaged",,"Review this module: fft queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#171022 In-place real-domain FFT operators for memory-efficient training (bf16 support), #188496 torch.fft.irfft2 ONNX export produces graph that onnxruntime cannot execute.",e936b8af789eb46e60050a85d251aff6e6242994229b966c3e949c89daa72d83
issue-collapse:d4c5f918f1fee4ed,module: dynamo needs maintainer decision  batch 3,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"170556, 173255",,,,"module: dynamo, module: performance, oncall: pt2","module: dynamo, module: performance, oncall: pt2, triaged",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#170556 [dynamo] Add linter rule to disallow `raise Unsupported` in torch/_dynamo, #173255 Torch.compile is decreasing training speed instead increasing it",5cc12fc02255e3653f4913543a78a941262d5a206b0268e614eaffec0dfdb516
issue-collapse:de0179a28d0ea9dc,module: dynamic shapes needs maintainer decision  batch 2,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"137093, 168279",,,,"module: dynamic shapes, module: inductor, oncall: pt2","module: dynamic shapes, module: inductor, oncall: pt2, triaged",,"Review this module: dynamic shapes queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#137093 Use SymInt instead of sympy.Expr uniformly in Inductor, #168279 Add flag to error on any Inductor specialization",95bd33ab959025bc8420a1b9c95e52fe44bfe4a8bf11c49c786794fff2082abe
issue-collapse:10935f1f5da9359a,module: dispatch needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"26207, 88109",,,,"module: dispatch, module: python dispatcher","enhancement, internals, module: dispatch, module: python dispatcher, triaged",,"Review this module: dispatch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#26207 Dispatch Tracing/Debugging, #88109 Python Dispatcher registrations beyond BackendSelect do nothing",e48e968ccb5230f2762d70c65c88ac61c53e92f020aa9def1a4df4e0eb9ad14f
issue-collapse:d6d770e14497b680,module: devicemesh needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"154787, 161776",,,,"module: devicemesh, module: dtensor, module: fsdp, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms","bot-triaged, module: DeviceMesh, module: dtensor, module: fsdp, oncall: distributed, oncall: distributed checkpointing, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,"Review this module: devicemesh queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#154787 Redistribute DTensor across different DeviceMeshes, #161776 FSDP&device_mesh Mesh_dim_names in init_device_mesh not work",295d1e765e1e9d8912bfe57245d3b2f18a3f62a7f82b72d60404600547dde84c
issue-collapse:d1d71eb383e68215,module: data needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"143471, 158714",,,,"module: data, module: dataloader, module: multithreading","module: data, module: dataloader, module: multithreading, triaged",,"Review this module: data queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#143471 NFS errors during DataLoader shutdown when num_workers > 1 when temporary directory is on NFS, #158714 Feature request: `DataLoader` using multithreading instead of multiprocessing",fe3adf7065df34b0e0d777b58b9be9abaa10161d867122eefa19a7433e2d841e
issue-collapse:40dda731a00b4f12,module: compile-time needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"108971, 181428",,,,"module: compile-time, module: ddp, module: dynamic shapes, module: functorch, module: inductor, oncall: distributed, oncall: pt2","bot-triaged, dynamo-functorch, module: compile-time, module: ddp, module: dynamic shapes, module: functorch, module: inductor, oncall: distributed, oncall: pt2, triaged",,"Review this module: compile-time queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#108971 Really slow compilation times for torch.compile causing distributed training errors, #181428 torch.compile + torch.func.grad: superlinear compile time with MaxPool2d → log/clamp → pow → AdaptiveAvgPool2d",cd8f0304b17a6208fa1fd28d60eefbfd96aad95a675b93b44e899fe37744dcfb
issue-collapse:5ae967308034b4fe,module: backend needs maintainer decision,batch action queue,needs maintainer decision,28,medium,bucket_component_queue,2,0,0,0,"178930, 187148",,,,"module: backend, module: compile ux, module: dispatch, module: dynamo, module: privateuse1, oncall: pt2","bot-triaged, feature, module: PrivateUse1, module: backend, module: compile ux, module: dispatch, module: dynamo, oncall: pt2, triaged",,"Review this module: backend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#178930 [torch.compile] Better way to override the default backend, #187148 [Proposal] Fix cpu_fallback returning tensors on wrong device for reverse binary ops",640f046a9cfd0466040b0a1e54ed61a2a0761cdd4437d8859240bc63c9732b85
issue-collapse:42d6f11dc8955035,module: docs has linked PR  batch 2,batch action queue,has linked PR,26,medium,bucket_component_queue,4,0,0,0,"67969, 71392, 99421, 129338",,,,"module: docs, module: nn, module: rnn","actionable, module: docs, module: nn, module: rnn, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#67969 [docs] F.multilabel_soft_margin_loss does not list arg reduction, but lists outdated size_average/reduce (true for other losses as well), #71392 Error in `torch.trapz` documentation, #99421 Question about GRU(RNN/LSTM) outputs shape, #129338 `linspace()` can also use `complex` and `bool` type for `start` and `end` argument against the doc",8ee1bfbe0b11e00f22cb3a755eefa01a4214b50c6ecd8a854a28ff30320ba01e
issue-collapse:d61f90119c75f6ed,oncall: pt2 has linked PR,batch action queue,has linked PR,26,medium,bucket_component_queue,2,0,0,0,"132929, 133078",,,,oncall: pt2,"oncall: pt2, triaged",,"Review this oncall: pt2 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#132929 Compile + torch.autograd.grad returns no gradients, #133078 `torch.compile` support for `CXX=""zig c++""`",3a9fba22a81d395e7a21e2c9a3bbf0cd3527d4ecd776461f153331e215af44a1
issue-collapse:a62662e10be6aa8e,oncall: export has linked PR,batch action queue,has linked PR,26,medium,bucket_component_queue,2,0,0,0,"134533, 147991",,,,"oncall: export, oncall: pt2","oncall: export, oncall: pt2, triaged",,"Review this oncall: export queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#134533 torch._assert does not work with torch.export, #147991 Verifier (in torch.export.export) does not make use of if-condition inside branches",e1f990d16a11d9f08697d9c6699e6a2c652c9a1db36c4905d34f220f6d5f549e
issue-collapse:52a16e5d8e32aad6,module: memory usage has linked PR,batch action queue,has linked PR,26,medium,bucket_component_queue,2,0,0,0,"105839, 167906",,,,"module: memory usage, module: mps, module: python frontend","better-engineering, module: memory usage, module: mps, module: python frontend, triaged",,"Review this module: memory usage queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#105839 MPS memory issue, MPS backend out of memory, but works if I empty the MPS cache, #167906 Avoid Exception Refcycle Problems",812b98c2a7eb742813fac60ad9244948ea28c9d413dc6d2cd3b9c106ec8a3535
issue-collapse:0c4c8aec90504078,module: optimizer needs design decision  batch 1,batch action queue,needs design decision,20,medium,bucket_component_queue,14,0,0,0,"44954, 80308, 83529, 84347, 87734, 93154, 94904, 103449, 108090, 108491, 127284, 127699, 147437, 160637",,,,"module: optimizer, module: performance, module: sparse","feature, module: optimizer, module: performance, module: sparse, needs design, needs research, triaged",,"Review this module: optimizer queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#44954 Optimization with constraints for torch.optim, #80308 CosineAnnealingWarmRestarts with initial warm up and weight decay applied on consecutive cycles without warm up, #83529 Adding Levenberg-marquardt optimizer in PyTorch, #84347 Lack of newly raised optimizers, #87734 Checkpointing Support for Modularized Optimizers, #93154 Support for VeLO optimizer.",033b21f92e605360c1f6dae3e2c7f7301a87718b51e32953d00e2484d2e09f5e
issue-collapse:cf8c43b014a813f4,module: nn needs design decision  batch 3,batch action queue,needs design decision,20,medium,bucket_component_queue,14,0,0,0,"67087, 83800, 84751, 98089, 102201, 102447, 102905, 105365, 108692, 109582, 110885, 115585, 115588, 117217",,,,"module: nn, module: performance","enhancement, feature, has workaround, module: nn, module: performance, needs research, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#67087 Feature Request: 1d grid sample, #83800 torch.var_mean is slower than layer norm, #84751 Allow passing dict (as opposed to OrderedDict) to nn.Sequential, #98089 GroupNorm cpu/gpu parity tests fail with pretty large differences, #102201 scipy.ndimage.find_objects, #102447 Add additional ""sigmoid"" approximation to GeLu activation?",28c1b5c3353a2523b087a822fdaa9e7f293ea1a953fee469279678b849684436
issue-collapse:44159d12b901a7e3,module: nn needs design decision  batch 2,batch action queue,needs design decision,20,medium,bucket_component_queue,14,0,0,0,"35458, 43101, 50566, 111538, 114462, 120072, 120700, 122551, 126762, 127932, 130480, 133466, 134325, 138952",,,,"module: nn, module: padding, module: rnn, module: serialization","enhancement, feature, has workaround, module: nn, module: padding, module: rnn, module: serialization, needs design, needs research, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35458 Shared-QK transformer for the transformer (nn.activation.MultiheadAttention) module in PyTorch?, #43101 New function for CTC decoding, #50566 Enhance supported fill value type for constant pad in functional.pad, #111538 Propose to add constant padding mode to the `torch.nn.functional.grid_sample` function, #114462 Transformer with convolutional network, #120072 Please consider adding more optimization support for multi-layer recurrent neural networks to improve fitting speed.",3996cebc99a0c64668c554d577e6362c395bb3d4ce2bf19643cd8922fbd5396c
issue-collapse:14c43dfb9a30c01c,module: autograd needs design decision  batch 2,batch action queue,needs design decision,20,medium,bucket_component_queue,14,0,0,0,"68879, 92330, 95916, 100628, 102337, 102417, 109948, 110780, 111563, 114650, 119475, 120107, 127996, 128793",,,,"module: autograd, module: bc-breaking, module: complex, module: docs, module: edge cases, module: functorch, module: nn, module: python frontend, module: typing, module: viewing and reshaping","Stale, enhancement, feature, has workaround, module: autograd, module: bc-breaking, module: complex, module: docs, module: edge cases, module: functorch, module: nn, module: python frontend, module: typing, module: viewing and reshaping, needs design, needs research, tensor subclass, topic: bc breaking, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68879 Stop gradient option for padding, #92330 Simplify module backward hooks to use multi-grad hooks instead, #95916 autograd.functional.jacobian : tensor instead of function as input for reverse mode?, #100628 Feature request: fast way to approximate the diagonal of the hessian, #102337 Tensors that share same underlying storage to also share gradient storage, #102417 _view_func but without keeping original view tensor alive",759c3ef1bb75e73a1ea4225dfc1839a54e110f88e4163993d08d02fc09d2a723
issue-collapse:3e7d87971c4da81a,module: nn needs design decision  batch 4,batch action queue,needs design decision,20,medium,bucket_component_queue,13,0,0,0,"8837, 14844, 37155, 48494, 52207, 68514, 68516, 69757, 73172, 80380, 83492, 89724, 91073",,,,"module: nn, module: norms and normalization, module: optimizer, module: ux","enhancement, feature, function request, module: nn, module: norms and normalization, module: optimizer, module: ux, needs design, needs research, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#8837 Inconsistency in implementation of _LRScheduler, #14844 Negative indexing for nn.Embedding inputs, #37155 Per-cluster biases in AdaptiveLogSoftmaxWithLoss, #48494 Inconsistent type of property stride of Conv1d and MaxPool1d, #52207 Wasserstein metric, #68514 Add common regularizations to PyTorch.",bd60675bb4ba964448bac99116243eaf07b53858bfedc6e4983de57d853cf447
issue-collapse:7962e5067a67748f,general needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,8,0,0,0,"42652, 45808, 57724, 108784, 121791, 128283, 128380, 170944",,,,,"has workaround, matrix multiplication, needs design, needs research, topic: performance, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#42652 Add non_blocking=True copy of torch.Storage, #45808 Support for Locality Sensitive Hashing Optimizations, #57724 Incorporate Posit Support, #108784 Collect more env variables in `collect_env.py`, #121791 [RFC] Set Float32 Precision for CONV/RNN, #128283 control output type in matmul",31a2ec780253a646b69e08776038dc0f4021bcab8ba0ab186182d4143a879fcd
issue-collapse:8f34c93e658b9c34,module: optimizer needs design decision  batch 2,batch action queue,needs design decision,20,medium,bucket_component_queue,7,0,0,0,"3904, 4660, 6657, 35801, 61343, 67587, 74904",,,,"module: optimizer, module: python frontend","enhancement, feature, function request, module: optimizer, module: python frontend, needs design, needs research, triaged",,"Review this module: optimizer queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#3904 Implement DE in pytorch.optim, #4660 [Feature proposal] Add MC-derived optimizers, #6657 [feature request] Stochastic Variance Reduced Gradient (SVRG) optimizer, #35801 Add KFAC optimizer, #61343 Feature Request: Remove Optimizer Lazy State Initialization, #67587 Re-implement `Optimizer.__repr__`",407e2ee76d03a289d44450f3cce205ef86b69e53ebe64867cd379b671079675b
issue-collapse:288036aa7a9231ca,module: linear algebra needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,5,0,0,0,"44591, 49890, 50204, 73143, 118225",,,,"module: linear algebra, module: performance","enhancement, feature, function request, module: linear algebra, module: performance, needs design, needs research, triaged",,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#44591 [Feature] Fused Matmul & Min/Max/Sum/Prod, #49890 GEMM with int8 datatype throws RuntimeError on GPU, #50204 Find the joint eigenvalue of two matrices., #73143 Add nan-safe einsum and bilinear, #118225 Support for banded matrix operations",6446a69020ad71305460a092597acdbee9fecf294c1cd4de74c5e376c15a6058
issue-collapse:9cc76a6fe02032c9,feature needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,5,0,0,0,"39780, 49338, 96555, 113730, 116354",,,,,"feature, needs design, needs research, triaged",,"Review this feature queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#39780 LRP-based explainability, #49338 a time series library, #96555 early stopping, #113730 Can the official provide cpu offload method for large model inference?, #116354 Abstract torch.device for GPU/NE/TPU computations in the cloud-based agent",9e4f3736ba44317953e88650a59d7e950d5f039c7396b79ae156aa84b508d355
issue-collapse:ba9d6ac52cbc39ac,module: serialization needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,4,0,0,0,"48747, 113530, 122099, 128934",,,,module: serialization,"enhancement, module: serialization, needs design, needs research, triaged",,"Review this module: serialization queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#48747 [bug][libtorch] torch::load seeks to the start of the input stream before reading, #113530 Make ""torch.load"" multi threaded process, #122099 No reduction of disk usage on slices of torch tensor, #128934 version inquiry",47dc9fe8f4d4415a1f5fd9923132ef142579dc8524cc837c3cc58921d1053fb8
issue-collapse:40b3042c9b94d001,module: docs needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,4,0,0,0,"41559, 62155, 68798, 150432",,,,"module: docs, module: nn, module: norms and normalization, oncall: releng","Stale, enhancement, module: docs, module: nn, module: norms and normalization, needs design, needs research, oncall: releng, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#41559 momentum in BatchNorm, #62155 [docs] Change code snippet formatting to be directly copy-pasteable into REPL or into source files, #68798 Clarify variables of BatchNorm*d functions, #150432 pytorch pip install instructions: always include the cuda index",3db0fcba2b6d642a2306788e3460a93a10ce0ce9996b8be403a6ec7d096e3d40
issue-collapse:dede4af12b711d42,module: ci needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,4,0,0,0,"83854, 142381, 160489, 165577",,,,"module: ci, module: tests","module: ci, module: tests, needs design, needs research, triaged",,"Review this module: ci queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#83854 Pytorch/Nova CI should monitor service outages for major dependencies, #142381 Add a linter for Docker images which are not included in docker-builds.yml, #160489 [CI] Mergebot tests are no longer updatable, #165577 CI: What is the purpose of `slow.yml`",aa9500728832c7eb7766347583ed0fde00fdf680c6cb402a304d57815ae6697f
issue-collapse:1a65a060c54314f5,module: internals good volunteer slice,batch action queue,good volunteer slice,20,medium,bucket_component_queue,3,0,0,0,"10756, 20704, 61654",,,,"module: internals, module: nn, module: partial aliasing","good first issue, module: internals, module: nn, module: partial aliasing, triaged",,Turn this module: internals slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#10756 Unexpected Behavior when Pointwise Operations Write to Expanded Tensors, #20704 Remove unpack() in torch/csrc/nn/type_checks.h and its caller functions in the codebase, #61654 Allow creation of pseudo devices for testing purposes",80d039c6b4569a78f85da190e197dd0e95fd8c557d980487e98ae37571193d35
issue-collapse:220fed2b8a3dc63c,module: functorch needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,3,0,0,0,"86641, 97248, 130451",,,,module: functorch,"actionable, module: functorch, needs design, triaged",,"Review this module: functorch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#86641 RFC(from users): nn.Module behavior with in-place changes, #97248 A more systematic API for resolving the ""vmap-incompatible in-place operation"" error, #130451 vmap over torch.allclose doesn't work",9115c1bfc6c4b4121c40e6b7c813dbaa6262a3d88c1213d604b5d62a7a5fe905
issue-collapse:5d45ecfad4d1610d,module: dataloader needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,3,0,0,0,"20822, 47037, 158704",,,,module: dataloader,"enhancement, feature, module: dataloader, needs research, triaged",,"Review this module: dataloader queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#20822 [Proposal] Data reading framework for PyTorch (Hive, MySQL, S3 etc.), #47037 Change ""_next_index()"" in DataLoader to a public and stable API, #158704 Enable DataLoaders to prefetch batches across epochs",9f8b7a3cd339065f720b62eef7d091dddf6dc8fa9debfa9132b5ca5068cb392f
issue-collapse:cfc8c948541abf36,module: cpu needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,3,0,0,0,"41556, 125223, 132446",,,,"module: cpu, module: distributions, module: nn, module: performance","enhancement, feature, module: cpu, module: distributions, module: nn, module: performance, needs design, needs research, triaged",,"Review this module: cpu queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#41556 Implement LSH Optimizations for Enhanced CPU-Only Performance, #125223 torch.uniform_() is single-threaded on CPU, #132446 [Feature Request] Support Custom CPU Allocator to Enable torch.cpu.change_current_allocator",206147dd33726b18b63d69b3fb9e533f3ec57395c20ce72ac4f4934dd8ac4ce0
issue-collapse:cc44fca0aa993ca5,module: complex needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,3,0,0,0,"67634, 95337, 138906",,,,"module: complex, module: cuda, module: interpolation, module: nn","feature, module: complex, module: cuda, module: interpolation, module: nn, needs design, needs research, triaged",,"Review this module: complex queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#67634 Feature request: Grid sample on complex tensors with float grid, #95337 Changing behavior of module.to() to better support mixed real- and complex-valued parameters, #138906 RuntimeError: ""addmm_cuda"" not implemented for 'ComplexHalf'",ccacebabff03afd32b80bc5926ae2c232fe35ecb109671fe1c3bc5c4470f67ee
issue-collapse:2f0ea2a87fc409d7,module: memory usage needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"43450, 52036",,,,"module: memory usage, module: nn","feature, module: memory usage, module: nn, needs research, triaged",,"Review this module: memory usage queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#43450 Training Large Neural Networks with Constant Memory using a New Execution Algorithm, #52036 Automatic splitting of batch during model feed-forwarding to avoid memory errors caused by the too big batch_size",827de8f54736fbbb9c5e234803d2a8cd33c1ca4c5f9882d051247cc54b08c3b3
issue-collapse:e9eb32f8b4c17a93,module: distributions needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"45133, 102938",,,,module: distributions,"feature, module: distributions, needs research, triaged",,"Review this module: distributions queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#45133 [Feature Request] Add support for Hidden Markov Models in torch.distributions, #102938 Support for efficiently processing categorical distributions with varying dimensions",85c9f958a6ec77c5e8044ee095c11ce871ba3450a240198eba451145f90998b4
issue-collapse:60d78cd5bc697b68,module: devx needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"115273, 118133",,,,module: devx,"module: devx, needs design, needs research, triaged",,"Review this module: devx queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#115273 Consider improving project management (12 pages of HighPriority tickets > 1 year old), #118133 Better test disabling system",4d92190e345758ceb8bf05769bff7e295b8fe1194ff88ff35afd4d4d124feb37
issue-collapse:b4a8368cdda1e517,module: cpp needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"100990, 107224",,,,"module: cpp, module: optimizer","feature, module: cpp, module: optimizer, needs design, needs research, triaged",,"Review this module: cpp queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#100990 Extending compatibility of LibTorch, #107224 [CPP API] Add Adadelta, Adamax, ASGD, NAdam, RAdam and Rprop",8508c820727b8f6e535e128b0eb2fcf1398fb9ec4b120400daf22f36830cf184
issue-collapse:be2a52e931599b40,module: c10d good volunteer slice,batch action queue,good volunteer slice,20,medium,bucket_component_queue,2,0,0,0,"84593, 97469",,,,module: c10d,"good first issue, module: c10d, triaged",,Turn this module: c10d slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#84593 Uneven and/or Dynamically sized collectives, #97469 Improve collectives fingerprinting",f0ca0f5d3900fa01095f21f58faf99cab8f2c0dd4ed28fe839e6702e05d0f918
issue-collapse:dba21f0e555a1cd1,module: bootcamp good volunteer slice,batch action queue,good volunteer slice,20,medium,bucket_component_queue,2,0,0,0,"58122, 79452",,,,"module: bootcamp, oncall: distributed","OSS contribution wanted, better-engineering, good first issue, module: bootcamp, oncall: distributed, pt_distributed_rampup, triaged",,Turn this module: bootcamp slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#58122 Generate special clear error messages for known common misuses in TorchScript, #79452 Use c10d broadcast_object in Zero",da82810382919f4a4f3ea8dada9df679bb9d6fbdccf4ccd2142b8265458433ae
issue-collapse:5a269076a041ca1a,module: autograd needs design decision  batch 3,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"68699, 79877",,,,"module: autograd, module: sparse, module: viewing and reshaping","Stale, module: autograd, module: sparse, module: viewing and reshaping, needs design, needs research, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#68699 Transpose of a sparse tensor is not a view operation, #79877 test_meta_vstack_cuda_int16 (__main__.TestMetaCUDA) Fails with DEBUG=1",8e1bd606fa96992cd92dceb4b177f27eef715a4bad4eb10577ce2f8f4a04ca7d
issue-collapse:77a2f4f5063d2005,module: arm needs design decision,batch action queue,needs design decision,20,medium,bucket_component_queue,2,0,0,0,"125767, 154260",,,,"module: arm, module: build, module: windows, oncall: releng","enhancement, module: arm, module: build, module: windows, needs design, needs research, oncall: releng, triaged",,"Review this module: arm queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#125767 RFC: Integrate Arm Compute Library (ACL) into PyTorch as a submodule and add build support in setup.py, #154260 Pypi Support for Windows arm64",96dc0acdd3e56240ff036804382b58bb001199dfd0959b5d2537dcce7228b9d3
issue-collapse:5d5ae32158bc1f55,module: cuda needs reproduction  batch 4,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"24398, 35754, 49832, 54487, 54799, 54975, 56771, 57794, 58135, 59883, 64560, 64889, 65027, 81257",,,,"module: cuda, module: cudnn, module: memory usage, module: numpy, module: performance, module: tensoriterator","module: TensorIterator, module: cuda, module: cudnn, module: memory usage, module: numpy, module: performance, needs reproduction, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24398 Significantly slower in latest version than in 0.4.0, #35754 pytorch 1.4.0 hangs when using with CUDA >= 10.1, #49832 RuntimeError: CUDA error: unspecified launch failure, #54487 RuntimeError: iter.device(arg).is_cuda() INTERNAL ASSERT FAILED at ""/opt/conda/conda-bld/pytorch_1607370116979/work/aten/src/ATen/native/cuda/Loops.cuh"":94, please report a bug to PyTorch., #54799 Vertices=torch.matmul(vertices.unsqueeze(0), rotations_init), RuntimeError: CUDA error: CUBLAS_STATUS_EXECUTION_FAILED when calling `cublasSgemmStridedBatched in CentOS, #54975 CUBLAS_STATUS_EXECUTION_FAILED error on torch >= 1.8.0 and CUDA 11.1",8ae1d42d9d0f11ddf28021f83921640080ef2dc36e145a1e831c064f300c5551
issue-collapse:ab5d6c91444f5f45,module: cuda needs reproduction  batch 3,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"25190, 63847, 70048, 73568, 86547, 90521, 91072, 97847, 109193, 110701, 112569, 113035, 114534, 125911",,,,"module: cuda, module: cudnn, module: dataloader, module: docker, module: error checking, module: linear algebra, module: multiprocessing, module: nn, module: performance","Stale, actionable, module: cuda, module: cudnn, module: dataloader, module: docker, module: error checking, module: linear algebra, module: multiprocessing, module: nn, module: performance, needs reproduction, quansight-nack, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25190 Error in python3: double free or corruption (fasttop), #63847 RuntimError:CUDA error: an illegal memory access was encountered when copying data to CUDA, #70048 Tensor transfer between gpus doesnt work, #73568 Bug when using `nn.Linear`, #86547 Segmentation fault (core dumped) in RTX3090, #90521 [A ERROR in Docker] RuntimeError: CUDA error: no kernel image is available for execution on the device",ae259e86c93fc003317a89feca01bd36cbdc7d00f8e834bc924bc87871b3b07a
issue-collapse:23d01d80c5c64e38,module: cuda needs reproduction  batch 2,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"90752, 121067, 126775, 128131, 130486, 130728, 138208, 139884, 140096, 141465, 143905, 149279, 152135, 159309",,,,"module: cuda, module: cudnn, module: half, module: memory format, module: memory usage, module: nans and infs, module: nn, module: numerical-reproducibility, module: numerical-stability, module: optimizer, module: performance, module: scatter & gather ops, module: windows","module: NaNs and Infs, module: cuda, module: cudnn, module: half, module: memory format, module: memory usage, module: nn, module: numerical-reproducibility, module: numerical-stability, module: optimizer, module: performance, module: scatter & gather ops, module: windows, needs reproduction, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90752 Adam (fused=True) issues, #121067 RuntimeError, #126775 The significant difference between the outputs of torch.reciprocal running on cpu and gpu, #128131 Unable to record Memory consumption with `torch.cuda.memory._record_memory_history()`, #130486 RuntimeError: NVML_SUCCESS == DriverAPI::get()->nvmlInit_v2_() INTERNAL ASSERT FAILED at ""../c10/cuda/CUDACachingAllocator.cpp"":813, please report a bug to PyTorch., #130728 torch.cuda.empty_cache() doesn't work correctly for the first instance of nn.Module set to cuda (and a potential method to deal with that).",740df31920f024feb5e8fa8c5e3477208fcbfa4760c5916ca31d2c3f7e6a54e2
issue-collapse:bb8ef1c57a981e77,module: build needs reproduction  batch 2,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"12641, 56019, 61341, 70038, 74670, 75963, 84370, 86877, 87597, 89041, 101321, 101404, 108013, 110588",,,,"module: build, module: cpp, module: cuda, module: jetson, module: mkldnn, module: performance, module: rocm, module: third_party","module: build, module: cpp, module: cuda, module: jetson, module: mkldnn, module: performance, module: rocm, module: third_party, needs reproduction, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12641 Install Jetson TX2 Max Regcount Error, #56019 when building apk with libtorch, #61341 inerror: reference to ‘DeviceType’ is ambiguous, #70038 comile error, #74670 Can't install Pytorch, #75963 Add build support for GCC 11.2",73af89546e21efaff04c558ce96bc80f93328b6eb705b3d92ac21b957d4fff7b
issue-collapse:294a127f57293e48,module: binaries needs reproduction  batch 2,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"85211, 104598, 111896, 116931, 117469, 118172, 120606, 121222, 121287, 121646, 127560, 128939, 134993, 138821",,,,"module: binaries, module: cpp, module: cuda, module: docker, module: docs, module: macos","actionable, module: binaries, module: cpp, module: cuda, module: docker, module: docs, module: macos, needs reproduction, security, topic: binaries, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#85211 OSError libstdc++.so.6 at import, #104598 ERROR: THESE PACKAGES DO NOT MATCH THE HASHES FROM THE REQUIREMENTS FILE., #111896 ImportError: cannot import name 'external_utils' from partially initialized module 'torch._dynamo', #116931 Previous Torch versions, #117469 Cannot import torch on Ubuntu 20.04 from clean pip install (shared library issues, libcufft and libcudart), #118172 ""PyTorch installation failed.""",33acd28325946ad8194e2601fd8e284e6021a7f9aa1e52c70f6671025be202b9
issue-collapse:302249e356dae5cc,general needs reproduction  batch 3,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"32584, 42581, 44024, 46153, 49870, 51275, 53464, 54827, 60936, 62152, 62644, 67835, 68864, 69448",,,,,"needs reproduction, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32584 Segmentation Fault (core dumped) with 1.4.0, #42581 Build script complains that gcc is < 6 even though gcc version is 8, #44024 When the pytorch training data reaches a certain epoch, the memory remains unchanged?, #46153 Getting this error while installing torch and torchvision!, #49870 Tensorboard Error, #51275 Error of running",a08e0c2967000d543853919e64e2488dc6269b25f0e72e1338bd634e19de8e02
issue-collapse:ff81b231b5b41e56,general needs reproduction  batch 2,batch action queue,needs reproduction,14,medium,bucket_component_queue,14,0,0,0,"72105, 75265, 76954, 76955, 76956, 76959, 76960, 90138, 91970, 92794, 93900, 95677, 98142, 101880",,,,,"low priority, needs reproduction, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72105 ImportError: cannot import name '_VF' from partially initialized module 'torch', #75265 Kernel Dies while using conv2d layer / function, #76954 ReduceSum causes the inconsistency of output between frameworks, #76955 GaussianNoise causes the inconsistency of output between frameworks, #76956 GlobalAvgPool2d causes the inconsistency of output between frameworks, #76959 When using Lambda, the output of the 1/x process is very likely to have nan/inf",2dc864776f283f2d6060de3966a48820e11af85c12ac99740eaa59790b4589c1
issue-collapse:45c848b2935d652f,module: autograd needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,12,0,0,0,"22597, 24397, 57796, 60923, 70702, 76798, 82217, 82218, 97135, 116589, 126141, 143259",,,,"module: autograd, module: convolution, module: determinism, module: interpolation, module: memory usage, module: nn, module: numerical-stability, module: serialization, module: sorting and selection, module: third_party, oncall: distributed","module: autograd, module: convolution, module: determinism, module: interpolation, module: memory usage, module: nn, module: numerical-stability, module: serialization, module: sorting and selection, module: third_party, needs reproduction, oncall: distributed, triaged",,"Review this module: autograd queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#22597 Autograd profiler memory leak when use_cuda=True, #24397 SyncBatchNorm error when using model.eval() with DistributedDataParallel, #57796 BatchNorm grad calculation is imprecise, #60923 RuntimeError: derivative for im2col_backward is not implemented, #70702 conv3d padding=`same` gradgradcheck fails on CUDA, #76798 `gradcheck` for `torch.solve` may trigger INTERNAL ASSERT FAILED",2b30f197526ed478a90ab1fe1ddb369b0d388fc868ea9854ea3d8785651e1c16
issue-collapse:f99f77cca21af399,module: nn needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,10,0,0,0,"64804, 76024, 80302, 104857, 108676, 109453, 139977, 142302, 143852, 165590",,,,"module: nn, module: numerical-stability, module: optimizer, module: performance, oncall: quantization","module: nn, module: numerical-stability, module: optimizer, module: performance, needs reproduction, oncall: quantization, triaged",,"Review this module: nn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#64804 RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation, #76024 The prediction results of different equipment are inconsistent, #80302 AttributeError: 'LinearPackedParams' object has no attribute '_modules', #104857 Torch's `LayerNorm` and Adam optimizer vs those in tensorflow, #108676 RuntimeError when calling conv_transpose2d with groups, #109453 Inconsistent Behavior of `ConvTranspose2d` on CPU and CUDA",ed9b964e27cba40e59a071208a0662d9789c1d30d18bf2e8f40779253457830b
issue-collapse:65624a3ebf5caec3,module: cuda needs reproduction  batch 5,batch action queue,needs reproduction,14,medium,bucket_component_queue,10,0,0,0,"4334, 26400, 28956, 30830, 31660, 32443, 35870, 37057, 49669, 52307",,,,"module: cuda, module: cudnn, module: deadlock, module: multi-gpu, module: multiprocessing, module: third_party","module: cuda, module: cudnn, module: deadlock, module: multi-gpu, module: multiprocessing, module: third_party, needs reproduction, triaged",,"Review this module: cuda queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#4334 torch.cuda.device_count() returns 1 using 4 TitanX setup., #26400 RuntimeError: tensor.ndimension() == static_cast<int64_t>(expected_size.size()) INTERNAL ASSERT FAILED, #28956 RuntimeError: CUDA error: unknown error, #30830 Torch getting stuck transfering model to GPU in multiple GPU setting, #31660 Cuda error 59 : device-side assert triggered, #32443 Segmentation fault in lazyInitCUDA -> CUDAHooks::initCUDA -> THCMagma_init -> magma_init",dc539652b291a46a4683991f2342ae7aebb1709e4fe68a234521435bfdd09aa4
issue-collapse:632c69a20c5e56c0,module: binaries needs reproduction  batch 3,batch action queue,needs reproduction,14,medium,bucket_component_queue,8,0,0,0,"31004, 56928, 60094, 86919, 90115, 97865, 101331, 106126",,,,"module: binaries, module: cpp, module: cuda, module: macos, module: openmp, module: performance","module: binaries, module: cpp, module: cuda, module: macos, module: openmp, module: performance, needs reproduction, triaged",,"Review this module: binaries queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#31004 The inference speed of the torch compiled manually is slower than the torch build from official binaries?, #56928 torch-1.8.1 wheel verification fails with distlib, #60094 bundled libiomp5 causing segfaults in other libraries that use libomp, #86919 Importing torch 1.12.0 breaks subprocess module, #90115 Couldn't install pytorch 2.0, #97865 Unable to install torch on python 3.8.16",254afff284b227e5ffc3419fa50f5cca3d45e7af85f15504daeeb726127630f4
issue-collapse:7eed98d331af5763,module: multiprocessing needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,7,0,0,0,"2482, 48579, 50261, 110594, 113840, 119546, 150921",,,,"module: multiprocessing, module: serialization, module: tests","module: multiprocessing, module: serialization, module: tests, needs reproduction, triaged",,"Review this module: multiprocessing queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#2482 ""Shared memory manager connection has timed out"", #48579 test_fs_pool fails, #50261 ModuleNotFoundError: No module named '__torch__', #110594 Multiprocessing takes forever after on .get() with mp.Queue() (Possible Deadlock), #113840 torch.einsum is stuck in mp.Process, #119546 Too many spawn process",33a34b6ce8fbe017497baea8aaf4e33e9bc6732e1ba90f498dfb0a9e14a11edc
issue-collapse:d801ebc52f5b4ac2,module: memory usage needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,6,0,0,0,"28943, 40133, 120167, 126388, 137393, 139709",,,,"module: memory usage, module: nn, module: serialization","module: memory usage, module: nn, module: serialization, needs reproduction, triaged",,"Review this module: memory usage queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#28943 The `@` operation uses too much memory on GPU, #40133 Why does Libtorch use more memory than Pytorch does in Python？, #120167 When loading a text file with torch.load, it exhausts the system's physical memory., #126388 RAM Not Freed on CPU After Moving Model with Multiple Transformers to CUDA, #137393 Cannot allocate memory for thread-local data: ABORT, #139709 less GPU memory used in Centos than ubuntu",bd7bcce082c69b128716998e54b6f74e85e02487271653185cdaa64076bbdeaa
issue-collapse:6aca7f2996b747e8,module: build needs reproduction  batch 3,batch action queue,needs reproduction,14,medium,bucket_component_queue,6,0,0,0,"14352, 19128, 20134, 21179, 36721, 46806",,,,"module: build, module: cuda","has workaround, module: build, module: cuda, needs reproduction, triaged",,"Review this module: build queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#14352 [libtorch] Catkin_make compilation error, #19128 pytorch.version.cuda is None when compiling with CUDA support, #20134 ERROR: Command ""python setup.py egg_info"" when dockerfile build, #21179 Failed to install pytorch from source on ubuntu., #36721 [1.4.1] Cuda build fails, #46806 Compile Pytorch with MAGMA Issue",ac8c1ab0ae40cb9ac74db992fa690c7dbb4a0f0d2a84152f46bf45340dcd979e
issue-collapse:f82bbc31756ab404,module: linear algebra needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,5,0,0,0,"69203, 70008, 88948, 140282, 157360",,,,"module: linear algebra, module: memory usage, module: performance","module: linear algebra, module: memory usage, module: performance, needs reproduction, triaged",,"Review this module: linear algebra queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#69203 Error in SVD cusolver on Linux, #70008 Torch function runtime seemingly dependent on scipy call, #88948 torch.linalg.matrix_rank memory leak, #140282 torch.norm produces wrong results, #157360 torch.Tensor.addmm_ The calculation result is inconsistent with the formula calculation result",b10142745f59581f5ae63e5fc7a5061944cadaa879c5db8ca8a037cbabb10c7f
issue-collapse:a975945c8eeb7b03,module: data parallel needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,5,0,0,0,"36649, 44855, 69966, 107133, 141482",,,,"module: data parallel, module: edge cases, module: linear algebra","Stale, module: data parallel, module: edge cases, module: linear algebra, needs reproduction, triaged",,"Review this module: data parallel queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#36649 RuntimeError: Expected cuda::check_device({sparse_, r_, t, dense}) to be true, but got false., #44855 RuntimeError: Caught RuntimeError in replica 1 on device 1., #69966 torch.nn.DataParallel caused inference failure with cpu set as device on NV machine, #107133 torch.inverse throws error when DP but not in DDP or single GPU, #141482 Can't use dataparallel",af313d06d733a8333dbea4c5139379574441936cbfe92c5505c8f76a84ac5c73
issue-collapse:058bdcee5e62ea84,oncall: jit needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,4,0,0,0,"20805, 34937, 34948, 35990",,,,oncall: jit,"needs reproduction, oncall: jit, triaged",,"Review this oncall: jit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#20805 LibTorch :About torch.jit.trace generate model.pt, #34937 [JIT]torch.jit.export invaild for pytorch1.4, #34948 TracerWarning: Output nr 1. of the traced function does not match the corresponding output of the Python function, #35990 [JIT] Tensor method API behavior discrepancy, Tensor.detach(..)",984dc681ef8e354d5002fb7607165783d36c515886df1e1db012ada3e967af1e
issue-collapse:906de8cd151cc802,module: nans and infs needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,4,0,0,0,"63322, 80588, 98204, 126760",,,,"module: nans and infs, module: numpy, module: rnn","module: NaNs and Infs, module: numpy, module: rnn, needs reproduction, triaged",,"Review this module: nans and infs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#63322 lstm's input h0 and c0 bug, #80588 Semi-reproducible random torch.baddbmm NaNs, #98204 NaN appears when initializing tensor, #126760 ""torch.orgqr"" performs differently on cpu and gpu",35249d802f106b035f7230f15dc97444964fd627e916c14a78d69b118205feec
issue-collapse:0e12abe04f4b2b9b,module: mps needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,4,0,0,0,"111547, 117210, 132226, 146769",,,,module: mps,"module: mps, needs reproduction, triaged",,"Review this module: mps queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#111547 Bug with as_strided_tensorimpl for MPS devices, #117210 When running the code model.to('mps'), the program gets stuck, but it runs fine on the CPU, #132226 关于yolov5在mac设备上使用pytorch的mps加速出现的各种问题, #146769 MPS Error on sequoia 15.3: NDArray dimension length > INT_MAX'",a2637498e2ca93dbeb895a87cff785d562e4d03dc8d330c7d71fb18c3430b2ac
issue-collapse:4564871178690c86,module: docs needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,4,0,0,0,"39832, 103423, 103424, 166303",,,,"module: docs, module: rpc","module: docs, module: rpc, needs reproduction, triaged",,"Review this module: docs queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#39832 Torch Distributed Asynch docs with invalid ENV args, #103423 The document does not emphasize hidden range in nn.MaxPool2d, #103424 The document does not emphasize hidden range in nn.Embedding, #166303 Pytorch Operators on older pytorch version",152dab3cf334bd4ba23d8bc684376f0edac5651b7c4ddb1ec421c51f5173abf3
issue-collapse:22f30c522c6de965,module: bfloat16 needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,4,0,0,0,"75458, 121190, 153033, 163347",,,,"module: bfloat16, module: correctness (silent), module: deadlock, module: intel, module: linear algebra, module: mkldnn, module: numerical-stability, module: padding","module: bfloat16, module: correctness (silent), module: deadlock, module: intel, module: linear algebra, module: mkldnn, module: numerical-stability, module: padding, needs reproduction, triaged",,"Review this module: bfloat16 queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#75458 Placing model on bfloat16 on CPU make it freeze/hang, #121190 The result of BF16 to FP32 is different in class LlamaRMSNorm training twice., #153033 missalignment with differenet shape in F.linear with bf16 dtype, #163347 Incorrect bf16 results in Conv",ac2085e3d4c987d31dccc94f96f14106e43058884ea99065886ff389319c8124
issue-collapse:34c57430bbb1a7fc,module: dynamo needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,3,0,0,0,"105700, 116766, 150912",,,,"module: dynamo, oncall: distributed, oncall: pt2","dynamo-triage-june2024, dynamo-variable-tracker, enhancement, module: dynamo, needs reproduction, oncall: distributed, oncall: pt2, triaged",,"Review this module: dynamo queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#105700 Don't use weak ref finalization for freeing resources when code objects die, #116766 [Dynamo][DeepSpeed] torch._dynamo.exc.InternalTorchDynamoError: NestedUserFunctionVariable() has no type, #150912 using torch.compile with torchao at the same time cause stack overflow error",cb4a14497e3d4cf920c201b271955d9e815c0c674a4a7a758e503202d2cb94dc
issue-collapse:886c4021528489be,module: determinism needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,3,0,0,0,"57491, 71222, 94855",,,,"module: determinism, module: inductor, module: linear algebra, module: random, oncall: pt2","module: determinism, module: inductor, module: linear algebra, module: random, needs reproduction, oncall: pt2, triaged",,"Review this module: determinism queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#57491 setting all seed still get different result., #71222 torch.linalg.lstsq is nondeterministic, #94855 torch.compile breaks reproducibility",92f39bc8edde876128c6efea982d6afe64340542f613cedf10f13db2bc72f1ca
issue-collapse:4c701b19e3ac7050,module: amp (automated mixed precision) needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,3,0,0,0,"45273, 47128, 58392",,,,"module: amp (automated mixed precision), module: data parallel, module: performance","module: amp (automated mixed precision), module: data parallel, module: performance, needs reproduction, triaged",,"Review this module: amp (automated mixed precision) queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#45273 Training fast with small dataset, slow with large dataset, #47128 AMP doesn't gracefully handle optimizers for disabled regions, #58392 torch.cuda.amp.autocast nan values when model.eval()",e99469c131b165848c77161fa8a78c9cd9403fe002165d56b44253b1b05e8899
issue-collapse:89cd871bd897f77f,general needs reproduction  batch 4,batch action queue,needs reproduction,14,medium,bucket_component_queue,3,0,0,0,"32662, 37067, 38155",,,,,"needs reproduction, triaged",,"Review this uncategorized queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#32662 Second-order gradient cause segfault over time, #37067 can't wrap two models in the same class, #38155 Python script name with ""profile.py"" will run twice",67caee65829b41afc823995bfe9f669d26af768db256e117fdc8a34cc5e91792
issue-collapse:0fb784c0c03479b3,module: python frontend needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"120791, 122548",,,,module: python frontend,"module: python frontend, needs reproduction, triaged",,"Review this module: python frontend queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#120791 about tensor.new(storage) issue, #122548 Unexpected device casting from CPU-pinned tensor to GPU tensor.",ef9f77befd971ad08c8c6983dffa48122f4de59d45572999dbfb716c91d8dce9
issue-collapse:74733775705da4c4,module: lrscheduler needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"66781, 71272",,,,"module: lrscheduler, module: optimizer","module: LrScheduler, module: optimizer, needs reproduction, triaged",,"Review this module: lrscheduler queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#66781 lr_scheduler.py / list index out of range, #71272 UserWarning: Seems like `optimizer.step()` has been overridden after learning rate scheduler initialization. Please, make sure to call `optimizer.step()` before `lr_scheduler.step()`. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate warnings.warn(""Seems like `optimizer.step()` has been overridden after learning rate scheduler",025b19ec9f75e337c484551233ce4a22a9449f10452d0a4393ee2eaa4c7f36c7
issue-collapse:e6e715a466ab2ddb,module: internals needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"30955, 129992",,,,module: internals,"module: internals, needs reproduction, triaged",,"Review this module: internals queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#30955 error C3203: “templated_iterator”: 未专用化的 类 模板 不能用作 模板 变量，该变量属于 模板 参数“_Ty1”，应为 real 类型, #129992 Got this error when just running my flask script",568f1255e12138dddb1bd8c0e953be96bfc5765d5403d28b3498ae7b7592ed3a
issue-collapse:eecfd71c4c6123b0,module: functorch needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"91942, 99573",,,,module: functorch,"module: functorch, needs reproduction, triaged",,"Review this module: functorch queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#91942 jacrev over huber function, #99573 Jacfwd become slower after update pytorch (""We’ve integrated functorch into PyTorch---Documentation"")",9dfc1078bb4f5b61dd8d8ea4cd33741b4280107024e40b6939d96b2a06f71e6d
issue-collapse:05be82c19f987b6f,module: cudnn needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"105941, 170524",,,,"module: cudnn, module: third_party","module: cudnn, module: third_party, needs reproduction, triaged",,"Review this module: cudnn queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#105941 Differences in the results of conv2d calculations in PyTorch 1.8, #170524 RuntimeError occurred when I train “Resnet”",51660a7b03541c4ba8dbe623b4350a96a2114a9a10381cc6579a78c746947587
issue-collapse:e4b8fa957d5e3cdc,module: 64-bit needs reproduction,batch action queue,needs reproduction,14,medium,bucket_component_queue,2,0,0,0,"119147, 131185",,,,"module: 64-bit, module: autograd, module: edge cases, module: error checking, module: nn","module: 64-bit, module: autograd, module: edge cases, module: error checking, module: nn, needs reproduction, triaged",,"Review this module: 64-bit queue together so the maintainer decides category, owner, and response pattern once.",Batch triage by component/action queue.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#119147 Incorrect Results from cdist_backward Operator with Input Dimensions Exceeding int32_max, #131185 RuntimeError: Expected grad_output.numel() <= std::numeric_limits<int32_t>::max() to be true, but got false.",6690e0dde3e860aae9264c6104e3c49abf21f18d7999454ca0fa6b64a14f2a94
issue-collapse:422ff1de50eefa44,oncall: releng stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"58533, 83257, 92310, 98006, 101798, 107611, 109958, 124974, 136691, 137370, 143206, 148984, 149660, 156085",,,,oncall: releng,"better-engineering, feature, oncall: releng, security, triaged",,Handle this oncall: releng stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#58533 Migrate wheels to manylinux2010, #83257 PyYAML not listed as a dependency, #92310 [Releng] Windows AMI needs to be pinned for release, #98006 Create a plan on removing conda dependency from CI/CD, #101798 Fix absolute links in pytorch repository and allow it to be proxied, #107611 Previous version not found",70c12fcf6299416c65032f082428300c12cabc3462bf6c0c1aa75fcb8d6bd23a
issue-collapse:04de42f389d646a8,oncall: quantization stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40759, 79709, 79739, 81287, 81381, 81667, 83521, 83672, 83941, 84052, 84148, 90289, 90394, 100051",,,,oncall: quantization,"low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40759 Quantization support for F.Softmax, #79709 ONEDNN testing is not done properly in quantization codebase, #79739 quantization: misleading backend config for linear_dynamic_fp16, #81287 Runtime error in Libtorch cpp project (Didn't find engine for operation quantized::conv2d_prepack NoQEngine), #81381 quantization: QConfigMapping should be easy to print, #81667 [Tracker] AO migration of quantization from `torch.nn` to `torch.ao.nn`",b82d40c51df4ddae5fc8a479c23e30a56612ca82b4461945d073756d3434e0f4
issue-collapse:5549b50f6f6fa2c0,oncall: quantization stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32262, 36508, 38569, 39058, 47363, 55945, 56978, 60008, 71721, 73738, 73941, 75440, 94371, 112833",,,,oncall: quantization,"better-engineering, low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32262 models after prepare/prepare_qat contains `qconfig` which prevents pickling, #36508 Quantized _out functions don't follow same conventions as other out functions in the codebase, #38569 [quantization] Version support for quantization BC tests, #39058 Failure when loading quantized pre-trained weights partially, #47363 FX quantization: we should preserve original model class name through the quantization passes, #55945 Quantizable LSTMCell does not work correctly.",58b16ca55a899bc8643359c229cb12855886ddc41b6b434a7f3caee5bb2d71ca
issue-collapse:df08607751a2b721,oncall: quantization stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"41149, 44676, 45635, 46604, 47535, 47857, 48106, 51152, 60210, 61046, 69158, 70480, 71889, 73507",,,,oncall: quantization,"Stale, enhancement, low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41149 eager mode quantization should remove qconfig in the (non-leaf module of) quantized model, #44676 Make quantized::prepack_fp16 op just do prepacking, #45635 fx graph mode quantization tutorials, #46604 torch.concat doesn't raise an error in a quantized model, #47535 per channel observer to work for weights of groupwise conv transpose, #47857 QAT with DDP should have documentation",2d1006087e1e2c4f54ec88856eec947643784e2439dade4562d5c6a1a9c40577
issue-collapse:946b417a0ca22c28,oncall: quantization stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"74953, 75785, 76304, 76755, 83929, 84064, 84178, 85745, 85851, 86848, 87126, 88505, 88579, 88800",,,,oncall: quantization,"low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#74953 LSTM quantization fails if proj_size > 0, #75785 [ONNX] Expected quantizer->qscheme() == kPerTensorAffine to be true, but got false., #76304 Quantization fails when padding parameter given as string, #76755 Quantization in Libtorch, #83929 ModuleNotFoundError: No module named 'torch.ao.quantization.experimental', #84064 Better error message for qlinear_prepack",eb3fd24ca96b06f6ea86ad4b37272f6a6679f5c0c16264dad5abf38875daa309
issue-collapse:d37d3289147e7604,oncall: quantization stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31316, 42594, 58969, 78722, 89303, 90288, 90553, 91263, 91604, 106780, 108399, 110515, 114928, 117508",,,,oncall: quantization,"low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31316 quantization - Missing operations needed for object detection, #42594 support LSTM for quantization aware training, #58969 Improper use of quantization API for MHA should fail fast, #78722 AlBert quantization, #89303 code sharing for fundamental ops in quantization, #90288 quantization observers: can we relax the default epsilon value?",b1c405d25d52b607807476f49d0c97ced45cfdd277368846e9f321c7631af0e2
issue-collapse:fb94dad41d611d18,oncall: quantization stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"54816, 67651, 68892, 71028, 74540, 74627, 82053, 95785, 118634, 118829, 127391, 144403, 157591, 170402",,,,oncall: quantization,"feature, low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54816 Per channel weight observer for ConvTranspose, #67651 MultiHeadAttention in quantizable seems incorrect with batch_first=True, #68892 fusion in fx graph mode did not take care of direct attribute access, #71028 Skip LSTM quantization by default in get_default_qconfig_dict and get_default_qat_qconfig_dict, #74540 No factory functions for strided quantized tensors, #74627 Feature request: INT4 format support",5b087edc95cf467f2d207e61e5ca9a295244170ba971b6c68c50b60e48fdb670
issue-collapse:04a7499fa5b6474e,oncall: pt2 stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"138299, 141560, 141727, 147239, 149094, 149096, 149318, 149452, 149453, 149454, 149455, 150961, 150963, 153242",,,,oncall: pt2,"compile-cache, empathy-day, low priority, oncall: pt2, triaged",,Handle this oncall: pt2 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#138299 Cache enablement should be tested inside of cache load() function, not at call site, #141560 [Feature Request] Add source Inductor Python file path to each Triton kernel, #141727 inductor and eager gives slightly different results for rms_norm, with `emulate_precision_casts=True`, #147239 `DeviceCopy in input program` source hint, #149094 How to skip backward specific steps in torch.compile, #149096 How to determine which part of torch.compile undergoes recompiling after caching",8af807065f48451126d323373aff3daee0894af609262a22040178c61296a950
issue-collapse:5d488920c64acf02,oncall: pt2 stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"102853, 105859, 106269, 106388, 111519, 124427, 126644, 154214, 154463, 155065, 155332, 157918, 162259, 167895",,,,oncall: pt2,"compile-cache, enhancement, feature, low priority, no-scrub, oncall: pt2, onnx-needs-info, triaged, upstream triton",,Handle this oncall: pt2 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#102853 pdb but for dynamo (and time travel debugging), #105859 Report model flop utilization (mfu) in benchmark, #106269 Make our source attribution debug prints more useful for Compiler Explorer, #106388 Better export story for autograd.Function?, #111519 [pt2+profiler] attach aot_id to CompiledFunction, _compiled_fn number without aot_autograd, #124427 Find a common home for decompositions, perhaps outside of the obliquely named _refs directory",4fcd1662e3d38c83a96e8763e0c80f7a52191ef5d6c6ec2b5f213a58a81f3eac
issue-collapse:835505ce676e514b,oncall: jit stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14219, 29108, 29190, 31545, 32287, 32641, 33876, 33887, 34483, 34814, 35408, 35788, 35820, 36370",,,,oncall: jit,"oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14219 C++ model load error, #29108 a retrained and saved jit module could not be reload., #29190 How to run two different jit models in two GPUs respectively in one scrip?, #31545 IndexExpressions (or slice) for jit.script functions, #32287 Python 3.8 Windows JIT test failure, #32641 changed format of trace graph in torch 1.4.0",6c766adb4115a9ee3942f97cb47b5e5e33387421a1d5da51d55d6cf692b653f6
issue-collapse:7a9491de950b7d49,oncall: jit stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18776, 33672, 36575, 36739, 36917, 36931, 37103, 37987, 37988, 38142, 38472, 38473, 38863, 39362",,,,oncall: jit,"oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18776 Value of torch.backends.cudnn.benchmark Baked into JIT-Traced Modules ( 150x slowdown on ConvTranspose2d() ) [jit] [libtorch] [cudnn], #33672 [dev] `RecursiveScriptModule` does not expose `jit.ignore`d methods, #36575 One confusion about the CompilationUnit destructuring process in torch/jit/__init__.py, #36739 Custom class type name is very wordy, #36917 jit trace failed due to ""failed to differentiate `prim::ListConstruct`"", #36931 Pull hacked twins out of prim ops",f274ef553f414ac4032173a21b58b0c7005afb257aeaabc1271bbbd1f2015563
issue-collapse:1799d1b4b98ab679,oncall: jit stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31565, 34126, 39356, 39463, 39719, 39745, 39806, 40027, 40070, 40071, 40679, 40882, 40886, 42366",,,,"oncall: jit, oncall: mobile","enhancement, jit-backlog, oncall: jit, oncall: mobile, triaged, weeks",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31565 Force libtorch to use CUDA context, #34126 torch::jit::script::Object::attr should throw AttributeError instead of RuntimeError, #39356 Torch script to support dictionary with keys of type tuple, #39463 JIT test suite has dependencies across tests, #39719 User Objects aren't recursively scripted as nn.Module attributes, #39745 [JIT] exported dunder methods are ignored",47c3f74f98c5010d7e0143c380d163ccf4503280ab87d3a370612c4855362bc9
issue-collapse:dbd70cfef0002ad3,oncall: jit stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18094, 22454, 22487, 28258, 32988, 33463, 35116, 35180, 37104, 39365, 42459, 42882, 42912, 48153",,,,oncall: jit,"days, enhancement, feature, oncall: jit, small, triaged, weeks",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18094 [JIT Script] Need support on distributions.Categorical., #22454 Tracing an RNN does not support torch.nn.utils.rnn.PackedSequence as input, #22487 ""Floating point exception"" after trying the method from the issue #22382, #28258 [jit] TorchScript classes don't work in notebooks, #32988 [jit] Dict set item type mismatch error doesn't say the type was inferred, #33463 torch._C.Node.scopeName() missing in pytorch 1.4",cea5da8f17f3013bacc38728b89fbfc742f6a52837e34a8c8513a554344803fc
issue-collapse:9e5f4bdbd79af8a7,oncall: jit stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17688, 18106, 26744, 27094, 27343, 28239, 35290, 35534, 38964, 47650, 55083, 56183, 61919, 63475",,,,oncall: jit,"jit-backlog, low priority, oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17688 Improve save() method in torch.jit.ScriptModule, #18106 copy.copy not working for ScriptModule, #26744 [jit] Default args don't work with TorchScript classes, #27094 [jit] Document what types can be traced, #27343 [jit] `ScriptFunction`s are loaded as `ScriptModule`s, #28239 support class annotations in __init__",78d54b076fcc2fe01aaabf5936c054d3d579eadb43c47210cdf0e16ed9983ce2
issue-collapse:30fb8721f9badfe6,oncall: jit stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17655, 18157, 19628, 19978, 20591, 21275, 30754, 32039, 32358, 32695, 32822, 40507, 58720, 66476",,,,oncall: jit,"enhancement, jit-backlog, oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17655 TracedModule 'to' attribute doesn't work for tensors created on forward., #18157 JIT does not batch linear layers in an ensemble, #19628 [jit] nn.LSTM errors in nn.ScriptModule, #19978 [JIT] traced model with optimization shows no performance improvement, #20591 @ignore annotation for user defined type, #21275 weight_norm is not supported in TorchScript",a49513dd9ece381ed3eb28ebddf7fef828b7c3bc2a2e7583e8dba4380aca0fa2
issue-collapse:7728d65b2e7aad2d,oncall: jit stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25387, 25661, 31141, 32268, 34839, 35074, 35520, 37384, 38138, 55230, 60156, 65253, 67687, 85229",,,,oncall: jit,"enhancement, jit-backlog, oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25387 [DRAFT] Auto-casting in JIT - Automatic mixed precision, #25661 finfo operator not bound into JIT, #31141 [JIT] tensor(device=...) and tensor.to(device = ...) does not work properly in traced functions and modules, #32268 Numpy array functionality in torchscript., #34839 [torch.jit.script] Allow `range` to index into Tensor, #35074 [JIT] If a python function type comment is referring to a wrong type, JIT frontend gives a not helpful error message",3102b2a5db116fb291348247058ab872434f640670bbc3f8be527b5772732e20
issue-collapse:b3b79d8c65a2ec82,oncall: jit stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"19117, 19349, 29637, 31815, 35881, 39421, 40024, 74112, 90000, 94441, 97595, 97638, 101716, 107321",,,,"oncall: jit, oncall: quantization","feature, jit-backlog, low priority, oncall: jit, oncall: quantization, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19117 [jit] Can't `torch.jit.script` a lambda, #19349 Different behavior when trace model., #29637 Compiled functions can't take variable number of arguments, #31815 ImportError: /home/xx/anaconda3/lib/python3.7/site-packages/torch/lib/libtorch_python.so: undefined symbol: _ZNK5torch3jit5Graph8toStringE, #35881 Error when loading jit traced FasterRCNN model in C++, #39421 torch::jit::script::Module::to(torch::kDouble) also casts buffers",1b38a71529189db19265c8864b7aaf527bbe3a2566177d36bbf1bb7565f4d98a
issue-collapse:fd4d0a73dc7ab45a,oncall: jit stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18850, 20119, 21442, 21544, 24102, 24786, 25329, 25410, 25648, 26730, 28220, 29254, 29377, 29419",,,,oncall: jit,"enhancement, feature, jit-backlog, oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18850 Allow tracing of models which output `None`, #20119 [JIT] Source highlighting doesn't line up when tabs are used for indentation, #21442 [JIT] kwarg with default doesn't work for class instantiation, #21544 [jit] In pickler, don't memoize if not necessary, #24102 [JIT] script doesn't convert dtypes back to torch.dtype from long, #24786 [jit] Dict iterator invalidation doesn't match Python",1f7dda6a5083652bed0648ed83ed1b97b670d58e9853bcc33237b94f64ef6572
issue-collapse:417749a5c2eec043,oncall: jit stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17970, 18005, 18212, 19581, 22911, 22954, 23355, 29662, 30161, 30182, 30489, 31191, 31259, 31391",,,,oncall: jit,"jit-backlog, low priority, oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17970 Binary not operator causes crash when Jit module is executed on different device, #18005 Support 'bytes' type in torchscript, #18212 [JIT] bitwise NOT does not handle tensor shapes correctly under JIT, #19581 [jit] Do aten::values dispatch at build time instead of runtime, #22911 JIT trace parameter sharing error if Module attributes happen to be the same, #22954 Support serializing IValue to bytes (and deserialize from bytes)",a74a1cedcbc1e994669b302f885a595036b59eb76bbcac79fe39f5eb6aa9dbd6
issue-collapse:62d122edddf63537,oncall: jit stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23890, 23958, 24225, 24229, 24248, 31866, 32357, 32375, 32390, 32488, 32511, 32840, 33114, 33396",,,,oncall: jit,"jit-backlog, oncall: jit, quantization_release_1.3, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23890 [JIT] Can't use ndim in script, #23958 [jit] Python @property's not supported in TorchScript, #24225 Allow forward method to be defined with .define() in new TorchScript API, #24229 Confusing error message for Custom Class type mismatch, #24248 Loading custom Torchscript C++ operators in python segfaults due to ABI compatibility issue between pytorch and libtorch, #31866 Negative indices in chunk could cause Out of Range access on loss.backward in JIT",2118f96fe49b2c9ef959eaca411ffc26b521e311a3f44083404b8bba9e18fd39
issue-collapse:4c615cc8b9ee68e2,oncall: jit stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25047, 25803, 27003, 27406, 27505, 28517, 28519, 29063, 29597, 29642, 29750, 29822, 30226, 30788",,,,"oncall: jit, oncall: mobile","jit-backlog, low priority, oncall: jit, oncall: mobile, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25047 ScriptModule and nn.Module parameter ordering difference, #25803 [jit] `random` module support, #27003 [JIT] Script doesn't preserve builtin torch named tuples upon python return, #27406 [jit] String frontend doesn't support default arg values, #27505 [JIT] ndimension not supported, #28517 [JIT] __repr__ support for ScriptModules",f61d91109b81ab413c463e3e7f778cdc11663787f49b32d7b4ec9dd3d857da3f
issue-collapse:fa5249613b475fa8,oncall: jit stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31202, 31386, 31430, 31551, 31574, 31882, 31953, 32253, 32337, 32690, 33353, 33501, 33782, 33844",,,,"oncall: jit, oncall: mobile","feature, oncall: jit, oncall: mobile, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31202 Request for Lint Pass to Detect Modification on Parameters/Attributes during TorchScript Inference, #31386 [jit] Python type hints in TorchScript classes don't work, #31430 [jit] `del` with slices doesn't work, #31551 TorchBind broken on rocm, #31574 attribute and register_buffer are not the same on gpu, #31882 Example cmakelists for custom cuda operator?",970c1a8f380d6d21c181ee82d640345f23ea33f7e10d070450d8c46b73ea3cad
issue-collapse:0f30554e7874b2fd,oncall: jit stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23850, 27970, 34837, 35108, 39357, 40635, 72263, 104195, 107503, 108224, 108432, 120820, 120822, 134860",,,,"oncall: jit, oncall: quantization","TSRootCause:TypeChecking, TSUsability, better-engineering, days, enhancement, feature, jit-backlog, low priority, oncall: jit, oncall: quantization, quantization_release_1.3, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23850 [quantization] jit::class_ for packed weights, #27970 can't load model on cuda after call cudaDeviceReset functions, #34837 [torch.jit.script] Support tensor indexing after an Ellipsis, #35108 zip(list, tuple) throws an non-actionable error message, #39357 Enable scripting modules with recursive functions, #40635 [JIT] Recursive compilation doesn't apply for types used in type expressions but not value expressions",ab0a48c37736729818a55cd8937e7fbd0c8f358222260e73f3d1378a0bf0fab4
issue-collapse:db953d8058dc170e,oncall: export stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"96386, 103483, 111902, 112079, 112095, 113507, 113552, 113985, 117791, 118922, 123971, 123973, 133262, 134180",,,,"oncall: export, oncall: pt2","export-triage-review, export-triaged, feature, low priority, oncall: export, oncall: pt2, onnx-needs-info, triaged",,Handle this oncall: export stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#96386 [export] ""strict subset of traced input/output"" error when huggingface `ModelOutput` is returned, #103483 torch._dynamo.exc.Unsupported: Tensor.backward with aten_graph=True, #111902 torchscript file can not be loaded if its saved form the export model produced by torch.export.export, #112079 The error type `UserError` in dynamo is confusing., #112095 torch.export fails on a model with optional parameter, #113507 Revisit test_edge_op_registration on PyTorch CI",b9440036cc1573281d00ae79455e1ad7188c8e2f008f26a5e5cbacbcb11ed6d0
issue-collapse:1f3d180b375ee4cc,oncall: export stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"112443, 113122, 136381, 137274, 139305, 139401, 142025, 142176, 148193, 148568, 152337, 153832, 156411, 160937",,,,"oncall: export, oncall: fx, oncall: pt2","enhancement, export-triage-review, export-triaged, oncall: export, oncall: fx, oncall: pt2, triaged",,Handle this oncall: export stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#112443 torch.export emits node outside of Core ATen IR, #113122 Request to create a tutorial for loading a model dumped by `torch.export`, #136381 [Meta] torch.export readiness for Recommendation models, #137274 Not decomposing CompositeImplicitAutograd ops in post-dispatch export changes autocast (and possibly more) behavior when compared with eager mode, #139305 torch.export torchaudio kaldi module, #139401 [export] Failure to export Mixtral-8x7B-Instruct-v0.1",a4b82fafcf6bba950ccc3123a749469fe39e2f27fc1022987fd7c4e5e0760764
issue-collapse:671b82949966f815,oncall: distributed stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22400, 29115, 50820, 57278, 75512, 120466, 120847, 121650, 123070, 123528, 124228, 125620, 125800, 126347",,,,oncall: distributed,"better-engineering, feature, oncall: distributed, pt_distributed_rampup, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22400 Sparse allreduce for ProcessGroupNCCL, #29115 crash when call dist.new_group(ranks=local_ranks, backend='gloo'), #50820 NCCL_BLOCKING_WAIT=1 makes training extremely slow (but if not set then OOM on one device will hang training), #57278 Attributes consistency in Sampler and DistributedSampler, #75512 Connection closed by peer when using dist.isend in gloo backend, #120466 faster communication",b79fabf12b3b27503b008d629768c2cf096372d1c9f0c317b88cf6d7d53dfae8
issue-collapse:0fe138f36f57988a,oncall: distributed stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"104507, 117883, 126216, 126773, 126899, 128066, 128128, 128696, 128699, 128706, 128821, 128963, 129477, 129672",,,,oncall: distributed,"oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#104507 [error] while Implementation of pytorch DistributedParallel, #117883 Flight Recorder - feature requests, #126216 Allow Lazy Initialization for PipelineStage, #126773 Give`wait_tensor()` a schema that reflects its side effect, #126899 function signature for multiprocessing.spawn is multiprocessing.spawn.spawn, #128066 reduce_scatter_tensor with strided inputs produces corrupted results",ee24b035479493bc400c5ab080ad4b930c355c5939dad20070d0097ad7bed710
issue-collapse:fb0e78b86a2be35d,oncall: distributed stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30723, 44544, 121519, 125173, 128204, 130621, 131691, 132428, 132649, 132696, 133119, 133224, 133380, 134190",,,,oncall: distributed,"oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30723 Distributed Package asynchronous send/receive not working as expected (Gloo), #44544 torch.distributed server back-connects to clients on random ports leading to firewall problems, #121519 Avoid shard 0 download on all ranks, #125173 Flight Recorder Sequence IDs are insufficient, #128204 Documentation for DDP-related environment variables, #130621 [Bug Report] Choosing the network interface to use",84edbb64d9aeb5999b79766ab45786e7c6cd3606e76b8e19b31d4a87bf1ff083
issue-collapse:2ba16f7506ba6378,oncall: distributed stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22788, 109401, 127517, 133761, 134668, 134803, 134945, 134960, 135267, 137006, 137245, 137610, 137969, 137974",,,,oncall: distributed,"actionable, oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22788 Pytorch deadlock from distributed multiprocessing, #109401 Interleaved isend and irecv causes hang, #127517 With pytest8.2.0 or later, test cases under test/distributed/ execute will meet issue ""object has no attribute 'runTest'. Did you mean: 'run_test'"", #133761 Incorrect GPU management and deadlocks without torch.cuda.set_device, #134668 Whether tensor parallelism supports the overlap of communication calculations for gradient computation, and how to implement it, #134803 enhanced debugging prints in torch.distributed",7f98350c09a887a76cedbf53fc4f2461c56333381b42031a5b0140b7414d613a
issue-collapse:21ec45591ef96287,oncall: distributed stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22049, 41485, 92132, 130137, 131450, 133092, 138811, 139304, 139324, 140712, 141335, 141746, 142229, 145043",,,,"oncall: distributed, oncall: pt2, oncall: r2p","feature, oncall: distributed, oncall: pt2, oncall: r2p, pt2d-triage-nov2024, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22049 Cannot update part of the parameters in DistributedDataParallel., #41485 GLOO infiniband with PyTorch, #92132 Torchrun seems to have problem with virtual environment, #130137 How to get stream operators in custom backend compiler ?, #131450 Make torch.compile support inter-op parallelism, #133092 Umbrella issue for Dynamo tracing through ProcessGroup related torch.distributed APIs",016f9150d36054c03e5991a9d6afcebdb15197f4207e9756ab5ae25285c90011
issue-collapse:e9743af8a35ec247,oncall: distributed stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"116393, 143969, 145263, 145592, 145978, 146119, 147263, 147732, 148150, 148162, 148302, 148858, 148976, 150928",,,,"oncall: distributed, oncall: distributed checkpointing","oncall: distributed, oncall: distributed checkpointing, release notes: distributed (checkpoint), triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#116393 Error: IndexError: map::at When using torch.distributed.all_reduce(tensor), #143969 [distributed] parallelize_module error with `SequenceParallel`, #145263 [Pipelining] Problem using `torch.distributed.pipelining` on `Gemma2ForCausalLM`, #145592 _pickle.UnpicklingError: invalid load key, ''., #145978 What is the recommended way to use Distributed Checkpointing Save/Load with HSDP?, #146119 File based multi gpu",48302825cfc776b0152a2000ec883955b73c8a43f84ab8e03697e29f1073134d
issue-collapse:968ddfaf4b3891f7,oncall: distributed stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"148689, 148837, 149049, 149516, 149934, 150585, 151159, 151222, 151240, 151457, 151556, 151759, 152127, 152269",,,,oncall: distributed,"oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#148689 NUMA Binding Integration with torchrun, #148837 NVLS support in Pytorch, #149049 `pytroch.distribute` should support ""meta"" device tensors, #149516 ```StateDictOptions``` in combination with ```cpu_offload=True``` and ```strict=False``` not working, #149934 inconsistant behavior to pass tensor/tensor_list from torch.distributed module to cpp API, #150585 [distributed] Crash when trying to use default PG after creating new PG",5914c88201108e8a5dac12360b96a5fd78d44c125ab935b6302bd0ee9fe276c4
issue-collapse:bc5bbfd758ec9f0b,oncall: distributed stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7471, 8290, 14632, 17041, 18920, 19330, 21015, 22128, 22340, 22455, 23230, 25417, 25552, 28552",,,,oncall: distributed,"enhancement, feature, newcomer, oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7471 Make c10d/FileStore cache file descriptor, #8290 [feature request] MPI init_method for torch.distributed, #14632 [c10d] Check that allgather/gather output tensors point to different storage, #17041 Errors running distributed example, #18920 [c10d] CUDA tests for C++ reducer, #19330 Multiprocessing on distributed Multi-nodes shutdown error: ‘spawn’ on slave node leads to semaphore_tracker leaked",09d49ba4a0efcd87305153fd4e165e8571f6de2d419cf65bdedeb1d5551ee12e
issue-collapse:d39100d84a24845b,oncall: distributed stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11793, 17703, 21144, 27955, 29033, 30469, 30505, 30718, 32372, 32490, 32924, 33684, 35160, 35671",,,,oncall: distributed,"enhancement, feature, oncall: distributed, todo, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11793 DataParallel: Parallel_apply assert len(modules) == len(inputs) AssertionError, #17703 Training hangs when using DistributedDataParallel in two pod on two nodes, #21144 Data Parallel Implementation Improvements, #27955 Expand Pytorch C10D backend to dynamic load third party communication library, #29033 Move allgather_coalesced functionality to comm.cpp., #30469 Add single input/output tensor scatter/gather to ProcessGroup base class",bca5f6e22acf7bc735c52a933f64f41a77458ca73e80aa4dbed429c7118bf85a
issue-collapse:c2c4da09d13d4736,oncall: distributed stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"114357, 124714, 138329, 148086, 150836, 152261, 152645, 152746, 152964, 153673, 155680, 157321, 157391, 158244",,,,oncall: distributed,"oncall: distributed, release notes: distributed (symm_mem), triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#114357 Pytorch DDP across nodes: self._store = TCPStore( # type: ignore[call-arg] RuntimeError: Stop_waiting response is expected, #124714 `torch.distributed` hangs when using `torch.distributed.barrier` before any other communication primitives., #138329 DTensor RNG state for non CUDA backends, #148086 FSDP2 without sharding works slower than DDP, #150836 `Aborted (core dumped)` in `torch.cuda.nccl.reduce`, #152261 Context Parallel -- unsharded output doesn't match output without CP.",afe6040babd2db8cf09eb459606b3b238f6d491fdc55cba65542201ec87165ca
issue-collapse:bebb876a862b2332,oncall: distributed stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4406, 11324, 32018, 36516, 36519, 37079, 37149, 38782, 39022, 40566, 41101, 41236, 41428, 42879",,,,oncall: distributed,"better-engineering, feature, oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4406 DistributedDataParallel doesn't converge well when using MPI, #11324 Why does DistributedDataSampler not use default RNG?, #32018 DistributedDataParallel non-floating point dtype parameter with requires_grad=False, #36516 DDP should divide bucket contents by the number of global replicas instead of world size, #36519 Restructure test_c10d.py and test_distributed.py, #37079 Result parameters from Single-Process Multi-GPU DDP training on RNN do not match local training",6e3d61d5fac7dc53d69bdbe70c890c3a49e57120e9230c72d50a1d31448066f5
issue-collapse:afd7f040bcb53586,oncall: distributed stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13865, 14372, 31710, 39662, 43238, 43510, 44403, 46753, 49347, 49481, 50303, 53427, 54823, 55448",,,,oncall: distributed,"better-engineering, distributed-backlog, feature, oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13865 [c10d] Coordinated file truncation for FileStore, #14372 [c10d] Configurable timeout per operation for MPI backend, #31710 AttributeError: module 'torch.distributed' has no attribute 'init_process_group' on torch 1.3 aarch64, #39662 [RFC][distributed] RFC: c10d ProcessGroup extension and C++ API change, #43238 Expose counters for torch.distributed to report communication overhead within a block., #43510 Make busy waiting time in DDP (kSynchronizeBusyWaitMillis) a configurable environment variable",359c5b1d33955e85780d7b6622d3a3cc00bdc1729e90b25bb4a31ade405da68c
issue-collapse:02ed133086c53a56,oncall: distributed stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15716, 19267, 50323, 51014, 56583, 56772, 57158, 57646, 57787, 58352, 58371, 61289, 61661, 61839",,,,oncall: distributed,"feature, oncall: distributed, pipeline parallelism, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15716 Nn.dataparallel with multiple output, weird gradient result None, #19267 torch.distributed.broadcast should default to current stream, #50323 running a job on multiple gpus with qsub, #51014 [pipeline] gpu util + peak mem reporting to tune partitions and chunks, #56583 All thread stack trace dumps interleave output in distributed tests, giving garbled output, #56772 The multi-fc losses calculating in DistributedDataParallel.",ea4a09782678abed126aa9e2dbc8eecc562d002a805b2996e4f669ccc59713fe
issue-collapse:e72de2b5a1ff98ca,oncall: distributed stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40147, 45042, 57136, 61591, 61636, 61659, 61662, 65038, 65428, 65569, 65628, 65650, 65664, 65670",,,,oncall: distributed,"better-engineering, oncall: distributed, pipeline parallelism, pt_distributed_rampup, sharded_tensor, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40147 SyncBatchNorm doesn't work when I set track_running_stats False, #45042 torch.distributed launch.py is hanged. (pid, sts) = os.waitpid(self.pid, wait_flags), #57136 Intermittent segmentation fault when using rpc_async/rpc_sync with CUDA tensors, #61591 Add `map_location` arg for c10d collective comms, #61636 Enhance Distributed `get_future()` docs, #61659 Python program locks up when using some innocuous code inside mp.spawn()",451e3cdc2119bbc739dc170f4648163829d173d044325d5c0c356d009e0a6a59
issue-collapse:0ec3a645c7bab08c,oncall: distributed stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17113, 47433, 49846, 56838, 70654, 72022, 75129, 75213, 76225, 76745, 76747, 78047, 79073, 79467",,,,oncall: distributed,"awaiting response (this tag is deprecated), better-engineering, feature, oncall: distributed, pt_distributed_rampup, sharded_tensor, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17113 Distributed training jobs do not terminate properly if there is a crash, #47433 unbalanced gpu memory when using DistributedDataParallel, #49846 torch init_process_group always hangs, #56838 Comprehensively test NCCL's `get_future()` API, #70654 [RFC] UCC integration in ProcessGroupNCCL, #72022 [PT-D] To make the ShardedTensor Reshard more generic",c5b3591ed285296c6fb688fbdd95f6d225141b093248788335a0e97a57324149
issue-collapse:9d7e8d01bfa65a69,oncall: distributed stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25767, 55967, 61781, 69663, 75480, 79468, 79847, 80241, 80242, 83193, 84468, 87555, 87634, 89686",,,,oncall: distributed,"Stale, feature, oncall: distributed, pipeline parallelism, pt_distributed_rampup, release notes: distributed (pipeline), triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25767 Model parallel with DDP get `Socket Timeout` error when using NCCL, while GLOO works fine, #55967 Cannot init, destroy, and then re-init process groups, #61781 Letting `_allgather_base` to support multiple tensors as inputs and outputs, #69663 After updated to pytorch1.10.0 cuda11.1, NCCL is not available, #75480 IndexError: Caught IndexError in replica 0 on device 0., #79468 Ensure the guards in distributed_c10d.py wrappers get executed in the replay of the graph",0efd9997cf99463a670151792c79fd26d7f3686a87f1a0559526c367771f048a
issue-collapse:580dde1ff5ebca05,oncall: distributed stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"52040, 57776, 76069, 91760, 91855, 91856, 91940, 91943, 94806, 95309, 95403, 97114, 97484, 97670",,,,"oncall: distributed, oncall: r2p","oncall: distributed, oncall: r2p, topic: improvements, topic: performance, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#52040 torch.distributed.TCPStore doesn't work with dual IPv4/IPv6 network interface, #57776 [rfc] Build a debug tool to help users find which parameters are unused during DDP training, #76069 `init_process_group` hanging on HPC multi-node system w GPU, #91760 Inconsistent rank among torch.distributed primitives, #91855 torchrun default value of command line options, #91856 torchrun --help is too slow",41ec2eed8e5c119fcd5d79974df5fdc1b14dcc7f429e0e568fe58a261ff95d95
issue-collapse:09f5712409b93d11,oncall: distributed stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"51429, 92128, 93947, 97991, 98286, 98617, 99432, 101177, 101505, 101506, 101586, 102435, 102812, 103306",,,,oncall: distributed,"oncall: distributed, topic: new features, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51429 dist.init_process_group('nccl', ...) hangs in some combinations of pytorch+python+cuda version, #92128 Forward arguments are not updated in DDP, #93947 [RFC] Support Huge Model Init Without mallocs for Compile/Distributed Use Cases, #97991 functional collective should respect the whole mesh, #98286 When I use the DDP model, I use a custom loss function, when the batch size changes during training, the process will be stuck., #98617 Add test/distributed/test_c10d_mpi.py",6d55e7b54e7bc6ed00a1e2ce5bb29c4d1198c205e8cbfa8f311c7f620948241d
issue-collapse:88da5108b9059384,oncall: distributed stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28244, 102953, 103258, 103372, 103498, 103829, 104389, 104510, 107575, 107855, 108016, 108047, 108231, 108976",,,,oncall: distributed,"oncall: distributed, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28244 ProcessGroupMPI reports incorrect world size, #102953 TypeError: (): incompatible function arguments, #103258 Warn / deprecate / remove ProcessGroupNCCL._group_start(), _group_end() APIs, #103372 ImportError: cannot import name 'Store' from 'torch.distributed', #103498 ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 6047) of binary: /home/win10-ubuntu/anaconda3/envs/vicuna-7b/bin/python, #103829 RPC Framework support custom backend",a7851cb9a5e3cc997a8751908302c7cb52e70e97be8a037d55d7d217bdc5b409
issue-collapse:1770064ef4e1dba1,oncall: distributed stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"43536, 47885, 88621, 108484, 109191, 109934, 110363, 111623, 111789, 115309, 116400, 116691, 116980, 118891",,,,"oncall: distributed, oncall: profiler","enhancement, oncall: distributed, oncall: profiler, release notes: distributed (fsdp), triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43536 How do I debug ""RuntimeError: trying to initialize the default process group twice!"", #47885 NCCL 2.7.8 errors on PyTorch distributed process group creation, #88621 [FSDP] FSDP produces different gradient norms vs DDP, and w/ grad norm clipping creates different training results, #108484 DistributedDataParallel to support __getattr__, #109191 Gradients across different ranks are not synchronized when using DDP, #109934 test test_2d_fsdp_integration_fsdp_nested_param_groups failed",a546c46f228e693bfdff5cded0f273c585eceebc588b6d0a0f6ece2b18cb9ff1
issue-collapse:7cd946dcfd814b07,oncall: distributed stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"107352, 122529, 132292, 146371, 149990, 155472, 157445, 157606, 157627, 164843, 164925, 165122, 167349, 171179",,,,"oncall: distributed, oncall: distributed checkpointing, oncall: pt2","enhancement, has workaround, oncall: distributed, oncall: distributed checkpointing, oncall: pt2, triaged",,Handle this oncall: distributed stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#107352 dist.scatter is incompatible with transpose/permute operation, #122529 Distributed checkpoint state_dict load may report nccl error and hide the real root-cause exception, #132292 Can we avoid RecordStream for P2P ops?, #146371 Torchrun does not handle worker failure gracefully, #149990 [Async TP] Fuse all-gather-matmuls for float8 rowwise training, #155472 [RFC][c10d] Make it easier to act on group",298fea13a3baf9c8e40e3ad1e7b4603e576f8399b01decbea3bd0cd6ac5c7167
issue-collapse:512b38422a73cd4b,module: vision stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4958, 6934, 10719, 38009, 38619, 38630, 38709, 39993, 40756, 42012, 43123, 49752, 53890, 79130",,,,module: vision,"caffe2, function request, module: vision, triaged",,Handle this module: vision stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4958 [Feature Request] Extract glimpses from a batch of images (as in tf.image.extract_glimpse), #6934 Feature request: SSIM/MS-SSIM, #10719 size mismatch when trying to reconstruct predifined network, #38009 Resnet Model always predicting same label, #38619 Can you add NMS,RoIAlign,RoIPool for libtorch?, #38630 Torchvision error TypeError: _resolve_type_from_object()",1db7c0393686685dec93a22af638625ac1ed6815eaf5b8b40ac98563731cb5ce
issue-collapse:6451c3f22fced7d5,module: typing stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38240, 39009, 39670, 40659, 42969, 47027, 50031, 60766, 61233, 76465, 81046, 82095, 83313, 98260",,,,"module: typing, oncall: jit","enhancement, module: typing, oncall: jit, triaged",,Handle this module: typing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38240 RuntimeError: arg_types.size() == param_names.size() - (moduleSelf_ ? 1 : 0) INTERNAL ASSERT FAILED, #39009 Missing OneCycleLR and MultiplicativeLR in lr_scheduler.pyi, #39670 Force JIT to do type inference even when mypy annotated, #40659 [typing] overly restrictive List[int], #42969 Enable torch.testing._internal typechecks during CI, #47027 Fix the way imports are done to be more correct for static type checkers",e973b7206f20d57882e0f6bca27857e5d44534fc841678d08a8741481623927d
issue-collapse:5c7d434ec0250f25,module: typing stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"45414, 73359, 101080, 115806, 117449, 125390, 131054, 131432, 141869, 148552, 150257, 157266, 167119, 169073",,,,module: typing,"actionable, better-engineering, feature, module: typing, triaged",,Handle this module: typing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#45414 Module __call__ typing, #73359 Pytorch Typing, for Tensor type annotations, #101080 Sequence annotation in type hints is wrong, #115806 Merge all 3 mypy lint jobs, #117449 Deprecate and remove usage of from __future__ import annotations in codebase, #125390 [typing] Make arguments to `__getitem__`/`__setitem__`/ etc. positional only.",6ba5d92908d1059ed83a4c2fb86538a606e21f1a31322c39b0f60d3ecee1065a
issue-collapse:25431e6b4bbb4e1c,module: tests stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3791, 29717, 29720, 30611, 32132, 33089, 33249, 35648, 35651, 38188, 38189, 38190, 38241, 38677",,,,"module: tests, module: type promotion","enhancement, module: tests, module: type promotion, small, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3791 Make pytest stop printing docstrings in its default diagnostic output, #29717 CPU and CUDA error messages are divergent in type promotion, #29720 Provide a mechanism to set global state per test in thread-safe manner, #30611 Refactor/consolidate code for generating test tensors, #32132 Slighty out of tolerance for `test_mv` and `test_cholesky_solve_batched_cuda_float64`, #33089 test_baddbmm_cpu_float32 fails locally for me when built with DEBUG=1",c02a1b1fea7a2cb3872d4a7dc46dd7f941111d60d50f2af2d909d00c1e11597b
issue-collapse:b2b93181f1a24bb1,module: tests stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5096, 5326, 23849, 35553, 40086, 40470, 41187, 42498, 43663, 43677, 46102, 46222, 46276, 53256",,,,"module: tests, module: xla, oncall: distributed, oncall: jit","better-engineering, caffe2, module: tests, module: xla, oncall: distributed, oncall: jit, pt_distributed_rampup, shadow review, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5096 [feature request]Add an env variable to cover different pathes when testing code with openmp, #5326 TestMultiprocessing.test_fd_sharing hangs with ASAN, #23849 Make MultiProcessTestCase pickable, #35553 caffe2 `DEPTHWISE3x3.Conv` test is broken, #40086 TestListwiseL2rOps::test_lambda_rank_loss fails, #40470 jit's default dtype is different in sandcastle and test_jit.py",b13abd7822065363dce32b25c312f1769e3838d06efb00f580b2a18cfd7665f8
issue-collapse:dd610f37e9b953f4,module: tests stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8577, 11496, 15142, 46272, 53627, 54278, 54282, 54629, 55583, 56023, 59185, 61825, 62449, 63781",,,,"module: tests, module: third_party, oncall: distributed","NNC, better-engineering, enhancement, feature, module: tests, module: third_party, oncall: distributed, pt_distributed_rampup, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8577 Update tests to no longer spew debug info, #11496 Tests with ""."" in the name cannot be run standalone, #15142 Update third_party/googletest - Ability to skip tests in GTEST, #46272 test_distributed_* does not work with run_test.py -i option, #53627 test_metal.py must skip when not compiled with metal support, #54278 [Static Runtime] StaticRuntime.EmbeddingBag test case broken",8c389543dda5d58f8c5b4d3666550b6cc6a99ee71d9e31ca55e166e39a1e7c80
issue-collapse:a97231b3f5e14814,module: tests stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"60748, 62354, 64048, 64106, 64284, 65522, 66224, 66403, 67008, 71575, 72915, 74187, 74544, 74911",,,,module: tests,"better-engineering, enhancement, feature, low priority, module: tests, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#60748 Record file/line number when creating test data, and then report it in backtraces associated with this data, #62354 Some way to specify expected failures for OpInfo-based tests, #64048 Be able to use pytest to run a ""core"" set of tests, #64106 modulefinder_determinator is incompatible with imports into the test/ directory, #64284 FR: Record results of OpInfo reference tests and detect when numerics of an operator change, #65522 pytorch framework tests using make_tensor hangs with pytest's boxed exec option ""--forked""",0a24fa70c8442884d48f51f16468edd23d86357c64e900377262814987cff0e0
issue-collapse:9fb883d6e3f8dcd7,module: tests stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"74052, 76127, 76128, 76611, 76733, 77380, 77589, 78729, 80221, 84014, 84538, 84972, 86381, 94705",,,,module: tests,"better-engineering, feature, module: tests, topic: not user facing, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#74052 assertEqual gives confusing error when comparing tuple with Tensor with Tensor, #76127 run_test.py option to write out failed tests, #76128 test_license_for_wheel always fails on my local dev copy, #76611 OpInfo CUDA bfloat16 support detection is buggy, #76733 test_python_reference_meta_functions takes too long to run, #77380 microbenchmark-style tests",4ee954b0fc75cdf75682f842c3731af41ddb3342c8c1f3932ed0a903c2883bb1
issue-collapse:4248c3f37d8ff195,module: tests stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"35636, 63337, 67773, 74613, 76609, 86694, 98904, 115202, 121162, 138673, 141963, 143965, 159510, 162004",,,,"module: tests, module: torchbind, oncall: pt2, oncall: quantization","better-engineering, feature, low priority, module: tests, module: torchbind, oncall: pt2, oncall: quantization, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35636 Caffe2 generate_proposals_op_gpu_test crashes on Windows, #63337 Property based testing like AFL, #67773 Instability in `test_input_weight_equalization_activation_values` test for random test values., #74613 OpInfo request for `nn.functional` and `unbind`, #76609 AttributeError: '_thread._local' object has no attribute 'rel_tol' (cannot use TestCase.assertEqual from other threads), #86694 Feature request: Deterministic test input generation",3a546562ca2e7ff148984f402b30490a86eab37b88d2faae113ab07a74d610dc
issue-collapse:658aa6381272d8df,module: testing stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55331, 61758, 72194, 74320, 74328, 74500, 74531, 74642, 78761, 82669, 83019, 97761, 98977, 99248",,,,"module: testing, module: tests","better-engineering, enhancement, feature, module: testing, module: tests, tracker, triaged",,Handle this module: testing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55331 Tests should be runnable without run_test.py, #61758 `make_tensor` tracking issue, #72194 `TestCase.assertEqual` has `equal_nan` default to `True`, #74320 Make OpInfo repr more useful, #74328 instantiate_device_type_tests is misnamed, #74500 One should be able to query if a DecorateInfo is a xfail or skip",e31ecfc36fc3396539844277e4df7bcace1e15ff8b2e300e39a27027ac665329
issue-collapse:ec85ae513d762cc3,module: testing stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"104875, 106828, 111126, 111564, 116762, 118352, 122017, 125028, 139252, 140560, 141758, 142029, 152647, 162964",,,,"module: testing, module: tests, module: xpu","actionable, better-engineering, module: testing, module: tests, module: xpu, tensor subclass, triaged",,Handle this module: testing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#104875 torch/testing/_comparison.py: If you are a user and see this message during normal operation please file an issue, #106828 Use expect tests for error inputs, #111126 type promotion test for torch.div variants is broken, #111564 misusing percision value in test_cuda function in torch/testing/_internal/common_nn.py., #116762 parametrized should fail to generate tests with weird names, #118352 Unify TEST_WITH_TORCHDYNAMO and PYTORCH_TEST_WITH_DYNAMO",c317fc2680e3649856f1fe0d7fed560216fe0ddf7657964ca08bba27ef51e1a5
issue-collapse:189b312554eaff67,module: tensorboard stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24468, 27030, 27300, 39370, 42861, 44931, 54269, 61186, 67989, 79120, 79647, 88304, 93017, 102501",,,,"module: tensorboard, oncall: visualization","enhancement, feature, module: tensorboard, oncall: visualization, triaged",,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24468 [Tensorboard] Write summaries to S3 or GCS bucket, #27030 Tensorboard logging image WITH LABEL, #27300 Tensorboard add image with boxes, labels, and confidence scores., #39370 Draw quantized tensors in tensorboard, #42861 Can not get pytorch working with tensorboard, #44931 Issue while writing scalars o tensorboard using writer.add_scalars(....)",6afe1cf61b282beb688d20c6149121f8a2024bbdc00b6b74fd730af2f8b2ac41
issue-collapse:7a398e134a168a23,module: tensorboard stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"46739, 49284, 73131, 73909, 95460, 104029, 107536, 107820, 113942, 125298, 139269, 144677, 147317, 154855",,,,"module: tensorboard, module: windows, oncall: visualization","OSS contribution wanted, enhancement, feature, module: tensorboard, module: windows, oncall: visualization, triaged",,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#46739 torch.utils.tensorboard.SummaryWriter.add_embedding fails for some label_img sizes, #49284 TensorBoard SummaryWriter to remote storage unreasonably slow, #73131 Support writing tensorboard traces to AWS S3 (and other cloud storage services) in profiler, #73909 SummaryWriter reports encoding error, #95460 torch.profiler.tensorboard_trace_handler Generates an incorrect JSON file, #104029 SummaryWriter.add_embedding not working for RGBA images",6d7e5736c040a8ef85ac894a1242e0f3b1160b0930e3f3fb1155e72b330a6387
issue-collapse:0d6a8e95ad6b0508,module: sparse stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3152, 4829, 14112, 16328, 30402, 43904, 46980, 48892, 52738, 55453, 55944, 58735, 58814, 64079",,,,"module: sparse, module: tensor creation, module: tests, module: ux","enhancement, feature, function request, module: sparse, module: tensor creation, module: tests, module: ux, open source, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3152 Sparse tensor .new(size) can be confusing, #4829 [Feature Request] clip_grad_norm for sparse gradients, #14112 [sparse] add descriptions and examples for methods at torch.sparse doc page, #16328 Should torch.arange take a layout parameter?, #30402 There is no support for `weight_decay`/`momentum` in SGD for sparse tensors., #43904 Documentation and `torch.sparse` alias for `torch.bmm` sparse-dense",457e17fcb5d449799c39632edb543ec2626403d4e9a695b111ae878e0fd6c8e3
issue-collapse:27907b171753d454,module: sparse stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1462, 4247, 6257, 16187, 34951, 39437, 54309, 65230, 65400, 69348, 69516, 69665, 74855, 76519",,,,module: sparse,"Stale, enhancement, feature, module: sparse, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1462 Change sparse_mask to take indexing mask, rather than entire sparse tensor, #4247 Feature request: sparse matrix max(axis), #6257 [feature request] adding a nonzero element ""in-place"" in sparse tensor, #16187 Sparse matrix multiplication is too slow, #34951 [feature request] Sparse (hybrid sparse-dense) output option for topk, min, max, #39437 error when specifying sparse=True in embedding",565713d3195338c2f51aa7e88f09ce7de48fccfbb615a9b69db0b87f2c4d78b2
issue-collapse:d3ade58608b33fb9,module: sparse stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"76638, 78422, 81648, 81650, 81979, 82140, 82494, 85642, 86111, 88053, 89080, 91599, 91615, 94185",,,,module: sparse,"module: sparse, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76638 Design API for accessing sparse tensor indices, #78422 Permutation of Sparse Tensor, #81648 `sparse_coo.to_dense()` produces different results between CPU and CUDA backends for boolean non-coalesced inputs., #81650 RFC: auto-generated plain Tensor argument only sparse primitives, #81979 test_sparse_spdiags_cpu_bool fails on my local working copy, #82140 Stop manually binding sparse factory functions",6830bf79fe82be3f28600ac1f81f2bf1db92e87d28948254739bbc71b11d10f7
issue-collapse:f67d0fb11603007c,module: sparse stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"71156, 71678, 87433, 95463, 96428, 98495, 99797, 101189, 102162, 103643, 104845, 105529, 105548, 105916",,,,module: sparse,"module: sparse, topic: new features, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#71156 Slow backward for matrix multiplication of two sparse COO tensors on CPU, #71678 torch.bmm backward with sparse input, #87433 index_select() applied in sparse tensor can't backprop, #95463 `add/add_` for CSC: errors when trying to access non-existent `crow_indices`., #96428 Compressed sparse constructor allows mixed `int32/int64` indices which leads to dtype promotion/demotion in conversions., #98495 Strided to batch BSR/BSC conversion fails when the number of zeros per block varies while the number of blocks per patch is constant",87ab4ba3a412080bb896049984e99c2822562201369cb09696b990261d594c49
issue-collapse:4a426a48059de51e,module: sparse stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31742, 31942, 46705, 70926, 98929, 100158, 100411, 107102, 107590, 108102, 108483, 109462, 114808, 114871",,,,"module: sparse, module: tensor creation","feature, function request, module: sparse, module: tensor creation, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31742 sparse tensor eliminate_zeros, #31942 Support sparse inputs for torch.block_diag, #46705 torch.eye not supported for sparse layout, #70926 Feature Request: Implement `torch.sparse.spdiags` and `torch.sparse.diags`, #98929 torch.sparse_csr_tensor() stops gradients, #100158 `torch.sparse_csc_tensor` matrix multiplication produces MKL error SPARSE_STATUS_ALLOC_FAILED when density is too high",b5d65056f9bf0b69f5c71807113eaac0950c8b12fd62fc22b7745c0479e46bbc
issue-collapse:d9c5ac82f632e393,module: sparse stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"67092, 109194, 111509, 116565, 116633, 117146, 117243, 117244, 117247, 117248, 117250, 117256, 117259, 119711",,,,module: sparse,"enhancement, feature, module: sparse, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#67092 Sparse matrix multiplication (torch.sparse.mm) NotImplementedError, #109194 `torch.mean` not supported for `torch.sparse_coo_tensor`, but `torch.sum` is supported (`scipy.sparse.coo_matrix` does support both `mean` and `sum`), #111509 Sparse Tensor Sum Still Does Not Work for PyTorch Geometric, #116565 `test_autograd_dense_output_addmm` fails if matrix sizes are square, #116633 Sparse tensor element-wise multiplication bug when broadcast needed, #117146 `test_triton_bsr_scatter_mm` fails on V100",bd2645cab5fdbaf602b7ac3ec65e857029f9e6ee4f3097848534cf4f314e8e43
issue-collapse:c017b28c04597f5b,module: sparse stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3898, 34375, 74454, 106141, 117242, 120989, 121797, 122711, 123500, 124778, 125493, 131318, 132028, 134953",,,,module: sparse,"module: sparse, tensor subclass, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3898 Sparse matrices in dataloader error, #34375 .detach() behaves differently for dense tensors vs sparse tensors, #74454 torch has no attribute sparse_csr_tensor, #106141 Sparse COO indices are torch.Int64 -- is this necessary?, #117242 ```torch.tensordot```: Internal assert failed, #120989 input.is_sparse() INTERNAL ASSERT FAILED",a60e50e2e4f4dec2c626510f63e78101cb8a578dafb1b432c82714934f25a7a2
issue-collapse:19937bbd12bc32a2,module: sparse stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72065, 75721, 86494, 122089, 134396, 135091, 137542, 138168, 140778, 141144, 141472, 144210, 144688, 146976",,,,module: sparse,"enhancement, feature, module: sparse, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72065 Batched sparse-sparse matrix multiplication/ sparse torch.einsum, #75721 Sparse tensors: indexing with `index` and/or being able run `sparse_tensor[index]`, #86494 Automatic broadcasting for batch addition for sparse tensors, #122089 Infinite cycle on BSC tensor: base -> values -> base, #134396 torch.sparse.softmax missing support for sparse CSR, #135091 Any support for sparse linear layer in the future?",5455f418e680c89df5bcbfdbeeee2d84e02545accbe57c5381cb306085753850
issue-collapse:6acfb3eec2a483e2,module: sparse stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"101385, 107451, 119076, 129843, 146910, 148548, 153497, 154419, 155333, 160517, 161357, 162482, 164707, 169119",,,,"module: sparse, oncall: pt2","enhancement, feature, matrix multiplication, module: sparse, oncall: pt2, topic: fuzzer, triaged",,Handle this module: sparse stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#101385 torch.Tensor.is_sparse returns false for non-COO sparse tensors, #107451 Conversion from COO with two sparse dimensions to CSR with dense_dim specified fails, #119076 torch.bmm fails when multiplying a batch of sparse matrices by a dense tensor, #129843 Concatenating CSR matrices fails, #146910 INTERNAL ASSERT FAILED at ""/pytorch/aten/src/ATen/native/sparse/SparseCsrTensorMath.cpp"":817, please report a bug to PyTorch., #148548 Issue with Sparse Tensor Matrix Multiplication and Broadcasting",1842012d1526ce17e38a07a47fb5d1f72d7789242c6cd6a95d0cde5611f111ba
issue-collapse:8d065a29b0ba1ce8,module: serialization stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30867, 32706, 43672, 44295, 47753, 53185, 53708, 57843, 58671, 60165, 60933, 61923, 62271, 62273",,,,"module: serialization, module: tests","enhancement, feature, module: serialization, module: tests, triaged",,Handle this module: serialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30867 save model definition file, #32706 Torch serialization does not restore tensors properly when custom __reduce__ is defined, #43672 Make `torch.Generator` picklable, #44295 Request: Add PyTorch version to state dicts, #47753 Attempted to read a PyTorch file with version 3, but the maximum supported version for reading is 2, #53185 [NIT] Unpickler persistent_load monkey patch should fall back to original persistent_load",bf6b0803f95597dad09664545d561122de7b810605925fb524a9604ecfd03a28
issue-collapse:7b3561b2e33f00b2,module: serialization stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25137, 63720, 64458, 65604, 65694, 79145, 87276, 91325, 94238, 98187, 100244, 103895, 109791, 112395",,,,module: serialization,"feature, module: serialization, quantization_release_1.3, triaged",,Handle this module: serialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25137 Serialization does not work for quantized modules, #63720 Torch.save with zip serialization causes memory bloat, #64458 Rename C++ Serialization APIs, #65604 pytorch serializes entire tensor when you try to pickle a slice, #65694 Plain Tensor serialization does not save the content of `__dict__`., #79145 Redundant info are saved when using torch.save to save part of torch.tensor",2cc1fe22c9a174dfd9d2ec5736818fc1770acf44b9c588f2bd758ff75136fc23
issue-collapse:7cb4b2883abb71f5,module: serialization stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24139, 32167, 76108, 78720, 120708, 124195, 125622, 126952, 127001, 127525, 129356, 132024, 132991, 134680",,,,"module: serialization, oncall: quantization","enhancement, module: serialization, oncall: quantization, tensor subclass, triaged",,Handle this module: serialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24139 Allow incompatible shapes in load_state_dict(strict=False), #32167 Support in-place pinning of memory, #76108 RuntimeError: [enforce fail at C:\actions-runner\_work\pytorch\pytorch\builder\windows\pytorch\caffe2\serialize\inline_container.cc:300], #78720 ValueError during `yaml.dump(dtype)`, #120708 ELU quantization parameters (scale + zp) missing from model's state_dict, #124195 torch.save and torch.load is slow. Slower than numpy. Slower even than pickle.",483f3d664edd3212b336af3330dd007e985788885ce189e7607867d8477d9e56
issue-collapse:8d042901d2245c95,module: serialization stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88320, 126154, 132067, 137409, 141350, 146885, 149224, 149313, 149545, 157624, 158187, 161960, 166686, 169763",,,,"module: serialization, module: windows, oncall: pt2","enhancement, module: serialization, module: windows, oncall: pt2, triaged",,Handle this module: serialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88320 `torch.load()` cannot load data saved at non-zero position in a file (`failed finding central directory`), #126154 Optimizer classes not `dill` picklable after using `torch.compile`, #132067 Pickle loads torch tensor memory leak, #137409 Unable to save model after removing weight normalization due to remaining hooks, #141350 Support saving pytorch models in OCI registries, #146885 Add numpy.core.multiarray.scalar to default safe globals",cbd7628b6b4b60fd3b73793ec56da007f5a3bb77e7351a9a17131e05e7c02775
issue-collapse:1336c54d568afebe,module: sdpa stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"116350, 125070, 135590, 137486, 138907, 144960, 145040, 146848, 147585, 152257, 158969, 162592, 163400, 167754",,,,"module: sdpa, module: sparse, module: windows, oncall: pt2","enhancement, module: sdpa, module: sparse, module: windows, oncall: pt2, rocm, triaged",,Handle this module: sdpa stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#116350 RuntimeError: derivative for aten::_scaled_dot_product_flash_attention_backward is not implemented, #125070 2.2.0+ regresses SDPA performance on Windows, #135590 Discussion about the testing methods in test_transformers.py after math backend always uses fp32, #137486 [feature request] [ux] Introduce special value `SDPBackend.ALL` (or similar functionality) to mean ""all available backends"" for `torch.nn.attention.sdpa_kernel(...)` (some special value `SDPBackend.ERROR` already exists), #138907 Select SDPA backend smartly by `sdp_params` and benchmark results, #144960 Memory-efficient attention is not selected if inputs's ndim != 4",cf848030def735da633e1b4258ac8c3a445329a96890ff0963343dba29734d0c
issue-collapse:7f51999dc6169787,module: rpc stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24130, 27517, 31752, 32565, 32752, 32784, 33115, 33363, 33482, 33738, 33982, 33983, 33989, 33990",,,,"module: rpc, oncall: distributed, oncall: jit","better-engineering, module: rpc, oncall: distributed, oncall: jit, triaged",,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24130 [RPC] Add type annotations for RPC-related Python files, #27517 Allow explicit gradients in torch.distributed.autograd.backward() API, #31752 Need a launch utility for Distributed RPC framework., #32565 Reduce RPC branches for Python/Built-inOp/TorchScript, #32752 RPC mock mode for unit tests., #32784 Implement Backend-Agnostic RPC functionality in RpcAgent",23004b5bffa3463020a73f28d6842dfa8e3e57372438a6b19f9821ca573f1fa6
issue-collapse:3db8b2dd244e64ad,module: rpc stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32130, 33991, 34080, 34754, 35758, 37136, 38118, 38245, 39124, 39280, 39355, 40266, 40880, 40974",,,,"module: rpc, module: tensorpipe, oncall: distributed, oncall: jit","better-engineering, enhancement, feature, module: rpc, module: tensorpipe, oncall: distributed, oncall: jit, triaged",,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32130 Idempotency Keys for RPC Retry, #33991 TensorPipes RPC Agent Default Args/Result Device Mapping, #34080 RPC API Changes for TensorPipes, #34754 General Purpose Faulty RPC Agent, #35758 UserRRef should store error if it sees any and prevent subsequent usage, #37136 [Design][RFC] RemoteModule API Design",abd6fe9fde130c7619bc077a0ef5108087afa49de23cff69f87a4394f5e17c6a
issue-collapse:8d0a7b3ee4c5f4f0,module: rpc stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40936, 41369, 41561, 42345, 42544, 42705, 43561, 44084, 44771, 45044, 49101, 49911, 50462, 50468",,,,"module: rpc, module: tensorpipe, oncall: distributed","better-engineering, feature, module: rpc, module: tensorpipe, oncall: distributed, triaged",,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40936 [RFC] [RPC] Automatic retries of all requests in TensorPipe agent, #41369 [RFC] Device Placement API for RPC, #41561 [RPC] Should we support users _not_ calling rpc.shutdown()?, #42345 Generate unique `worker_id` for each node in the RPC framework for the store., #42544 torch rpc lose device information between rpc calls, #42705 torch.distributed.rpc package not work well with generator and lambda",9601ae53f55d459e8f8af3a772667ccd0c60718787143b757a87c564a3171bdb
issue-collapse:27cc7ac8cfc087fd,module: rpc stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30807, 39411, 46904, 50346, 51778, 53696, 57475, 57960, 65022, 71029, 73721, 73722, 78367, 104620",,,,"module: rpc, module: serialization, module: tensorpipe, oncall: distributed","better-engineering, feature, module: rpc, module: serialization, module: tensorpipe, oncall: distributed, pipeline parallelism, pt_distributed_rampup, triaged",,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30807 Sending sparse tensors over RPC not yet supported, #39411 Allow to_here to interrupt blocking wait in the case of rpc.remote timeout, #46904 Authentication for RPC, #50346 how to save weights when using RPC framework, #51778 Assertion on the existence of RRef user-facing methods' docstrings, #53696 Support GPU/CPU communication in RPC",163b604a7cd8814e206d74d69a017e0350a500fc47fbde7a4fc567841b01fec1
issue-collapse:117ba46f0ba13bf9,module: rocm stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"84044, 91630, 96643, 100795, 103608, 104147, 106529, 106608, 138360, 145608, 159226, 162606, 168935, 169378",,,,"module: rocm, module: tensorpipe, oncall: profiler","module: rocm, module: tensorpipe, oncall: profiler, rocm, triaged",,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#84044 hipErrorNoBinaryForGpu, but reversed, #91630 Periodic ROCM distribtued jobs are broken, #96643 PyTorch SGEMV is using 1 single core on AMD CPUs (very slow), #100795 Quickstart notebook fails to train properly with ROCm, #103608 Installing Torch on AMD Platform Leads to Huge Docker Image, #104147 PyTorch2.0 ROCM LayerNorm HIP error: invalid configuration",8d23bdb3720ece3c6dcb8f79914aa6dbaf95170ee7818798eb3247fc0f206db7
issue-collapse:1dcb7bb9a91a2e02,module: rnn stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1794, 47439, 64635, 65890, 68377, 80605, 82577, 88643, 89357, 135198, 145368, 149194, 149227, 169213",,,,"module: rnn, oncall: quantization","actionable, enhancement, feature, low priority, module: rnn, oncall: quantization, release notes: nn, topic: performance, triaged",,Handle this module: rnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1794 Feature request: reverse_padded_sequence, #47439 how to use torch.utils.checkpoint + gru with variable length sequence?, #64635 'LSTM' object has no attribute '_flat_weights_names', #65890 `pad_packed_sequence` is not exactly the inverse of `pack_padded_sequence`, #68377 Misleading Error Message from nn.RNN when Passing Incorrect Data Type, #80605 pad_sequence and pack_sequence should support length zero tensors",4bddb037f03dea78fc1bff01bce0ee7fb89fe3c7b93e9b92e29890f02906b42d
issue-collapse:387261b4289abaf0,module: random stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33388, 43816, 56805, 66924, 67571, 79018, 88775, 89438, 94454, 106563, 114945, 128126, 133453, 157662",,,,"module: random, oncall: distributed","csprng, enhancement, feature, module: random, oncall: distributed, triaged",,Handle this module: random stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33388 Sobol engine generates out-of-bounds samples after drawing too many samples, #43816 Move randperm() to DistributionTemplates, #56805 Implement random SeedSequence, #66924 [BUG] Inconsistent initialization on different machines (tensor.uniform_()), #67571 [discussion] Have default size/shape equal to a scalar, #79018 Random functions should infer device from user-specified Generator",a70bdd4af00c4fc934f1529de824dc5d6b275b3d0e7de19cba597d08e0a2b6f1
issue-collapse:e636f0881cbcaa23,module: python frontend stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"90284, 92251, 96615, 97189, 97823, 98948, 104935, 105062, 107701, 111819, 115614, 118488, 122081, 126006",,,,"module: python frontend, module: serialization, module: testing, module: typing","actionable, better-engineering, feature, module: python frontend, module: serialization, module: testing, module: typing, tensor subclass, topic: docs, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90284 Public API definition is not compatible with `torch.testing`, #92251 `torch.load(..., map_location=""cuda:0"")` allocates memory on both CPU and GPU, #96615 Tensor Permutation Along Given Axis, #97189 torch.randint range for torch.int64 dtype seems wrong, #97823 Using `param in param_list` can trigger `non-singleton dimension` error?, #98948 behaviour of `torch.tensor()` changes after editing `Tensor.__getitem__`",9b614df725c704d7d1e97ee554f6a0d0419d13f9b4f94a6a30bc8c90b768293e
issue-collapse:26ebb6e8e767b372,module: python frontend stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"126014, 126021, 126034, 126041, 126244, 126247, 126250, 126253, 126404, 126947, 127009, 127010, 127019, 127020",,,,module: python frontend,"module: python frontend, topic: docs, topic: fuzzer, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#126014 The unexpected behavior of `can_cast()`, #126021 Incorrect output from `torch.histc` when min and max are equal., #126034 A scalar with `tensor` argument of `result_type()` doesn't work, #126041 A scalar with `input` argument of `remainder()` doesn't work, #126244 `dtype` of `torch.tensor()` also accepts `int()`, `float()` and `bool()` which are python built-in functions., #126247 `type1` and `type2` of `torch.promote_types()` also accept `int()`, `float()` and `bool()` which are python built-in functions.",684315c1227ff5d097145a4d34f3e5214638fbb6a845bafba2d50a229bfbfc4f
issue-collapse:db19901e688ac566,module: python frontend stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"36380, 101185, 126045, 127083, 130512, 130693, 130807, 131463, 131769, 132483, 133564, 133985, 133998, 134978",,,,module: python frontend,"actionable, enhancement, feature, module: python frontend, topic: deprecation, topic: fuzzer, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36380 Population Count Op, #101185 Pure virtual function call exception on Python interpreter exit when using debug wheel, #126045 A scalar with `elements` or `test_elements` argument of `isin()` doesn't work, #127083 Setting a `float`, `complex` or `bool` type value to `precision` argument of `set_printoptions()` has problem when creating a `float`, `complex` or `bool` type tensor, #130512 The argument of `take()`, `take_along_dim()` and `gather()` should be `index` or `indices` for consistency., #130693 The doc of `vstack()` should say `tuple` or `list` of tensors for `tensors` argument",c84b88c53e93a385ef9cc1042848175d96bf8a204dbcdb506ceda9ecf6e89a29
issue-collapse:a7dfabf375736ba3,module: python frontend stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"139200, 141295, 142129, 142852, 143208, 143306, 143334, 143440, 144671, 146765, 148536, 149815, 150124, 150284",,,,"module: python frontend, module: third_party, module: typing, module: viewing and reshaping","actionable, enhancement, module: python frontend, module: third_party, module: typing, module: viewing and reshaping, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#139200 find_spec does something weird to Python Path when loading modules, #141295 Memory efficient reshape for stride-0 tensors, #142129 Feature Request - Backport torch.special.gammainc for Backward Compatibility, #142852 Deprecate `torch.squeeze()` with unspecified `dim`, #143208 We should switch Tensor object (and other perf-critical objects) to use the new managed dict feature from CPython, #143306 Add CPU scalar support in addcdiv",17b4a639d728a26a31aadb5320c58840544da2a90becda0e06d4204a0da2b572
issue-collapse:5ccd6e82c5491fa7,module: python frontend stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"95882, 101930, 117762, 154557, 156701, 156956, 158125, 158738, 158871, 159739, 161886, 162790, 165799, 170573",,,,module: python frontend,"actionable, enhancement, low priority, matrix multiplication, module: python frontend, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#95882 Proposal: accept ellipsis as index in `torch.index_put()` to enable broadcasting, #101930 [feature request] [minor] Inplace torch.flip_, #117762 PyEval_AcquireThread can crash process if called from another thread at interpreter shutdown, #154557 grouped_mm optional zero initialization of the output, #156701 Add generator argument to torch.rand_like, #156956 Allow clamp to work on absolute values and preserve sign",9a9586b7b3c44c43c3440970e77f400d3e222f3cdb65ddb0b834574ece341372
issue-collapse:e5b12949010bf63c,module: primtorch stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"81649, 81669, 81959, 83931, 84335, 84412, 84529, 84717, 85121, 85438, 101669, 102609, 102610, 107494",,,,"module: primtorch, module: tests, module: type promotion","better-engineering, module: primTorch, module: tests, module: type promotion, triaged",,Handle this module: primtorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#81649 Idiom for PrimTorch refs for Tensor methods, #81669 Register refs for CompositeImplicitAutograd ops as decompositions, #81959 Overloading multiple signatures for a single ref, #83931 primTorch: support refs and decompositions when ATen and Python disagree, #84335 Should enable skipped tests for `to` OpInfo, #84412 Odd type-casting behaviour in prims.div",b0f07529ce068de05534ebba7de8cc5910e9ee3a0e1ee698a77ede815268a598
issue-collapse:196e3a4f916a15ac,module: performance stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"39358, 39852, 40916, 41383, 44274, 44504, 45764, 47912, 50852, 51393, 52465, 52578, 53194, 53491",,,,"module: performance, module: pooling, module: sorting and selection, module: vectorization, oncall: distributed","NNC, enhancement, module: performance, module: pooling, module: sorting and selection, module: vectorization, oncall: distributed, triaged",,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39358 SyncBatchNorm doesn't scale well across multiple nodes on large data sizes, #39852 Shared file-system initialization in pytorch distributed is slow, #40916 Vectorized torch.eig(), #41383 Large overhead (7 microseconds) for PyTorch operation, #44274 Slower speeds when using half()., #44504 FP16 inference latency after sleeping",85f6788cc017738ef7377b07d580eaa68efa0164385cc46db762a4417fb11e37
issue-collapse:7b6c658c5e8e1b23,module: performance stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"19953, 28761, 37487, 41162, 50328, 57611, 65512, 66249, 66787, 74951, 78800, 82793, 85147, 92223",,,,"module: performance, module: proxytensor, module: serialization, module: sparse, module: tensor creation, module: tensoriterator","feature, has workaround, module: ProxyTensor, module: TensorIterator, module: performance, module: serialization, module: sparse, module: tensor creation, triaged",,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19953 Performance difference between 0.4.1 and 1.1.0, #28761 torch.tensor() is very slow when it is passed an h5py Dataset., #37487 Log-linear version of cumsum and cumprod, #41162 Advanced indexing gradient is extremely slow when there are many duplicate indices, #50328 RTX3090 performs no better than 1080ti, #57611 [perf] 10x improvement on element-wise operations with manual broadcast",e5f3964b6d4cc29e7cf59b2bcfe8fd692e591c1c455136d15429cbc8671a932f
issue-collapse:4d2b64112f56cad9,module: performance stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"82397, 101070, 104296, 120667, 121661, 124465, 132612, 134388, 137872, 138773, 142875, 148045, 151066, 158080",,,,"module: performance, module: python frontend, module: rocm, module: serialization, oncall: distributed, oncall: pt2","empathy-day, feature, matrix multiplication, module: performance, module: python frontend, module: rocm, module: serialization, oncall: distributed, oncall: pt2, triaged, upstream triton",,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#82397 Bilinear interpolation with antialiasing is slow in performance, #101070 Lazily format C++ stack trace if it is not used, #104296 affine_grid and grid_sample operators merge/accelleration, #120667 Triton kernel unexpectedly gets 1.35x slower by more specializaiton, #121661 Look at gather + gemv performance., #124465 Matmul with broadcast makes copy because of reshape to bmm, and it is slow",1da2ba41aa4ae6910c63a52e769867f29c01098776528c526680618bb7fc667d
issue-collapse:c01d82d9bead6eb3,module: optimizer stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2545, 2849, 2854, 5737, 8741, 20086, 20138, 20405, 28194, 39625, 40492, 40997, 41322, 41623",,,,module: optimizer,"enhancement, feature, module: optimizer, todo, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2545 Counter-intuitive Patience & Cooldown of ReduceLROnPlateau, #2849 Learning rate scheduler have different APIs, #2854 Add the new lr_scheduler which called poly, #5737 [feature request] Callback on learning rate drop in torch.optim.lr_scheduler.ReduceLROnPlateau, #8741 [Feature Request] Add to() method for optimizers/schedulers, #20086 CosineAnnealingLR giving unexpected learning rates on PyTorch 1.1.",cd6c3f5fadfc2d0efb3fd5ecd25e939b54063288119832efd35a1198abecde7a
issue-collapse:306b49e7dc56256d,module: optimizer stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6138, 7342, 36989, 38853, 41006, 42428, 43706, 46947, 47307, 47560, 48518, 49082, 49300, 49885",,,,module: optimizer,"enhancement, feature, module: optimizer, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6138 [feature request] SIGNUM an optimizer that takes the sign of gradient or momentum., #7342 [feature request] [PyTorch] More flexible optimizer API, #36989 pytorch latest update(1.4) broke CosineAnnealingWarmRestarts: T_cur is not define, #38853 Weight decay in AdamW, #41006 Simplify Adam Optimizer, #42428 KeyError: xxxxxxxxxx when calling optimizer.state_dict()",65eb7a13d973a0c0a1a224b5960ea1f6d6e1fce516dcbba810949e56b94d88f5
issue-collapse:1a2a9d06852570b9,module: optimizer stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7066, 9560, 9952, 14467, 46822, 50366, 50694, 51407, 51794, 52236, 52911, 56237, 56843, 58127",,,,"module: optimizer, module: serialization","awaiting response (this tag is deprecated), better-engineering, enhancement, function request, module: optimizer, module: serialization, todo, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7066 LBFGS contribution, #9560 Incorrect term in _LRScheduler., #9952 [feature request] Add COCOB Optimizer, #14467 Optimizer warning when parameters ""change"", #46822 C++ Optimizer: remove warning on Optimizer::size method, #50366 Optimizer Closure: Enable skip by returning None loss.",7644b74a4d4c48fa632c30d234b6a96cdd3533463a16ae6afb5b08e35107b4d6
issue-collapse:e526654fc442f921,module: optimizer stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18414, 34660, 38471, 41839, 49630, 50402, 70352, 81162, 86339, 97004, 101291, 101370, 104789, 111025",,,,"module: optimizer, module: performance, module: sparse","actionable, better-engineering, enhancement, feature, module: optimizer, module: performance, module: sparse, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18414 Add layer-wise adaptive rate scaling (LARS) optimizer, #34660 Add ""strict"" flag to ignore missing parameters in Optimizer.load_state_dict, #38471 Immediate mode API (with functional flavor) for optimizers, #41839 [Feature Request] Add `to`, `cpu`, and `cuda` method to optimizer, #49630 Bug in CosineAnnealingWarmRestarts, #50402 ValueError: optimizer got an empty parameter list",d002c03465a0cf214040aa9a420cd9612cf43839327ae78345a570aa99563d36
issue-collapse:377d6bcc64bef263,module: optimizer stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1329, 1489, 12146, 106121, 116052, 116592, 116707, 118370, 123238, 124133, 124523, 126638, 127189, 128208",,,,"module: optimizer, module: serialization, module: tests, oncall: distributed","actionable, feature, has workaround, module: optimizer, module: serialization, module: tests, oncall: distributed, todo, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1329 [proposed feature] Eve: Improving Stochastic Gradient Descent with Feedback, #1489 Optimizer should track parameter names and not id, #12146 Misleading step method in lr_scheduler.ReduceLROnPlateau, #106121 Got error when train models with more than one param_group in torch2.0, #116052 Add a capturable impl to ASGD single tensor, #116592 Support float64 step in fused optimizers",ef015dbcd555780437e2456633b23a412087ebbfccec3711a42f3793024167fe
issue-collapse:c8c7483660c391c9,module: optimizer stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"46809, 106221, 122550, 133268, 137387, 138248, 141216, 143836, 143947, 146157, 147288, 149269, 168967, 169170",,,,"module: optimizer, oncall: distributed checkpointing, oncall: pt2","actionable, enhancement, module: optimizer, oncall: distributed checkpointing, oncall: pt2, topic: docs, topic: performance, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#46809 Including AdaBound in the list of Optimizers., #106221 RProp improvement tracker, #122550 Adam optimizer：RuntimeError: t == DeviceType::CUDA INTERNAL ASSERT FAILED, #133268 Adafactor compile support tracker, #137387 Simple explanation should be added to `capturable` parameter of `optim.RMSprop()`, #138248 Adam optimizer without first moment estimate, for less vram",e47db90da94ab85e35e07a6ed56d1c7bfb2342f4ce8e4a26c3372bdcad6b0795
issue-collapse:52771e4404ac5d1c,module: onnx stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"44878, 70446, 94164, 103089, 105299, 107016, 108342, 119442, 121629, 122107, 122359, 124915, 124956, 124973",,,,"module: onnx, oncall: pt2, oncall: quantization","OSS contribution wanted, feature, module: onnx, oncall: pt2, oncall: quantization, onnx-triaged, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#44878 [ONNX] Support Tuple(List(Tensors),...) in input/ouput, #70446 Exporting the operator prim_DictConstruct to ONNX opset version 13 is not supported, #94164 Pytorch 2.0: Detection models from torchvision don't work with onnx and tensorrt backends, #103089 [OOM] Unable to convert 30B model to ONNX, using 4x A100's, #105299 Support ONNX opset 20 to export GELU to one single op, #107016 cov onnx error",e142e2ce81484e7b7fdcd6fa124ecf4898bbd2fed4a29d16c0ae3ea234b758de
issue-collapse:58bacdf397f0a39e,module: onnx stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"68944, 98727, 109292, 110131, 116965, 122321, 125835, 125893, 125902, 126065, 126549, 126660, 127518, 127606",,,,"module: onnx, module: windows","module: onnx, module: windows, onnx-triaged, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68944 Pytorch to ONNX model conversion, #98727 Pytorch member variable not working after converting to onnx format, #109292 aten::squeeze exported to ONNX as an `If` node, #110131 Some ONNX tests have been disabled because of new tensor.split signature, #116965 Support GridSample 5D, #122321 Batch size can't be varied if you export self attention to onnx",305d2148763ae52ba272333bed07fc6f818b2dcd9cbda5fa08f25a19c638c984
issue-collapse:7c59402ae77cc112,module: onnx stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"112831, 117209, 122957, 123353, 123555, 126856, 126972, 127875, 128325, 129200, 131349, 134664, 134827, 135398",,,,"module: onnx, module: optimizer, module: protobuf, module: third_party","OSS contribution wanted, low priority, module: onnx, module: optimizer, module: protobuf, module: third_party, onnx-needs-info, onnx-triaged, security, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#112831 dynamo_export successfully export model but fails at onnx.checker.check_model, #117209 torch.onnx.export the torch,nn.TransformerEncoderLayer, the dynamic_axes work not right, #122957 How to export torch.optim.LBFGS using torch.onnx.export, #123353 Type parameter (T) of Optype (Where) bound to different types (tensor(int64) and tensor(float) in node (/Where), #123555 torch.onnx.errors.UnsupportedOperatorError: Exporting the operator 'quantized::linear_dynamic' to ONNX opset version 17 is not supported. Please feel free to request support or submit a pull request on PyTorch GitHub: https://github.com/pytorch/pytorch/issues, #126856 TorchDynamo ONNX Export does not work as expected with masking (ScatterElements)",0c40cd629cb2ac4a369ac81a2c99baab1ae61e60416b3a64b08c834848f4d6ca
issue-collapse:48fb6819238fce94,module: onnx stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"81084, 110866, 116182, 123628, 130366, 130539, 130644, 136083, 136592, 136918, 137419, 140237, 143497, 147859",,,,"module: onnx, oncall: quantization","OSS contribution wanted, low priority, module: onnx, oncall: quantization, onnx-triaged, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#81084 [ONNX] Exporting the operator `::svd` to ONNX opset version 13 is not supported., #110866 Gradients (Jacobian) in inference, #116182 [ONNX] GFPGANv1.pth to onnx conversion error, #123628 `torch.onnx.export` doesn't correctly constfold constants and DequantLinear, #130366 Deadcode emitted when using `jit.script` with `onnx.export`., #130539 RuntimeException: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running Reshape node. Name:'/Reshape_5' Status Message: /onnxruntime_src/onnxruntime/core/providers/cpu/tensor/reshape_helper.h:45 onnxruntime::ReshapeHelper::ReshapeHelper(const onnxruntime::TensorShape&, onnxruntime::TensorShapeVector&, bool) input_shape_size == size was false. The input tensor cannot be reshaped to the requested shape. Input shape:{1,32,512}, requested shape:{1,1,8,64}",c52f143ef6c645c733d0749db179eb45f76d7544d37bb3f99561825336b9c2a0
issue-collapse:380fb34be3190a5d,module: onnx stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"66200, 110056, 113817, 124711, 130114, 143474, 147720, 148098, 150544, 151064, 151402, 151793, 152097, 156811",,,,module: onnx,"OSS contribution wanted, low priority, module: onnx, onnx-needs-info, onnx-triaged, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#66200 An exported onnx model can't reduce on dim with value of 0 if 'keepdims' is false, #110056 torch.onnx.export causes floating point exception with core dump for empty slice assignment, #113817 [ONNX] ONNX export of simple quantized model fails, #124711 Output Discrepancy between PyTorch Model and Converted ONNX Model, #130114 [feature request] [discussion] Baseline ONNX interpreter / executor in python / PyTorch, #143474 [ONNX] Failed to export PyTorch-2-Export-Quantized model to onnx",b20f3c1d78599712b27f38dcd9e461ae7dd07012a2512836eab8cddb15546412
issue-collapse:59caf6bd43cbb9ad,module: onnx stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"90607, 122290, 126774, 128323, 136622, 140130, 144464, 155997, 156873, 157220, 161365, 162449, 164116, 167063",,,,"module: onnx, oncall: quantization","OSS contribution wanted, actionable, feature, low priority, module: onnx, oncall: quantization, onnx-needs-info, onnx-triaged, release notes: onnx, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90607 Export to ONNX of `as_strided()` hard codes stride in the graph, although it should be dynamic, #122290 In a simple model, exporting onnx loses model operations, #126774 [ONNX] Implement aten._trilinear and aten.bilinear.default, #128323 UnsupportedOperatorError: Exporting the operator 'quantized::batch_norm2d' to ONNX opset version 14 is not supported., #136622 torch.onnx.errors.UnsupportedOperatorError: Exporting the operator 'aten::linalg_inv' to ONNX opset version 18 is not supported, #140130 [onnx] [njt] [feature request] Export NJT-enabled SDPA / MHA ops to ORT's PackingMode Attention",898e63305f0b2006d76de771b17e3281a4014662677883a262967c7a4b9fc259
issue-collapse:63afe41c17482933,module: nvfuser stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"79244, 79261, 79325, 80189, 80230, 80231, 80541, 80551, 80986, 81307, 81358, 81531, 83726, 84311",,,,"module: nvfuser, module: primtorch","module: nvfuser, module: primTorch, triaged",,Handle this module: nvfuser stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#79244 [NVFuser] bad performance on pyhpc_equation_of_state, #79261 [NVFuser] bad performance on pyhpc_isoneutral_mixing, #79325 [NVFuser] hitting fallbacks on demucs (from torchbench + lazy tensor), #80189 [NVFuser] Investigate models without any fusion groups found, #80230 [Prims+NVFuser] Aten2Prim refs tracking items, #80231 [Prims+NVFuser] Supports 0-sized inputs",98448b87927abbb173537d45f8246a2b5dfd0eaa977da32773771a87d7ed4945
issue-collapse:e521d5dfdcfdccee,module: nvfuser stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"80851, 83388, 85081, 85258, 85520, 85544, 86529, 86554, 86601, 88194, 88931, 88991, 89609, 94161",,,,"module: nvfuser, module: primtorch, module: tests, oncall: jit","feature, module: nvfuser, module: primTorch, module: tests, oncall: jit, triaged",,Handle this module: nvfuser stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#80851 [bug][nvfuser] Applying nvfuser to the model leads to runtime error, #83388 zero-numel tensor has ""RuntimeError: strides[cur - 1] == sizes[cur] * strides[cur] INTERNAL ASSERT FAILED"" in multi-thread., #85081 primTorch/nvfuser: have a way to check that refs are added to __all__, #85258 nvFuser support for {ceil,floor,round,trunc}(int), #85520 Implement `rand_like` ref and implement nvfuser_impl for `uniform` prim, #85544 NvFuser single mode changes the output",200fe6d29314ad570f26ed577acb1363a7b49647256daf713b7424b0c574e6fe
issue-collapse:1ba32b32ea55be31,module: numpy stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7610, 9186, 16703, 18315, 32589, 36309, 38323, 38328, 42671, 42805, 43064, 44968, 47841, 48338",,,,"module: numpy, module: pybind, module: tensor creation, module: type promotion, module: ux","enhancement, feature, function request, module: numpy, module: pybind, module: tensor creation, module: type promotion, module: ux, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7610 Inconsistent interactions of PyTorch tensors and NumPy ops, #9186 [pytorch] Make dtype second positional argument of tensor factory methods, #16703 split_with_sizes should accept a LongTensor as the split_sizes parameter, #18315 [FR] support default size (scalar) in torch.randint, #32589 User-friendly handling of types and devices, #36309 torch.ceil, torch.floor should accept a dtype argument",7cd5d622fb4084585b13747a16c92e2e1ab3741e012bb3d043c981f98b059a09
issue-collapse:0e9331c4a5253273,module: numpy stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"37809, 50333, 50360, 51406, 51849, 52334, 52552, 52722, 53090, 53284, 54116, 54620, 55286, 55356",,,,"module: numpy, module: tensor creation, module: testing, module: type promotion","enhancement, feature, function request, has workaround, module: numpy, module: tensor creation, module: testing, module: type promotion, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#37809 Add Numpy-like ""order"" argument to reshape, #50333 Function Request: scipy.stats.pearsonr, #50360 error: conversion from ‘std::vector<at::Tensor>’ to non-scalar type ‘at::Tensor’ requested, #51406 [numpy] `round` and `trunc` not supported for Integral Type while Python and NumPy supports them, #51849 .numpy() array failes to keep original storage around, #52334 Normal_like operator",60a55a98eaf0b764221058f1cc03b6ff6c718d3965f5bdbb3cc8b31cea4cc33e
issue-collapse:9354adb1983025a2,module: numpy stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"19143, 29372, 35499, 42053, 56671, 56774, 57018, 59247, 62026, 64107, 64502, 66489, 71204, 78413",,,,"module: numpy, module: primtorch, module: python array api, module: reductions, module: tensor creation, module: viewing and reshaping","enhancement, feature, function request, module: numpy, module: primTorch, module: python array api, module: reductions, module: tensor creation, module: viewing and reshaping, small, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19143 Support memoryview() method on torch.Tensor, #29372 torch.std() returns nan for single item tensors., #35499 Support tensor.cumsum() for 1-dim tensors, #42053 Function request: np.copy (alias of clone?), #56671 {h / v / d}split methods are missing, #56774 Support `expand_dims`",ff8ad3c8f864f833e9e75e115be4423465ffa3ecfde8eb5bd6cde999421ad80c
issue-collapse:aa7237fe050b97c2,module: numpy stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16330, 51199, 51280, 61816, 70555, 84539, 85606, 86315, 89419, 90412, 91447, 92778, 97760, 114679",,,,"module: numpy, module: python frontend, module: type promotion, module: ux","feature, module: numpy, module: python frontend, module: type promotion, module: ux, todo, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16330 support `unique_indices` option for `unique`, #51199 [numpy] torch.ceil and torch.floor don't support integer inputs while Numpy does, #51280 [ux] Proposal to have t() === transpose(-1, -2), since batches are very frequent, #61816 Better argument names for torch.atan2 and other math functions, #70555 [feature request] Exponential moving average (EMA) of a tensor across a dimension, #84539 list of tensors can't be converted to a torch tensor while list of lists gets easily converted to a pytorch tensor",8515cb47686aef64d1897c6adb6d6371eb86164efac3f029a098f6343bfc73fe
issue-collapse:ee83ac5d8362bd0c,module: numpy stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"39656, 42223, 53095, 67011, 94909, 109873, 120209, 121926, 127194, 129085, 129634, 132026, 135484, 136305",,,,"module: numpy, module: performance, module: power, module: python array api, module: python frontend, module: reductions, module: special","actionable, enhancement, feature, module: POWER, module: numpy, module: performance, module: python array api, module: python frontend, module: reductions, module: special, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39656 Broadcasting for torch.cross, #42223 Implementing packbits, #53095 Reset mask for torch.cumsum?, #67011 Importing numpy interacts with `tensor.sum` perf, #94909 [numpy] mean & nanmean should support int dtypes, #109873 Allow try except check for numpy bfloat16 representation",562e744978c29198b6aacc3afdb6d5cf45ef02e433fce08a2f115339afce0728
issue-collapse:dfb315c3617f2c8d,module: numpy stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31895, 36748, 43255, 51284, 55289, 61292, 90820, 110285, 135611, 137045, 137182, 139281, 141343, 142174",,,,"module: numpy, module: python frontend, module: safe resize, module: tests, oncall: pt2","enhancement, feature, function request, module: numpy, module: python frontend, module: safe resize, module: tests, oncall: pt2, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31895 torch.masked_select out argument can easily be misused, because output shape is dynamically computed, #36748 return_index option for torch.unique, #43255 Adding entropy function analogous to SciPy, #51284 result_type doesn't take dtypes and doesn't match numpy, #55289 Feature Request: Add a rounding mode to round, #61292 torch.linspace tensor support",bfe0be03296deec615317d758417f26e88f441a43229c7731d0b3c77abcd3f7a
issue-collapse:a5bb1267315d958f,module: numerical-stability stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38051, 50473, 53263, 53758, 56698, 57690, 65050, 66460, 78741, 95934, 101666, 121789, 141802, 165906",,,,"module: numerical-stability, module: numpy, module: optimizer, module: python frontend, module: rnn, module: testing, module: tf32, module: type promotion","module: numerical-stability, module: numpy, module: optimizer, module: python frontend, module: rnn, module: testing, module: tf32, module: type promotion, triaged",,Handle this module: numerical-stability stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38051 Div by zero error not triggered and inf not returned when dividing by 0 for some dtypes, #50473 Make use of eps for numerical stability more consistent, #53263 Different gradients in torch.matmul depending on input shape, #53758 ZeroDivisionError: float division by zero in Adam (bias_correction1 is zero), #56698 Convolution2D may have a numerical error., #57690 torch.cdist returns high diagonal values with CUDA",8e33eb459423f8fb28321154210964ee7aa47e8741cf6ba12db1a9aec0c85172
issue-collapse:466ba1dc55103284,module: nn stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28884, 31277, 34453, 34760, 41070, 43865, 46031, 49555, 49791, 127736, 138320, 139701, 139910, 140703",,,,"module: nn, module: performance","enhancement, feature, function request, module: nn, module: performance, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28884 Why attn_mask is not 3D tensor in nn.MultiheadAttention?, #31277 nn.MultiHeadAttention with different similarity measures, #34453 [Feature request] Query padding mask for nn.MultiheadAttention, #34760 incomplete implementation doubt, #41070 Cannot define v_dim in MultiheadAttention, #43865 Allow custom kwargs for forward method of nn.TransformerEncoderLayer, nn.TransformerDecoderLayer.",3e7a701efc91d625d33ea0ad3a109b915668720e66cf557db8a65a705003ce5a
issue-collapse:7366bd0fef16f33e,module: nn stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"60939, 64509, 66678, 67999, 69321, 80738, 82919, 90842, 91435, 92990, 95210, 96633, 96883, 97128",,,,"module: nn, module: performance","Stale, enhancement, feature, module: nn, module: performance, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#60939 Functional multi_head_attention_forward softmax get nan for fp16 mode, #64509 Return attention weights in nn.Transformer, #66678 Adding modern linear transformer variants to `transformer` module, #67999 Improved Transformer and MultiHeadAttention design, #69321 Poor Transformer inference scalability on CPU, #80738 Output for `aten::_native_multi_head_attention` appears inconsistent with entry in `native_functions.yaml`",92658f9237a03398c675c100525d2ee16c47aa961c4b1d78a4a94a74d32f698c
issue-collapse:5dedda2336d4be45,module: nn stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"98203, 98291, 99012, 100087, 102333, 103484, 103668, 104739, 105925, 106630, 109987, 110154, 114796, 115472",,,,"module: nn, module: onnx","module: nn, module: onnx, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#98203 AssertionError: was expecting embedding dimension of 22, but got 1320, #98291 Make BetterTransformer implementation non-blocking, #99012 [BUG]Float32 attention mask not working with torch.autocast(""cpu""), #100087 TransformerEncoderLayer behavior inconsistent between training and evaluation mode, #102333 transformer encoder-layer, the sample-Independent attn_mask(dim=3) has different behaviors when training and validating, #103484 No backward implementation for `torch._native_multi_head_attention`",8025616a3ec73b5d72d591fe1d64b32179c4035cc566b248fd8c2bf3e4966426
issue-collapse:9eeebc26f990db12,module: nn stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"116077, 116472, 118628, 118853, 119371, 120790, 120793, 123642, 124937, 126973, 131254, 135150, 137173, 137192",,,,"module: nn, module: performance","module: nn, module: performance, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#116077 TransformerEncoderLayer and TransformerDecoderLayer do not work, #116472 Using a 4D attention mask in MultiheadAttention forward, #118628 TransformerEncoderLayer fast path predicts NaN when provided attention bias, #118853 MultiheadAttention returns different results for binary mask and float mask, #119371 Nan output after masked TransforrmerDecoder, #120790 MultiheadAttention returns incorrect results for windowed long causal mask for long input",6fa01abfc06199c2c848e8fb7bc97405f8e57fcadf24705e6c773006a2a74bb8
issue-collapse:89378228f80cb057,module: nn stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"115024, 128412, 133833, 138420, 139191, 139381, 140881, 141018, 141801, 141828, 142301, 143293, 143424, 143669",,,,"module: nn, module: python frontend, module: sdpa, module: typing","feature, module: nn, module: python frontend, module: sdpa, module: typing, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#115024 InstanceNorm1d throws bad warning, #128412 A Permute Layer for torch.nn.Sequential, #133833 Unexpected behavior of `CrossEntropyLoss()`, #138420 batch_first support for multi_head_attention_forward, #139191 When I use register_full_backward_hook and register_forward_hook in nn.MultiheadAttention, The MultiheadAttention executed wrong conditional branch and get wrong backward grads., #139381 Improved definition for Hardsigmoid",a09261bab208ca7df158d509850a5fbc887a35a79ec7f0dc9c1b9ffe4fc9bd63
issue-collapse:580f439c10de89eb,module: nn stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"86274, 132136, 139582, 143754, 143860, 144116, 144223, 145468, 145851, 146034, 146708, 146711, 146831, 147389",,,,"module: nn, module: norms and normalization, module: performance, module: rnn, module: sdpa, module: typing, oncall: pt2","feature, module: nn, module: norms and normalization, module: performance, module: rnn, module: sdpa, module: typing, oncall: pt2, topic: docs, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#86274 TransformerEncoder/TransformerDecoder has same initial parameters for all layers, #132136 `torch.nn.transformer.forward` returns incorrect value inside `torch.no_grad()` blocks., #139582 nn.LSTM documentation, #143754 nn.LayerNorm is slower than naive implementation when dimension is low, #143860 _transform_bias_rescale_qkv cpu op get error on debug build, #144116 RNN batch_first argument only works on the input not h_0 when if should work on both",c8daaa368c0f037b2b5162659cdee1aa9ead61f3d1dfd12604424fa56718dca7
issue-collapse:549143c1dab45185,module: nn stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"37386, 99304, 102374, 107573, 119482, 147495, 147506, 148232, 149040, 149120, 149841, 150402, 152152, 152188",,,,"module: nn, module: padding, module: python frontend, module: rnn, module: typing","actionable, feature, module: nn, module: padding, module: python frontend, module: rnn, module: typing, topic: new features, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#37386 Add BufferDict container, #99304 Need TransformerEncoder to output attention map, #102374 Hooks not working in version 2.0.1+cu118, #107573 caching keys+values in TransformerDecoderLayer for faster inference, #119482 `pylint`: `torch.nn.functional.linear` is not callable, #147495 Native Node Sleep/Wake Functionality for Neural Network Layers",e6ec431d1cba91f1f4c38ae5195a4249e5bb7246a2b3690db3095df80615b791
issue-collapse:b4a0d801d686ae36,module: nn stale/low urgency  batch 26,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6442, 10006, 20967, 30413, 30440, 30535, 30589, 30899, 30947, 30962, 31110, 31367, 31651, 32264",,,,"module: nn, module: rnn, module: sparse","enhancement, module: nn, module: rnn, module: sparse, todo, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6442 [feature request] dropout1d, #10006 RNN gradients in eval mode in pytorch 0.4, #20967 Track running stats regardless of track_running_stats=False, #30413 nn.functional should maintain API parity with nn where possible, #30440 Redundant counter in batchnorm impl, #30535 [FR] nn.init.* accept None as input tensor",d5cddc5e02210924afb34317e92765a340e61451ab3c4d2c402f77c6e9680940
issue-collapse:c8de5e4d801466ae,module: nn stale/low urgency  batch 25,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1686, 1932, 4927, 9046, 28746, 33241, 33360, 33436, 34765, 36317, 38000, 38184, 39524, 40763",,,,"module: nn, module: numpy, module: rnn, module: tests","enhancement, feature, low priority, module: nn, module: numpy, module: rnn, module: tests, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1686 Feature Request: load_state_dict should take filenames, #1932 Feature Request: ReLU on LSTMs and GRUs, #4927 [Feature Request]would PackedSequence support unsorted sequences?, #9046 [feature request] freeze() for nn.Module, #28746 super().__init__() not called in torch.nn.Module.__init__, #33241 [feature request] Add ""groups"" argument to nn.Fold and nn.Unfold",a77d1f89b7ca4d43e63f7bf630f58d2a80d5ac89bd463ac9d20d9ca2b184030d
issue-collapse:7553d36db7e792cd,module: nn stale/low urgency  batch 24,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2400, 5435, 17313, 18728, 26527, 27839, 30658, 31856, 34874, 35775, 38955, 41528, 42625, 43402",,,,"module: nn, module: numerical-stability, module: pooling, module: vision, oncall: jit","enhancement, feature, function request, module: nn, module: numerical-stability, module: pooling, module: vision, oncall: jit, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2400 Add CRF Layer, #5435 Gaussian Sampling, #17313 Implement Adaptive Input Representations for Neural Language Modeling, #18728 MaxPool with n-dimensional tensors, #26527 Multilinear map, #27839 RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 1. Got 24 and 195 in dimension 0 at /opt/conda/conda-bld/pytorch_1565272271120/work/aten/src/TH/generic/THTensor.cpp:689",1ab2283d59963416a163dac8841d395923cd8c1ea2e23afaca3a50907f18fe8d
issue-collapse:9aae5c1e39d1d5ad,module: nn stale/low urgency  batch 23,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7773, 7865, 9681, 21411, 38394, 46149, 46971, 47588, 47699, 48235, 48511, 48526, 48645, 49723",,,,"module: nn, module: norms and normalization, module: optimizer, module: pickle, module: rnn","enhancement, feature, function request, low priority, module: nn, module: norms and normalization, module: optimizer, module: pickle, module: rnn, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7773 [feature request] Add Local Contrast Normalization, #7865 [PyTorch] weight tensor dimension assumption, #9681 [feature request] Support for 0-length sequences in packed_sequences, #21411 Implementation of Group equivariant convolutions, #38394 4D tensor support for torch.nn.functionnal.fold() (col2im), #46149 seeding does not work when I initialize a linear model even if I do not use it in the code",43cfb468bb2aea1ef4f87c1192db8b0ee60565183abb11bf97d5b70b2fe9656b
issue-collapse:ca48bcaad4d6485b,module: nn stale/low urgency  batch 22,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5790, 11624, 12159, 12460, 12675, 20058, 39674, 41256, 43876, 47175, 51320, 52011, 55030, 55804",,,,"module: nn, module: norms and normalization, module: optimizer, module: performance, module: tests, module: ux","enhancement, feature, function request, module: nn, module: norms and normalization, module: optimizer, module: performance, module: tests, module: ux, todo, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5790 Add hookable weights, #11624 Add min mode to embedding bags, #12159 clip_grad_norm_ does not work on grads of different types, #12460 input_device, output_device, devices_used properties, #12675 [feature request] ignore_index and size_average in nn.AdaptiveLogSoftMaxWithLoss, #20058 Implement noise_shape keyword for Dropout layers",4628d09d2bcfefdd9816d89c6d0221b7e432a81fd63780a90a9dd6ca77e546c5
issue-collapse:ff81e13cedd08fd7,module: nn stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5912, 9678, 13162, 20424, 41099, 57406, 57631, 58093, 59925, 61149, 61452, 65163, 65520, 65622",,,,"module: nn, module: norms and normalization, module: padding, module: performance, module: pooling, module: pruning","enhancement, feature, module: nn, module: norms and normalization, module: padding, module: performance, module: pooling, module: pruning, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5912 Conv-RNN combination slow in backward pass, #9678 Remove BatchNorm layers once the training is completed., #13162 circular module reference raises RecursionError, #20424 pos_weight argument in torch.nn.BCELoss, #41099 max_pool2d always compute indices even when it's not required, #57406 nn.Module custom setattr leads very confusing behaviors",1de4eca4a47d50f2a344fb3f07c62fd894f6a96fa7a710221c0b9f2c0d9b1bfe
issue-collapse:605b73d96809fc48,module: nn stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18796, 63004, 65688, 66874, 67481, 68179, 68337, 68420, 68512, 68655, 68752, 69207, 69721, 70995",,,,"module: nn, module: norms and normalization, module: padding, module: performance, module: pooling, module: rnn, module: tests, module: typing, module: ux","Stale, actionable, enhancement, feature, module: nn, module: norms and normalization, module: padding, module: performance, module: pooling, module: rnn, module: tests, module: typing, module: ux, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18796 [Feature Request] Flattened indices option for max pooling, #63004 ignore_index does not work as described with nn.CrossEntropyLoss, #65688 Parametrization silently disables RNN weight updates, #66874 Error using _stateless version of Module, #67481 Batchnorm2D does behaves differently with different shapes when batch_size=1, #68179 torch.nn.Linear wrong type annotation for bias",9537f49154307554c674006a0e6935133491b0b010a4b18e6475b9a47b9da950
issue-collapse:9daaca12896ab277,module: nn stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1410, 58846, 72643, 72821, 121160, 148065, 152607, 154241, 155193, 156177, 158832, 159138, 161869, 163821",,,,"module: nn, module: norms and normalization, module: python frontend, module: rnn","enhancement, feature, module: nn, module: norms and normalization, module: python frontend, module: rnn, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1410 BatchNorm should use Bessel's correction consistently, #58846 Simplification of pruned models, #72643 upstream `apex.normalization.FusedRMSNorm`, #72821 `{Batch,Instance}Norm{1,2,3}d` works when `num_features != C`!, #121160 RFCS review request: fast Viterbi Decoding, #148065 Support alpha=inf consistently for torch.celu",e7f71b5ed76b047304796131031f0642b4f0a5b70d03b61cdbdef3702008d070
issue-collapse:725287d1f4689e0e,module: nn stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"65774, 69325, 69756, 70095, 70100, 70455, 70500, 71211, 71877, 72452, 72580, 72672, 73222, 73484",,,,"module: nn, module: nn.utils.parametrize, module: norms and normalization, module: random, module: rnn, module: serialization, module: tests","Stale, actionable, enhancement, feature, function request, low priority, module: nn, module: nn.utils.parametrize, module: norms and normalization, module: random, module: rnn, module: serialization, module: tests, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#65774 Non-persistent modules, #69325 `nn.functional.fractional_max_pool2d` and `nn.functional.fractional_max_pool3d` produce incorrect output on non-contiguous inputs, #69756 Introduction of nn.LazyRNNs, #70095 Add support for a `default` arg in `ModuleDict.pop`, #70100 Feature Request: dim parameter of torch.nn.functional.normalize should accept tuples, #70455 Make it possible to remove all hooks on a specified module without needing the hook handles",263f6e4951fa8bd5562c6a11ee09e66e753945581e4876f7a5955bc15b6ace33
issue-collapse:0c266d9353b28267,module: nn stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"285, 67565, 70348, 74143, 74296, 74411, 74554, 75251, 75680, 75923, 76011, 76636, 76906, 77067",,,,"module: nn, module: norms and normalization, module: padding, module: python frontend, module: rnn, module: tests, module: type promotion, oncall: jit","Stale, better-engineering, enhancement, feature, has workaround, low priority, module: nn, module: norms and normalization, module: padding, module: python frontend, module: rnn, module: tests, module: type promotion, oncall: jit, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#285 Keyword arguments passed to module's __call__ aren't forwarded to the hooks, #67565 Adding a new kwarg to a torch.nn.functional function breaks FC for JIT, #70348 nan return by nn.CrossEntropyLoss when all the labels are ignore_index in torch 1.11, #74143 Tracing model parameter shapes without instantiating the model parameters, #74296 Time Gated Lstm, #74411 UserWarning: size_average and reduce args will be deprecated, please use reduction='mean' instead.",04aefea0a7ec448161ca89ec49e706d32e11ed9a7be518fbefb55389d1971025
issue-collapse:88a1378f837d326b,module: nn stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8921, 35798, 48626, 62545, 76920, 78450, 78744, 78786, 80458, 80932, 80951, 80954, 83152, 83169",,,,"module: nn, module: pooling, module: primtorch, module: sparse, module: testing","actionable, enhancement, feature, module: nn, module: pooling, module: primTorch, module: sparse, module: testing, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8921 [feature request] More methods for PackedSequence, #35798 Dropout on sparse tensors, #48626 Missing super().__init__() call in nn.Module, #62545 Incosistency with args for `nn.functional.max_poolNd` vs `nn.MaxPoolNd` functions, #76920 Avoid Self-loops on Module Creation, #78450 Softmax, LogSoftmax are over parameterized",ed190d697df6165f4437449d5b5b2a069d80667c26eecc518d29ca7802b74b74
issue-collapse:101d90b8e56e13da,module: nn stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1927, 26781, 31818, 53103, 76613, 83032, 83221, 83229, 83232, 83238, 86356, 89344, 89597, 95290",,,,"module: nn, module: norms and normalization, module: pooling, module: sparse, module: typing","actionable, enhancement, feature, module: nn, module: norms and normalization, module: pooling, module: sparse, module: typing, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1927 [feature request] time-distributed layers for application of normal layers to sequence data, #26781 Matrix corresponding to convolution by a 2D kernel (convmtx2), #31818 How to distinguish different layers in hook？, #53103 ModuleList not recognised as reversible by mypy, #76613 MeanVarianceNormalization, #83032 Support for CSR Tensor with NN layers",400d84f0d68668fcd4ce961b2e6b9b7c57ef983c409623df92087ef3f6eb5adb
issue-collapse:4356399115c85035,module: nn stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1362, 18220, 25092, 34772, 37092, 55457, 67913, 72283, 97915, 103316, 104698, 104788, 104823, 106545",,,,"module: nn, module: pooling, module: tests, module: vision","feature, has workaround, module: nn, module: pooling, module: tests, module: vision, tracker, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1362 Feature Request: noise contrastive estimation/negative sampling, #18220 Context Manager that disables training mode with in a nn.Module., #25092 No way to correctly reset weights of a model with spectral norm, #34772 Unsupported ONNX op (Upsample 3D, bicubic) contrary to documentation, #37092 nn.Bilinear cannot be used inside nn.Sequential, #55457 [feature request] TupleSequential module or an option for Sequential for unpacking tuples (i.e. support varargs)",1f231dd40a0db8973f4000b7ef0fa1e87c922b2edd1b1db48ec53e812b16b3a1
issue-collapse:a4c6083bb419d16f,module: nn stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31423, 45352, 106011, 109586, 110356, 110959, 111368, 114093, 114107, 114131, 115064, 116502, 116861, 117437",,,,"module: nn, module: optimizer, module: python frontend, module: random, module: rnn, module: serialization, module: tests","actionable, better-engineering, enhancement, feature, module: nn, module: optimizer, module: python frontend, module: random, module: rnn, module: serialization, module: tests, topic: docs, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31423 Add option in LSTM layer to access all cell states of all time steps, #45352 torch.nn.functional.one_hot should gracefully skip negative and out-of-range indices, #106011 Misleading error message in multilabel_margin_loss when passing incompatible tensor dimensions, #109586 Max pool with negative integer inputs and channels_last memory layout gives the wrong values, #110356 Dropout signature inconsistent between `torch.dropout`, `torch.nn.Dropout` and `torch.nn.functional.dropout`, #110959 `model.named_buffers()` fails if module not hashable.",43b87ff28c56af3ea09362daca3e2f13c5c19140d77dc8f97624976bc7e44d15
issue-collapse:ac679586d728a40f,module: nn stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7733, 24089, 37554, 102999, 117275, 118294, 119368, 120396, 122803, 124194, 125116, 126000, 126755, 127701",,,,"module: nn, module: nn.utils.parametrize, module: rnn, module: rpc, module: testing, oncall: distributed","enhancement, module: nn, module: nn.utils.parametrize, module: rnn, module: rpc, module: testing, oncall: distributed, todo, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7733 Inserting a tensor into a python dict causes strange behavior, #24089 torch.nn.functional.grid_sample with 'circular' border conditions, #37554 Add name to Class Parameter(), #102999 Preserve weight_g/weight_v accessors on new weight_norm, #117275 deepcopy of `nn.Parameter` does not transfer __dict__, #118294 Failures in test_nn depending on test order",167d20cf9320790154c0d681aae882778a74efe75e02737cb70fee7f4ddcbdb8
issue-collapse:915f3c9bc1feae28,module: nn stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40859, 53785, 71203, 116608, 127535, 127732, 127733, 127734, 128165, 128493, 129564, 130342, 131110, 132034",,,,"module: nn, module: python frontend, module: rnn, module: serialization, module: xla","enhancement, module: nn, module: python frontend, module: rnn, module: serialization, module: xla, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40859 Option to allow loading state dict with mismatching shapes., #53785 Support for one-hot of dtypes besides torch.int64, #71203 Keys of a `ModuleDict` cannot have the same name as existing `ModuleDict` class attributes., #116608 unstable test_GroupNorm_general_cpu in test/test_nn.py, #127535 Can we have a Y-Split module for torch.nn.Sequential?, #127732 torch.nn.ConvTranspose2d/torch.nn.ConvTranspose1d lacks a check for the parameter dilation in the model building process",ea4d20e5685c85b02548b840a72204ed8ce9ca07dc2f9aa4ad533be6ecd348dd
issue-collapse:ad455da9c2c7f0e5,module: nn stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55910, 119474, 125389, 128779, 131307, 131688, 132231, 133522, 133527, 133535, 133818, 134948, 136071, 136072",,,,"module: nn, module: rnn","enhancement, feature, module: nn, module: rnn, topic: docs, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55910 Allow variable intermediate hidden dimensions for stacked RNN/LSTM layers, #119474 Extend functionality of torch.nn.functional.fold to apply to sequences (1D tensors + batch & channels) not just images (2D tensors + batch & channels), #125389 Support for Kolmogorov-Arnold Networks (KANs), #128779 The name of the function `nn.L1Loss()` should be `nn.MAE()` or the name of the function `MSELoss()` should be `nn.L2Loss()`, #131307 allow tuples as keys for ModuleDict, #131688 Drop Connect Layer",e51d0710809d87f717ef5bdeadd39c3cbe34b7bb11775320bfe66b5c1470b36c
issue-collapse:019dfe66bc566cc6,module: nn stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"117546, 136073, 136096, 136098, 136110, 136138, 136409, 136926, 136937, 136938, 136955, 136994, 137086, 137089",,,,"module: nn, module: optimizer, module: rnn","enhancement, module: nn, module: optimizer, module: rnn, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#117546 interpolate::trilinear has different result between 1.8.0 and the newer version, #136073 The real AttributeError information, #136096 Each parameter of `nn.MaxPool1d()`, `nn.MaxPool2d()` and `nn.MaxPool3d()` should have `required` or `optional`, #136098 `int`, `float` and `complex` type with `return_indices` of `nn.MaxPool1d()` also work, #136110 Setting `True` to `return_indices` argument of `nn.MaxPool1d()` with the tensor of `int` works, #136138 `num_features` of `nn.BatchNorm1d()` with a tuple of `int` works",4fe9c494e817e48fec77dfcce49b6c41a09c098be6e968cd6f7eedd4df2f00a7
issue-collapse:fcc45ff49e44ecdd,module: nn stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88810, 105448, 120587, 122250, 152308, 156062, 165312, 165982, 166080, 166378, 168010, 170382, 170691, 170851",,,,"module: nn, module: sdpa, module: typing","actionable, enhancement, module: nn, module: sdpa, module: typing, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88810 Return the attention weights using the Transformer Encoder class., #105448 Will nn.unfold support non-4D-tensor input in future version?, #120587 Type hints on nn.Module.__setattr__ value, #122250 Proposal to make `torch.nn.modules.utils._pair` public, #152308 bizarre behavior with torch module's Attribute Error, #156062 torch.nn.functional.conv_transpose3d return inconsistent results when weight containing inf between CPU and GPU",c09e74881c4e925c486e542428bbbfd4901bcb141fdcaa059c918cacc3d7b335
issue-collapse:4d2af97cfe6e6dc3,module: nestedtensor stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"82534, 95485, 95562, 95815, 104163, 108859, 110516, 118580, 125940, 126453, 129191, 131967, 133138, 135930",,,,"module: nestedtensor, module: serialization, module: sparse","actionable, enhancement, has workaround, module: nestedtensor, module: serialization, module: sparse, triaged",,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#82534 Use NestedTensor in RNN models, #95485 Enrich shape operations with nested tensors, #95562 test_layer_norm_backward and test_layer_norm_backward_5d run OOM in slow gradcheck, #95815 Let Nested Tensor Metadata be cached on GPU, #104163 Nested Tensor with PyG dataset custom class, #108859 Certain torch functions are not handled by torch func wrapper",5fc2a9535905a284ea6f231a4f4ca1b0f9a71f208215a7b9ef48af7550158961
issue-collapse:36b01a34f77e0fac,module: nestedtensor stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"78754, 90866, 97111, 122701, 124216, 130470, 136656, 136747, 137167, 137336, 138180, 138552, 138993, 145214",,,,"module: nestedtensor, module: nn, module: sdpa, module: tensor creation","actionable, feature, low priority, module: nestedtensor, module: nn, module: sdpa, module: tensor creation, triaged",,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#78754 Getting NotImplementedError when trying to implement E2E support for `prim::is_nested` Op in torch-mlir., #90866 Support arbitrary masks for _nested_tensor_from_mask in nn.TransformerEncoder, #97111 TransformerEncoder truncates output when some token positions are masked by `src_key_padding_mask` across batch, #122701 Unable to call len on nested tensor, #124216 [NestedTensor] NJT with lengths set fails torch.compile(), #130470 nested_tensor jagged_from_list doesn't correctly handle corner case of empty list",ac1428d36ad5ff87808bf24a8c89f270daaf374621e8a165e4ca0b03c8f58e0b
issue-collapse:6e30771f4b4240ca,module: nestedtensor stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"81405, 132695, 137514, 138831, 145158, 145837, 148379, 149720, 150242, 151717, 152962, 153238, 153472, 161287",,,,"module: nestedtensor, module: performance, module: tests","actionable, module: nestedtensor, module: performance, module: tests, triaged",,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#81405 Implement shape/size functions for nestedtensor, #132695 Backward on nested jagged tensor vector multiplication throws error on 0 dim (but not last dim), #137514 Support multiple ragged dims for NJT, #138831 Add mean and var operation for Nested Tensors, #145158 Pytorch matmul for nested 4D tensors in jagged layout doesn't work, #145837 torch.nested.narrow() or torch.nested.to_padded_tensor() breaks backwards pass - invalid gradient",f548c845dd37322a30103511091278f985554635e22844beffb22d60a83c4303
issue-collapse:2689ab9ba97c9a31,module: nccl stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8386, 42332, 67158, 113128, 115301, 118472, 130739, 132747, 135095, 139245, 146021, 146118, 149984, 159559",,,,"module: nccl, oncall: distributed","enhancement, feature, module: nccl, oncall: distributed, triaged",,Handle this module: nccl stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8386 Properly release NCCL resources, #42332 RuntimeError: ProcessGroupNCCL does not support recv (torch-nightly), #67158 Make streams used for NCCL operations configurable, #113128 NCCL watchdog thread terminated with exception, #115301 inconsistency for enableTiming / FlightRecorder / duration_ms, #118472 [c10d] Avoid busy looping in C10D_NCCL_CHECK_TIMEOUT",045281ab95c4df7b2bf51e91a42fefadf1d2189e36dd1036db3ce6ae0048cd19
issue-collapse:59a1147a518fa0f0,module: nans and infs stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"19126, 51906, 51911, 64525, 65663, 71475, 78484, 102400, 106604, 132614, 135992, 138570, 161500, 165537",,,,"module: nans and infs, module: nn, module: norms and normalization, module: python frontend, module: special, module: testing","low priority, module: NaNs and Infs, module: nn, module: norms and normalization, module: python frontend, module: special, module: testing, topic: fuzzer, triaged",,Handle this module: nans and infs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#19126 weight_norm doesn't support eta and returns nan for zero weights, #51906 torch.nn.functional.binary_cross_entropy(_with_logits) outputs NaN, #51911 torch.nn.functional.grid_sample outputs NaN, #64525 nn.TransformerEncoder - all nan values issues when src_key_padding_mask provided, #65663 Model Evaluation Returns Nan Values Sometimes for the Same Input, #71475 The signature of `torch.nanmedian` in the doc is wrong",0c5812c82d1cb416e1f8fb88c87ea0758255689b74bf06de8fc4293960d1f2f2
issue-collapse:391f72bb9ea43a11,module: named tensor stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28742, 37995, 41667, 49427, 61117, 61136, 63297, 64743, 67168, 70735, 79091, 83902, 119464, 121725",,,,"module: named tensor, module: python frontend, module: tensor creation, module: tests","actionable, enhancement, module: named tensor, module: python frontend, module: tensor creation, module: tests, triaged",,Handle this module: named tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28742 torch.as_tensor fails to create named tensors, #37995 unsqueeze support for named tensors, #41667 [Feature Request] named tensor support for `tensordot`(tensor contract), #49427 Inserting named tensor into other fails, #61117 flatten renames in place, #61136 rename(**{}) should throw no error",a8675457064697e3fd7ffee36c596688fa5a7af03e893018a3958aa8993aa452
issue-collapse:4d5a3c5af09e9b6c,module: multithreading stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20189, 25601, 33327, 33367, 38510, 42094, 64864, 65229, 70940, 71383, 76966, 110032, 138468, 140012",,,,"module: multithreading, module: openmp, module: performance, module: rpc, module: sanitizers, oncall: jit","better-engineering, enhancement, module: multithreading, module: openmp, module: performance, module: rpc, module: sanitizers, oncall: jit, triaged",,Handle this module: multithreading stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20189 Unify Caffe2 and PyTorch OpenMP initialization, #25601 Avoid non-POD data in thread_local, #33327 Creating Torch tensors slows OpenCV video reading a lot, #33367 Expose `internal::GRAIN_SIZE` through Python API., #38510 [DISCUSSION] RPC server-side ThreadLocalState, #42094 Setting threads number to the number of default by torch.set_num_threads is faster than not setting it",7cc14f3c2cff2b33c888f952476a2e383911b9f562a5beb500cc8ff761777859
issue-collapse:b988b80243d232f7,module: multiprocessing stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11786, 13300, 14047, 16954, 31571, 32322, 36922, 42444, 49672, 59546, 68935, 77050, 78075, 80765",,,,"module: multiprocessing, module: performance, module: power, module: random, module: serialization, module: tensorboard, module: xnnpack","feature, module: POWER, module: multiprocessing, module: performance, module: random, module: serialization, module: tensorboard, module: xnnpack, small, triaged",,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11786 Mysterious error due to num_workers: 1, #13300 Use a dill-based multiprocessing library and serialization, #14047 Make torch.multiprocessing.SpawnContext usable, #16954 torch.multiprocessing.pool.Pool broken, #31571 Bug in ForkingPickler for multiprocessing spawn context for shared storages on Linux, #32322 deadlock when using mp.spawn multiprocessing",667a9b30c8491c687296077c1f32575efa9a8fa488cd40cb34bb47987cc4977f
issue-collapse:63ae7c48674d57e3,module: multiprocessing stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8126, 10996, 35472, 60401, 65418, 86467, 87056, 91889, 97902, 103397, 106989, 120354, 121430, 122186",,,,module: multiprocessing,"feature, module: multiprocessing, todo, triaged",,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8126 PyTorch multiprocessing using single CPU core, #10996 Multiprocess Deadlock when using np.transpose and torch.stack, #35472 Multiprocessing map gets stuck if doing inference on loaded model, #60401 Need workaround to support multiprocess CUDA tensor sharing on Jetson Platforms, #65418 Sending tensors from short-lived multiprocessing process fails on Linux, #86467 torch.tensor obj automatically moved to shared memory upon Process launch",dd534f21efd856cd0f33f8e0425716b0c78d48b921721f17ce39c44b131ae655
issue-collapse:8f9708f556e3a53b,module: multiprocessing stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20248, 68861, 75662, 84994, 101293, 122626, 128351, 129656, 133010, 133210, 134755, 134926, 135705, 137777",,,,"module: multiprocessing, module: privateuse1, module: python frontend, module: serialization, module: sparse, module: third_party","Stale, feature, module: PrivateUse1, module: multiprocessing, module: python frontend, module: serialization, module: sparse, module: third_party, triaged",,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20248 Sparse tensors can't be used in DataLoader running many workers, #68861 [In Gunicorn & multiprocessing environment] Cannot re-initialize CUDA in forked subprocess, #75662 multiprocessing and torch.tensor, Cannot allocate memory error, #84994 Cuda tensor is zero when passed through multiprocessing queue, #101293 Operations to shared tensors in the forked process could lead to silent crash, #122626 Significant performance degradation with multiprocessing in PyTorch 2.x compared to 1.13.1",9f6f1d98a457649ae1f98a0dc0e37423aef613e7aac377bf2d6bc8a4a039e05b
issue-collapse:848859d555b5b8ca,module: mps stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"78210, 88893, 92148, 95622, 99160, 101251, 104188, 104315, 104761, 106408, 112849, 115152, 115350, 118517",,,,"module: mps, module: multiprocessing, module: optimizer, module: performance, module: random, module: scatter & gather ops, module: sparse","actionable, enhancement, module: mps, module: multiprocessing, module: optimizer, module: performance, module: random, module: scatter & gather ops, module: sparse, triaged",,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#78210 Performance with MPS on AMD GPUs are worse than CPU, #88893 RuntimeError: derivative for aten::mps_max_pool2d_backward is not implemented, #92148 Occassional OverflowError with mps running yolov7, #95622 Testing InvokeAI 2.3.1.post1, using mps, with PyTorch nightly dev20230226 yields RuntimeError cross-device copies are not allowed!), #99160 Run ChatRWKV on MBP(intel CPU)+eGPU[rx6800 16G], returna a very big number -9223372036854775808, looks like overflow, #101251 Mac m2 MPSNDArray.mm:78: failed assertion `[MPSNDArrayDescriptor sliceDimension:withSubrange:] error: dimension index (2) not within number of dimensions (2) Dimension indices are 0-based'",f38e41bf2881c05a4db4a119bda6735941d8c81f24c4702ece0980819a73c199
issue-collapse:92304f0afb02fc03,module: mps stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"87129, 109943, 112956, 116857, 121632, 122973, 124056, 125098, 127676, 127743, 138199, 138898, 141789, 143955",,,,"module: mps, module: nestedtensor, module: nn, module: performance, module: sparse","actionable, enhancement, feature, module: mps, module: nestedtensor, module: nn, module: performance, module: sparse, triaged",,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#87129 [MPS] sum on a size=1 dim is ~5x slower than squeeze, #109943 Problems when loading PT files und Linux - Duda which are created under Mac Apple Silicon MPS, #112956 Optim.step() is significantly SLOW on MPS, #116857 Operator 'aten::_weight_norm_interface' not implemented MPS device., #121632 [BUG] Given boolean tgt_mask, TransformerDecoder produces wrong results with MPS backend, #122973 mps training is slower than cpu",c2cc4fb3ce6afc806325698ffc528974077a5c93265e9c7db3a607183f4d8286
issue-collapse:8670f8ee35bcbd1e,module: mps stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"86060, 111374, 117949, 119108, 119968, 143864, 148219, 149184, 152736, 153976, 156792, 161283, 169290, 170175",,,,"module: mps, module: performance, module: pooling, module: tests, oncall: pt2","actionable, enhancement, module: mps, module: performance, module: pooling, module: tests, oncall: pt2, triaged",,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#86060 [MPS] Pin_memory on MPS backend raises exception of setting wrong device storage, #111374 Tensor.lerp inconsistent when using -Infinity between MPS and CPU, #117949 aten::slow_conv3d_forward still missing from MPS, #119108 [MPS] Tracking issue for ModuleInfo failures when enabling testing for `torch.float16`, #119968 Error on trying to run detectron2 on Mac M1 with device set to ""mps"" (Metal Performance Shaders), #143864 Adaptive pool MPS",aaeb0298ed2a4638855e60aff435361ca408b28182958f4442026391ed674046
issue-collapse:3aeaea7d55cd5820,module: meta tensors stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72650, 73065, 74329, 79848, 80067, 80921, 89301, 99414, 105105, 108671, 108870, 127754, 140602, 147643",,,,"module: meta tensors, module: numpy, module: primtorch, module: python frontend, module: tests","feature, module: meta tensors, module: numpy, module: primTorch, module: python frontend, module: tests, release notes: Meta API, triaged",,Handle this module: meta tensors stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72650 Some inplace ops don't raise on incompatible shapes and meta device, #73065 `index_fill_` accepts wrong dtype for meta tensors, #74329 CompositeImplicitAutograd addr implementation does not have correct behavior for bool (meta gives wrong dtype), #79848 Automatically calculate output_shape of sequential model (or any other fCNN), #80067 FakeTensor: Support torch.tensor([FakeTensor, 0]), #80921 [primTorch] `|` operator does not work with FakeTensor in _refs",c9ed1732d4fcebee9c298cd6a090bb9b5c3a5d92f4d7c8fa8f8be9195e2fedaa
issue-collapse:435d82d7a98a7b45,module: memory usage stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23382, 31655, 40400, 44182, 46084, 48377, 49139, 49394, 50835, 54713, 55260, 58980, 68422, 71266",,,,"module: memory usage, module: numpy, module: tensor creation, module: testing, module: tests, oncall: distributed","better-engineering, enhancement, feature, module: memory usage, module: numpy, module: tensor creation, module: testing, module: tests, oncall: distributed, quansight-nack, triaged",,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23382 Tensor from mmaped storage loads the entire file into memory, #31655 pytorch out of memory when calculation squared difference of unfold, #40400 New ProcessGroups created with dist.new_group may leak memory, #44182 Support keep stride for neg with requires_grad=False, #46084 In PyTorch Tutorials, RuntimeError: CUDA error: out of memory happen, #48377 Management of Gpu memory to avoid memory errors",37a3c156280de0b72d0bb109a8692e6258cd45c5eaf0a523035aa623c6761a75
issue-collapse:9a9316cb3460528c,module: memory usage stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5434, 45321, 55607, 66162, 73709, 73883, 77839, 78274, 83914, 84944, 87090, 92252, 103422, 117467",,,,"module: memory usage, module: mps, module: multiprocessing, module: numpy, module: python frontend, module: testing, module: vision","feature, module: memory usage, module: mps, module: multiprocessing, module: numpy, module: python frontend, module: testing, module: vision, triaged",,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5434 RuntimeError: $ Torch: not enough memory: you tried to allocate 72GB. Buy new RAM!, #45321 Memory leak when using share_memory_ on cuda device, #55607 pytorch inference lead to memory leak in cpu, #66162 Cuda Out of Memory - MMDMatic, #73709 DropIT: Dropping Intermediate Tensors for Memory-Efficient DNN Training, #73883 Possible issue with memory allocation.",158888ce5e0dfc28f13117b0c391b54c410ed944aae3be749e01c89758b622c3
issue-collapse:d4faad5135c4d814,module: memory usage stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33864, 64043, 72450, 100656, 102334, 109445, 124487, 125217, 132596, 134051, 152550, 154128, 159622, 165319",,,,"module: memory usage, module: mps, module: nn, module: numpy, module: rocm, module: sdpa, module: serialization, module: sparse, oncall: distributed","actionable, enhancement, has workaround, matrix multiplication, module: memory usage, module: mps, module: nn, module: numpy, module: rocm, module: sdpa, module: serialization, module: sparse, oncall: distributed, triaged",,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33864 Memory leak in embedding layer and LSTM, #64043 Uncleared memory use after torch.load(), #72450 Memory parity with JAX attention, #100656 On UMA systems, pytorch fails to reserve memory exceeding the initial memory size, #102334 There is a memory leak in torch.load, #109445 Memory usage steadily increasing when using back propagation with sparse CSR parameter matrices on CPU",93d98707ca2e92b6f75044d0cc6765d0bf94f862cab871f1181cab0b0e9e2797
issue-collapse:decba1aa08f56729,module: memory format stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33571, 43195, 45823, 51558, 53510, 56525, 58979, 60063, 60092, 62925, 63299, 65687, 68430, 69213",,,,"module: memory format, module: nn, module: sparse, module: tensor creation, module: tests, module: viewing and reshaping","enhancement, module: memory format, module: nn, module: sparse, module: tensor creation, module: tests, module: viewing and reshaping, triaged",,Handle this module: memory format stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33571 empty_sparse shouldn't be called with memory layout but is, #43195 one_hot tensors are channels_last but marked as contiguous, #45823 Traceback for ""Warning: Mixed memory format inputs detected while calling the operator."", #51558 torch.reshape fails to keep the memory format, #53510 Improve memory format testing, #56525 Can't deepcopy memory format objects",a9ae31f46981c1c7667c19681c72c94704be0c74fc31868f7308d1bdb6de634f
issue-collapse:ab948eacb86b789f,module: memory format stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50549, 60028, 70171, 71630, 71978, 74442, 74935, 87451, 88612, 90708, 96743, 110387, 133665, 144711",,,,"module: memory format, module: nn, module: numerical-reproducibility, module: performance, module: pruning","Stale, enhancement, feature, module: memory format, module: nn, module: numerical-reproducibility, module: performance, module: pruning, triaged",,Handle this module: memory format stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50549 Channels last doesn't improve speed when using SyncBatchNorm, #60028 RReLU and PReLU do not propagate input strides to outputs, #70171 Channels last performance problem, #71630 More than 4 Dimensions for Channel Last Memory Format, #71978 from_blob / make_tensor support MemoryFormat, #74442 kron has unnecessary and undocumented dependence on memory layout",08876331f8a668891d4c629ce970b62102f98758de9f1ca707eceec304132529
issue-collapse:d871a92bfccaa69a,module: macos stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7686, 54655, 57185, 79382, 83540, 87041, 95708, 97106, 103868, 117094, 127973, 148440, 159781, 171323",,,,"module: macos, module: mps, module: multiprocessing, module: multithreading, module: openmp, module: optimizer, module: performance, module: tests, module: third_party, oncall: cpu inductor, oncall: distributed, oncall: pt2, oncall: releng","has workaround, module: macos, module: mps, module: multiprocessing, module: multithreading, module: openmp, module: optimizer, module: performance, module: tests, module: third_party, oncall: cpu inductor, oncall: distributed, oncall: pt2, oncall: releng, triaged",,Handle this module: macos stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7686 Only one thread is used on macOS (super slow on CPU), #54655 qsize not implemented error, #57185 Segmentation fault with ITIMER_REAL, #79382 torch failure to open libcuda.so.1 on macOS, #83540 I have the same issue as @samgelman on my MacOS., #87041 Segmentation fault: 11 when running ""import torch"" on Mac OS X",11e77de8b9fa297f21bda9839d2ec3cef5863807ab6194e33ebaaed23ae702be
issue-collapse:334d16b33b006b88,module: lrscheduler stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"62640, 63488, 69753, 72714, 74884, 78253, 95122, 106339, 106359, 113222, 121271, 129565, 153842, 156846",,,,"module: lrscheduler, module: optimizer, oncall: distributed checkpointing","Stale, enhancement, has workaround, module: LrScheduler, module: optimizer, oncall: distributed checkpointing, topic: deprecation, triaged",,Handle this module: lrscheduler stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#62640 add gamma to CosineAnnealingWarmRestarts so max lr can decrease in cycles, #63488 MultiStepLR with different gammas for each parameter group, #69753 last_epoch parameter of CyclicLR and OneCycleLR is not the number of epochs, #72714 Add `pct_end` parameter to `OneCycleLR`, #74884 Problematic ASGD Optimizer, #78253 Any plan to add Noam scheduling?",44910e35a706fb2ef7b63936719f91d61c26e0e1d3b0da1d9d28df90ba33c7b0
issue-collapse:24740a34b44aca90,module: loss stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2629, 5157, 8062, 9250, 9286, 16148, 16599, 48621, 54843, 58351, 59216, 65329, 66083, 67136",,,,"module: loss, module: nn, module: shape checking","actionable, enhancement, feature, function request, module: loss, module: nn, module: shape checking, triaged",,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2629 [feature request] add pairwise ranking loss, #5157 BCELoss - weight parameter shape incorrect, #8062 CrossEntropyLoss mishandles weights, #9250 [feature request] SSIM-based cost function as part of the standard set of loss functions, #9286 Accumulate into accreal instead of real for CPU loss functions, #16148 CategoricalCrossEntropy Loss runs with wrong tag",44178a8e3331a8cdaeec610162951d2aebb0667ec42be9246b3101814a753871
issue-collapse:cdf64a9a7d064641,module: loss stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"842, 69900, 76646, 76732, 77256, 77529, 89369, 89408, 89907, 104711, 104776, 105304, 106467, 107680",,,,"module: loss, module: nn, module: optimizer, module: performance, module: sparse, module: type promotion","Stale, actionable, enhancement, feature, module: loss, module: nn, module: optimizer, module: performance, module: sparse, module: type promotion, triaged",,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#842 BCELoss doesn't accept LongTensor targets, #69900 Port MarginRankingLoss to TensorIterator, #76646 `torch.nn.HuberLoss` backwards unexpectedly fail, #76732 Adding Polyloss to `torch`, #77256 CrossEntropyLoss computes SoftMax always across the second dimension, #77529 softmarginloss should use `log1p` and has an incorrect out= behaviour.",52141a761db750b0559ec044d0f2af0b416813bf9735b6e77262e650cc222138
issue-collapse:175d21e8cc174c6b,module: loss stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"49844, 49851, 54137, 55902, 61309, 69858, 71367, 97539, 118918, 129222, 133084, 136145, 139916, 142146",,,,"module: loss, module: nans and infs, module: nn","actionable, enhancement, feature, function request, module: NaNs and Infs, module: loss, module: nn, triaged",,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#49844 BCEWithLogitsLoss gives out nan with -inf logits, #49851 [Feature Request] SuperLoss (NeurIPS 2020), #54137 Add additive angular margin loss, #55902 Warn users when cross entropy is called after softmax, #61309 The class weights implementation for loss functions is incorrect, #69858 Feature Request: Bayesian Personalized Ranking Loss",249a4ec3fd6b9fc383ee6fcb09f40243b81bcb2c1cdbd6726456b1c40683a940
issue-collapse:db6a5e3d55a60a03,module: lint stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9032, 55764, 59854, 63919, 63920, 76618, 76827, 77253, 77548, 81259, 90827, 94504, 114155, 124857",,,,"module: lint, module: macos, module: typing","better-engineering, enhancement, module: lint, module: macos, module: typing, triaged",,Handle this module: lint stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9032 Lint check for non-Unicode characters in diffs / Unicode characters without coding, #55764 clang format on OS X ssl verification failure, #59854 Misreported linting messages, #63919 quicklint mypy step fails with assertion, #63920 quicklint clang-tidy step fails with many unrelated errors, #76618 pylint segfault",b1ac2512d560f1351f8f1e62472618ff5dcfa52d429b3af7df2fadf568f0ad31
issue-collapse:a9b587e49762a446,module: lint stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"123062, 123465, 126386, 127352, 129410, 144217, 146699, 148451, 153638, 156073, 158783, 158785, 159122, 168907",,,,"module: lint, module: macos, module: typing, module: windows","actionable, better-engineering, enhancement, module: lint, module: macos, module: typing, module: windows, triaged",,Handle this module: lint stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#123062 Enable UFMT on all files in PyTorch, #123465 lintrunner-noclang job may diverge from `lintrunner` locally, #126386 lintrunner failures on main, #127352 `flake8: noqa` disables flake8 linter for the whole file and it's not obvious, #129410 line-length of ruff, black and isort are not the same in pyproject.toml, #144217 PyTorch Enum PYI stubs are invalid. Fix typing of PYI stubs",0cd4491b1232ea57263116cfd9c86ef797066010665befe8dd9b5b4d3d9977a3
issue-collapse:ec0fa5569a19ed04,module: linear algebra stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1178, 12435, 22848, 31546, 55907, 60264, 62955, 64818, 68696, 70391, 72537, 72711, 72909, 74389",,,,"module: linear algebra, module: memory format, module: mkl, module: nans and infs, module: numpy, module: pickle, module: python array api, module: serialization, module: sparse, module: tensor creation, module: tests","Stale, feature, function request, has workaround, module: NaNs and Infs, module: linear algebra, module: memory format, module: mkl, module: numpy, module: pickle, module: python array api, module: serialization, module: sparse, module: tensor creation, module: tests, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1178 Batched sparse QR factorizations and solves with cusolver, #12435 `pstrf` on positive semi-definite matrices, #22848 [feature request] Log-determinant for symmetric positive definite matrices, #31546 [Feature Request] Make torch.solve output NaN for singular matrix, #55907 test_variant_consistency_eager_addbmm fails on both cpu and cuda, #60264 Subtensor operations like `tril/triu` to have an option to respect the strides of the input",ae9133cf383d232c42e978df4f591122c09f33f5cd0466e319118cba90144456
issue-collapse:6f0ed627c4a60f77,module: linear algebra stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"69080, 75701, 76300, 76612, 76649, 76705, 76779, 78808, 80132, 82088, 83493, 84835, 86814, 87170",,,,"module: linear algebra, module: magma, module: nans and infs, module: performance, module: sparse, module: tests","better-engineering, feature, module: NaNs and Infs, module: linear algebra, module: magma, module: performance, module: sparse, module: tests, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#69080 Generalized matmul, #75701 Einsum should have an `out=` parameter, #76300 Unexpected _LinAlgError appeared only on my device, #76612 OpInfo incorrectly advertises lu_solve support on CUDA even when compiled without magma, #76649 Add `ldl_unpack` functionality, #76705 Torch `x += y.bmm(z)` is faster than `x.baddbmm_(y, z)`",9758ccac50349cc6bf5a784d3b7fb6a169f2cd47dddfbcddc9a8ada52ae636de
issue-collapse:4a51c69b2794cd12,module: linear algebra stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"47918, 61649, 88942, 90613, 95304, 101106, 102772, 103313, 105068, 105697, 109497, 111908, 121683, 122699",,,,"module: linear algebra, module: meta tensors, module: mkl, module: multithreading, module: python array api, module: python frontend, module: sanitizers, module: sparse, module: type promotion","actionable, enhancement, function request, module: linear algebra, module: meta tensors, module: mkl, module: multithreading, module: python array api, module: python frontend, module: sanitizers, module: sparse, module: type promotion, topic: deprecation, topic: docs, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47918 Function request: Sparse matrix inverse, #61649 Move `tensordot` into `torch.linalg`, #88942 `torch.svd_lowrank` Trigger RuntimeError under UndefinedBehaviorSanitizer, #90613 `torch.inverse` multi-threading RuntimeError: lazy wrapper should be called at most once, #95304 `torch.cartesian_prod` returns inconsistent dimensions with only one input, #101106 addmv doesn't do type promotion correctly,",9e6c6dd7a5439f8f1539055bf6df692e5557d30b2337411673b4c8b5884c8e96
issue-collapse:6e0cd0c53db37345,module: linear algebra stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"121482, 125182, 127223, 128952, 132693, 132940, 134326, 135514, 136452, 136913, 137464, 138353, 142815, 143189",,,,"module: linear algebra, module: mkl, module: nestedtensor, module: performance, module: sparse","enhancement, module: linear algebra, module: mkl, module: nestedtensor, module: performance, module: sparse, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#121482 torch.linalg.eigh is very slow compared to tf.linalg.eigh, #125182 QR Decomposition for Sparse Matrix, #127223 [BUG] torch.linalg.lstsq returning wrong result., #128952 Question about torch.lstsq and torch.linalg.lstsq, #132693 INTERNAL ASSERT FAILED in BatchLinearAlgebra.cpp, #132940 Enabling GPU mixed precision algorithms",45f7dcac990238e73b2281ae4c8f9a88f7731590c0e9e9db656658cb56928ff7
issue-collapse:ef9fe03a3bcf7feb,module: linear algebra stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25785, 55261, 78809, 84530, 122454, 131774, 138452, 146732, 148775, 148954, 154791, 154842, 158428, 165251",,,,"module: linear algebra, module: python frontend, module: sparse, module: type promotion, module: windows","actionable, enhancement, feature, has workaround, module: linear algebra, module: python frontend, module: sparse, module: type promotion, module: windows, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25785 [Feature request] modified Cholesky decomposition, #55261 Batched multi_dot / chain_matmul + let it accept a tensor instead of tuple, #78809 Feature Request: Hessenberg and Schur decompositions, #84530 `tensordot` not working for dtype int32 and lower when there is only 1 element in the given axis, #122454 Triangular matrix storage + matmul, #131774 Different Cholesky results between Windows & Linux",9db5c8fa00c66bea046882a54f4d2a062591b50085fa539f372a401ebbd07136
issue-collapse:e77d8d246e9c9ac9,module: lazy stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"65576, 65640, 67937, 70505, 73989, 73994, 74134, 79020, 80226, 80942, 83826, 94132, 99299, 147661",,,,"module: lazy, module: linear algebra, module: meta tensors, module: molly-guard, module: nn, module: tests, module: xla, oncall: jit","enhancement, lazy, module: lazy, module: linear algebra, module: meta tensors, module: molly-guard, module: nn, module: tests, module: xla, oncall: jit, tensor subclass, triaged",,Handle this module: lazy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#65576 [LTC] Code-gen IRs and TorchScript lowerings needed by running full TorchBench, #65640 [LTC] Memory growing up during training until OOM, #67937 Add support for LTC ops needed by backwards/autograd/autocast, #70505 `UninitializedParameter.to(device='meta')` creates a zero-sized meta tensor instead of remaining uninitialized, #73989 Invalid example tensor: detectron2_maskrcnn, #73994 Lazy: add support for index ops",18aaa9c796b0cf18bb3e99821ffb8e7da35e49e3367a44d81ba5d3d340a0bdde
issue-collapse:0e191b939c5cccae,module: internals stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20006, 24193, 24264, 24956, 26139, 27608, 28577, 30780, 30953, 32427, 33476, 34939, 38973, 38974",,,,"module: internals, module: mkldnn, module: pybind","better-engineering, enhancement, module: internals, module: mkldnn, module: pybind, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20006 RuntimeError: invalid argument 10: ldb should be at least max(1, 0), but have 0 at ../aten/src/TH/generic/THBlas.cpp:36, #24193 TensorIterator stubs are designed for merge conflicts., #24264 TensorIterator ""builder"" options should be documented., #24956 [RPC] Fix logging initialization warning in ProcessGroupAgent, #26139 c10 List API hard to use, #27608 Unify warning logging mechanism",f995e70390e8c81511f35ff10213cc3c7a9e9419416f0401fbe34661a16f2db0
issue-collapse:6869b9bd2257daec,module: internals stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4123, 4148, 4636, 22024, 24261, 26428, 37142, 40215, 40355, 40564, 41478, 43385, 44421, 44721",,,,"module: internals, module: memory format, module: mkl, module: performance, module: pybind","better-engineering, csprng, enhancement, module: internals, module: memory format, module: mkl, module: performance, module: pybind, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4123 Use the int64 version of MKL calls, #4148 ATen explicitly differentiated native function resolution hazard (call is ambiguous), #4636 Better header hygiene in ATen, #22024 Consolidate definition of operators/gradients where possible, #24261 ""PyTorch core"" thread local flag, #26428 Remove TensorOptions logic from generated code",02ccc6f7343c7806f8a59ebce767203a52f4f157aaecc60a1349f9e202fb9e04
issue-collapse:2ced4359091a6ac0,module: internals stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8602, 12069, 46311, 46319, 46382, 46760, 48650, 48667, 48684, 51054, 51229, 52569, 52680, 56399",,,,"module: internals, module: performance, module: structured kernels","better-engineering, feature, module: internals, module: performance, module: structured kernels, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8602 Come with a better strategy for TensorArg (error reporting), #12069 Stop using make_intrusive directly; provide some make_tensor, #46311 Compile error when use c10::RegisterOperators::Options::kernel with the return type ""std::tuple<Tensor&, Tensor&>"", #46319 Remove header declarations for CPUType/TypeDefault, #46382 Code asserts when register new aten operations' implementation thru ""c10::RegisterOperators::op"" API., #46760 How to define a new data type in native_functions.yaml?",2d6702d2a7c138db03e85c5ea6fb50cf52ccaab59dae153febe0a15b01f3f1b7
issue-collapse:b50a7c96188a68d9,module: internals stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14797, 20005, 42155, 57059, 62500, 62845, 62876, 63645, 73134, 73506, 75172, 76936, 77047, 78737",,,,"module: internals, module: meta tensors, module: printing, module: tensorpipe","enhancement, module: internals, module: meta tensors, module: printing, module: tensorpipe, proposal accepted, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14797 Modern interface for Storage, #20005 install error from source, #42155 BusError memory-mapped tensor, #57059 Get rid of weak_intrusive_ptr, #62500 meta tensor `set_` seems fishy, #62845 allow specification of variadic arguments in native_functions.yaml",e847b305efc3fac089ac69a53d38ee2fed06b8e8f011caadd49fb4a13a276a84
issue-collapse:19ecd79045f31b6a,module: internals stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"49298, 50953, 77165, 81750, 81801, 90398, 91395, 99359, 102081, 105281, 117645, 123284, 126519, 147867",,,,"module: internals, module: logging, module: performance, module: structured kernels, module: typing","better-engineering, enhancement, feature, has workaround, module: internals, module: logging, module: performance, module: structured kernels, module: typing, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#49298 [question] How hard would it be to implement 4-bit precision training?, #50953 Dispatch-less structured wrapper / composite / alias kernels, #77165 Automate cleanup of header includes, #81750 Modernize logging tensor in torch.testing._internal, #81801 Rename DispatchKey Dense/Sparse/etc to DenseFunctionality/SparseFunctionality, use original name for alias, #90398 IValue(c10::List<IValue>) constructor is confusing and undocumented",a6ffa5f0631392f819e773819d582af2cc743c5f7a12fadc1fabf069f225ce6a
issue-collapse:6b3a8600f5febf87,module: inductor stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"99807, 109213, 110448, 121311, 125683, 133383, 135305, 136118, 136733, 137012, 138743, 140592, 141507, 143453",,,,"module: inductor, module: nccl, module: performance, module: tests, oncall: cpu inductor, oncall: distributed, oncall: pt2","actionable, feature, module: inductor, module: nccl, module: performance, module: tests, oncall: cpu inductor, oncall: distributed, oncall: pt2, triaged",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#99807 AOTAutograd/Inductor file system cache, #109213 inductor/test_max_autotune having timeout issues, #110448 Explore Hybrid (CPU+GPU) Graphs in Scalar parameters, #121311 In-place native functional collectives right after a custom torch op fail to codegen in inductor, #125683 [RFC] Add Cpp Template for GEMM related ops via max-autotune for Inductor CPU, #133383 [Inductor] Remove dead code from triton kernels",71f2b3a299e15c81cc15601c9b6eb71543c8ec7c2ed299601dbf369e301ed64a
issue-collapse:16117349a4c9a483,module: inductor stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"130769, 139602, 143931, 144375, 145242, 145326, 146629, 147999, 148052, 148073, 148116, 148196, 148960, 149319",,,,"module: inductor, module: user triton, oncall: pt2","enhancement, feature, module: inductor, module: user triton, oncall: pt2, rocm, triaged, upstream triton",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#130769 [Inductor] Add an API to register external callable candidates for inductor's Matmul/Conv tuning choices, #139602 [RFC]: Adding Triton Backend for Aten operators, #143931 The model compiled with torch.compile encounters an error when run., #144375 Unable to compile models using tensorrt backend: CUDNN_STATUS_BAD_PARAM_STREAM_MISMATCH, #145242 Expose configurable path instead of using fixed path in the inductor module for serialized pattern generation, #145326 [inductor][triton] refactor ASTSource.make_ir integration",b431a8a18e075e6bece025f7b3dad1aa353e7779fe3ff784afb6dbb75a9832b7
issue-collapse:113ea6d6c4af2f45,module: inductor stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"118175, 121968, 139268, 147062, 149440, 149603, 149879, 150476, 150871, 151555, 153103, 153241, 153532, 155838",,,,"module: inductor, module: user triton, oncall: pt2, oncall: quantization","actionable, better-engineering, feature, module: inductor, module: user triton, oncall: pt2, oncall: quantization, triaged",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#118175 Different behaviors in `torch.nn.functional.hinge_embedding_loss` between eagermode and torch.compile, #121968 [RFC] Use CUDA graphs by default on torch.compile, #139268 Torch Inductor should have a way for new backend to provide build options, #147062 [Feature Request] Release original parameters by layer when turning on `freezing_discard_parameters`, #149440 [pt2] Support statically launching triton compiled cuda kernels, #149603 [inductor] API for user-controlled fusion ""escape-hatch"" mechanism in inductor",6a12f96df2f2c2871d9ebf8584ad9509c83732823be5b883934c92d6f0d68673
issue-collapse:c094a8b637e39273,module: inductor stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"102319, 103990, 120057, 120116, 121371, 121631, 125672, 126913, 128012, 158240, 159769, 160142, 160359, 162198",,,,"module: inductor, module: performance, module: windows, oncall: cpu inductor, oncall: pt2","actionable, enhancement, feature, internal ramp-up task, matrix multiplication, module: inductor, module: performance, module: windows, oncall: cpu inductor, oncall: pt2, triaged, weeks",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#102319 Re-enable Low Memory Dropout, #103990 [Inductor] Freezing Add support for Caching Parameter Conversions, #120057 Torch compile fx graph is not removing constant propagation, #120116 Use Incremental Fake Tensor Updater more uniformly across torch.compile compilation, #121371 Match HuggingFace T5 SDPA pattern in Inductor, #121631 [inductor] Incorrect handle of `autocast` results in type mismatch",4203c1d15945d995667af0911354a75f194a4951fce5dee218456db31ff488d8
issue-collapse:c007d4d8350d1d7a,module: inductor stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"94167, 118573, 127271, 127500, 138280, 153737, 158699, 164041, 164813, 167027, 167520, 170345, 171184, 171534",,,,"module: inductor, module: pallas, module: python version, module: rnn, module: vllm, oncall: pt2","feature, inductor_pattern_match, intel, internal ramp-up task, module: inductor, module: pallas, module: python version, module: rnn, module: vllm, oncall: pt2, triaged, upstream triton, vllm-compile",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#94167 Faster `pad_sequence` and `tensor_split` function with CUDA kernel, are they possible?, #118573 [Inductor] Unite all possible way to track mutation, #127271 ImportError: cannot import name 'triton_key' from 'triton.compiler.compiler' when using triton 2.3.x, #127500 Avoid Having to Register Op For ExternKernelChoice of Aten Refs, #138280 Refactor FlexibleLayout to separate out ""this stride can be changed"" and ""how this buffer is allocated can be changed"", #153737 SystemError: PY_SSIZE_T_CLEAN macro must be defined for '#' formats - PyTorch compile fails with Python 3.12",f44542feeb829ac7e4f7ad872118d9d27a552ed0ee217174a56470fff77e1b13
issue-collapse:7776f15aacfc2035,module: hub stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"52596, 56067, 65881, 68916, 71205, 73466, 74368, 75577, 81297, 84620, 90126, 90147, 91608, 134947",,,,"module: hub, module: models, module: pickle, module: serialization, module: vision","feature, module: hub, module: models, module: pickle, module: serialization, module: vision, topic: security, triaged",,Handle this module: hub stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#52596 pickle is a security issue, #56067 A more flexible torch.hub search strategy, #65881 torch.hub.load can confuse external python package with local python package., #68916 Crash loading fairseq based model, #71205 Possible security issue of `torch.hub.load`, #73466 [feature request] torch.hub.load_state_dict_from_url to be replaced by a new good general download-a-file function and to also support local paths and google drive links / private github release links",5bac62a1b55b9a8355a68840e4ec2aee5569b7ad894e70f53ce82731c7a183dd
issue-collapse:7ad76f641d2c2dd8,module: half stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32137, 55442, 65447, 74634, 83281, 84348, 107514, 109753, 111454, 121960, 123911, 135969, 136680, 162674",,,,"module: half, module: memory format, module: mkldnn, module: nn, module: norms and normalization, module: numerical-stability, module: onnx, module: sdpa, module: sparse, module: vectorization, oncall: export, oncall: jit","actionable, enhancement, low priority, module: half, module: memory format, module: mkldnn, module: nn, module: norms and normalization, module: numerical-stability, module: onnx, module: sdpa, module: sparse, module: vectorization, oncall: export, oncall: jit, triaged",,Handle this module: half stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32137 torch.nn.functional.normalize epsilon too small for half precision, #55442 More AVX2 vectorization support for half (float16), #65447 No support for torch.trace on CPU for float16 tensors, #74634 Sampling Issue With Distributions, #83281 addcdiv_ (in and out of place) not implemented for torch.float16 and cpu, #84348 RuntimeError: ""slow_conv2d_cpu"" not implemented for 'Half'",4f4ef3f22a8e878e98345ee0b22b45dee84e7da423202cdbe783540739751fc4
issue-collapse:32fb532daf4b7d84,module: fx stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50143, 50177, 52022, 52538, 52564, 53121, 53732, 53794, 53935, 54071, 54503, 54574, 54785, 54913",,,,"module: fx, module: tests","module: fx, module: tests, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50143 FX call_function does not work if target is a class member function, #50177 [FX] Static specialization of callsites (e.g. chunk), #52022 [FX] Interpreter doesn't GC values, #52538 Helper methods not preserved during AST rewrite, #52564 Submodules not rewritten during AST rewrite, #53121 [fx] Handle models with Optional inputs",2efd2f5c50864458403cf79289d2e3f799507806551ead775236730b51720378
issue-collapse:d74568d80884d781,module: fx stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55394, 55395, 55397, 55400, 55401, 55544, 55668, 55694, 56688, 56778, 56938, 57124, 58051, 58402",,,,module: fx,"FX-TorchScript Compatibility, module: fx, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55394 [FX][testing] Test tracing into all the standard `nn.Module` instances, #55395 [FX][testing] Test tracing into all the standard `torch.nn.functional` instances, #55397 [FX][testing] Run symbolic tracing tests over torch benchmark in CI, #55400 [FX][testing] Add symbolic tracing tests for torchtext, #55401 [FX][testing] Add symbolic tracing tests for torchaudio, #55544 [FX] See if we can type-annotate attribute accesses on proxy",40b10a7e21dba677e9b4fa2ca53cfc5cf29083c07dba22c6a0c579f8c0c333b1
issue-collapse:df41458785147fcb,module: fx stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"63076, 63490, 63815, 64497, 64997, 65096, 66197, 67030, 67364, 67402, 67543, 67662, 68190, 68301",,,,module: fx,"feature, module: fx, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#63076 Tracing models all the way down to basic functions, #63490 [FX] Data type of target should be validated on Node construction and lint, #63815 [FX] Documentation for Graph.flatten_inps and Graph.unflatten_outs, #64497 FX tracing on Fairseq-based roberta fails., #64997 FX failed to normalize op `block_diag` and `broadcast_tensors`, #65096 [FX] `passes.split_module` doesn't have documentation",34a27e2a9fe44ac4fd05afd4e6dc41b97ea061fe7ec46a1ad93438edc21519a7
issue-collapse:775131d971908e65,module: fx stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"68513, 69095, 69804, 70449, 70892, 71715, 73304, 73992, 74058, 74218, 75090, 75342, 76186, 77016",,,,module: fx,"Stale, module: fx, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68513 torch.fx cannot trace torch.Size() properly, #69095 [FX] `concrete_args` with unpacking breaks `fx.Interpreter`, #69804 typing error in the signatures of `torch.unbind` when getting them using `torch.fx`., #70449 [FX] Detect attribute mutation during tracing, #70892 Memory leak while training model generated by torch.fx.symbolic_trace() in data parallel mode, #71715 [FX] Support call_method in NormalizeArgs",e6f0b6186fe1c019ba369175608cb0de624408fe02f5a407333e27ac918e2de4
issue-collapse:dd4dc635fa1d3058,module: fx stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"69025, 69577, 73469, 78201, 78435, 80172, 80427, 80903, 81229, 82727, 83292, 84316, 84515, 84840",,,,"module: fx, module: pytree","Stale, fx, module: fx, module: pytree, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#69025 [subgraph_rewriter] Support skipping matching for some ops, #69577 [subgraph_rewriter] Add match_filter for the matches, #73469 torch.fx failed when tracing functions from other Libs., #78201 nn.Sequential causes fx.replace_pattern to not find any match., #78435 fx.Tracer with param_shapes_constant=True not working for RobertaForMaskedLM, #80172 Torch.fx: add reporting of the name of a module not found during tracing",1406191395b17cf53a1b5f1b3749ed504319f250d27b06fc9ff8189432d1d1c9
issue-collapse:4b70988af6dda077,module: fx stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55399, 56862, 62395, 67350, 74901, 84415, 89421, 94374, 95276, 98942, 99023, 99407, 103060, 104593",,,,"module: fx, oncall: fx","FX-TorchScript Compatibility, fx, module: fx, oncall: fx, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55399 [FX][testing] Test symbolic tracing of detectron2, #56862 Non-symbolic FX tracer, #62395 [FX] Does PyTorch want to support for dynamic control flow for torch.fx?, #67350 [FX] autowrap_functions doesn't work for module-scoped functions, #74901 torch.fx.wrap will not work, when encapsulate the code, #84415 Deepcopy of FX graph fails with nested make_fx and constant tensors",cd0b5df7cd0de8b5fbbf205cb3383c0d8b2b00691ad4d8d3982930d2eb7d020c
issue-collapse:2b50901e149de19f,module: fx stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"53026, 54979, 61246, 71919, 74842, 79307, 87777, 105445, 107538, 107678, 109386, 112775, 113034, 125581",,,,"module: fx, module: multiprocessing, module: proxytensor, oncall: fx, oncall: quantization","low priority, module: ProxyTensor, module: fx, module: multiprocessing, oncall: fx, oncall: quantization, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#53026 Ensure that nn.MHA and functional multi-head attention is supported by FX graph mode quantization, #54979 test_quantize_fx.py test_resnet_18_dpp test failure, #61246 torch.quantization.fx.fuser forget to pass ""additional_fuser_method_mapping"" to fuse method, #71919 Graph Mode Quantization does not keep NamedTuple Information, #74842 error in quantization by quantize_fx.prepare_fx, #79307 `prepare_qat_fx` docstring doesn't run",207e0c0efaff2e960f81b2135f82b3d995b6d8a96b1ed38800455a5fe322601b
issue-collapse:089aca3d2eb6fb25,module: fx stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"60069, 109539, 127282, 130701, 139163, 139185, 139876, 147884, 150922, 153930, 156733, 165277, 165634, 169797",,,,"module: fx, module: fx.passes, module: python frontend, oncall: profiler","module: fx, module: fx.passes, module: python frontend, oncall: profiler, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#60069 Symbolic trace with *args, #109539 Torch FX SubgraphMatcher Any / Oneof Patterns, #127282 torch.fx.symbolic_trace doesn't support many Callable types, #130701 fx.wrap() doesn't really work for things in torch/*, #139163 SubgraphMatcher may fail to match when the matching pattern having call_module IR, #139185 Operators being traced as method calls in torch.fx",4b9cc79d0d6f673998a6325538fb2ced8c5e78ec59edbffc4a0d0dc9829e74bb
issue-collapse:e769285f334533f9,module: functorch stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"85656, 86298, 86819, 87222, 91199, 91244, 91509, 91810, 93231, 93317, 93319, 93523, 95718, 96337",,,,"module: functorch, module: fx, module: pytree","feature, module: functorch, module: fx, module: pytree, triaged",,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#85656 [functorch] conv.{1, 2, 3}d should raise errors, #86298 AOT Autograd Device Partitioning, #86819 Could not run select_backward [vmap] [dlrm] [functorch], #87222 `custom_jvp` and `custom_vjp`, #91199 functorch.functionalize doesn't work with torch.autograd.grad, #91244 first class dims leak memeory",42c1e2985511086d56ebccb1cef887bc324931c6923707adaba49c72d4fa72fa
issue-collapse:62d49f46ee984b67,module: functorch stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"96855, 97283, 97286, 98338, 98515, 98557, 99287, 99311, 99558, 99836, 100801, 101073, 101415, 104704",,,,"module: functorch, module: memory format, module: pytree, module: sparse, module: vmap, oncall: jit","actionable, better-engineering, enhancement, module: functorch, module: memory format, module: pytree, module: sparse, module: vmap, oncall: jit, triaged",,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#96855 Performance Drop for linalg_ldl_factor and ldl_solve, #97283 `jacfwd` fails when computing the gradient for `channels_last` tensor, #97286 `jacrev` and `jacfwd` raise an error that `Sparse CSR tensors do not have strides`, #98338 [functorch] [vmap] tests fail when `_set_vmap_fallback_enabled(False)`., #98515 torch.cond should work with expressions involving SymInt, #98557 torch.jit.script codegen warning with cuda and vmap",0b956785ac6712265e9229608e27e1d6e1e4f84ac1a18f3a39a69b3c52d969f8
issue-collapse:a4b55e30bf7f56b2,module: functorch stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"85533, 98724, 102085, 103879, 105326, 107125, 109938, 111311, 112013, 114604, 115342, 118861, 127750, 129459",,,,"module: functorch, module: typing, oncall: distributed","actionable, feature, module: functorch, module: typing, oncall: distributed, triaged",,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#85533 [functorch] transforms like jacrev, jacfwd, grad, etc don't work with BatchNorm, #98724 Conflict between ``torch.func`` transformations and ``torch.jit.trace``, #102085 BatchNorm can't be symbolically traced with torch.fx as a standalone module, #103879 Can ``torch.vmap`` add ``grad_fn``= SelectBackward when maping over some dimension of the inputs?, #105326 Can't vmap over torch.tensor constructor, #107125 combining `vmap` with NN containing `MaxPool2d' leads to discrepancies in output",8664b429069118a057e42bc10bf99e4464f9ca72ae0839270033d920f062d395
issue-collapse:e8c28fb23c0b5d08,module: functorch stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"116102, 132052, 134038, 136357, 139241, 140704, 145262, 149250, 149609, 149620, 149788, 149804, 151113, 152295",,,,"module: functorch, module: linear algebra, module: mps, module: nestedtensor, module: sparse, module: tests, module: vmap","actionable, enhancement, module: functorch, module: linear algebra, module: mps, module: nestedtensor, module: sparse, module: tests, module: vmap, tensor subclass, triaged",,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#116102 torch.func.grad with respect to intermediate outputs, #132052 Make C++ torch::autograd::Function support functorch, #134038 vmap tests aren't erroring on fallback usage, #136357 torch.func.vjp and sparse tensors: ""Sparse CSR tensors do not have strides"", #139241 Please implement batching rule for torch.nn.functional.multi_margin_loss, #140704 vmap with mps tensor fails to recognized batched shape",9591b8e6ec2b88c49c34d3f4daee14a64924ebf743cb144fb119666ddf2cfe5a
issue-collapse:a759619a8366db1d,module: functorch stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"97395, 101241, 102109, 103329, 105912, 114605, 144850, 149694, 151591, 153861, 158169, 161296, 162346, 166511",,,,"module: functorch, module: linear algebra, module: numerical-stability, module: pt2-dispatcher, module: vmap, oncall: jit, oncall: pt2","actionable, module: functorch, module: linear algebra, module: numerical-stability, module: pt2-dispatcher, module: vmap, oncall: jit, oncall: pt2, topic: performance, triaged",,Handle this module: functorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#97395 Expanded weights tests broken, #101241 Different results with vmap when using torch.jit.script, #102109 Can't vmap over a slice expression, #103329 RuntimeError: torch.vmap a function that includes in-place arithmetic operations on a zero-initialized tensor, an error ""vmap: inplace arithmetic(self, *extra_args) is not possible"" is raised., #105912 Batching rule for aten::bincount., #114605 vmap: Transform single-element tensor to integer",8f3620da8bb81137f3059ce491112312f8998e97c53403aa22c22588cf1c3cf0
issue-collapse:d73595b8df61dd9b,module: functionalization stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"87019, 87571, 89068, 90992, 92912, 94471, 97852, 107961, 115706, 120585, 130021, 135382, 159784, 165409",,,,"module: functionalization, module: functorch, module: meta tensors, module: pt2-dispatcher, module: viewing and reshaping, module: xla, oncall: pt2","has workaround, inference mode, module: functionalization, module: functorch, module: meta tensors, module: pt2-dispatcher, module: viewing and reshaping, module: xla, oncall: pt2, triaged",,Handle this module: functionalization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#87019 Rewrite `narrow_copy_dense_cpu_out` using `copy_` and `narrow`, #87571 Testing insufficient to catch incorrect dispatch key for bernoulli.p re functionalization, #89068 view_copy out= does not reshape zero element tensors, #90992 as_strided_scatter : INTERNAL_ASSERT_FAILED for requires_grad=True and non-config input, #92912 functorch.functionalize doesn't error out with logcumsumexp.out, #94471 [Functionalization] `index_reduce_` op tests with functionalization enabled",b8dcb99571cc4bece5360103625fcee6ea1dcfccf0b9f4f15f6a4dec914fa015
issue-collapse:34ace4f65844613f,module: fsdp stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"101335, 103962, 108023, 109002, 109440, 117601, 118095, 118245, 119431, 120879, 125738, 133849, 148257, 148504",,,,"module: fsdp, oncall: distributed","actionable, feature, module: fsdp, oncall: distributed, triaged",,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#101335 fsdp training with the seq2seqTranier module gets stuck during evaluation., #103962 How to unwrap after auto_wrap in FSDP?, #108023 [FSDP] Ignored modules on meta device seem to be initialized on CUDA device, #109002 [FSDP] How can I wrap a model that has both nn.Parameter and nn.Module, #109440 [FSDP] supports QLora finetuning, #117601 [FSDP] more actionable error logging in meta init",64506a73c8074ce5b9f022adcd934dc80d1638ac02f6f338f95525e13a996505
issue-collapse:be495bf39228576d,module: fsdp stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"77724, 140514, 148756, 148831, 149100, 150096, 150140, 150799, 151030, 154272, 154836, 154960, 158819, 159498",,,,"module: fsdp, module: optimizer, oncall: distributed, oncall: distributed checkpointing","module: fsdp, module: optimizer, oncall: distributed, oncall: distributed checkpointing, topic: new features, triaged",,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#77724 FSDP: enhanced shared parameter support, #140514 Error with fused AdamW and FSDP, #148756 FSPD ValueError: expected to be in states [<TrainingState.FORWARD_BACKWARD: 2>] but current state is TrainingState.IDLE, #148831 FSDP2 and autocast compatibility issue, #149100 Memory leak when using get_model_state_dict with FSDP-sharded models, #150096 FSDP OOM during `sync_params_and_buffers`",1e578bbb476357dbd1d4e6f10b36c704a20081adcf4b62c3f1125ce8f310f1e6
issue-collapse:5012575b659879a4,module: flaky-tests stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9563, 14478, 16458, 17354, 20488, 59875, 61614, 62602, 87087, 88883, 88902, 146307, 149016, 160387",,,,"module: flaky-tests, module: inductor, module: lint, module: mps, module: rocm, module: serialization, module: unknown, oncall: pt2","better-engineering, caffe2, module: flaky-tests, module: inductor, module: lint, module: mps, module: rocm, module: serialization, module: unknown, oncall: pt2, rocm-skipped-tests, skipped, todo, triaged",,Handle this module: flaky-tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9563 batch_sampler/test_worker_seed intermittently fails with address already in use on OS X, #14478 Flaky download from files.pythonhosted.org when installing botocore, #16458 TensorRTOpTest.test_vgg19 is flaky, #17354 testConvnetBenchmarks intermittently segfaults, #20488 RoiAlignTest.CheckCPUGPUEqual is still flaky, #59875 [ROCm] test_gather_stress_cuda is flaky",4a240be4feb27ce91cdbb863112538561ca3105c6ac4692f649e0d7e5695a41e
issue-collapse:7de270d92ea8dd02,module: fft stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"51965, 69627, 87353, 96518, 99515, 106622, 113525, 114581, 114672, 128044, 145943, 145944, 165425, 165739",,,,"module: fft, module: flop counter, module: nans and infs, module: onnx, module: performance, module: tests","feature, function request, low priority, module: NaNs and Infs, module: fft, module: flop counter, module: onnx, module: performance, module: tests, triaged",,Handle this module: fft stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51965 Feature Request:Fast Cosine Transform, #69627 c2r fft input generation, #87353 Speed of torch.istft, #96518 fft should ignore dims with shape 1, #99515 Support polyphase channelizer, #106622 FFT Samples Inputs with More than Three Dimensions",f2619b006152908cfb3b071b4a524020b8fda0ce09393d843692a438c1fb72a1
issue-collapse:ea4f0a06ffaf3cf2,module: error checking stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11426, 13782, 27936, 34880, 47610, 49224, 55161, 56330, 58380, 58657, 60277, 62865, 63282, 65375",,,,"module: error checking, module: linear algebra, module: nn, module: numpy, module: partial aliasing, module: serialization, module: sparse, module: tensor creation","better-engineering, module: error checking, module: linear algebra, module: nn, module: numpy, module: partial aliasing, module: serialization, module: sparse, module: tensor creation, triaged",,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11426 [Feature request] Intuitive error message when input to Linear is not cudarized, #13782 torch.linspace does not check for infinity and nan, #27936 Specifying `pos_weight` in F.binary_cross_entropy_with_logits leads to RuntimeError: class size not match, #34880 Expanded tensors don't work as registered buffers, #47610 Input dimension check for `torch.gather`, #49224 Torch abs op crashes on -128 int8 tensor with ASAN",94e2cc4febc548c4a428728aadd467b2b97789bacd0f2b0e759b41e72d04b6aa
issue-collapse:9102068c9712fc7f,module: error checking stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10298, 20056, 54135, 66868, 69800, 69912, 71059, 71078, 71084, 71541, 72332, 74295, 75733, 76362",,,,"module: error checking, module: linear algebra, module: loss, module: nn, module: padding, module: reductions, module: scatter & gather ops, module: tensor creation, module: ux","Stale, function request, module: error checking, module: linear algebra, module: loss, module: nn, module: padding, module: reductions, module: scatter & gather ops, module: tensor creation, module: ux, triaged",,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10298 [feature request] Runtime warning for inappropriate labels (among others), #20056 Creation of too big multidimensional array returns empty tensor., #54135 torch.kron of a transposed input error, #66868 torch.triu behaves differently when diagonal out of range, #69800 Incomplete error message at tensor indexing (when indexing with set), #69912 torch.nn.functional.ctc_loss with invalid input produce NaN or infinity gradient, while the batch entries are fine",7ec38d4410f0c8d054a0dcf7ecfead87713dfc9fb16280341c7b111178e71c96
issue-collapse:49a02f8b3b54f5f8,module: error checking stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"73638, 81459, 87031, 89218, 89254, 89381, 89418, 94704, 105629, 105860, 106124, 120973, 121226, 126758",,,,"module: error checking, module: loss, module: mps, module: nn, module: norms and normalization, module: padding, module: python frontend, module: tests, module: type promotion, module: viewing and reshaping","actionable, module: error checking, module: loss, module: mps, module: nn, module: norms and normalization, module: padding, module: python frontend, module: tests, module: type promotion, module: viewing and reshaping, release notes: python_frontend, triaged",,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#73638 addcdiv is failing the ASAN test for zero divisors, #81459 Missing corner case handling in ATen ctc_loss implementation, #87031 Improve Readability of error(s) when provided unexpected keyword arguments., #89218 `torch.nn.LayerNorm` Abort with ""invalid device ordinal"" Error, #89254 `torch.nn.ReplicationPad2D` Report ""invalid configuration argument"" Error under Compute Sanitizer, #89381 torch.nn.ReplicationPad1d:The description of the exception information thrown is not accurate",79ff241e787ab806527ffb6a83217ca63ec818de41f09ae87b8ff564dd3bdb8a
issue-collapse:fe1ee0da349af448,module: error checking stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"78929, 89394, 130608, 135640, 135648, 138192, 138418, 149724, 152437, 159740, 162712, 164878, 171536, 174986",,,,"module: error checking, module: fft, module: linear algebra, module: molly-guard, module: mps, module: nn, module: onnx, module: performance, module: python frontend, oncall: jit","actionable, bot-mislabeled, bot-triaged, enhancement, low priority, module: error checking, module: fft, module: linear algebra, module: molly-guard, module: mps, module: nn, module: onnx, module: performance, module: python frontend, oncall: jit, topic: fuzzer, triaged",,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#78929 Add type() support for mps backend, #89394 torch.nn.TransformerEncoderLayer missing exception description information., #130608 Setting `shape` and `dtype` without `dtype=` to `view()` works against error message but an `int` tensor is returned instead of a `float` tensor, #135640 Issue UserWarning for missing ModuleList/ModuleDict in Module classes, #135648 Setting a `float` value to `in_channels`, `out_channels` and `kernel_size` argument of `nn.Conv1d()` gets indirect errors and in_channels and kernel_size with with non-bools, #138192 Inconsistent behavior of `torch.tensor` in converting `NaN` to int32 on normal list input and numpy array, and different architectures.",df2f23bb7a05d1ed7e40b1e5d33482fa4a24ef3767f9c7bfba9121119bf07d82
issue-collapse:aefde3b0ea51d039,module: elastic stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"63223, 63846, 73380, 85300, 99684, 109747, 113515, 114097, 122694, 124906, 130656, 132951, 147563, 158326",,,,"module: elastic, module: tests, oncall: distributed, oncall: r2p","enhancement, feature, has workaround, module: elastic, module: tests, oncall: distributed, oncall: r2p, topic: bug fixes, topic: docs, triaged",,Handle this module: elastic stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#63223 [bug] torch.distributed.elastic logging: Failed to print the first info statement, #63846 [torch.distributed] Make dynamic_rendezvous log handler configurable, #73380 elastic/rendezvous: _matches_machine_hostname doesn't resolve hostnames fully, #85300 torchrun substitutes host names for IP addresses, #99684 In torchelastic support running worker rank 0 on agent rank 0 consistently, #109747 [RFC][TorchElastic] topology info in training apps/ranks",9f2a4619d90d8f32f33af2cf61165b5038c6bf2af8454eb9925761a0515c192c
issue-collapse:baee41b9d3399346,module: edge cases stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"77514, 78127, 78128, 78129, 78130, 78131, 78153, 78255, 83175, 83624, 84560, 86918, 87551, 88941",,,,"module: edge cases, module: linear algebra, module: nn, module: optimizer, module: pooling, module: sanitizers","actionable, module: edge cases, module: linear algebra, module: nn, module: optimizer, module: pooling, module: sanitizers, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#77514 `max_pool1d` can succeed when padding is negative for tensor requiring grad, #78127 Segfault in torch._C._nn.reflection_pad2d, #78128 Segfault in torch._C._nn.thnn_conv2d, #78129 Segfault in _embedding_bag_forward_only, #78130 Segfault in _grid_sampler_2d_cpu_fallback, #78131 Segfault in _pad_packed_sequence",c835a0d6b01a5f579315446096169f47487db0ed4f3407fda6b814b6928db0c1
issue-collapse:8c05b73a6e8d633c,module: edge cases stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"89393, 89420, 91552, 91553, 91556, 91557, 91593, 92835, 94115, 94590, 96236, 96613, 97087, 97088",,,,"module: edge cases, module: forward ad, module: linear algebra, module: nn, module: numpy, module: pooling, module: type promotion","module: edge cases, module: forward ad, module: linear algebra, module: nn, module: numpy, module: pooling, module: type promotion, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#89393 Edge case: CPU bool abs is not supported, #89420 Edge case: torch.baddbmm supports double x int8 x int8 inputs on CPU but not CUDA, #91552 Overflow when running torch.nn.AdaptiveMaxPool2d, #91553 Segmentation fault when running torch.nn.AdaptiveMaxPool2d, #91556 Overflow when running torch.nn.AdaptiveMaxPool3d on torch 1.12.0 and 1.13.1, #91557 Segmentation fault when running torch.nn.AdaptiveMaxPool3d",0a049f7635a4af319f875807367a5c6676fb1200efdec055a501a3d7ae3ecb15
issue-collapse:442cc7271cb59869,module: edge cases stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88943, 101653, 113370, 114216, 126658, 126659, 126776, 126777, 131273, 134183, 134244, 135330, 136699, 136717",,,,"module: edge cases, module: empty tensor, module: error checking, module: linear algebra, module: nn, module: python frontend, module: sanitizers, module: scatter & gather ops, module: type promotion","actionable, module: edge cases, module: empty tensor, module: error checking, module: linear algebra, module: nn, module: python frontend, module: sanitizers, module: scatter & gather ops, module: type promotion, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88943 `torch.vander` Trigger RuntimeError with UndefinedBehaviorSanitizer, #101653 Unexpected behavior comparing uint8 tensor to value greater than 255, #113370 Incorrect stride when permuting shapes where a zero dimension is present., #114216 torch.nn.functional.max_pool2d outputs inf, #126658 `kthvalue()` cannot be used with `complex` or `bool` type of a 1D or more D tensor but `kthvalue()` can be used with `complex` or `bool` type of a 0D tensor, #126659 `topk()` cannot be used with `complex` or `bool` type of a 0D(only `cuda`) or more D tensor but `topk()` can be used with `complex` or `bool` type of a 0D(only `cpu`) tensor",effb9675de55beba3373e012b1678799adec1caade3b17780d82d9a6746d49f2
issue-collapse:4a7092df6983d7f7,module: edge cases stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"120986, 135048, 135050, 137484, 141203, 141449, 141450, 144643, 146570, 149013, 153329, 153337, 154474, 156020",,,,"module: edge cases, module: empty tensor, module: error checking, module: fft, module: intel, module: nans and infs, module: nn, module: numpy, module: python frontend, module: sanitizers, module: sparse","actionable, low priority, module: NaNs and Infs, module: edge cases, module: empty tensor, module: error checking, module: fft, module: intel, module: nn, module: numpy, module: python frontend, module: sanitizers, module: sparse, topic: fuzzer, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#120986 Intel MKL DFTI ERROR, #135048 Setting a `bool` tensor both to `min` and `max` argument of `clamp()` gets error while setting a `bool` tensor only to `min` or `max` argument of `clamp()` works, #135050 Setting a scalar and a 0D tensor or a 0D tensor and a scalar to `min` and `max` argument of `clamp()` respectively works, #137484 Better error message in `torch.linalg.vector_norm`, #141203 SEGV in `torch.native_batch_norm`, #141449 Segmentation fault (core dumped) in `_fft_r2c`",dae6ab3e80860a8d0660e3f31ee6b3865ce3328ccc45a7b68675c2b7f40d2f4a
issue-collapse:8335f040ef533432,module: edge cases stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"85791, 102894, 153187, 153234, 154311, 158412, 158419, 159333, 159363, 160856, 161524, 161853, 162489, 169058",,,,"module: edge cases, module: error checking, module: loss, module: mps, module: mta, module: nans and infs, module: nn, module: onnx, module: python frontend, module: reductions","actionable, low priority, module: NaNs and Infs, module: edge cases, module: error checking, module: loss, module: mps, module: mta, module: nn, module: onnx, module: python frontend, module: reductions, topic: fuzzer, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#85791 nn.CrossEntropyLoss overflow with FP16 and minibatch, #102894 BCELoss and BCEWithLogitsLoss differ when one of the input logits is float(""inf""), #153187 Inconsistent overflow handling in `torch.clamp_min` between CPU and CUDA for float16 tensors, #153234 `torch.quantile` perform differently on cpu and cuda, #154311 torch.inf.to(torch.int32) produces different values on CPU vs CUDA, #158412 Inconsistent torch.absolute results on complex128 between CPU and CUDA",922eba5f938ff8ad9bbbae9b8f6a9a06fac2730fc03b83214a12aa30710a58f3
issue-collapse:eca9878887dba015,module: dynamo stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"93737, 93743, 109309, 112727, 133040, 137811, 140416, 141125, 141422, 141690, 141964, 142171, 142414, 144478",,,,"module: dynamo, module: inductor, module: onnx, oncall: distributed, oncall: pt2","better-engineering, dynamo-ctx-manager, dynamo-polyfill, dynamo-triage-june2024, enhancement, feature, internal ramp-up task, module: dynamo, module: inductor, module: onnx, oncall: distributed, oncall: pt2, onnx-needs-info, pt2d-triage-nov2024, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#93737 Improve handling of generators, #93743 Support data-dependent control flow operators, #109309 Support the `ExitStack` context manager (or a simplified version), #112727 [dynamo] Implement iter fallback (and possibly all iters/generators) as polyfill, #133040 Add Test for Inductor and Dynamo Config BC Breakages, #137811 Mechanism to attempt to create a VariableTracker but graph break and turn Dynamo off if we fail",bde5adf71a9f7437115263e8d06e88ce2869a9e3ea8a22544169bdcdebd585db
issue-collapse:de87f0e8c2ede909,module: dynamo stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"128095, 129482, 129937, 129942, 133117, 138226, 141589, 142395, 146930, 146964, 147597, 148103, 148144, 148828",,,,"module: dynamo, module: graph breaks, module: guards, oncall: distributed, oncall: export, oncall: pt2","dynamo-triage-jan2025, dynamo-triage-june2024, dynamo-user-empathy-day, empathy-day, enhancement, module: dynamo, module: graph breaks, module: guards, oncall: distributed, oncall: export, oncall: pt2, pt2d-triage-nov2024, topic: not user facing, tracker, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#128095 [inductor] Graph breaks in CohereForAI/aya-23-8b, #129482 Dynamo graph breaks on boxed ProcessGroup, #129937 [tracker] Dynamo guard validation exposed issues, #129942 Torch Compile mode giving Graph Break for Higher Order Op for ProcessGroup param, #133117 [dynamo] Improve support for subclasses of popular builtin types like dict, list, set, #138226 No guards for frozen dataclasses and readonly configs",4e64b43aa82fc0a3f1eb47eca0b899ad47990b9eb49973142295f83b304518d1
issue-collapse:3e1a11443a90683a,module: dynamo stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"128149, 149872, 149906, 149908, 149956, 149962, 149970, 150864, 151015, 151251, 151776, 152829, 152869, 152994",,,,"module: dynamo, module: flaky-tests, module: graph breaks, module: user triton, oncall: distributed, oncall: pt2","dynamo-functools, dynamo-tensor-subclasses, dynamo-triage-june2024, feature, module: dynamo, module: flaky-tests, module: graph breaks, module: user triton, oncall: distributed, oncall: pt2, skipped, tensor subclass, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#128149 torch.compile with Custom tensor subclass doesn't inline the tensor subclass methods, #149872 Do all lazy imports for torch.compile in one place?, #149906 [ued][chatTTS][dynamo] graph break on should_compile_partial_graph=False, #149908 [ued][ChatTTS][guards] Too many recompilations, #149956 [ued][flux][dynamo] Wrong error message with dynamo.disable, #149962 [ued][wan] Dynamic shape issue with tolist",66a62f34f579b0d6f420a9aa2643b4acfd00c29f684a8d29ddf78da710e4ad7c
issue-collapse:1cf0403e985884ad,module: dynamo stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"93480, 93511, 101107, 115122, 115526, 116013, 116575, 124749, 148106, 148289, 155266, 155825, 157075, 160071",,,,"module: dynamo, module: flex attention, module: fx, module: graph breaks, module: higher order operators, module: inductor, module: pt2-dispatcher, module: tests, module: user triton, oncall: pt2","awaiting response (this tag is deprecated), compile-cache, dynamo-nn-modules, dynamo-symbolic-analysis, dynamo-triage-jan2025, dynamo-triage-june2024, feature, low priority, module: dynamo, module: flex attention, module: fx, module: graph breaks, module: higher order operators, module: inductor, module: pt2-dispatcher, module: tests, module: user triton, oncall: pt2, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#93480 Umbrella issue for only populate real_value_cache in export test suite fails, #93511 support setattr of arbitrary user provided types in tracing, #101107 Make compiled models serializable, #115122 [dynamo] Graph breaks from copy.deepcopy, #115526 Can PyTorch support using custom compilers for model's components and the Inductor compiler for the model?, #116013 [Dynamo] bytecode transformed by Dynamo is not serializable by marshal",2e77383d0a366c2fdaff861b26d9aff383fb2067c7a86172de447a5038a22ad1
issue-collapse:30c6364c33e5f3c7,module: dynamo stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"93923, 102063, 102898, 111674, 117321, 118884, 126566, 136833, 149062, 161105, 161330, 162375, 164062, 167257",,,,"module: dynamo, module: inductor, module: pt2-dispatcher, oncall: export, oncall: pt2","better-engineering, dynamo-logging, dynamo-must-fix, dynamo-polyfill, empathy-day, enhancement, feature, low priority, module: dynamo, module: inductor, module: pt2-dispatcher, months, no-scrub, oncall: export, oncall: pt2, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#93923 Dynamo uses CONSTANT_MATCH guards for string inputs, #102063 [inlined-inbuilt-nn-modules][dynamo][BE] Revisit call_method of NNModuleVariable, #102898 After dynamo minifier generates repros that don't entirely match what we minified over, #111674 Dynamo Compile samples should record file/line that raised exception, #117321 [C++-polyfill] Pt2 - Discussion around user defined type->behavior dispatching, #118884 [Export] [TorchVision] Export Support for TorchVision Detection Models",d862ca92c9c1c258006797649a219076d685d08eaa21949c326dcb4d3cca18ca
issue-collapse:2b49fd3d3e6419b8,module: dynamic shapes stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"119549, 128150, 130534, 136119, 136605, 136681, 137217, 138813, 141594, 141675, 141934, 144969, 147264, 149743",,,,"module: dynamic shapes, module: dynamo, module: fsdp, module: inductor, module: minifier, module: user triton, oncall: distributed, oncall: export, oncall: pt2","export-triaged, ezyang's list, module: dynamic shapes, module: dynamo, module: fsdp, module: inductor, module: minifier, module: user triton, oncall: distributed, oncall: export, oncall: pt2, pt2d-triage-nov2024, recompilations, triaged",,Handle this module: dynamic shapes stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#119549 Way of mocking collectives, especially data-dependent ones like all2all_single, #128150 More economical use of FX nodes when generating symbolic shapes related manipulations, #130534 Clarify invariants established by torch.fx.passes.runtime_asserts or factor so we can have a pre-pass that reestablishes them, #136119 [torch.export] Dim discrete/finite set, #136605 automatic_dynamic_shapes for mark_unbacked, #136681 [triton_op] Automatically `tl.constexpr` user-written kernel params when they are static integers",adc0463063fae5832196fd396becc25377dc3de6c3bffff9809130b911666f32
issue-collapse:bc631f86d5cedaad,module: dynamic shapes stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"107873, 113232, 121286, 124686, 126738, 126846, 144068, 148564, 154489, 159942, 163785, 165421, 165912, 166186",,,,"module: dynamic shapes, module: dynamo, module: faketensor, module: inductor, oncall: distributed, oncall: export, oncall: pt2","better-engineering, dynamo-triage-jan2025, feature, module: dynamic shapes, module: dynamo, module: fakeTensor, module: inductor, no-scrub, oncall: distributed, oncall: export, oncall: pt2, pipeline parallelism, recompilations, tensor subclass, triaged",,Handle this module: dynamic shapes stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#107873 [BE] Consolidation of SymNode methods constant_int, maybe_as_int, etc, #113232 [Torch Inductor] Torch Inductor Better Support for GNN workload and Inductor Sparse Compiler, #121286 Assertion error (`assert op == ""==""`) in symbolic_shapes.prettify_results for operator `<=`, #124686 Unbacked SymInts: Should backwards graph with unbacked SymInts be recompiled with hints, #126738 FakeTensor `forward()` on `bert` attention when using FakeTensor returns a DataDependentOutputException, #126846 propagate_real_tensors can cause deadlocks for programs with collectives",0b50357d665b53ae7e8fdef7d9f5177b7a938c2535abd12e2ce7d003e18be549
issue-collapse:899d5df407c3f7f9,module: dtensor stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"115058, 117728, 118461, 119343, 119714, 122485, 124989, 124990, 126795, 134462, 144109, 147980, 148269, 150336",,,,"module: dtensor, module: functorch, module: nestedtensor, oncall: distributed","actionable, module: dtensor, module: functorch, module: nestedtensor, oncall: distributed, release notes: distributed (dtensor), triaged",,Handle this module: dtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#115058 [DTensor] enforcing foreach for optimizer causes jittering, #117728 [DTensor] addmm strategy cannot handle 0-element input tensors, #118461 [DTensor] DTensor fails to execute backward of index_put op., #119343 Use Dtensor and ColwiseParallel api to do Tensor Parallel for llama2 model but got communication error., #119714 [NestedSubTensor] add communication operators support, #122485 DTensor doesn't support expand_as dispath mode",5bbc80ab28e5fa7dfb2301875c0e8b208e746187b63f47d38aa0cb92398283ad
issue-collapse:2f606ce253d86cf3,module: dtensor stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"90585, 113761, 133221, 136946, 137392, 147570, 147814, 151858, 152717, 157545, 157976, 164891, 168304, 170936",,,,"module: dtensor, module: dynamo, module: nccl, module: random, module: threaded pg, oncall: distributed, oncall: pt2","better-engineering, feature, llm-amenable, module: dtensor, module: dynamo, module: nccl, module: random, module: threaded pg, oncall: distributed, oncall: pt2, triaged",,Handle this module: dtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90585 [threaded pg] All threads share one Random Number Generator, #113761 [DTensor][TP] 1-way TP raises index out of range, #133221 Tensor parallel for convolutions and groupnorm, #136946 DTensor does not support `nn.init.eye_`, #137392 segfault when using DTensor with nonblocking nccl comm, #147570 `view()` + modify-in-place fails silently with DTensor",f76fef61805a6dec37cfe0cf26ee18e9d9f9b94e645f34e00987ad8d128dd455
issue-collapse:9beceeb04eb8eabc,module: docs stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40967, 51623, 71552, 83244, 97118, 97597, 97966, 98409, 98499, 98805, 98871, 99310, 99316, 100584",,,,"module: docs, module: functorch, module: linear algebra, module: mps, module: nn, module: optimizer, module: pooling, module: python frontend, module: random, module: rnn","actionable, enhancement, module: docs, module: functorch, module: linear algebra, module: mps, module: nn, module: optimizer, module: pooling, module: python frontend, module: random, module: rnn, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40967 Inconsistent behaviour when parameter appears multiple times in parameter list, #51623 [docs] FractionalMaxPool2d : _random_samples arg is present in signature but undocumented., #71552 [docs] Tensor.uniform_ docs are not clear about whether from/to boundary values are included in sampling or not, #83244 torch.nn.Upsample's error message is inconsistent with the documentation, #97118 nn.Conv function to compute conv formula, #97597 Insufficient MPS Documentation",4f2edf5b433c67632d0d681b30c5ddc3552bdf9a08040733098eebcb73b351aa
issue-collapse:117a5f874869c273,module: docs stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"36217, 52262, 54209, 77053, 102437, 102515, 103640, 103846, 103847, 103960, 104033, 104154, 104872, 104962",,,,"module: docs, module: nans and infs, module: nn, module: numerical-reproducibility, module: numpy, oncall: java, oncall: mobile","actionable, module: NaNs and Infs, module: docs, module: nn, module: numerical-reproducibility, module: numpy, oncall: java, oncall: mobile, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36217 Rectify docs for MultiLabelSoftMarginLoss, #52262 tolist called on torch scalar does not return a list => proposal to support new arg force = True to provide a work around, #54209 [docs] Incomplete examples for torch.Generator, #77053 Incorrect documentation in ``gumble_softmax`` function., #102437 discuss.pytorch.org signup issue, #102515 pytorch java api documentation is not clear and does not cover example",2e3a29013e540c3e8bf2cf3107346ab09ae78682adfee68dd95ec0f00ff5b1e0
issue-collapse:24171c426d52d6e1,module: docs stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28906, 36091, 46253, 56262, 98566, 98600, 107703, 108197, 110084, 111621, 112715, 114364, 116569, 120255",,,,"module: docs, module: edge cases, module: nn, module: pooling, module: python frontend, oncall: quantization","actionable, enhancement, feature, low priority, module: docs, module: edge cases, module: nn, module: pooling, module: python frontend, oncall: quantization, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28906 Font used in documentation is not always sharp, #36091 [quantization] torch.quantized_lstm and torch.quantized_gru not documented, #46253 Add short guide for updating `torch.norm` calls to `torch.linalg.norm`, #56262 torch.nn.quantized.functional.conv_transpose1d/2d/3d support, #98566 `F.interpolate` and `F.grid_sample` - documentation error and bug, #98600 Wrong illustration in README.md",8f46eb5f4aa5164b072ff4488c83b66d9a271711c91699b09664fe4cf5b3e54e
issue-collapse:5aec04edde32075e,module: docs stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8069, 41571, 54393, 76510, 96073, 105460, 123113, 123344, 127031, 127497, 128495, 129050, 130254, 130448",,,,"module: docs, module: macos, module: nn, module: optimizer, module: python frontend, oncall: pt2","module: docs, module: macos, module: nn, module: optimizer, module: python frontend, oncall: pt2, topic: docs, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8069 Document torch.acos() behavior near -1 and 1, #41571 torch.cuda.BoolTensor uses 8 bits per element, not 1 bit as reported by element_size(), #54393 [docs] torch.optim.lr_scheduler, #76510 Docs source link broken for all methods defined on Tensor in _tensor.py, e.g. `Tensor.register_hook()`, #96073 Pytorch 2.0 installation tutorial does not work under Macbook, #105460 Specify version",0767035b5164d2aa312bd39d2840ed020b3c35a281abdf74dcc00d432f20d2ac
issue-collapse:4db7be0c2ec099da,module: docs stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"21518, 47917, 56136, 64006, 66806, 68009, 108522, 120735, 124395, 129122, 134691, 138017, 141277, 141417",,,,"module: docs, module: library, module: lint, module: lrscheduler, module: nn, module: numpy, module: serialization, oncall: releng","actionable, feature, module: LrScheduler, module: docs, module: library, module: lint, module: nn, module: numpy, module: serialization, oncall: releng, security, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21518 `attn_mask` in nn.MultiheadAttention is additive, #47917 LibTorch cannot load PyTorch exported model, #56136 Please add a tab of ""all"" packages in `Docs` option, #64006 Missing device and dtype description in torch.nn.MultiheadAttention, #66806 Missing doc for `torch.segment_reduce`, #68009 Properly document the `to.dtype_layout` overload of `Tensor.to`",22b499ebc7ff773751814a5452680a3ac7048191e168834b45f28ada49d24e8b
issue-collapse:279bc227de0f2fba,module: docs stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4474, 68352, 82430, 107130, 120407, 126518, 140296, 141616, 145357, 145907, 146086, 149284, 150252, 152381",,,,"module: docs, module: floatx (formerly float8), module: library, module: loss, module: memory format, module: nans and infs, module: nestedtensor, module: nn, module: python frontend, oncall: pt2","actionable, enhancement, module: NaNs and Infs, module: docs, module: floatx (formerly float8), module: library, module: loss, module: memory format, module: nestedtensor, module: nn, module: python frontend, oncall: pt2, topic: docs, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4474 [docs] Return value type hint to be specified in docs for torch.is_grad_enabled (i.e. -> bool), #68352 `Delta` not defined in the docs of `HingeEmbeddingLoss`, #82430 Minor inconsistency in description of `attn_output_weights` in MultiheadAttention docs, #107130 [docs] Document dtype conversions dtype.to_complex() dtype.to_real(), #120407 Dark mode please 🙏🏻, #126518 missing docs for torch.Tag",d8f44de6d29e682fc3d1336fa275d8ed41dbcfdddac28699de3f88f98d47fc90
issue-collapse:a5dce29be52dc8fb,module: docs stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"90633, 112585, 112587, 112588, 112600, 112638, 112641, 112644, 112646, 113186, 113187, 113188, 113190, 113192",,,,module: docs,"medium, module: docs, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90633 torch.min document not up to date, #112585 Fix docstring errors in _VF.py, _appdirs.py, hub.py, _classes.py, _storage_docs.py, _linalg_utils.py, torch_version.py, quasirandom.py, random.py, __future__.py, _lowrank.py, _vmap_internals.py, _sources.py, __config__.py, _lobpcg.py, _namedtensor_internals.py, #112587 Fix docstring errors in __init__.py, _tensor_docs.py, _meta_registrations.py, _tensor.py, #112588 Fix docstring errors in _torch_docs.py, serialization.py, overrides.py, _utils.py, #112600 Fix docstring errors in loss.py, #112638 Fix docstring errors in spectral_ops_fuzz_test.py, simple_timeit.py, timer_interface.py, op_benchmark.py, _stubs.py, fuzzer.py, compare.py, compile.py, interp.py, hipify_python.py, common.py, end_to_end.py, timer.py, __init__.py, sparse_fuzzer.py, blas_compare_setup.py",fec10202489aa6e5aa79c3c5844232824e440eae451efece110bf125b8dc55dd
issue-collapse:895e54d02f01ea7f,module: docs stale/low urgency  batch 27,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14318, 16257, 16268, 21290, 21682, 22220, 22232, 22494, 23054, 23072, 23079, 23787, 25991, 26338",,,,"module: docs, module: nn, module: rnn","enhancement, low priority, module: docs, module: nn, module: rnn, small, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14318 No code example for AdaptiveLogSoftmaxWithLoss, #16257 why check ArgumentInfo is_pod? suffer bugs, #16268 Complete dtype support for torch.norm, #21290 Title of docs page includes ""PyTorch master documentation"" even for non-master branches., #21682 C++ module API footgun: assigning to parameter doesn't update `parameters()` list, #22220 [doc] nn.Module.forward documentation unclear",c5077ea835eb6865551820c65b49d211dd88a14a2c1764a4ffb8d93667749465
issue-collapse:87b85ade15b3b1a8,module: docs stale/low urgency  batch 26,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20151, 26990, 27687, 28110, 28558, 28721, 28757, 28938, 29021, 29202, 29800, 29860, 30162, 30384",,,,"module: docs, module: named tensor, module: serialization, module: tensorboard, oncall: distributed","has workaround, low priority, module: docs, module: named tensor, module: serialization, module: tensorboard, oncall: distributed, quantization_release_1.3, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20151 RuntimeError: Given input size: (2048x1x1). Calculated output size: (2048x-5x-5). Output size is too small at /pytorch/aten/src/THNN/generic/SpatialAveragePooling.c:48, #26990 torch.quantized.modules.floatFunctional -- add a little color commentary to the doc, #27687 `nn.Sequential.__setattr__` appends to the execution list, #28110 Stable docs show dead __config__ section link, #28558 We should suggest using as_strided_ instead of set_, #28721 [docs] Improve docs of nn.MultiheadAttention",998488a102490c20c60b50e9eb22c3d4675d4e8b8ad4a8c52d5b68e907c24564
issue-collapse:bc51b3895cd4d256,module: docs stale/low urgency  batch 25,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30380, 30403, 30533, 30762, 31048, 31321, 32087, 32141, 32767, 34063, 34210, 34300, 34412, 34833",,,,"module: docs, module: numpy, module: sparse, module: type promotion","enhancement, module: docs, module: numpy, module: sparse, module: type promotion, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30380 Add Google Colab as an option on the 'Get Started' page., #30403 last updated timestamp, #30533 Document memory characteristics of in-place ops, #30762 Page for `torch.__config__` 404s, #31048 MathJax too small in Firefox, #31321 Ability to download docs HTML for offline use",d63b46da2b77efd5b7a87497bc3a30b88e3a145df72e62b625061b61fecdbd0c
issue-collapse:b2683b11bb0b00f1,module: docs stale/low urgency  batch 24,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"21779, 33055, 35766, 36059, 36307, 37416, 37440, 37441, 38595, 38720, 39130, 39399, 39619, 39705",,,,"module: docs, module: rpc, oncall: distributed","feature, module: docs, module: rpc, oncall: distributed, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21779 Use of word ""elements"" in `torch.utils.data` samplers, #33055 [docs] Strange order of items in docs contents in left pane, #35766 Clarify tensor storage communication behavior in RPC, #36059 Some tutorials cannot be found from both side panel and tutorial welcome page, #36307 [docs] Matrix transpose missing in formula in nn.functional.bilinear docs, #37416 [docs] Unclear return type of torch.randint and extra comma in arg spec",da020740b9b4b3243e528e085d975979f4fe6445b715d1eafc742aa8efbff419
issue-collapse:d4efa9dab24d4e9f,module: docs stale/low urgency  batch 23,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4400, 4530, 36386, 39426, 39799, 40417, 40969, 40971, 41106, 41148, 41282, 41392, 42112, 42159",,,,"module: docs, module: optimizer, module: tensorboard","enhancement, module: docs, module: optimizer, module: tensorboard, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4400 Installation Optimise For Chinese Users Who Behind the Wall, #4530 descriptor 'add' of 'torch._C._VariableBase' object needs an argument, #36386 Docs of distributed module do not include the full documentation for torch.distributed.launch, #39426 Format issue in `torch.quantization.add_quant_dequant` documentation parameter section, #39799 Tensor.size() API not correctly documented?, #40417 Missing explanation in torch.utils.tensorboard.add_histogram()",2978f2ce1a44fd5e18cd15b1c9f895cb5b0f280bbf1b7408ff4751c62db82e1e
issue-collapse:fa88c59493dd873f,module: docs stale/low urgency  batch 22,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5524, 42177, 42188, 42228, 42447, 42539, 42729, 42774, 42829, 42830, 43109, 43245, 43503, 44179",,,,"module: docs, module: multiprocessing, module: nn, module: optimizer","module: docs, module: multiprocessing, module: nn, module: optimizer, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5524 Redo torch.nn.functional docstring strategy, #42177 record_function no longer appears in the docs, #42188 Eliminate warning when cloning a tensor using `torch.tensor(x)`, #42228 [docs] Bitwise ops miss arguments, #42447 OneCycleLR argument `pct_start` used as a proportion not percentage, #42539 Docs for Tensor.copy_ non_blocking should mention pinned memory",cd8e3d4a15bac49d2de95eba13fdb2f3ffcde0f0b9bd2c24d6774e006910f1a0
issue-collapse:ba8c09b5df8bd89c,module: docs stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6165, 10735, 19160, 20935, 23305, 29510, 32872, 42529, 44415, 44552, 44917, 45483, 45717, 45767",,,,"module: docs, module: error checking, module: hub, module: memory format, module: nn, module: optimizer, module: scatter & gather ops, module: ux, module: vectorization","better-engineering, enhancement, internals, module: docs, module: error checking, module: hub, module: memory format, module: nn, module: optimizer, module: scatter & gather ops, module: ux, module: vectorization, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6165 Note about unusual stride situations in dev docs / make it easier to test for this in the library, #10735 Have all C++ modules expose a __file__ attribute, #19160 Suggest model.eval() in torch.no_grad (and vice versa), #20935 torch.full_like missing documentation for out input variable, #23305 make torch.utils._download_url_from_file public and add a docstring, #29510 torch.stack: bad shape error message",f189f049aa2e4461dc2cc34ac372de435a042dd9ef47b148b9dafd5c4ba1e82e
issue-collapse:53a4de7d541b7e10,module: docs stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9882, 43352, 46597, 46819, 47357, 48466, 48648, 48649, 49475, 50076, 50094, 50118, 50890, 51008",,,,"module: docs, module: linear algebra, module: loss, module: nn, module: numpy, module: optimizer, module: sorting and selection, oncall: distributed","module: docs, module: linear algebra, module: loss, module: nn, module: numpy, module: optimizer, module: sorting and selection, oncall: distributed, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9882 Unintuitive reduction of mini-batch loss for NLLLoss, #43352 Remove warning, and update documentation., #46597 Suggestion: link individual APIs in docs with tutorials that use them, #46819 Specify input dimensions where they are not obvious, #47357 Conv1D formula in docs is wrong, #48466 [master] new default signature: op=<ReduceOp.SUM: 0>from `op=<ReduceOp.SUM>",e1987869f285420ad635e8b286439c47a9976f2df11e166e7e5d046f63962743
issue-collapse:746b4e2b73878236,module: docs stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88327, 127886, 127915, 127917, 142290, 149393, 153531, 156072, 156258, 157101, 158895, 159234, 159536, 159886",,,,"module: docs, module: dynamo, module: fft, module: lint, module: nccl, module: nn, module: onnx, module: rpc, oncall: distributed, oncall: pt2","actionable, better-engineering, enhancement, medium, module: docs, module: dynamo, module: fft, module: lint, module: nccl, module: nn, module: onnx, module: rpc, oncall: distributed, oncall: pt2, topic: docs, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88327 MSE documentation is weak, #127886 Document the torch.fx.experimental.proxy_tensor.make_fx function, #127915 Document the torch.fx.traceback.preserve_node_meta function, #127917 Document the torch.cuda.profiler.start function, #142290 RFC: Display Supported Versions in API Documentation#, #149393 Torch RPC examples from docs say usage is deprecated.",a12da3eaab01c699326c5c7cabc1400201891fa58a7311be792253e95d23d812
issue-collapse:19d6508f531be58e,module: docs stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3920, 11551, 22755, 51197, 51256, 52688, 53360, 53477, 53642, 54213, 54368, 54387, 55450, 56328",,,,"module: docs, module: interpolation, module: loss, module: nn, module: optimizer, module: rnn, module: type promotion, module: viewing and reshaping","enhancement, feature, module: docs, module: interpolation, module: loss, module: nn, module: optimizer, module: rnn, module: type promotion, module: viewing and reshaping, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3920 [docs] Tensor.new is not documented, #11551 CrossEntropyLoss, ignore_index does not prevent back-prop if the logits are -inf, #22755 [RFC] InstanceNorm default affine value, #51197 [docs] use `versionadded`, `versionchanged` and `deprecated` directive, #51256 [search] duplicated entries in results, #52688 [BE] Consolidating scripts/ and tools/ folder",96bd8fe7aff4a5079719fe26a4ac2ee0054729711906df2acbbdb38d60863e68
issue-collapse:7c660a03ef807743,module: docs stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12532, 14790, 15259, 15298, 38639, 56329, 57112, 57712, 58395, 58585, 59627, 60096, 60400, 60561",,,,"module: docs, module: linear algebra, module: macos, module: multithreading, module: nn","enhancement, module: docs, module: linear algebra, module: macos, module: multithreading, module: nn, todo, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12532 Provide better documentation for torch.Size, #14790 Tensor.copy_() seems to work improperly with numpy/list indices, #15259 err:torch.nn.CrossEntropyLoss, #15298 Momentum problem (1-momentum is correct?) in BatchNorm2d, #38639 Difference between using a python list and nn.ModuleList, #56329 Change `other` to `src` in torch.Tensor.scatter_add_",45cc4d1cc06d2ae04069538c8bf546ab3f6a34b9e30384b86ba1f468dbf5f1c7
issue-collapse:7c126a5570d65c7f,module: docs stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16523, 20359, 28794, 37063, 60372, 60381, 60520, 60843, 61515, 62343, 62399, 62606, 62989, 63207",,,,"module: docs, module: nn, module: norms and normalization","feature, module: docs, module: nn, module: norms and normalization, todo, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16523 The word ""ints"" is not rendered correclty in docs., #20359 [feature request] Run examples from docs as tests, #28794 torch.from_file not documented, #37063 Misleading documentation in torch.nn.functional.fold, #60372 Docs for `Installing C++ Distributions of PyTorch` needs update, #60381 broken source link / missing anchors",af8e8aade614b4b24a2b9ee26dd4cd1622bcafdc557d63b78a8cfa0abacd9da5
issue-collapse:ff9951bbdcf2efc1,module: docs stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16896, 17612, 18434, 57272, 62162, 63257, 63559, 64133, 64138, 64292, 64384, 64539, 65069, 65166",,,,"module: docs, module: interpolation, module: lazy, module: linear algebra, module: nn, module: serialization, module: sparse, module: tensor creation","lazy, module: docs, module: interpolation, module: lazy, module: linear algebra, module: nn, module: serialization, module: sparse, module: tensor creation, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16896 `nn.Linear` allows 1d input tensors, #17612 The documentation to Module._version isn't visible., #18434 improve jit error message for legacy constructor, #57272 Consistent treatment of non-differentiability in linear algebra operations, #62162 Clarify sparse COO tensor coalesce behavior wrt overflow + how to binarize a sparse tensor, #63257 `help(torch.finfo)` doesn't give useful information even though class has documentation",167de52f94b9596f5b7dde73e397eb983e7be6924ba9b551b74a62fbd4450133
issue-collapse:b7db2d7edcf16b22,module: docs stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10377, 16899, 36426, 47351, 55752, 65526, 65702, 65739, 65949, 66364, 66907, 66921, 67482, 69363",,,,"module: docs, module: multithreading, module: nn, module: optimizer, module: performance, module: random","actionable, better-engineering, module: docs, module: multithreading, module: nn, module: optimizer, module: performance, module: random, todo, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10377 [Docs] Update broadcasting documentation for scalars / n-dimensional empty tensors, #16899 Give clearer guidance about multithreading in PyTorch, and how to disable it, #36426 Any reference to LPPool2d, #47351 no member named 'beta1' in 'torch::optim::AdamOptions', #55752 [question] Influence of divisibility of B, C, T by 16 on Conv1d (and Conv2d perf) with CuDNN, including presence of padding, #65526 [docs] Better explain signature of toch.randn",9a0f2fc3ea7b001fdde7fc919d973317876f620bdb1dcba8d091b7169ec3cb17
issue-collapse:a05044291f3101dc,module: docs stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16014, 16737, 69408, 69442, 69984, 70265, 70454, 70498, 70970, 71076, 71366, 71633, 72240, 73007",,,,"module: docs, module: loss, module: nn, module: norms and normalization, module: reductions, module: rnn, module: scatter & gather ops, module: viewing and reshaping","Stale, enhancement, module: docs, module: loss, module: nn, module: norms and normalization, module: reductions, module: rnn, module: scatter & gather ops, module: viewing and reshaping, todo, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16014 [docs] Missing argument description (value) in scatter_ function documentation, #16737 torch.save overwrite, #69408 `torch.hstack` should raise an error when tensor is 0 dimensional, #69442 The formula for KL-divloss is wrong in the document, #69984 Docs for torch.nn.MSELoss are confusing, #70265 Shape parameter inconsistency in torch.Tensor.view, torch.reshape, torch.Tensor.reshape",2194cbd066af315fda977af4aa422ed45ee21e647e24f93787f39e3823f76cc6
issue-collapse:6a24b088368da93b,module: docs stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"21567, 47201, 56396, 68017, 70419, 72788, 73699, 74937, 75510, 75659, 75737, 76138, 76338, 76627",,,,"module: docs, module: linear algebra, module: macos, module: nn, module: numpy, module: partial aliasing, module: python frontend, module: random, module: serialization, module: sorting and selection, module: tensor creation, module: testing, module: tests, module: viewing and reshaping","Stale, actionable, function request, module: docs, module: linear algebra, module: macos, module: nn, module: numpy, module: partial aliasing, module: python frontend, module: random, module: serialization, module: sorting and selection, module: tensor creation, module: testing, module: tests, module: viewing and reshaping, topic: docs, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21567 torch.bernoulli's parameter generator not documented, #47201 torch.save has a maximum size regardless of RAM, #56396 Tests in CI are run from the test/ directory, #68017 MacOS CI result is not in /README.md, #70419 LazyModules `cls_to_become` field exposes implementation detail, #72788 Docs bug: type annotations for linspace (and logspace) start and end arguments is wrong",732f4b956d42ece4c5fc40e4a7e760155f0ac27fa7bec6537fa04df2c4b61421
issue-collapse:b66c0c6d42414076,module: docs stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"42303, 63140, 77454, 78742, 79518, 80771, 80940, 82041, 82081, 82377, 82802, 82960, 83060, 83135",,,,"module: docs, module: elastic, module: loss, module: nn, module: numerical-reproducibility, module: primtorch, module: rocm, module: scatter & gather ops, oncall: r2p","actionable, better-engineering, module: docs, module: elastic, module: loss, module: nn, module: numerical-reproducibility, module: primTorch, module: rocm, module: scatter & gather ops, oncall: r2p, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42303 frame_length reference in pytorch.stft documentation, #63140 [documentation] torch.distributed.elastic: illustrate how to write load_checkpoint and save_checkpoint in Train Script, #77454 More clarity in doc for `torch.cuda.Event.record`?, #78742 Add a gallery of examples with sphinx-gallery, #79518 Display a ""reference"" link for ops that points to primTorch implementations, #80771 Anaconda is not a package manager",2fe2a09ef42bb2217f30ea1b7a7b327de90c836f69587a4124ef21c633968384
issue-collapse:1bd18451b5a3a266,module: docs stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"39716, 53261, 68169, 74022, 83234, 84140, 84537, 85558, 86020, 86563, 87388, 87494, 87902, 88192",,,,"module: docs, module: functorch, module: nn, module: numpy, module: optimizer, module: random, module: sorting and selection, module: sparse, oncall: distributed","actionable, module: docs, module: functorch, module: nn, module: numpy, module: optimizer, module: random, module: sorting and selection, module: sparse, oncall: distributed, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39716 Do not modify global random state, #53261 Dimension argument names in torch.diag_embed/diagonal vs. transpose/transpose_, #68169 torch.histogram: Sum of PDFs is num of bins instead of 1, #74022 Inconsistent implementation on SWA, #83234 torch.nn.ReplicationPad{1|2}d supports more input dimension than are written on documentation, #84140 Please include virtual/physical batch sizes in the tutorials",97a8d26f1628f9f3ac58f671c2e9fcb6665194143566ea94c51ce46dd087d728
issue-collapse:5dbaff026f08c047,module: docs stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"68058, 88423, 88992, 90114, 90485, 90832, 91210, 91618, 91842, 94654, 95462, 95953, 96523, 97026",,,,"module: docs, module: lrscheduler, module: memory usage, module: mps, module: numpy, module: optimizer, module: sparse","actionable, inference mode, module: LrScheduler, module: docs, module: memory usage, module: mps, module: numpy, module: optimizer, module: sparse, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#68058 Some lr-schedulers docs seems to have typos/missing information, #88423 view_as_real and split_with_sizes links in Tensor Views docs are broken, #88992 Incorrect version in the instructions on official website, #90114 documentation need to be as per python version, #90485 Tensor indexing and slicing documentation should explicitly state that indexing follows numpy semantics and link to the numpy indexing documentation., #90832 Inference Mode docs",1888718f5dc6d177dfa8d66284039f8cac05cd61859c34a5d7639ee25d2931e3
issue-collapse:4a5bf76cd9237c90,module: docs stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"98456, 142840, 161509, 162711, 163789, 165578, 165621, 165626, 166823, 167229, 167721, 167950, 168128, 171401",,,,"module: docs, module: infra, module: linear algebra, module: nn, module: nn.utils.parametrize, module: privateuse1, module: rnn, module: testing","actionable, enhancement, feature, module: PrivateUse1, module: docs, module: infra, module: linear algebra, module: nn, module: nn.utils.parametrize, module: rnn, module: testing, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#98456 README could use link to governance, #142840 Document spectral norm non-support on non-diagonalizable inputs., #161509 Update install commands in c++ Debugging, #162711 `scatter` / `gather` index type unclear, #163789 [docs] instructions to locally build docs are underspecified, #165578 Out of tree backend documentation does not seem accurate",2ee9970cfdfa5d3eb3a9d7931480a9cd1b513d4100af6d9307ee29c330773612
issue-collapse:634267a3aaddfc34,module: docker stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32030, 69685, 72784, 73920, 79604, 87196, 87701, 103832, 112876, 115421, 118508, 119843, 146701, 150049",,,,"module: docker, module: docs, oncall: distributed, oncall: releng","Stale, enhancement, module: docker, module: docs, oncall: distributed, oncall: releng, security, triaged",,Handle this module: docker stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32030 Dockerfile for people to quick-start contributing, #69685 Wrong PyTorch version in Docker image, #72784 Torch version in docker container does not match tag, #73920 pytorch/pytorch:1.8.1-cuda10.2-cudnn7-devel docker container contains cudnn 8.2, not 7.x as the name implies, #79604 Some unit tests are failing, #87196 Missing `docker` directory in `tools/`",5245617789ed22b2afb2b3e6fdd74943321d44a511b09a28bd783a838e230bdb
issue-collapse:84d0cd120ee85bd7,module: doc infra stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4954, 44593, 48287, 51575, 52256, 52747, 64541, 64711, 71760, 74152, 84932, 85351, 101369, 127530",,,,"module: doc infra, module: docs","better-engineering, enhancement, module: doc infra, module: docs, triaged",,Handle this module: doc infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4954 [docs] Docs website search finds duplicates and produces bad snippets, #44593 Docs include the verbose arg of `CosineAnnealingWarmRestarts` but the actual docstring in code doesn't, #48287 [doc] Error ""You've reached a dead end"" when opening torch.__config__ in the docs, #51575 [doc] being able to switch version while keeping the same doc, #52256 [docs] Bad docs SEO, #52747 [proposal] Make torchvision and other libraries' docs searchable via main docs search",c4371b1436590d304a3686d6e2a135c34a781663326137c09f12a1a349e0d73d
issue-collapse:1620ddfab6e4882d,module: distributions stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3364, 16291, 20343, 20682, 21673, 25783, 28441, 32406, 33898, 34949, 38474, 38611, 38612, 38613",,,,"module: distributions, module: docs, module: memory usage, module: random","enhancement, feature, module: distributions, module: docs, module: memory usage, module: random, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3364 type of torch.bernoulli and torch.multinomial inconsistent, #16291 Confusing documentation with distributions.Categorical about logits, #20343 torch.distributions.Binomial.sample() uses a massive amount of memory, #20682 [distribution] Support for various domain for AffineTransform, #21673 [FR] Diagonal Transform for Distributions, #25783 Detaching a distribution's `log_prob` to block gradients only w.r.t its parameters",17c2e6cd557a76faf595978c5ceba9bb23fe29b9c1b1c9c0176ba3d9b78d274a
issue-collapse:9fdd4f4add190dff,module: distributions stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1747, 4159, 4574, 11937, 12716, 12895, 18730, 32552, 36036, 41284, 46350, 50616, 52298, 56030",,,,"module: distributions, module: linear algebra, module: numerical-stability, module: ux","enhancement, feature, function request, module: distributions, module: linear algebra, module: numerical-stability, module: ux, todo, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1747 In-place bernoulli_ has more functionality than torch.bernoulli with output parameter, #4159 Variable outputs of stochastic functions should never require grad, #4574 `from` keyword in `random_` gives error, #11937 [feature request] Kumaraswamy distribution, #12716 cdf in torch.distributions.bernoulli throws NotImplementedError, #12895 RelaxedOneHotCategorical not implementing entropy (and other abstract methods)",9d26da761da9e6ce94a762fa85e236f0390a5d798136e938c243f9eb47fb6607
issue-collapse:4d4b804432e9c870,module: distributions stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15288, 29044, 54394, 56881, 59822, 60916, 61727, 65811, 67390, 68746, 69281, 71689, 71951, 73585",,,,"module: distributions, module: docs, module: half, module: nans and infs, module: numerical-stability, module: random, module: tests","Stale, enhancement, feature, module: NaNs and Infs, module: distributions, module: docs, module: half, module: numerical-stability, module: random, module: tests, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15288 possible unsafety in torch.distributions.kl_divergence for Bernoullis, #29044 Add Wishart and inverse-Wishart distributions, #54394 Half Normal Log_Prob not defined for 0, #56881 [FR] integral discrete distributions should support dtype=int64, #59822 TransformedDistribution's log_prob gradient is inconsistent w.r.t. cache_size, #60916 Scipy 1.7.0 may cause some test failures",ae8113ec5b32296dc12fe48413ec9b2701ed881dd61b7784598921776e165ed7
issue-collapse:5b81a719a4451ad6,module: distributions stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7857, 42407, 45115, 62178, 72545, 73070, 73826, 73848, 73916, 76030, 76160, 76709, 83564, 85375",,,,"module: distributions, module: docs, module: error checking, module: molly-guard, module: numpy, module: tests","feature, module: distributions, module: docs, module: error checking, module: molly-guard, module: numpy, module: tests, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7857 Feature Request: Logistic Distribution, #42407 torch.distributions.multinomial.Multinomial cannot be used in batch, #45115 Standardized Distributions, #62178 Wrong implementation of method log_prob in torch.distributions.negative_binomial, #72545 Split up torch.distributions docs into multiple pages, #73070 `enumerate_support` for continuous distributions",849d7740dd50a4635e22b988e5103ba9ef4c1decc6158f8d4c4e2c4ff35921ab
issue-collapse:b65d38a9c05314f1,module: distributions stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"43250, 52625, 87468, 89684, 91574, 92260, 95129, 95548, 96494, 97656, 99630, 102811, 107821, 108152",,,,"module: distributions, module: edge cases, module: error checking, module: linear algebra, module: macos, module: nans and infs, module: python frontend","enhancement, feature, module: NaNs and Infs, module: distributions, module: edge cases, module: error checking, module: linear algebra, module: macos, module: python frontend, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43250 Support for Multi-Categorical in torch.distributions, #52625 Add slicing to distributions, #87468 Categorical fails simplex validation after its own normalisation on CUDA, #89684 `positive_semidefinite` constraint fails on CUDA 11.7, #91574 Add BlockWise Distribution Support to the torch.distributions Package, #92260 distributions.Beta returning incorrect results at 0 and 1",61f2ce2e6f72c7a2384731d3f80ff9f72c1239d48be538e287fb5c5caa00ffcf
issue-collapse:f66a85f58e666f75,module: distributions stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15738, 21553, 49291, 91863, 97156, 98939, 110366, 113579, 119524, 122021, 122185, 127387, 130950, 133754",,,,"module: distributions, module: docs, module: nn","enhancement, feature, module: distributions, module: docs, module: nn, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15738 Beta Distribution values wrong for a=b---> 0, #21553 Feature Request: beta cdf, #49291 add inverse cdf for Chi-square Distribution, #91863 torch.Categorical samples indexes with 0 probability when given logits as argument, #97156 Implement `torch.distributions.Poisson.cdf()`, #98939 torch.dist with minus norm returns tensor(0.), while with -inf can return result",c8028e1d3198063f170b642a472bb48d9caa9c04961caf95d708885cab9dc5d8
issue-collapse:51550ffffe98bfe5,module: distributions stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"26334, 37162, 39178, 48273, 75173, 121220, 136750, 142825, 144660, 145729, 148247, 149718, 164061, 166243",,,,"module: distributions, module: docs, module: random","enhancement, low priority, module: distributions, module: docs, module: random, triaged",,Handle this module: distributions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#26334 Supporting ""cdf"" for Student-T distribution, #37162 torch.distributions bug in RelaxedOneHotCategorical.log_prob, #39178 Add support for rsample to MixtureSameFamily Distribution, #48273 torch.multinomial example is incorrect, #75173 LowRankMultivariateNormal doesn't work with 0 diagonal, #121220 Misleading `log_abs_det_jacobian` shape for `StackTransform`?",b24a799aaf2242bab6fe864f8e9eb6a39a62a6c04b871d916f7ecc624c30af8e
issue-collapse:5835e0ebf302cd13,module: dispatch stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28532, 32549, 32937, 35038, 35040, 42818, 42858, 42902, 43429, 44657, 44809, 54096, 54283, 54356",,,,"module: dispatch, module: distance functions, module: internals, module: performance","better-engineering, module: dispatch, module: distance functions, module: internals, module: performance, triaged",,Handle this module: dispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28532 TestTorch.test_c10_layer_norm fails if you run it on a build of PyTorch with BUILD_CAFFE2_OPS=0, #32549 ""Tried to register multiple operators with the same name and the same overload name"" error is confusing, #32937 Operation Registration Error, #35038 Rename Dispatcher::findSchema to Dispatcher::findOperator, #35040 Deprecate and remove RegisterOperators, #42818 Can't registered boxed kernel for operator that doesn't support boxing",505ecfe700f61f007e47f7c6dcfda05149bcbd41b6a3ace2a8a801ff8112ce66
issue-collapse:740fc99016464464,module: dispatch stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"29844, 45203, 54082, 54614, 57326, 57369, 57957, 60997, 65426, 65825, 66563, 67279, 76175, 78519",,,,"module: dispatch, module: internals, module: lazy, module: norms and normalization, module: performance","better-engineering, enhancement, feature, inference mode, module: dispatch, module: internals, module: lazy, module: norms and normalization, module: performance, triaged",,Handle this module: dispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#29844 Fold DispatchStub into c10 dispatcher, #45203 Staged backend boxed fallback (per-operator precomputation / precompute), #54082 Dispatch table for linalg_norm is fishy, #54614 Dispatcher TLS to bypass loads of dispatch key from tensor arguments, #57326 Renaming Autograd dispatch keys to ADCreateGraph, #57369 Delay errors of inference tensor to backward pass?",224a1790d9e51087716a91f03e02482124832056c757a5e580213a43831cd9cd
issue-collapse:287ec0454212ad51,module: dispatch stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"78743, 81722, 85122, 85815, 102423, 109724, 112666, 114440, 119971, 124083, 124630, 140322, 148029, 164872",,,,"module: dispatch, module: docs, module: higher order operators, module: library, module: pt2-dispatcher, oncall: pt2, oncall: quantization","inference mode, internal ramp-up task, module: dispatch, module: docs, module: higher order operators, module: library, module: pt2-dispatcher, oncall: pt2, oncall: quantization, tensor subclass, triaged",,Handle this module: dispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#78743 Expose docs from the yaml for each torch.Tag in Python, #81722 Inconsistent naming convention for end of enum in DispatchKey, #85122 PT Dispatcher confusing error message ""There were no tensor arguments to this function"", #85815 string interning for dispatcher operator names, #102423 Add default device function interface for device-aware apis, #109724 assert_is_valid_input_type is too weak",0342ae36a0f44d3da1f3f9d913f2ad54e3663f9b5b8defa38896ac09cc8ee931
issue-collapse:b77bb5a35e4487c8,module: devx stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"102227, 102559, 103354, 116539, 125297, 125467, 127788, 131357, 141299, 144890, 147387, 152190, 153273, 167558",,,,"module: devx, module: tests, oncall: pt2","enhancement, feature, module: devx, module: tests, oncall: pt2, triaged",,Handle this module: devx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#102227 lintrunner should fail on badly formatted docstrings, #102559 Mergebot should merge non-stacked PR, #103354 Add a requirements.txt for windows pip packages, #116539 Auto bisect for broken periodic CI jobs, #125297 Make the `sccache` cache easily available to all pytorch contributors in readonly mode, #125467 Inaccurate filename due to test class metaprogramming that doesn't set __file__ and __name__",c4db286cbe2adb861cd528826a95300ff50cd32a1222c1ec1460e7d93fb29346
issue-collapse:94cab38a35f74dde,module: determinism stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17798, 23151, 27542, 51493, 53352, 58471, 59783, 78249, 85096, 88591, 94388, 97236, 101246, 107011",,,,"module: determinism, module: distributions, module: docs, module: loss, module: mps, module: nn, module: numerical-stability, module: performance, module: pooling, module: random, module: rnn, module: scatter & gather ops, module: sorting and selection, module: tests","OSS contribution wanted, enhancement, feature, function request, module: determinism, module: distributions, module: docs, module: loss, module: mps, module: nn, module: numerical-stability, module: performance, module: pooling, module: random, module: rnn, module: scatter & gather ops, module: sorting and selection, module: tests, triaged",,Handle this module: determinism stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17798 Feature Request: deterministic CUDA torch.nn.CTCLoss, #23151 Deterministic mode for scatter_add operation, #27542 Make topk sort stable, #51493 Add branch predictor hints to prefer `Context::deterministicAlgorithms() == false`, #53352 OpInfo mechanism to test for nondeterminism, #58471 Why the same training code running on different GPUs (V100 vs P100) brings a significiant difference in outputs and thus train loss / model accuracy?",ebcb80f33036b27a95d88b29a9adc3c4308bcc90ad35970e4d5c61abe8110d6b
issue-collapse:118ef4e1ac7352b4,module: determinism stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72766, 82843, 98208, 107999, 108341, 110698, 120689, 130968, 140513, 143655, 147595, 155552, 161029, 166114",,,,"module: determinism, module: floatx (formerly float8), module: intel, module: mkl, module: mkldnn, module: mps, module: multiprocessing, module: nn, module: pooling, oncall: jit, oncall: pt2","actionable, enhancement, feature, low priority, module: determinism, module: floatx (formerly float8), module: intel, module: mkl, module: mkldnn, module: mps, module: multiprocessing, module: nn, module: pooling, oncall: jit, oncall: pt2, shadow review, triaged",,Handle this module: determinism stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72766 Feature Request: Deterministic MaxPool3d and AvgPool3d, #82843 Tensor operation hangs when used with multiprocessing, #98208 Add a deterministic version of reflection_pad2d_backward_cuda, #107999 `upsample_bilinear2d_backward_out_cuda` is nondeterministic, #108341 `adaptive_avg_pool2d_backward_cuda` is not deterministic, #110698 max_pool3d_with_indices_backward_cuda and avg_pool3d_backward_cuda does not have a deterministic implementation",6b7b7ad87e26d0958665de6d4e36a065f3a28c8df27ccbb6c2d38556fca2de39
issue-collapse:21b2d6481728afbe,module: deprecation stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50273, 50274, 50275, 50276, 55948, 60548, 65752, 65910, 65936, 86968, 97014, 117540, 120584, 162782",,,,"module: deprecation, module: docs, module: error checking, module: fft, module: lrscheduler, module: nn, module: numpy, module: special, module: testing","better-engineering, module: LrScheduler, module: deprecation, module: docs, module: error checking, module: fft, module: nn, module: numpy, module: special, module: testing, triaged",,Handle this module: deprecation stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50273 torch.cross is divergent from np.cross, #50274 torch.equal is divergent from np.equal, #50275 torch.transpose is divergent from np.transpose, #50276 torch.meshgrid is divergent from np.meshgrid, #55948 Deprecate torch.stft returning real-valued tensors and torch.istft accepting real-valued inputs, #60548 Ensure warnings relate to user code with stacklevel",0b5ab49c740210e625853fc0877c6ce23327435f07e8291a2296b989611b7724
issue-collapse:9516f54b632e2c3c,module: ddp stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"45943, 57593, 62862, 63041, 63211, 63290, 64637, 65436, 65637, 68781, 70222, 71330, 71377, 73600",,,,"module: ddp, module: docs, module: lazy, module: memory usage, module: optimizer, module: partial aliasing, oncall: distributed","Stale, better-engineering, feature, module: ddp, module: docs, module: lazy, module: memory usage, module: optimizer, module: partial aliasing, oncall: distributed, pt_distributed_rampup, triaged",,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#45943 Improve distributed documentation for NCCL_BLOCKING_WAIT, #57593 [RFC] Provide an API for Structural Performance Tips in DDP, #62862 Convert unset variables in `Reducer::Timer` to use c10::optional, #63041 [RFC] Should DDP support custom reduction logic for registered module buffers?, #63211 Memory leak in `ddp_zero_hook.py` hooks, #63290 DistributedDataParallel documentation",6d304b4fe0c740bfe97ec8ab955ceaebc329ac3b5f65f9eafc3180682ea53851
issue-collapse:8e2b55276beae6b7,module: ddp stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"58813, 74185, 74907, 78157, 78159, 79164, 79171, 79464, 79542, 79629, 80080, 80420, 83074, 83245",,,,"module: ddp, module: mta, module: optimizer, module: performance, module: sparse, module: xla, oncall: distributed","better-engineering, module: ddp, module: mta, module: optimizer, module: performance, module: sparse, module: xla, oncall: distributed, pt_distributed_rampup, triaged",,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#58813 Single-Process Multi-GPU is not the recommended mode for DDP, #74185 [DDP] Parallelize initialization collectives, #74907 [Proposal] Use batched oprations to accelerate PowerSGD, #78157 [Optimizer Overlap] Custom optimizer registration, #78159 [Optimizer Overlap] Parameter group support, #79164 PyTorch/XLA's DDP XLABackend is broken by upstream change",5f16b7d5a42b0e62f20527ae4af76e335669037fcda4eec150b6f04ce88751c6
issue-collapse:514529cbe8bff0d5,module: ddp stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"79299, 83320, 84414, 89942, 90742, 95872, 97030, 109774, 127873, 130932, 149762, 150734, 156680, 158936",,,,"module: ddp, module: dynamo, module: floatx (formerly float8), module: fsdp, module: functorch, module: nccl, oncall: distributed, oncall: pt2","enhancement, feature, module: ddp, module: dynamo, module: floatx (formerly float8), module: fsdp, module: functorch, module: nccl, no-scrub, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, triaged",,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#79299 PyTorch gets stuck when using an NVLink/A6000 and more than two GPUs, #83320 Torch1.10.2 is slower than torch1.9.1, #84414 several questions about pytorch DDP, #89942 Random K compression hook in PyTorch DDP, #90742 Adopt full_backward_pre_hook in DDP, #95872 corrupted size vs prev size error",13b798bed06d4492a501a575293cd5074667f80205319617b80a68fcb8a852aa
issue-collapse:c2a57c9d60964a25,module: dataloader stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2407, 2724, 3281, 5231, 6350, 21016, 24984, 31232, 33413, 37409, 41329, 43467, 44528, 44631",,,,"module: dataloader, module: docs, module: multiprocessing, module: performance, oncall: distributed","enhancement, feature, module: dataloader, module: docs, module: multiprocessing, module: performance, oncall: distributed, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2407 Hard-negative mining using __getitem__ directive in Dataset class, #2724 DataLoader gets stuck after model initialization, #3281 DataLoader ""casting"" non statndard objects to lists, #5231 [feature request] Stratified splits in random_split function, #6350 worker assignments in torch.utils.dataloader.py, #21016 [utils.bottleneck] throws initialization error for cuda profiling",e007587c07a5fc713653bc5899b78f155dee7d7ca266796b2c1e5d169e06f3cd
issue-collapse:bb917ee632d68528,module: dataloader stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7423, 9171, 9559, 9753, 10165, 46530, 47563, 48702, 49727, 49849, 50831, 50939, 51418, 51663",,,,"module: dataloader, module: docs, oncall: distributed","enhancement, feature, module: dataloader, module: docs, oncall: distributed, todo, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7423 Multi queue for dataloader when workers > 1, #9171 Mismatch in behaviour of WeightedRandomSampler and other samplers, #9559 [feature request] Rename `Subset` -> `Resample` to reflect wider use, #9753 dataloader stuck at sched_yield =0, #10165 torch.utils.data.random_split() returns dataset index as tensor, #46530 [Request] Batched Dataset->DataLoader interface",80444da519c02c8d32a6f871f8f6090ece588a1c5dec057855773a4eed579cc9
issue-collapse:f02c1c50255606c9,module: dataloader stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3356, 11062, 11727, 11790, 18317, 52850, 53337, 54871, 58847, 59438, 61503, 61665, 65665, 66251",,,,"module: dataloader, module: ddp, oncall: distributed","enhancement, feature, module: dataloader, module: ddp, oncall: distributed, todo, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3356 Data sampling seems to be more complicated than necessary, #11062 non-shuffling data loaders can affect random states, thus the results of shuffling data loaders., #11727 Semaphore leaks in dataloader, #11790 set num_workers on the dataloader make the jupyter kernel crash at the almost end of the epoch, #18317 Add a RandomBatchSampler ?, #52850 How to skip the images in a custom dataset and deal with None values?",33c33107b13bd5a9e3a643d93fab7c36536f9112d6046e5b0d2c30689c6dea94
issue-collapse:62af1bf30f409cc8,module: dataloader stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"934, 21645, 42402, 46802, 65216, 66223, 67439, 68519, 70158, 70386, 73417, 74337, 76887, 78921",,,,"module: dataloader, module: multiprocessing, module: multithreading, module: performance, module: tests","Stale, enhancement, feature, module: dataloader, module: multiprocessing, module: multithreading, module: performance, module: tests, todo, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#934 dataloader parallels over elements vs over batches, #21645 Batch Dataloader and Dataset, #42402 Reading data speed slower than tensorflow, #46802 DataLoader gives ""Broken pipe"" error on Linux platform, #65216 Add function that retrieves a batch from a DataLoader, #66223 Data Loader tests hang when run in ASAN test job",4bbdc5b269cd03972b1049fa51cce26525405b6cec61f3206b8b9917d125cc31
issue-collapse:adbfb8225bea8ca9,module: dataloader stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50201, 61281, 75147, 78987, 80594, 81412, 82185, 82219, 82583, 85775, 87003, 87955, 94131, 96629",,,,"module: dataloader, module: deadlock, module: memory usage, module: multiprocessing, module: openmp, oncall: distributed","module: dataloader, module: deadlock, module: memory usage, module: multiprocessing, module: openmp, oncall: distributed, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50201 Excessive memory usage caused by Samplers storing lists of indices, #61281 DataLoader with IterativeDataset throws an error when providing a BatchSampler, #75147 Dataloader hangs. Potential deadlock with `set_num_threads` in worker processes?, #78987 DataLoader leaking resources?, #80594 RuntimeError: DataLoader worker (pid 22822) is killed by signal: Aborted., #81412 num_worker and prefetch_factor in DataLoader do not scale",68b79a3d0002ad2954db51d6e8501a4cdf2a7232063b3255669a968c03fa6f73
issue-collapse:e0172f53ba6cdc4e,module: dataloader stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23048, 51011, 61457, 97498, 98792, 99584, 100012, 102511, 105799, 107392, 107443, 116117, 116479, 116558",,,,"module: dataloader, module: determinism, module: error checking, module: multiprocessing, module: performance, module: typing, oncall: distributed","enhancement, feature, low priority, module: dataloader, module: determinism, module: error checking, module: multiprocessing, module: performance, module: typing, oncall: distributed, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23048 Add automatic tuning flags to utils.data.dataloader, #51011 DataLoader is slow in spawned processes, #61457 using DistributedSampler occur RuntimeError: Expected a 'cuda:0' generator device but found 'cpu', #97498 Function Registry for extending collate_fn, #98792 DataLoader doesn't accept non-cpu device for loading., #99584 Training Faster R-CNN model with COCO dataset has been consistently unsuccessful.",a53ada3c0c854fcf77fec4c74a62d092cfff2dc29bfa0af90aef25df81cc50c1
issue-collapse:af1f500b4e44d3f2,module: dataloader stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"27421, 40972, 62066, 77159, 77646, 91353, 98002, 110029, 110820, 113600, 119281, 123262, 125844, 126620",,,,"module: dataloader, module: memory usage, module: mps, module: nn, oncall: distributed","module: dataloader, module: memory usage, module: mps, module: nn, oncall: distributed, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#27421 [dataloader] Sampler abstract constructor API minor proposal, #40972 len of dataloader when using iterable dataset does not reflect batch size, #62066 Memory Leak Found in Persistent DataLoader, #77159 A somewhat cryptic error message (for newcomers) - ""Cannot re-initialize CUDA in forked subprocess"" - report and suggestion for a possible solution, #77646 Werror=nonnull in dataloader.cpp (part of tests), #91353 Perf reduction due to munmap with dataloader pinning thread ?",e2aa6b025505838c42ca401842f643ce51dcadc6fd5c1a0f5bf2727397dceb05
issue-collapse:34f3c0b8adc055e4,module: dataloader stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11372, 31689, 32034, 65198, 71872, 106634, 119123, 127072, 130610, 131631, 131883, 131931, 132919, 136588",,,,"module: dataloader, module: docs, module: multiprocessing, module: performance, module: typing","enhancement, module: dataloader, module: docs, module: multiprocessing, module: performance, module: typing, small, todo, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11372 Suggest: DataLoader add device parameter, #31689 DataLoader does not consider default floating point type, #32034 Extend `ConcatDataset` to return dataset index, #65198 Multi-processing leaking file descriptors, #71872 Clarify the behavior of DataLoader sampler and batch_sampler parameters, #106634 Confusing error message for DataLoader with num_workers=0 and non-zero timeout",06e850f2c079e844a6022fd149451d3ae6a60daf8e3721f4f3a0148856f77d26
issue-collapse:38b9652b281b406f,module: dataloader stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"42654, 67602, 76750, 92134, 97432, 117586, 135409, 137882, 142884, 149837, 153594, 154810, 155157, 169078",,,,"module: dataloader, module: memory usage, module: multiprocessing, module: performance, module: windows","enhancement, feature, module: dataloader, module: memory usage, module: multiprocessing, module: performance, module: windows, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42654 RuntimeError: each element in list of batch should be of equal size, #67602 [feature request] Add no-op set_epoch on Sampler class, #76750 Bug in dataloader iterator found by mypy, #92134 [Bug][Dataloader] unable to mmap 2048 bytes from file <filename not specified>: Cannot allocate memory (12), #97432 suspicious memory leak when increase DataLoader's prefetch_factor and enable pin_memory, #117586 PyTorch 2.1.0 High memory utilization compared to PyTorch 2.0.1",1ee2f1ccc22f15fb6eb9663a457c6ce085addbb2a1b168d96878c566909ac99c
issue-collapse:7e4b3f660422cd5c,module: data stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"73141, 75582, 80134, 88062, 102368, 106540, 107217, 107854, 108861, 109968, 110156, 113748, 116403, 123172",,,,"module: data, module: dataloader, module: deprecation","actionable, enhancement, module: data, module: dataloader, module: deprecation, triaged",,Handle this module: data stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#73141 libtorch: `DistributedRandomSampler` uses the same random order in every epoch, #75582 Add `balance` flag to `random_split`, #80134 [feature request] Add support for a custom DatasetFetcher in DataLoader, #88062 Cannot import `traverse_dps` from torch.data.utils.graph, #102368 Issue with ShufflerIterDataPipe in torch 1.13.1, #106540 Dataset with Queue issue",0ffa2757f7d27be7ad68f14978cb790190e961b3fa1f0c6888ef45bb233422b0
issue-collapse:ac132a4bcbd1e923,module: data stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"56360, 91252, 128015, 128497, 129224, 129225, 131113, 132492, 138957, 148807, 153536, 160849, 164862, 169080",,,,"module: data, module: dataloader, module: tensorboard","feature, module: data, module: dataloader, module: tensorboard, triaged",,Handle this module: data stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#56360 Support for stereo audio data in from torch.utils.tensorboard.SummaryWriter, #91252 Open file leak when dataloader is using persistent_workers and pin_memory AND you create multiple dataloaders., #128015 Best approach for dynamically changing batch sizes on iterable datasets, #128497 [bug]When the variable in the dataset is of type cuda, calling the dataloader will result in 0, #129224 _If pin_memory_thread is alive but dataqueue is empty, it will fall into a dead loop,, #129225 Why need to transpose when collate a sequence data in dataloader?",9e465738103eecb10ab996c6f1a17ea83b571fb303742f601ed62e6c08a8aa98
issue-collapse:b089cc1a053741a0,module: data parallel stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2312, 7890, 31478, 32444, 33342, 34001, 34516, 34941, 35118, 35208, 37444, 37484, 37814, 38543",,,,"module: data parallel, module: memory format, module: nn, module: performance, oncall: distributed","awaiting response (this tag is deprecated), enhancement, module: data parallel, module: memory format, module: nn, module: performance, oncall: distributed, small, triaged",,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2312 DataParallel is not compatible with pack_padded_sequence, #7890 [feature request] batch_first of RNN hidden weight for Multi GPU training, #31478 DataParallel has different tensor copy behavior if batch size = 1, #32444 DataParallel does not work with sparse parameters, #33342 malloc(): memory corruption (fast), #34001 _broadcast_coalesced_reshape doesn't respect zero-size tensor",c5f737c9d36d4333359875a819ec38fe0ee5cad46ae9378bfd9fdf25e0554a63
issue-collapse:579a6084e9d90e98,module: data parallel stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7358, 7365, 14461, 24155, 33785, 38936, 38940, 40978, 42183, 42478, 45840, 46381, 50170, 51385",,,,"module: data parallel, module: dataloader, module: nccl, module: nn, module: performance, module: rnn, oncall: distributed","enhancement, module: data parallel, module: dataloader, module: nccl, module: nn, module: performance, module: rnn, oncall: distributed, small, triaged",,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7358 DataParallel on list inputs, #7365 nn.DataParallel fills None grads with 0, #14461 The speed of scatter is influenced by the data size while using nn.DataParallel, #24155 ""To compact weights again call flatten_parameters()"" is printed every step for every GPU, #33785 Deepcopy fails with nn.parallel.replicate, #38936 [FR] DataParallel arg rename device_ids->devices",ad82d446685e663eb6cd74c00d87b79a606f0f383174a84ade170ff3cb1db1ac
issue-collapse:6d776c20792b5172,module: data parallel stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11645, 15161, 15260, 16700, 33800, 43120, 43754, 48714, 52597, 53405, 55304, 58218, 62130, 68931",,,,"module: data parallel, module: multi-gpu, module: performance, module: rnn, oncall: distributed","enhancement, module: data parallel, module: multi-gpu, module: performance, module: rnn, oncall: distributed, triaged",,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11645 One GPU is more memory efficient than Multiple GPUs, #15161 Maybe a bug when using DataParallel, #15260 MultiGPU for gru, #16700 [feature request] Store accumulated gradients in separate GPU or on CPU memory, #33800 Can't get module gradient in autograd.Function's custom backward when DataParallel is used, #43120 When modified the model python file, the pytorch will raise the KeyError of this file",9dc0fb04da3a7d614d19230e3405cd71c34f90749a8793fac26eef4146be474c
issue-collapse:7842d587212bd9ce,module: data parallel stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7457, 31460, 37246, 41327, 41681, 43259, 44938, 50210, 70447, 73218, 77576, 83315, 122080, 122568",,,,"module: data parallel, module: deadlock, module: nn, oncall: distributed","Stale, actionable, enhancement, module: data parallel, module: deadlock, module: nn, oncall: distributed, triaged",,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7457 [feature request] torch.nn.DataParallel should work nicely both for cpu and gpu devices, #31460 DataParallel doesn't properly handle kwargs, #37246 `max_norm` parameter on nn.Embedding will fail inside nn.DataParallel, #41327 torch.nn.parallel.scatter_gather.gather can't gather outputs that are dataclasses, #41681 TypeError: cannot create 'generator' instances, #43259 RuntimeError: Expected to have finished reduction in the prior iteration before starting a new one.",e2c6cc60f0e2783906678beb2e021c9f5969b83ca84f96d82e1b4e7c8aee80be
issue-collapse:a807bc1d14448ab3,module: custom-operators stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"124645, 127337, 130282, 139494, 142353, 143466, 144344, 150972, 152720, 155386, 155616, 156622, 161001, 162870",,,,"module: custom-operators, module: dispatch, module: dynamic shapes, module: inductor, module: library, module: pt2-dispatcher, module: vllm, oncall: pt2","actionable, dynamo-triage-jan2025, feature, internal ramp-up task, llm-amenable, module: custom-operators, module: dispatch, module: dynamic shapes, module: inductor, module: library, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged, vllm-compile",,Handle this module: custom-operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#124645 Registering function that takes `const SymInt&` to op that accepts `SymInt` leads to cryptic error, #127337 Add option for custom ops to automatically get a FakeTensor kernel (during static shapes), #130282 [custom_ops] We should be able to mark an operator as non-differentiable., #139494 Mutating custom ops slower than non-mutating custom ops., #142353 Extend custom op fwd/bwd functions to support data types beyond the basic set, #143466 Support Dict Parameter Type for custom_op",f23aa90f693dd265be979e18c09ad0a107ebe28a3d01d27fcf6715a4d083327a
issue-collapse:2ac8502aa230ff29,module: cudnn stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4241, 4930, 34977, 45053, 51044, 51776, 51933, 57921, 62034, 62926, 64025, 70487, 77844, 107751",,,,"module: cudnn, module: data parallel, module: dependency bug, module: memory format, module: performance, module: rnn","feature, function request, module: cudnn, module: data parallel, module: dependency bug, module: memory format, module: performance, module: rnn, triaged",,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4241 single-gpu works but multi-gpu hangs, #4930 [feature request] Type-1 Multi-layer bidirectional RNN, #34977 `conv2d` is slow with specific shapes of channels_last tensors, #45053 about benchmark issue, #51044 CuDNN 8 with benchmark=True takes minutes to execute for certain configurations, #51776 BatchNorm3d error : CUDNN_STATUS_NOT_SUPPORTED",2231c3d180293bf2e8f8184156305bfec04f6ad128d7c5f17951344959361004
issue-collapse:6f01dcd0f31ee1f6,module: cudnn stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28777, 40134, 77955, 80259, 104697, 109010, 113590, 133780, 134977, 144913, 146096, 149090, 166557, 169146",,,,"module: cudnn, module: dependency bug, module: determinism, module: memory format, module: nccl, module: nn, module: numerical-stability, module: random, module: sdpa, module: tests","module: cudnn, module: dependency bug, module: determinism, module: memory format, module: nccl, module: nn, module: numerical-stability, module: random, module: sdpa, module: tests, topic: build, triaged",,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28777 cudnn.determinstic=True causes dilated convolution to be >10x slower, #40134 About torch.backends.cudnn.deterministic issue, #77955 NaN tensor values problem for GTX16xx users (no problem on other devices), #80259 Cudnn batch norm kernel (batchnorm_bwtr_nhwc_semiPersist) gets blocked by overlapping NCCL all_reduce calls, #104697 LSTM built-in dropout not reproducible on GPU, #109010 The following will always fail on NixOS",7d7acb9279d265a170aca2082231e219535ae2c4b999dacc4d44fcdadbfb0e77
issue-collapse:f3797a385727f291,module: cuda stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"28829, 74235, 98925, 115618, 118991, 119131, 126344, 129523, 132218, 140214, 140687, 140930, 141915, 142127",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: determinism, module: linear algebra, module: memory usage, module: nestedtensor, module: numerical-stability, module: padding, module: sdpa, module: sparse, module: tests","actionable, enhancement, feature, module: CUDACachingAllocator, module: cuda, module: cudnn, module: determinism, module: linear algebra, module: memory usage, module: nestedtensor, module: numerical-stability, module: padding, module: sdpa, module: sparse, module: tests, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#28829 torch.cuda.close request to be able to reset communications, #74235 RuntimeError: CUDA error: unspecified launch failure, #98925 Request for deterministic support for reflection_pad2d_backward_cuda, #115618 RuntimeError: handle_0 INTERNAL ASSERT FAILED at ""../c10/cuda/driver_api.cpp"":15, #118991 Memory Leak in v2.0.1, #119131 `scaled_dot_product_attention` produces NaNs for math backend, but not memory efficient backend when used in mixed precision.",382b42666c798742e0d4fbdd7ad0873de62b2725a3ba386d5c2c67d77711eb58
issue-collapse:66dd82a0de5d52fb,module: cuda stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"59692, 104591, 113560, 137581, 144025, 144611, 144612, 145145, 145349, 145350, 145453, 146040, 146056, 147284",,,,"module: cuda, module: cudacachingallocator, module: docs, module: edge cases, module: nn, module: performance, module: python frontend","actionable, enhancement, module: CUDACachingAllocator, module: cuda, module: docs, module: edge cases, module: nn, module: performance, module: python frontend, topic: fuzzer, topic: performance, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#59692 cuda streams run sequentially, expected to run parallel, #104591 version libcudnn_ops_infer.so.8 not defined in file libcudnn_ops_infer.so.8 with link time reference, #113560 t.contiguous() ~10 slower in eager mode compared to torch.compile, #137581 Expose underlying cudastream_t for torch.cuda.Stream, #144025 torch.cuda.empty_cache() causes extra memory usage on 'cuda:0', #144611 [CUDA] Illegal Memory Access with `ConvTranspose2d`",4a3e2a7bc817297b329aa100ded7f513917d41acc40aad679382c6919c09da44
issue-collapse:73f627e358dca594,module: cuda stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"37406, 37664, 67978, 68272, 104152, 121857, 127225, 147682, 147874, 148885, 149802, 150574, 150700, 153210",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: edge cases, module: error checking, module: int overflow, module: memory usage, module: nccl, module: optimizer, module: performance, module: testing, module: third_party, oncall: distributed","actionable, enhancement, function request, module: CUDACachingAllocator, module: cuda, module: cudnn, module: edge cases, module: error checking, module: int overflow, module: memory usage, module: nccl, module: optimizer, module: performance, module: testing, module: third_party, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#37406 3D grouped & depthwise convolution very slow on backward pass, #37664 Remove everything from a GPU (including drivers), #67978 c10::CUDAError, #68272 Query CUDA version LibTorch has been compiled with, #104152 Error 101: invalid device ordinal (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:109.), #121857 AdamW(fused=True) slower than unfused AdamW",f3fbf507237c5e834b2efe24ca4f3cb41fa89dba77bb96aec985a8dabe5103f9
issue-collapse:ac911319b1737160,module: cuda stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"64491, 76071, 123548, 123834, 147011, 149155, 149196, 152150, 152922, 153649, 154428, 154602, 156238, 157502",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: docs, module: edge cases, module: fft, module: floatx (formerly float8), module: memory usage, module: multiprocessing, module: optimizer, module: performance, module: python frontend, module: sdpa, module: windows, oncall: profiler","actionable, better-engineering, module: CUDACachingAllocator, module: cuda, module: cudnn, module: docs, module: edge cases, module: fft, module: floatx (formerly float8), module: memory usage, module: multiprocessing, module: optimizer, module: performance, module: python frontend, module: sdpa, module: windows, oncall: profiler, topic: build, topic: performance, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#64491 TypeError when using torch.cuda.list_gpu_processes() on Windows with the WDDM driver, #76071 HIPFFT_EXEC_FAILED when using AMD GPU run FFT, #123548 An error occurs when ‘max_split_size_mb ’and ‘expandable_segments ’ are enabled at the same time., #123834 RuntimeError: NVML_SUCCESS == r INTERNAL ASSERT FAILED at ""/opt/pytorch/pytorch/c10/cuda/CUDACachingAllocator.cpp"":830, please report a bug to PyTorch., #147011 Deprecation of NVTX 2 (`nvToolsExt`): Recommended to move to NVTX 3, #149155 torch.multiprocessing.Queue Zeroes Out Tensors on Retrieval",41fada29309549492f13af1005200ee7e7bef68ae30e5e60a64b4f8cbb313358
issue-collapse:e34f883d3c818433,module: cuda stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38671, 68959, 81717, 145471, 152756, 153534, 157950, 158172, 158444, 158641, 159301, 159564, 160308, 160891",,,,"module: cuda, module: cudacachingallocator, module: determinism, module: edge cases, module: embedding, module: floatx (formerly float8), module: linear algebra, module: norms and normalization, module: rocm, module: rpc, module: tests","actionable, feature, matrix multiplication, module: CUDACachingAllocator, module: cuda, module: determinism, module: edge cases, module: embedding, module: floatx (formerly float8), module: linear algebra, module: norms and normalization, module: rocm, module: rpc, module: tests, topic: new features, topic: performance, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38671 Add CUDA callback to Python API, #68959 Deterministic implementation for grid_sampler_2d_backward_cuda, #81717 PyTorch Embedding Op with max_norm is not working as expected, #145471 Is there a PyTorch version that can work properly on the Thor platform based on the Blackwell architecture?, #152756 Cuda-12.9 removed libnvToolsExt.so.* and is now purely header nvtx3, #153534 `pca_lowrank` on CUDA is significantly slower than on CPU",df43054be94ae990def851cdd26be1ed52a9f7f8112704889e2a5e04e3d09770
issue-collapse:e7adfcebcc13cde1,module: cuda stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"29961, 119420, 153502, 155798, 156618, 158831, 159674, 159914, 160816, 162127, 162582, 163062, 163709, 163982",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: dependency bug, module: fft, module: floatx (formerly float8), module: linear algebra, module: memory usage, module: multiprocessing, module: performance, module: python frontend, module: tf32, module: wsl","feature, mempool-bug, module: CUDACachingAllocator, module: cuda, module: cudnn, module: dependency bug, module: fft, module: floatx (formerly float8), module: linear algebra, module: memory usage, module: multiprocessing, module: performance, module: python frontend, module: tf32, module: wsl, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#29961 Add support for integer matrix multiplication (particularly for dtype = torch.int8 ), #119420 torch.stft sometimes raises RuntimeError: cuFFT error: CUFFT_INTERNAL_ERROR on low free memory, #153502 Add MXFP8 Support to scaled_grouped_gemm, #155798 `nn.RNN(...).to('cuda')` fails with `cuDNN error: CUDNN_STATUS_BAD_PARAM` on GPU, but works on CPU, #156618 THPStorage_shareCuda zeroes CUDA tensor, #158831 Severe Performance Degradation in Batched SVD on A100-40G",829024d7c533bb530a77be9060007f6f48de8963453399eb365492437369ead1
issue-collapse:b603670881b54ad5,module: cuda stale/low urgency  batch 35,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18849, 21061, 21780, 22383, 22785, 23026, 24016, 25752, 25820, 27024, 27138, 27297, 27540, 27926",,,,"module: cuda, module: dependency bug, module: flaky-tests, module: loss, module: macos, module: memory usage, module: nn, module: numerical-stability, module: performance, module: tests","caffe2-op, enhancement, feature, low priority, module: cuda, module: dependency bug, module: flaky-tests, module: loss, module: macos, module: memory usage, module: nn, module: numerical-stability, module: performance, module: tests, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18849 More efficient STFT on CUDA, #21061 Misleading Error when doing Large Batch Matrix Multiplication, #21780 cumsum cuda numerical instability, #22383 Storage operation failing on second GPU, #22785 Getting cuda runtime error (48) with Jetson TX2 when running simple program, #23026 BatchNorm1d fails on first run through GPU",d5a94ccde7ee66cc612fa57728270ebaee7137c479fb48e402ae39b7fdd283fc
issue-collapse:f70848fd4bc3a81f,module: cuda stale/low urgency  batch 34,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20864, 24963, 25297, 28791, 29198, 29545, 29553, 30404, 30569, 31038, 31425, 32504, 32510, 33051",,,,"module: cuda, module: cudnn, module: dataloader, module: error checking, module: memory usage, module: performance, module: printing, module: random, module: tensorboard, module: third_party","enhancement, module: cuda, module: cudnn, module: dataloader, module: error checking, module: memory usage, module: performance, module: printing, module: random, module: tensorboard, module: third_party, small, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20864 Import warning when using the wrong version of CUDA, #24963 torch.cuda.synchronize blocks CUDA execution on other threads using other devices., #25297 torch.as_tensor(bytearray(...)) seems to leak memory, #28791 [CUDA-MEMCHECK] TestTorchDeviceTypeCUDA.test_pin_memory_from_constructor_cuda fails, #29198 RuntimeError: CUDA out of memory with available GPU memory, #29545 Tensorboard GPU Problems",fc5b490d21ebe7fa2cd64c4a135501e8db6f6acdb0c578c3a4b5f1a0fa825a98
issue-collapse:2c1f739d812f6c0d,module: cuda stale/low urgency  batch 33,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3390, 4987, 22961, 24176, 27975, 30528, 32370, 33979, 34323, 34878, 36160, 36638, 37153, 37549",,,,"module: cuda, module: cudnn, module: data parallel, module: macos, module: memory usage, module: multiprocessing, module: performance, module: rpc, module: tensoriterator, oncall: distributed","enhancement, feature, module: TensorIterator, module: cuda, module: cudnn, module: data parallel, module: macos, module: memory usage, module: multiprocessing, module: performance, module: rpc, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3390 High CPU use by clock_gettime syscall, #4987 MultiGPU hangs Titan Xp in multiprocessing/queue.py, #22961 performance much worse on 2080ti than 1080ti, #24176 torch.fft crash when used with nn.DataParallel, #27975 Sending CUDA tensors via queue between processes, memory of Consumer process grows infinitely, #30528 Warn about driver CUDA mismatch in torch.cuda.is_available()",5f1e501851dffe9776cfcd4fa8fba2682059e198ff191b97bb0a41d427908ea3
issue-collapse:05e5d66be23e13d1,module: cuda stale/low urgency  batch 32,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3600, 13786, 29516, 29710, 32868, 38911, 38996, 38997, 39317, 40020, 40548, 40667, 42074, 42301",,,,"module: cuda, module: data parallel, module: docs, module: internals, module: memory usage, module: nccl, module: nn, module: numba, module: numpy, module: performance, oncall: distributed","caffe2, enhancement, feature, module: cuda, module: data parallel, module: docs, module: internals, module: memory usage, module: nccl, module: nn, module: numba, module: numpy, module: performance, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3600 making .cuda() falls back to an identity function when gpu is not available, #13786 PyTorch streams are not cuda-memcheck clean, #29516 RuntimeError: CUDA error: invalid device ordinal, #29710 Reliable way to identify RuntimeErrors (CUDA), #32868 __cuda_array_interface__ conversion does not support readonly arrays, #38911 [FR] cuda.comm.broadcast/reduce_add support `out=`",184f1e6d057b770e1aeeee52921c27552fce2da84d336b2e6cf5db1b6ba82f48
issue-collapse:c35d81791e18e615,module: cuda stale/low urgency  batch 31,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1328, 3396, 14996, 16542, 23061, 31570, 37048, 37585, 39632, 43281, 43685, 44547, 44624, 44943",,,,"module: cuda, module: data parallel, module: docs, module: error checking, module: linear algebra, module: loss, module: nn, module: performance, module: rnn, module: sorting and selection, oncall: distributed","module: cuda, module: data parallel, module: docs, module: error checking, module: linear algebra, module: loss, module: nn, module: performance, module: rnn, module: sorting and selection, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1328 Unhelpful CrossEntropyLoss dimension error message, #3396 CUDA topk is slow for some input sizes, #14996 as_tensor does not use the device of the default tensor type, #16542 Error with setting tensors to use cpu in packed_padded_sequence when CUDA tensor is set as default, #23061 The speed of `torch.einsum` and `torch.matmul` when using `fp16` is slow, #31570 Can't pin storage memory",ec8e72a7e02ad74bd484f1a23cceebc0ba22d1460276b530d9c146a567875ce9
issue-collapse:2ca5ac12035c9d56,module: cuda stale/low urgency  batch 30,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1750, 25045, 29528, 31261, 37682, 44248, 45447, 45578, 46379, 46899, 47111, 48202, 48706, 49332",,,,"module: cuda, module: distance functions, module: docs, module: half, module: internals, module: memory usage, module: mkl, module: multiprocessing, module: performance, module: pruning, module: rnn","better-engineering, function request, module: cuda, module: distance functions, module: docs, module: half, module: internals, module: memory usage, module: mkl, module: multiprocessing, module: performance, module: pruning, module: rnn, newcomer, shadow review, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1750 Add support for colors (and maybe other attributes) to NVTX API, #25045 [Distance functions] F.pdist backward CUDA invalid configuration, #29528 Half precision cdist, #31261 pinned memory requires DeviceGuard in multi-process envs, #37682 LSTMCell consumes x1.5 more memory on CUDA on pytorch >=1.3 comparing to pytorch 1.2, #44248 Additional clutter memory allocated on GPU 0 when training on GPU 1+",b535c8252cd3efd76dcf2c1cf31d6c00e924b2512da37fbeb7d72baeae2b0a7a
issue-collapse:0cea410fd835b4af,module: cuda stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"27780, 77049, 122274, 129649, 140696, 142232, 151433, 160838, 162402, 164249, 164799, 165715, 165741, 165826",,,,"module: cuda, module: docs, module: jetson, module: memory usage, module: multithreading, module: nn, module: numerical-stability, module: performance, module: sorting and selection, module: tests, module: vllm, oncall: distributed, oncall: profiler, oncall: pt2","enhancement, module: cuda, module: docs, module: jetson, module: memory usage, module: multithreading, module: nn, module: numerical-stability, module: performance, module: sorting and selection, module: tests, module: vllm, oncall: distributed, oncall: profiler, oncall: pt2, triaged, vllm-compile",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#27780 torch.cuda.default_generators documentation referenced but don't exist., #77049 Peak GPU-memory usage extremely huge when sorting with torch.sort, #122274 CUDA: Performance issues with `torch.unique`, #129649 Low GPU precision, #140696 inconsistency in ```torch.nn.functional.adaptive_avg_pool3d``` on CPU and GPU, #142232 torch.tensor.to(‘cuda’) so slow in jetson orin",6c94b048a45475744d703227c88697babe3b6472bd2ee30880865f0d2c243fff
issue-collapse:63e102e194f05c00,module: cuda stale/low urgency  batch 29,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"37713, 49023, 50185, 50268, 50354, 50829, 50925, 51224, 52041, 52145, 52266, 52570, 52644, 53130",,,,"module: cuda, module: cudnn, module: deadlock, module: half, module: lint, module: memory usage, module: nccl, module: nn, module: performance, module: tests","enhancement, module: cuda, module: cudnn, module: deadlock, module: half, module: lint, module: memory usage, module: nccl, module: nn, module: performance, module: tests, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#37713 Builtin FusedLayerNorm is slower than apex one, #49023 Automatically rerun tests with CUDA_LAUNCH_BLOCKING=1 when they fail with CUDA errors in CI, #50185 CUDACachingAllocator is not GC-aware, #50268 CUDA error when using `binary_cross_entropy_with_logits`, #50354 for CNN in fp16 execution time depends on input scale, #50829 pin_memory leads to GPU memory being used but it doesn't initialize CUDA",236b6f065e06a781bd8c2d2423ed75c9d1498fa7a04373c0b01f7ba7c27f4a6e
issue-collapse:f68a0a56b656de6a,module: cuda stale/low urgency  batch 28,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12060, 24401, 30246, 53957, 54307, 54564, 55643, 55736, 55756, 56350, 56694, 58385, 58474, 58626",,,,"module: cuda, module: cudnn, module: dataloader, module: memory usage, module: multiprocessing, module: performance, module: rnn, module: safe resize, oncall: profiler","feature, module: cuda, module: cudnn, module: dataloader, module: memory usage, module: multiprocessing, module: performance, module: rnn, module: safe resize, oncall: profiler, todo, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12060 cuda test hangs if GPUs in Exclusive Process mode, #24401 When running model forward with large batch size, it reports the error: THCudaTensor sizes too large for THCDeviceTensor conversion, #30246 CUDA masked_select uses way too much memory, #53957 Is pytorch 1.8.0 incompatible with cuda 11.2 or what is the reason for this error?, #54307 cuDNN error upon the backward method with a sliced tensor output from nn.GRU., #54564 why single pytorch process is displayed on each GPUs with memory 0MB",9695e9691c918f54fcaf8c32f8f91df3c503fd3a20a550b257f4673f24059b01
issue-collapse:614e75c671760fce,module: cuda stale/low urgency  batch 27,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23780, 36066, 41852, 45428, 48832, 51644, 59071, 59778, 60306, 61525, 62031, 62265, 63069, 63226",,,,"module: cuda, module: cudnn, module: distributions, module: internals, module: memory usage, module: performance, module: serialization, module: sparse, module: tests, module: vmap","awaiting response (this tag is deprecated), caffe2, enhancement, module: cuda, module: cudnn, module: distributions, module: internals, module: memory usage, module: performance, module: serialization, module: sparse, module: tests, module: vmap, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23780 Construction of MultivariateNormal much slower on GPU than CPU, #36066 Add Specific Warning/Error For Unsupported GPU or Systems, #41852 Difference in inference time between CUDA 10.0 & 10.2, #45428 Use libtorch for online inference: an illegal memory access was encountered, #48832 cudnn cannot be pickled by cloudpickle, #51644 Segmentation Fault: Garbage collector, cuda memory",27ed68fddaeb5e19c5e70cf873e686c38cae78c5a6515a0545d465a17310aac7
issue-collapse:efca55d98312b71d,module: cuda stale/low urgency  batch 26,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"46024, 59309, 61890, 62648, 63837, 63998, 65100, 65301, 66076, 66473, 66936, 67208, 67388, 68165",,,,"module: cuda, module: cudnn, module: determinism, module: distributions, module: docs, module: loss, module: memory format, module: memory usage, module: mta, module: multiprocessing, module: nccl, module: numerical-reproducibility, module: performance, module: rnn, module: scatter & gather ops, module: structured kernels, module: tensor creation, module: tests, module: tf32","feature, module: cuda, module: cudnn, module: determinism, module: distributions, module: docs, module: loss, module: memory format, module: memory usage, module: mta, module: multiprocessing, module: nccl, module: numerical-reproducibility, module: performance, module: rnn, module: scatter & gather ops, module: structured kernels, module: tensor creation, module: tests, module: tf32, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#46024 Reproducibility breaks down with weighted Cross Entropy loss, #59309 CUDA memory error when using torch.device('cpu'), #61890 The matrix multiplication operator can't get correct results on 3090 !!, #62648 Feature request: Generate Sobol points directly on the GPU, #63837 `torch.scatter` doesn't fail correctly on CUDA (memory overlap), #63998 cat out= variant memory_format inconsistent for cpu and cuda",7ac7e420ad8f8d2288e0f49880dfefec0f1dc947364707ea54918f5834d4b1be
issue-collapse:6020ab97d5cfe101,module: cuda stale/low urgency  batch 25,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"54047, 63992, 68909, 69772, 70490, 70901, 71414, 71495, 71697, 73196, 73461, 74146, 74391, 74679",,,,"module: cuda, module: cudnn, module: distributions, module: embedding, module: half, module: linear algebra, module: memory usage, module: nccl, module: nn, module: numerical-stability, module: performance, module: static linking, module: tests, module: tf32","Stale, module: cuda, module: cudnn, module: distributions, module: embedding, module: half, module: linear algebra, module: memory usage, module: nccl, module: nn, module: numerical-stability, module: performance, module: static linking, module: tests, module: tf32, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54047 RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED in pytorch lightning, #63992 Direct inversion and linear systems solutions for small matrices, #68909 Bug about Dropout CUDA Kernel, #69772 RuntimeError: t == DeviceType::CUDAINTERNAL ASSERT FAILED when trying to calculate gradients, #70490 Cuda sync mode input checking is wrong for non-string/int inputs., #70901 Accuracy problem of `torch.batch_norm_gather_stats_with_counts` when `running_mean` is half tensor",0d66ffbca189fde9f139493f3a8560e7343a996d6da84a3ae4a5309933a6c590
issue-collapse:5d9f452c7ab716c3,module: cuda stale/low urgency  batch 24,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32524, 32881, 43546, 55277, 62695, 75025, 75198, 75494, 76043, 77397, 77951, 78518, 78581, 78884",,,,"module: cuda, module: cudnn, module: ddp, module: half, module: jiterator, module: linear algebra, module: magma, module: memory usage, module: multiprocessing, module: nccl, module: numerical-stability, module: performance, module: python frontend, module: tf32, oncall: distributed","module: cuda, module: cudnn, module: ddp, module: half, module: jiterator, module: linear algebra, module: magma, module: memory usage, module: multiprocessing, module: nccl, module: numerical-stability, module: performance, module: python frontend, module: tf32, oncall: distributed, topic: performance, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32524 Pytorch 1.4 does not detect gpu, #32881 CUDNN_STATUS_EXECUTION_FAILED, #43546 NCCL WARN Your program may be hanging, this may be caused by a collective mismatch around rank 2. Please check your collective calls at and around this rank. You can use NCCL_DEBUG=INFO and NCCL_DEBUG_SUBSYS=COLL to see the collective logs, #55277 Small model occupies too much GPU in CUDA11.1 + Torch1.8.1 but is normal in Torch 1.6 + CUDA10.1, #62695 Use fast gpuAtomicAdd for FP16 data type in CUDA kernels, #75025 torch.cuda.init() unstacks existing CUDA contexts",9ab25844231de11bfde0f74e6da9f9141dab751d0c3b7fb0e3cd63715f036913
issue-collapse:f78d5cd83ca7d050,module: cuda stale/low urgency  batch 23,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"36560, 46386, 72709, 76743, 78018, 81620, 81988, 82061, 82200, 83148, 83392, 84194, 84445, 85387",,,,"module: cuda, module: cudnn, module: dataloader, module: determinism, module: docs, module: linear algebra, module: memory usage, module: multiprocessing, module: numerical-stability, module: numpy, module: performance","better-engineering, enhancement, feature, has workaround, module: cuda, module: cudnn, module: dataloader, module: determinism, module: docs, module: linear algebra, module: memory usage, module: multiprocessing, module: numerical-stability, module: numpy, module: performance, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36560 Allow `__array__` to automatically detach and move to CPU, #46386 GPU memory leak when registering a forward hook with ""self"" access, #72709 Feature: Use iterative refinement algorithm from cuSOLVER for the least-squares solver on CUDA, #76743 Supporting torch.tensor.apply_ over GPU, #78018 Three memory copies of every dataloader cpu tensor, #81620 PyTorch 1.12 cu113 Illegal Memory Access or Internal Error instead of Out of Memory cases",81167c5547d6cb29fefa2104ddc19fb6d9698a03c3daf165f17a0e9a1eb0d6fe
issue-collapse:9e8ef09fd5f3a119,module: cuda stale/low urgency  batch 22,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"84336, 85227, 85377, 85841, 85889, 86449, 86493, 86551, 86782, 86798, 87539, 87841, 88027, 88081",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: embedding, module: memory usage, module: multiprocessing, module: numerical-stability, module: nvfuser, module: performance, module: pooling, module: random, module: sanitizers, module: sorting and selection, module: tests, module: tf32, oncall: jit","module: CUDACachingAllocator, module: cuda, module: cudnn, module: embedding, module: memory usage, module: multiprocessing, module: numerical-stability, module: nvfuser, module: performance, module: pooling, module: random, module: sanitizers, module: sorting and selection, module: tests, module: tf32, oncall: jit, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#84336 torch 1.12.1 cuda 10.2 runs slower than torch 1.8.2 cuda 10.2, #85227 topk returns different results with the same input in cuda and cpu, #85377 CI fails for test_compare_cpu_nn_functional_embedding_cuda_float32 which is not reproducible locally, #85841 Setting the cuda device when using start_processes in Jupyter on Ampere leads to CUDA reinitialization error, #85889 High occupation on GPU 0 when converting Tensor to multi GPU, #86449 torch.cuda.empty_cache() is not working",3bdda01c5df40793288f9bc66a0ef7611a6ce6c87e369e88994ed6b29979a916
issue-collapse:7a2dcd5cf68b976f,module: cuda stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"36706, 62661, 71301, 73003, 87992, 88189, 88191, 88410, 88563, 88639, 89245, 90999, 91375, 91582",,,,"module: cuda, module: cuda graphs, module: cudnn, module: dataloader, module: numerical-stability, module: performance, module: random, module: rocm, module: tensoriterator","module: TensorIterator, module: cuda, module: cuda graphs, module: cudnn, module: dataloader, module: numerical-stability, module: performance, module: random, module: rocm, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36706 Reduction for `torch.int8` is super slow on CUDA, #62661 Rewrite PyTorch CUDA backend in Triton, #71301 Why AMP make backward speed more slow?, #73003 pin_memory hangs instead of throwing, #87992 torch.rand(...) is not consistent for large shape dimensions across GPUs (with the same random seed), #88189 Moving tensor to GPU by .cuda() gets stucked when AMD Secure Encripted Virtualization (SEV) is activated",c642c2d3d66f9a2ed922ec0e1deb046bd6419fb502deafeb3460c705bf51841a
issue-collapse:2b8d9330447926ff,module: cuda stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24081, 27588, 49095, 50036, 66482, 68323, 70669, 84309, 90793, 91536, 95058, 95077, 95169, 95237",,,,"module: cuda, module: cudnn, module: data parallel, module: dataloader, module: deadlock, module: dependency bug, module: edge cases, module: linear algebra, module: memory format, module: memory usage, module: mta, module: multi-gpu, module: multiprocessing, module: nn, module: performance, module: sparse","has workaround, matrix multiplication, module: cuda, module: cudnn, module: data parallel, module: dataloader, module: deadlock, module: dependency bug, module: edge cases, module: linear algebra, module: memory format, module: memory usage, module: mta, module: multi-gpu, module: multiprocessing, module: nn, module: performance, module: sparse, quansight-nack, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24081 Multi-gpu example freeze and is not killable, #27588 RuntimeError: cuDNN error: CUDNN_STATUS_MAPPING_ERROR, #49095 NCCL Error when training with 2x 3090s., #50036 channels_last format convolution is slower than normal NCHW, #66482 BAR1 memory of GPU is not released when main process is killed., #68323 sparse.mm: CUDA error: internal error when calling `cusparseSpGEMM_workEstimation [...]`",393ed6f322efa6168486677d2109e9b06997ce75bf634afa333638044cd07a48
issue-collapse:0993ac2b1b2bb2fa,module: cuda stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"103495, 123505, 133629, 148681, 150517, 152731, 166234, 166409, 166611, 166651, 167409, 167420, 167724, 168158",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: edge cases, module: fft, module: flaky-tests, module: half, module: linear algebra, module: memory usage, module: nans and infs, module: performance, module: rocm, module: windows, module: wsl","actionable, enhancement, low priority, matrix multiplication, module: CUDACachingAllocator, module: NaNs and Infs, module: cuda, module: cudnn, module: edge cases, module: fft, module: flaky-tests, module: half, module: linear algebra, module: memory usage, module: performance, module: rocm, module: windows, module: wsl, rocm-skipped-tests, skipped, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#103495 DISABLED test_mem_get_info (__main__.TestCuda), #123505 RuntimeError: Unrecognized CachingAllocator option: expandable_segments, #133629 RuntimeError: free_upper_bound + pytorch_used_bytes[device], #148681 foreach CUDA tests flaky on CUDA 12.6+ due to flaky profiler results, #150517 Pinned memory doubles memory usage for tensors slightly over 128MB, #152731 Inconsistent float16 overflow behavior between CPU and CUDA devices",b010ef2a4999ef9c583d0a45d5869e0bbe465ddede76e33c49f7ac948ae6d3fd
issue-collapse:bf32a6b387890015,module: cuda stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"44551, 58812, 62350, 67680, 89342, 89730, 95648, 95696, 95768, 98193, 99138, 99812, 100145, 100425",,,,"module: cuda, module: cudnn, module: data parallel, module: docs, module: error checking, module: memory usage, module: performance, module: vision","module: cuda, module: cudnn, module: data parallel, module: docs, module: error checking, module: memory usage, module: performance, module: vision, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#44551 PyTorch 1.6 DataParallel causes CUDNN_STATUS_BAD_PARAM in backward pass, #58812 Transferring tensor to the gpu and converting dtype in a single call to .to() is slower than first transferring, then converting., #62350 NotImplementedError: Could not run 'torchvision::nms' with arguments from the 'CUDA' backend., #67680 OOM with a lot of GPU memory left, #89342 torch.mm: Exceptions thrown on the CPU and GPU are inconsistent, #89730 torch.addbmm throws different exception differences on CPU and GPU.",25c1faccd6d907b8925364111d7d5f59f823ef5e6cd7455db2afbe7f71e9d4f1
issue-collapse:2bf2876ee111d590,module: cuda stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30711, 90261, 95121, 98481, 101890, 103589, 104107, 104322, 104867, 104925, 105726, 106520, 106606, 107201",,,,"module: cuda, module: cudacachingallocator, module: determinism, module: embedding, module: fft, module: half, module: internals, module: interpolation, module: linear algebra, module: logging, module: memory format, module: nn, module: performance, module: sparse, module: third_party","module: CUDACachingAllocator, module: cuda, module: determinism, module: embedding, module: fft, module: half, module: internals, module: interpolation, module: linear algebra, module: logging, module: memory format, module: nn, module: performance, module: sparse, module: third_party, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30711 Sorting in embedding_dense_backward_cuda takes very long time, #90261 Why torch.mode return different value between CPU and GPU, #95121 cuda 12 support request., #98481 Bring CudaPluggableAllocator to feature parity with the Native Allocator, #101890 Can't reproduce/non-deterministic results with CUDA, #103589 `interpolate` with `antialias=True` on CUDA doesn't work if the difference of spatial size is large",007a246c35562c042cbedda8b8a277dfc0dc92af3f686b7f2655bf70e3e161fa
issue-collapse:e58751437f4e3738,module: cuda stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"42080, 65528, 70701, 81883, 107374, 107528, 107864, 108407, 108494, 108645, 109383, 109494, 109863, 110014",,,,"module: cuda, module: cuda graphs, module: data parallel, module: distributions, module: jiterator, module: linear algebra, module: performance, module: random","feature, module: cuda, module: cuda graphs, module: data parallel, module: distributions, module: jiterator, module: linear algebra, module: performance, module: random, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42080 To have single cuda context across multiple processes, #65528 I cannot use x.to(GPU) or x.cuda(GPU), #70701 AdaptiveAvgPool1d - RuntimeError: shmem_size <= sharedMemPerBlockINTERNAL ASSERT FAILED, #81883 RuntimeError: CUDA error: no kernel image is available for execution on the device, #107374 DataParallel scatter method split tensor wrong, #107528 Inconsistent results when running torch.arctanh",9d01ca165715d887ebe3a56ddf0acc500b1ee9900c13809c6f5edd78ff183e6f
issue-collapse:8899b1747aff3621,module: cuda stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20655, 30401, 41970, 50198, 87696, 108277, 110116, 110194, 110252, 111024, 111349, 111363, 111466, 112847",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: determinism, module: docs, module: memory usage, module: multiprocessing, module: nn, module: performance, module: pooling, module: random, module: tf32, oncall: distributed","enhancement, feature, function request, module: CUDACachingAllocator, module: cuda, module: cudnn, module: determinism, module: docs, module: memory usage, module: multiprocessing, module: nn, module: performance, module: pooling, module: random, module: tf32, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20655 nn.Embedding backwards slow under high row contention, #30401 Unexpected difference torch.multiprocessing.manager.queue and torch.multiprocessing.queue, #41970 torch.random.fork tries to initialize cuda even when no cuda devices are available, #50198 deterministic implementation for adaptive_avg_pool2d_backward_cuda, #87696 Enable PostLocalSGDOptimizer on CUDA tensors, #108277 Transformer performance drop due to slow PyTorch GEMMs",0a79464e1365cb3a47f212ef7005876e8647cd585c198269b240ccf9797e5945
issue-collapse:6287c292559a2c88,module: cuda stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16266, 43623, 58155, 74393, 113326, 113408, 113449, 113847, 114943, 115340, 115418, 115735, 116543, 116870",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: docs, module: memory usage, module: multiprocessing, module: nccl, module: numerical-stability, module: performance, module: pooling, module: sparse, module: testing, oncall: distributed, oncall: quantization","feature, module: CUDACachingAllocator, module: cuda, module: cudnn, module: docs, module: memory usage, module: multiprocessing, module: nccl, module: numerical-stability, module: performance, module: pooling, module: sparse, module: testing, oncall: distributed, oncall: quantization, todo, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16266 support for multiple torch.cuda.max_memory_allocated() counters, #43623 [FR] Raise an exception when constructing non-empty index and empty values sparse tensor, #58155 ""uncorrectable NVLink error"" making the tests fail, #74393 Quantized cannot inference with cuda, #113326 `torch._C._cuda_getDeviceCount` inflates system memory usage, #113408 Early testing stop logic for CUDA error looks wrong for instantiated_test with pytest",4c3c9a71b944583beb7eff2279053c301c1243f27210f01352e6a8bcffa9a1a4
issue-collapse:8cf2f8b74f79dd21,module: cuda stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11340, 44156, 68961, 72117, 75391, 82073, 94966, 115331, 117910, 118373, 118859, 118863, 119624, 121203",,,,"module: cuda, module: cudacachingallocator, module: determinism, module: docs, module: linear algebra, module: memory usage, module: multiprocessing, module: performance, module: python frontend, module: random, module: sparse","feature, matrix multiplication, module: CUDACachingAllocator, module: cuda, module: determinism, module: docs, module: linear algebra, module: memory usage, module: multiprocessing, module: performance, module: python frontend, module: random, module: sparse, todo, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11340 Better user experience for using Generator object, #44156 CUDA memory leak in multi-processing, #68961 Deterministic implementation for upsample_bilinear2d_backward_cuda, #72117 Force PyTorch to clear CUDA cache, #75391 Pytorch linalg test failure with cuda 11.6, #82073 CUDACachingAllocator should be cuda memory merge/compact friendly",5de1409c4b7cd1e317f8c7e20e6bb4b7b02ab10938418cd19ab7af5fa387030c
issue-collapse:fd63d098f110c785,module: cuda stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22573, 48163, 72791, 99372, 104981, 114040, 122068, 122170, 123604, 124456, 124838, 126748, 127196, 128296",,,,"module: cuda, module: cudacachingallocator, module: dataloader, module: half, module: interpolation, module: linear algebra, module: memory usage, module: multiprocessing, module: multithreading, module: nans and infs, module: nn, module: performance, module: sorting and selection, module: sparse","actionable, module: CUDACachingAllocator, module: NaNs and Infs, module: cuda, module: dataloader, module: half, module: interpolation, module: linear algebra, module: memory usage, module: multiprocessing, module: multithreading, module: nn, module: performance, module: sorting and selection, module: sparse, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22573 Batched symeig and qr are very slow on GPU, #48163 nn.functional.interpolate backward in fp16 is extremely slow, #72791 why ram memory surges while loading model, with change in torch load device from CPU to GPU, #99372 PyTorch 2.0.0 encountered CUDA error: an illegal memory access was encountered, #104981 NotImplementedError: Could not run 'aten::_spdiags' with arguments from the 'CUDA' backend., #114040 IndexError: map::at with MPI CUDA collectives",a5fbf03abd07fa7eecc7c97ec03b9d7298e3cb3a11fed6415c008fdd183c64f6
issue-collapse:e61888f2da064117,module: cuda stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13058, 113053, 117130, 121641, 122092, 128601, 128791, 128988, 129670, 129954, 130078, 130124, 130193, 131491",,,,"module: cuda, module: cuda graphs, module: cudacachingallocator, module: dataloader, module: error checking, module: half, module: loss, module: nn, module: performance, module: python frontend, module: sparse","actionable, enhancement, module: CUDACachingAllocator, module: cuda, module: cuda graphs, module: dataloader, module: error checking, module: half, module: loss, module: nn, module: performance, module: python frontend, module: sparse, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13058 Backward pass over torch.nn.functional.pad is extremely slow with half tensors, #113053 [CUDA-12.2] cuSPARSE deprecated support for sparse BSR, #117130 `torch.cuda.memory_summary()` can give `KeyError`, #121641 torch.isin on GPU vs numpy isin: perf way slower than expected, #122092 masked_index_add, #128601 Retry on CUBLAS_STATUS_ALLOC_FAILED",26ed09f3b82d23849d5c258422749cb0eadfcaddb1c69844fec5c5b13dde19ef
issue-collapse:0e04e25cdcad103a,module: cuda stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32575, 37324, 52718, 113743, 120996, 124351, 131655, 131656, 133280, 133281, 134403, 134441, 134607, 134851",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: error checking, module: fft, module: memory usage, module: multiprocessing, module: performance, oncall: distributed","enhancement, feature, matrix multiplication, module: CUDACachingAllocator, module: cuda, module: cudnn, module: error checking, module: fft, module: memory usage, module: multiprocessing, module: performance, oncall: distributed, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32575 Interrupting a DDP worker while using the CUDA MPS server causes CUDA to hang until reboot, #37324 Multiprocessing: model shared between processes hangs during copy.deepcopy, #52718 Discrepancy between CPU->GPU and GPU->CPU data transfer speeds, #113743 The cuda batched GEMM has a poor performance for bigger batch size with smaller matrix size, #120996 Unnecessary CUDA synchronization, or missing warnings, #124351 BUG Report using `backend:cudaMallocAsync`",95738f478c2828d3c9b901bcfd26d34458e393a39e931fb6f803673f354c01d9
issue-collapse:515e0111691089c8,module: cuda stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"47996, 51075, 116242, 135837, 136812, 136815, 137174, 138453, 138829, 139115, 139333, 139342, 139382, 139648",,,,"module: cuda, module: cuda graphs, module: cudacachingallocator, module: memory usage, module: mta, module: multiprocessing, module: nn, module: pooling, module: python frontend, module: tests, module: vulkan","feature, module: CUDACachingAllocator, module: cuda, module: cuda graphs, module: memory usage, module: mta, module: multiprocessing, module: nn, module: pooling, module: python frontend, module: tests, module: vulkan, onnx-triaged, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47996 GPU Vendor-Agnosticism via Vulkan, #51075 RFC: Private CUDA memory pools, #116242 Building project using libtorch results in ""Failed to find nvToolsExt"", #135837 cudaFree will be called when the reserved memory > 70G for A100 80G, #136812 `CUDNN_BACKEND_OPERATION: cudnnFinalize Failed cudnn_status: CUDNN_STATUS_BAD_PARAM` when using `float16`, works for `bfloat16` / `float32`, #136815 Aborted (core dumped) in `torch.cuda.caching_allocator_delete`",08debb8c9ba21c21f9337e238b76c701fba3659d9eeaea2d6a8c94612b4e6493
issue-collapse:811363815d5203c1,module: cuda stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88380, 149187, 149821, 167526, 169237, 169266, 169390, 169909, 169956, 170694, 171039, 171286, 171697, 171771",,,,"module: cuda, module: cudacachingallocator, module: cudnn, module: docs, module: linear algebra, module: multiprocessing, module: norms and normalization, module: sparse, module: tests, oncall: profiler","module: CUDACachingAllocator, module: cuda, module: cudnn, module: docs, module: linear algebra, module: multiprocessing, module: norms and normalization, module: sparse, module: tests, oncall: profiler, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88380 cuDNN error (CUDNN_STATUS_NOT_SUPPORTED) for torch.nn.functional.grid_sample(), #149187 Issue with Shared CUDA Tensor Reference Counting in Multi-Processing, #149821 `Aborted` error when using `torch.cuda.memory.caching_allocator_delete`, #167526 Missing documentation for CUTLASS backend, #169237 Large discrepancy between CPU and GPU norms for identical float32 tensor, #169266 Translation units with thrust::pair should include thrust/pair.h",881f4d4e23d4027a91fa68233467c40cd61a1feb021e49f7fb7d76baee4ed60a
issue-collapse:d92340b87d2f5295,module: cuda graphs stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"115339, 126753, 132856, 133984, 148927, 149944, 150482, 150484, 150488, 153025, 156820, 159384, 160281, 171263",,,,"module: cuda graphs, module: dynamo, module: inductor, module: privateuse1, module: vllm, oncall: distributed, oncall: pt2","feature, module: PrivateUse1, module: cuda graphs, module: dynamo, module: inductor, module: vllm, oncall: distributed, oncall: pt2, triaged, vllm-compile",,Handle this module: cuda graphs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#115339 Add support for measuring elapsed time in cuda-graphs, #126753 [cudagraph] simplify usage of how cudagraph dumps debug file, #132856 Enhance the compiler module to use cudagraph with PrivateUse1, #133984 cudagraph capture p2p ops fail, #148927 default cudagraphable policy for custom op, #149944 include cudagraph skip reasons in tlparse output",28d803057c061f7c56cc0ef52b5613a5cc53ba70f6a7ad8e535380186d4c962c
issue-collapse:1b326f95e3c38733,module: cublas stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"51351, 100850, 101428, 102459, 103499, 107800, 116184, 118774, 120240, 132300, 137389, 141210, 154248, 157054",,,,"module: cublas, module: cuda, module: dependency bug, module: determinism, module: floatx (formerly float8), module: half, module: inductor, module: numerical-reproducibility, module: performance, module: wsl, oncall: pt2","feature, matrix multiplication, module: cublas, module: cuda, module: dependency bug, module: determinism, module: floatx (formerly float8), module: half, module: inductor, module: numerical-reproducibility, module: performance, module: wsl, oncall: pt2, triaged",,Handle this module: cublas stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#51351 Run time error, #100850 [BUG] Poor torch.bmm performance on H100, #101428 2D inputs to linear layers run up to 25% slower than 4D ones on some Nvidia GPUs, #102459 Matrix multiplication performance regression in case of an additional dimension of size 1, #103499 CUBLAS_WORKSPACE_CONFIG can not be parsed, #107800 [feature request] [discussion] Include basic `ctypes` bindings for `cudart`/`cublasLt`/`cublas`/`nvrtc`/`cudnn` with stock PyTorch",5d12599ef4dcaf38b28e36012678e8a7e2ac046b098a20e2a55bda1c90621c55
issue-collapse:a4f5cdb34ca60403,module: cpu stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2180, 4703, 4825, 18987, 19163, 22803, 24303, 32023, 33725, 34728, 35937, 41399, 47447, 48641",,,,"module: cpu, module: cuda, module: loss, module: mkldnn, module: nans and infs, module: nn, module: numerical-reproducibility, module: performance, module: porting, module: random, module: rnn, module: sleef, module: sorting and selection, module: trigonometric functions, module: vectorization","better-engineering, feature, module: NaNs and Infs, module: cpu, module: cuda, module: loss, module: mkldnn, module: nn, module: numerical-reproducibility, module: performance, module: porting, module: random, module: rnn, module: sleef, module: sorting and selection, module: trigonometric functions, module: vectorization, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2180 Discrepancy in BCEWithLogitsLoss and ClassNLLLoss, #4703 Very slow on CPU, #4825 [feature request]Support AVX512F intrinstics to vectorize operations, #18987 Performance issue of unique on CPU, #19163 LayerNorm is very slow (almost frozen) in CPU of multiprocessing, #22803 Port `fmod` operator from the TH code to Aten",5a9089cd0c732d39c3095c71296e7242823d5598fe453c5cc42ea1d17b5fddf1
issue-collapse:b99e99dabeb39263,module: cpu stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13018, 23509, 30388, 48245, 68394, 72408, 74876, 88565, 88805, 91936, 94457, 95945, 99690, 104602",,,,"module: cpu, module: fft, module: half, module: hub, module: linear algebra, module: memory usage, module: multiprocessing, module: multithreading, module: nn, module: numerical-stability, module: performance, module: vectorization","enhancement, intel, module: cpu, module: fft, module: half, module: hub, module: linear algebra, module: memory usage, module: multiprocessing, module: multithreading, module: nn, module: numerical-stability, module: performance, module: vectorization, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13018 Improved performance for torch.multinomial with small batches, #23509 Enable PyTorch Bfloat16 for CPU and add MKL-DNN bfloat16 optimization for Cooper Lake, #30388 Increasing memory usage on CPU, #48245 RuntimeError: ""threshold_cpu"" not implemented for 'Half', #68394 INTERNAL ASSERT FAILED at ""../aten/src/ATen/MapAllocator.cpp"":323, please report a bug to PyTorch. unable to mmap 68 bytes from file </torch_530808_7992>: Cannot allocate memory (12) ---------------------------------------------------------------------------, #72408 Matrix multiplication is 30 times slower for integers than floats on CPU",e839fdb977a6446467fe7ddae34034c4bacff0c01690b701b732705fbf76d332
issue-collapse:9e95019642ca7e97,module: cpu stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22331, 44343, 52291, 70239, 90695, 94322, 105778, 107433, 108496, 112714, 114080, 117429, 124934, 125744",,,,"module: cpu, module: cuda, module: deadlock, module: docs, module: half, module: intel, module: loss, module: memory usage, module: nn, module: numerical-stability, module: performance, module: sparse, module: windows","actionable, enhancement, feature, intel, matrix multiplication, module: cpu, module: cuda, module: deadlock, module: docs, module: half, module: intel, module: loss, module: memory usage, module: nn, module: numerical-stability, module: performance, module: sparse, module: windows, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22331 Integer division by Zero giving large number results instead of NaN/inf on Windows, #44343 Asynchronous Execution on CPU, #52291 ""LayerNormKernelImpl"" not implemented for 'Half' - CPU, #70239 `pytorch` hangs during interaction with `ray` package, #90695 `torch.empty` produces incorrect tensors with `layout=sparse_csr|sparse_csc` on the CPU, #94322 RFC: Enabling AVX512 dispatch for compute-intensive ATen ops",1d7722fc4b4c2ab65a3196627ce27b2adda103cfdcdf336819659a2a04339feb
issue-collapse:f4f0a62f5ce99f6c,module: cpu stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9873, 93188, 118499, 123047, 125547, 126115, 131011, 132386, 132579, 134459, 134996, 135542, 138337, 139472",,,,"module: cpu, module: cuda, module: docs, module: edge cases, module: half, module: linear algebra, module: multithreading, module: nn, module: performance, module: special, oncall: pt2","actionable, module: cpu, module: cuda, module: docs, module: edge cases, module: half, module: linear algebra, module: multithreading, module: nn, module: performance, module: special, oncall: pt2, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9873 Pytorch is slow when only using CPU, and cannot utilize multicore of CPU, #93188 Why is AvgPool2D taking longer than Conv2D for the same input?, #118499 Autocast on cpu dramatically slow, #123047 Alternate algorithm for computing MaxPool2D under specific condition., #125547 PyTorch model slows down 30x when binding threads with fewer than half the CPU cores., #126115 filtering on CPU slower in pytorch than scipy",2453aa1ec455da84e19a0340b26d8b066e063f24ef65156fc33853464eecf3db
issue-collapse:a46c8b1a35d41bac,module: cpu stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"110858, 132395, 133304, 140112, 149304, 151037, 151463, 152294, 153919, 154017, 155671, 156959, 162719, 163052",,,,"module: cpu, module: cuda, module: distributions, module: edge cases, module: error checking, module: intel, module: linear algebra, module: mkldnn, module: multiprocessing, module: numerical-stability, module: performance, module: privateuse1, module: python frontend, module: rocm, module: testing, module: xpu, oncall: quantization, oncall: releng","actionable, has workaround, module: PrivateUse1, module: cpu, module: cuda, module: distributions, module: edge cases, module: error checking, module: intel, module: linear algebra, module: mkldnn, module: multiprocessing, module: numerical-stability, module: performance, module: python frontend, module: rocm, module: testing, module: xpu, oncall: quantization, oncall: releng, rocm, topic: performance, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#110858 Broadcasting matmul is much slower than corresponding einsum, #132395 `torch.multinomial` generates incorrect distribution, #133304 The performance regression of `conv_transpose1d` on CPU with variable-shaped inputs, #140112 inconsistency in ```torch.nn.BatchNorm1d``` on CPU and GPU, #149304 Machine with isolated cores is not respected by torch, #151037 [release] CPU perf benchmark latency increase for 2.6->2.7 on c5.24xlarge and A100 instances",e49263b2a460757535f70e371736eb2a7b1fcadd51595e489ee4c53f015b2619
issue-collapse:c26b2a45b240f988,module: cpp-extensions stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55363, 81899, 82132, 95786, 99143, 113550, 114935, 114962, 119190, 119206, 119633, 121131, 132130, 136664",,,,"module: cpp-extensions, module: cuda, module: docs, module: mta, module: pybind, module: third_party","actionable, better-engineering, enhancement, module: cpp-extensions, module: cuda, module: docs, module: mta, module: pybind, module: third_party, triaged",,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55363 it is lack of compatibility of CUDA_HOME (or other env), #81899 Dependency header directory is not properly expanded in the utils.cpp_extention in ninja mode, #82132 External libraries cannot have a requirements.txt that needs to install a cpp_extension, #95786 Static asserts on accessor templates, #99143 No documentation to show how to implement aten::view for custom backend, #113550 CUDA extension error message doesn't look correct",f607bfb618c560fc4a813c58bbfaa1b4293e7063dca3c08d5f9628fa9b1dbd0f
issue-collapse:fb7c2901bceb080f,module: cpp-extensions stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"90775, 118842, 124454, 140015, 142145, 144690, 148920, 149798, 154187, 155378, 161068, 163659, 166219, 170542",,,,"module: cpp-extensions, module: cuda, module: dispatch, module: library, module: privateuse1, module: pt2-dispatcher, module: tests, module: windows, oncall: pt2","better-engineering, enhancement, module: PrivateUse1, module: cpp-extensions, module: cuda, module: dispatch, module: library, module: pt2-dispatcher, module: tests, module: windows, oncall: pt2, topic: binaries, topic: bug fixes, triaged",,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#90775 Extend ""torch.utils.cpp_extension.load"" for both lib64 and **lib**, #118842 CUDA Extension Bug for Pytorch 2.2.0, #124454 torch.utils.cpp_extension.load recompiling every time, #140015 Find ninja from local environment in `torch.utils.cpp_extension.load`, #142145 Fix disabled 3.13 cpp_extension test failing due to registration error, #144690 Support new CUDA conda package layout natively in cpp_extension.CUDAExtension",b3288766d1963648f5dd1d98231b38bd471053d0575d5dcf1d1b58224f08a20a
issue-collapse:e798e53f1d255dd0,module: cpp stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14959, 28929, 32614, 39522, 40729, 40959, 41525, 41903, 41955, 42848, 43490, 45855, 46712, 47591",,,,"module: cpp, module: cpp-extensions, module: docs, module: internals, module: nn, module: performance, module: pruning, module: pybind, module: serialization","better-engineering, enhancement, feature, function request, module: cpp, module: cpp-extensions, module: docs, module: internals, module: nn, module: performance, module: pruning, module: pybind, module: serialization, small, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14959 get/set device in c++, #28929 torch::tensor(scalar) behaves differently from at::tensor(scalar), #32614 Logical AND and OR for Tensors in C++ API., #39522 Valgrind leak checking flags losses in libtorch, #40729 Custom c++ extension build process doesn't preserve color from compiler, #40959 Will the model run slower when deployed using libtorch ?",0b724cb1ef31d44df3a45a6b17ed5154353ca3300bedab7bf0e8457d34e7883d
issue-collapse:d90693b7d6fed9e2,module: cpp stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9912, 9950, 25979, 31178, 49320, 52744, 53157, 53584, 53622, 53983, 54058, 54201, 54815, 54874",,,,"module: cpp, module: cpp-extensions, module: cuda, module: docs, module: internals, module: language binding, module: logging, module: nn, module: random","enhancement, feature, function request, module: cpp, module: cpp-extensions, module: cuda, module: docs, module: internals, module: language binding, module: logging, module: nn, module: random, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9912 Stop ifdef'ing out scatter/gather (comm) in libtorch, #9950 Stop passing inplace/out arguments as (non-const) Tensor& to functions, #25979 [C++] `Module::pretty_print` is broken, #31178 torch::nn::functional::interpolate crash, #49320 how could I print the log in source code, #52744 Gradient checkpointing support in C++ API",7c882b8e3a52fcbe70296705d11395f469f298b92c0ef65d2395f79ff86c9fa2
issue-collapse:e3cc9672c14cc1de,module: cpp stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12868, 17554, 17879, 26456, 35723, 41177, 50375, 52575, 56522, 56781, 59418, 62566, 63897, 64206",,,,"module: cpp, module: cpp-extensions, module: memory usage, module: molly-guard, module: named tensor, module: nn, module: optimizer, module: performance, module: rnn, module: tensor creation, oncall: jit, oncall: mobile","module: cpp, module: cpp-extensions, module: memory usage, module: molly-guard, module: named tensor, module: nn, module: optimizer, module: performance, module: rnn, module: tensor creation, oncall: jit, oncall: mobile, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12868 [CAPI] Increase of memory usage when exporting a Adam optimzer, #17554 Add ASSERT for calling accessor on a GPU tensor, #17879 C++ nn::Sequential push_back() copies module if the module is concrete type, #26456 [libtorch]Same model in CUDA and CPU got different result?, #35723 LibTorch API on Mobile, #41177 The new version of the libtorch become slow",cedc9cb0ebd8c985f34f45df4a4daae90b09570d89e45176b870e3f0937729bf
issue-collapse:0b2f0ebc2124bbb1,module: cpp stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17614, 63479, 65984, 66755, 70416, 71069, 74101, 74942, 75562, 77161, 77241, 78876, 79004, 81625",,,,"module: cpp, module: cpp-extensions, module: data, module: dataloader, module: determinism, module: docs, module: jetson, module: lint, module: nn, module: vision, oncall: jit","Stale, actionable, enhancement, module: cpp, module: cpp-extensions, module: data, module: dataloader, module: determinism, module: docs, module: jetson, module: lint, module: nn, module: vision, oncall: jit, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17614 Document whether it is possible to train TorchScript modules, #63479 Cannot include extension.h under Windows - linker error THPVariable_Wrap, #65984 libtorch: collate_fn equivalent, #66755 Domain Transformation APIs for LibTorch and LibTorch-Lite, #70416 libtorch cuda use too much system memory, #71069 C++ torch::nn::Sequential clone() method overwrites child module names",929adcf0c5eca3809d553a75ba47600f05fcb7c63d48d70ec257df85eec152ec
issue-collapse:c1c384147e3e1228,module: cpp stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18630, 32166, 36914, 73171, 78880, 82542, 84473, 85078, 87706, 90751, 94827, 97802, 97849, 98907",,,,"module: cpp, module: deadlock, module: docs, module: mkl, module: mps, module: optimizer, module: performance, module: testing","actionable, enhancement, module: cpp, module: deadlock, module: docs, module: mkl, module: mps, module: optimizer, module: performance, module: testing, topic: binaries, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18630 documentation for C++ / libtorch autograd profiler, #32166 C++ randint returns float32, python returns int64, #36914 Problem with c10/utils/variant.h, #73171 libtorch need operator= in torch::Device, #78880 fatal_signal_asan_no_sig_test in current master hang., #82542 Is there Doc that explains how to call an extension op in another extension implementation?",7b7e13735766556c33ac2b8b6fc8eba9c2eb393bcdcc318ba60aed1d15d83533
issue-collapse:de1a40f2ce9ce187,module: cpp stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25646, 35872, 59074, 80580, 90424, 99225, 101210, 103570, 103671, 103837, 106533, 108651, 115034, 117938",,,,"module: cpp, module: cuda, module: memory usage, module: pybind, module: sorting and selection, oncall: quantization","module: cpp, module: cuda, module: memory usage, module: pybind, module: sorting and selection, oncall: quantization, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25646 libtorch forward memory leak, #35872 torch.autograd.set_detect_anomaly(True) does not exist in C++?, #59074 c++ use pybind11 to import torch free(): invalid pointer, #80580 `torch.ops.aten.find` inconsistent with `str.find`, #90424 A few functions in fbgemm_utils.cpp are defined in global namespace, #99225 Libtorch consumes too much memory as 16225",15e0bcce8da4b96ac70d34c5176fe893241fb4f24dc33102a555b07cfc2c953c
issue-collapse:1f60169648cdf2de,module: cpp stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17984, 19453, 35441, 42665, 45139, 111704, 117930, 118478, 118723, 119486, 120085, 125471, 128509, 130514",,,,"module: cpp, module: cpp-extensions, module: cuda, module: dataloader, module: docs, module: multithreading, module: optimizer, module: performance, module: serialization, oncall: jit, oncall: mobile","actionable, enhancement, module: cpp, module: cpp-extensions, module: cuda, module: dataloader, module: docs, module: multithreading, module: optimizer, module: performance, module: serialization, oncall: jit, oncall: mobile, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17984 ""unknown builtin op"" error with static library, #19453 How to load PyTorch model with LSTM using C++ api, #35441 torch::normal only supports (double, double), but at::normal supports (double, double) / (double, Tensor) / (Tensor, double) / (Tensor, Tensor), #42665 Optimizer support via Libtorch C++ on Android, #45139 Mention accessor/data_ptr for raw memory access in Libtorch index API document and discuss performance implications, #111704 Add more flexibility on print / output console",cdda078628ba379afd500acbdc5b54490fb85182bf7868b6aa6b7d538054932b
issue-collapse:4d3eb1ebdd563388,module: cpp stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23859, 32421, 79272, 89051, 121485, 134098, 134637, 138387, 138940, 138981, 143564, 144955, 145493, 147581",,,,"module: cpp, module: dependency bug, module: lint, module: memory usage, module: multithreading, module: nn, module: privateuse1, module: rnn, oncall: jit","module: PrivateUse1, module: cpp, module: dependency bug, module: lint, module: memory usage, module: multithreading, module: nn, module: rnn, oncall: jit, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23859 CUDA: THTensor code complains about devices not matching when creating tensor from blob, #32421 Segmentation fault in C++ API torch::from_blob(...).clone(), #79272 PyTorch leaks a macro definition called ""CHECK"" in the C++ version, #89051 caffe2_interface_library CMake macro prevents linking to LibTorch as a transitive dependency, #121485 libtorch memory leakage when inference by Multi-thread, #134098 libtorch reasoned that the crnn model was out of speed the first few times",82fe59478eabc545ff2f8e2ab4f00becfcf665522870c75a2b5ab5f7505e4381
issue-collapse:5832a364f5ce4250,module: cpp stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15149, 17797, 19125, 20315, 22022, 25635, 26659, 27055, 28930, 31688, 32237, 32415, 33301, 33303",,,,"module: cpp, module: data parallel, module: doc infra, module: named tensor, module: nn, oncall: distributed, oncall: jit","feature, module: cpp, module: data parallel, module: doc infra, module: named tensor, module: nn, oncall: distributed, oncall: jit, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15149 Use std::variant to represent C++ side enumerations (with binding support), #17797 Non-coherent result for C++ with multithreading and GPU, #19125 Multi-gpu via torch::nn::parallel::data_parallel, #20315 libtorch+opencv Mat result error: different from the python ones, #22022 nn.modules.functional.h does not support optional arguments, #25635 [C++] Support negative index in `torch::TensorAccessor::size()`",b058e88376346fc90877f6c8b65132bd64c188013845b5b55207b102ef5d91fe
issue-collapse:001c423ba8defd5b,module: cpp stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18283, 26086, 27958, 33514, 34052, 34068, 34881, 35078, 35640, 35770, 36004, 36099, 36891, 40098",,,,"module: cpp, module: docs, module: error checking, module: memory usage, module: nn, module: optimizer, module: printing, module: random, module: sanitizers, oncall: jit","module: cpp, module: docs, module: error checking, module: memory usage, module: nn, module: optimizer, module: printing, module: random, module: sanitizers, oncall: jit, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18283 Doubly freed pointer in torch::cat error handling when called via pybind11., #26086 [Feature Request] Trace / Script C++ models, #27958 how to use libtorch library in cuda file with nvcc compiler(c++)?, #33514 Make setter non-optional, e.g., TensorOptions::device(optional<Device>) -> device(Device), and add a device_opt setter, #34052 Model loaded in C++ runtime is not thread safe, #34068 [C++ API Parity] Incorrect documentation for optim initialization in serialization docs",dfae0aaa963d08c923e9d3a837f53e087f5ac9bfab09a0c6b74c7e651e8f742a
issue-collapse:121abca215679ff9,module: cpp stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16375, 33811, 35713, 41213, 73016, 145337, 145656, 150548, 153463, 155011, 163502, 167762, 168376, 169893",,,,"module: cpp, module: cpp-extensions, module: cuda, module: debug-build, module: docs, module: error checking, module: mkldnn, module: pybind, module: serialization, module: sparse, module: windows, module: xpu","feature, module: cpp, module: cpp-extensions, module: cuda, module: debug-build, module: docs, module: error checking, module: mkldnn, module: pybind, module: serialization, module: sparse, module: windows, module: xpu, release notes: python_frontend, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16375 Python-bound C++ frontend modules don't handle attributes well, #33811 Models saved in C++ LibTorch with torch::save, cannot be loaded in python using torch.load, #35713 libtorch for Windows. MNIST example does no work., #41213 libTorch cpp docs missing for Tensor::item(), #73016 Unable to build and use libtorch function via pybind11: undefined symbol error upon import, #145337 [libTorch] Model initialization on multi-device is slow. It seems to run sequentially in multi-thread",fb84ba97e5c641e80f4d8906164d9c43d22c6ab76d1b06b54286825aeae4fd1c
issue-collapse:bfa89db3b40c1d26,module: correctness (silent) stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"70398, 76232, 79788, 80776, 81996, 84202, 86074, 86465, 89320, 97631, 101934, 113956, 114350, 125051",,,,"module: correctness (silent), module: cpu, module: cuda, module: docs, module: edge cases, module: error checking, module: fft, module: intel, module: linear algebra, module: loss, module: magma, module: masked operators, module: mps, module: nn, module: numerical-stability, module: python frontend, module: shape checking, module: sparse","actionable, module: correctness (silent), module: cpu, module: cuda, module: docs, module: edge cases, module: error checking, module: fft, module: intel, module: linear algebra, module: loss, module: magma, module: masked operators, module: mps, module: nn, module: numerical-stability, module: python frontend, module: shape checking, module: sparse, triaged",,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#70398 `torch.broadcast_to` can create tensor with negative dimension., #76232 Observing a strange behavior - Row parallelism, #79788 Parameter.__deepcopy__ doesn't preserve view relationships, #80776 `torch.inverse()` crash in cuda, #81996 linspace cpu and sometimes cuda is wrong on integral types, #84202 Gradient value calculation error in MultiLabelMarginLoss",16a61498d375ef0eb6b0a1c07f318cf9a393858b8c2dd96e9cfdb7c7a8b8cace
issue-collapse:e3be9505a30e42ed,module: correctness (silent) stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"57721, 80494, 83015, 96225, 96982, 100334, 101167, 101878, 131736, 132605, 133100, 136031, 137951, 137968",,,,"module: correctness (silent), module: cpp, module: cuda, module: docs, module: mps, module: nn, module: special","module: correctness (silent), module: cpp, module: cuda, module: docs, module: mps, module: nn, module: special, triaged",,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#57721 nn.TransformerEncoder cannot deal with large negative value even when this value is masked by src_key_padding_mask, #80494 [bug] libtorch bug in nn::MultiheadAttention and nn::Transformer, #83015 Incorrect tensor conversion to m1 MPS., #96225 [MPS] F.conv1d and F.conv2d produce incorrect gradients when minibatch >= 2^16, #96982 [mps] conv1d outputs zeros, #100334 MPS device inference all same value",65038aa6bfd1d3b447e1c036aa6f2a4d0fbd0fa2b36765832b950b81f703392c
issue-collapse:fc59ca7700fbf2bf,module: correctness (silent) stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"78064, 103023, 104106, 104832, 107214, 107455, 107694, 111479, 121961, 124029, 134430, 134534, 147510, 153957",,,,"module: correctness (silent), module: memory format, module: mps, module: nn","module: correctness (silent), module: memory format, module: mps, module: nn, triaged",,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#78064 nn.CosineSimilarity returns value larger than 1, #103023 MPS bug: padding_idx in nn.Embedding does not prevent gradient accumulation, #104106 batch size unexpectedly affects model inference on Mac M1, #104832 MPS matmul with sliced (strided) out argument produces wrong output, may corrupt memory, #107214 The difference between input grad computed by channels last backward and the input grad computed by channels first backward of Hardswish on MPS is too large, #107455 Moving tensor to MPS using .to(torch.device('mps') deletes entries from tensor",ed69c9912b746515e9d570167f829dd66b7182ab6737210201d68dc6140a176f
issue-collapse:8521c6ca4324eac3,module: correctness (silent) stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"48900, 65083, 67470, 71774, 94183, 94186, 107381, 107478, 121439, 122030, 136623, 158199, 162362, 169738",,,,"module: correctness (silent), module: docs, module: internals, module: linear algebra, module: mps, module: partial aliasing, module: performance, module: sdpa, module: sparse, module: structured kernels, oncall: pt2","actionable, module: correctness (silent), module: docs, module: internals, module: linear algebra, module: mps, module: partial aliasing, module: performance, module: sdpa, module: sparse, module: structured kernels, oncall: pt2, triaged",,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#48900 addmm with out= argument returns incorrect result, #65083 THPVariable_subclass_dealloc (i.e. tp_dealloc for torch.Tensor) should handle Python error state gracefully, #67470 integer (and possibly float16) matmul fails test_noncontiguous_samples on CPU (and all backward dtypes, too?), #71774 matmul returns uninitialized memory for int64 tensors with inner dimension of zero, #94183 Addition of batch CSR tensors produces incorrect and invalid CSR tensor, #94186 Addition of hybrid CSR tensors produces incorrect and invalid CSR tensor",2f84807518c5a9904f1713314c1c88b9f6eb73687fc46714ca00e6d8e6243f4e
issue-collapse:e13dc419001360df,module: convolution stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7740, 22330, 22577, 22919, 29369, 29713, 30399, 30934, 35655, 41311, 42663, 47341, 49207, 50190",,,,"module: convolution, module: cuda, module: cudnn, module: dependency bug, module: docs, module: double backwards, module: error checking, module: nn, module: numerical-stability, module: performance, module: vision","enhancement, function request, has workaround, module: convolution, module: cuda, module: cudnn, module: dependency bug, module: docs, module: double backwards, module: error checking, module: nn, module: numerical-stability, module: performance, module: vision, quansight-nack, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7740 Conv3D can be optimized for cases when kernel is spatial (probably), #22330 Training CNNs with deconvolution, #22577 Double backward 3 times slower for conv2d with padding = 1, #22919 Unify tensor shape formatting in shape checks, #29369 nn.Conv(n)d constructor doesn't check for the number of kernel dimensions, #29713 [FR] F.pad support syntax sugars for specifying the padding amount",d583e652f8c9d9deb8ab17e09e752474fc854aa963b380ff1c9c9457bac059da
issue-collapse:d0bb8113a2deb924,module: convolution stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10685, 11974, 13268, 17150, 40275, 48439, 51711, 54906, 55045, 56224, 56246, 56263, 58343, 61772",,,,"module: convolution, module: cuda, module: cudnn, module: dependency bug, module: docs, module: error checking, module: memory format, module: nn, module: padding","enhancement, feature, module: convolution, module: cuda, module: cudnn, module: dependency bug, module: docs, module: error checking, module: memory format, module: nn, module: padding, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10685 [feature request] convtbc with group convolution, #11974 Specify out= argument to convolution, #13268 Non-Zero Padding in Convolution Module, #17150 Conv2d layers should accept 4-tuple for padding argument, #40275 Conv3d with specific kernel size outputs inconsistent results between FP16 and FP32 in V100 GPU, #48439 cudnn convolution modifies the input Tensor metadata inplace when it tries to `.resize_()` it",6e8e2dcbd9da83c414c7759cdb94bf52438037bdb3dfc0f92762c49c181f8dd8
issue-collapse:591d282fafde018c,module: convolution stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23547, 64145, 65661, 66963, 67071, 70184, 72850, 73413, 75495, 77838, 77900, 77901, 79222, 82093",,,,"module: convolution, module: cpp, module: cuda, module: cudnn, module: edge cases, module: error checking, module: forward ad, module: macos, module: numerical-reproducibility, module: performance","enhancement, module: convolution, module: cpp, module: cuda, module: cudnn, module: edge cases, module: error checking, module: forward ad, module: macos, module: numerical-reproducibility, module: performance, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23547 model use dilated conv backward in v1.1.0 is ~3x slower than in v0.4.1 on 1080Ti, #64145 F.conv2d: confusing error message when using uint8 input instead of float32, #65661 [BUG] ConvTranspose with out_channels=1, #66963 Significantly difference in execution time when convolution is run as nn.Conv2d and as nn.Sequential, #67071 Forward results vary depending on batch size on A100 machine, #70184 Optimization: convolution_backward doesn't always need to call .contiguous on certain inputs",6f0a8fb827e76d49571aba0d66ab59e160ee458f36c21e22023035258eaca0a3
issue-collapse:2c57b08552729e50,module: convolution stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7806, 13505, 35603, 53191, 59730, 62672, 75500, 76244, 83579, 88147, 101871, 106450, 115610, 117963",,,,"module: convolution, module: cpp, module: cuda, module: cudnn, module: dependency bug, module: fft, module: lazy, module: linear algebra, module: memory usage, module: nn, module: nnpack, module: padding, module: performance, oncall: quantization","feature, function request, has workaround, low priority, module: convolution, module: cpp, module: cuda, module: cudnn, module: dependency bug, module: fft, module: lazy, module: linear algebra, module: memory usage, module: nn, module: nnpack, module: padding, module: performance, oncall: quantization, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7806 OOM Exception when using torch.nn.grad.conv2d_weight (apparently because CuDNN backwards is not used), #13505 [feature request] Singular values and spectral norm for convolutional layers, #35603 Performance bug with convolutions with weights and inputs of similar spatial size, #53191 RuntimeError: cuDNN error: CUDNN_STATUS_INTERNAL_ERROR, #59730 Quantized conv2d with dilation and groups much slower than float32, #62672 Add Convolution support for lazy tensor",a1995603f36643ca2c86067018527bbf09cb888559069324af7960bf825e8d83
issue-collapse:e1b5f89c6f04c39b,module: convolution stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23854, 65510, 77821, 121261, 121715, 123533, 127221, 127375, 128936, 133114, 134641, 136086, 136479, 139066",,,,"module: convolution, module: cuda, module: cudnn, module: docs, module: edge cases, module: error checking, module: forward ad, module: functorch, module: initialization, module: memory usage, module: nans and infs, module: nn, module: norms and normalization, module: onnx, module: performance, module: tf32, module: vmap","module: NaNs and Infs, module: convolution, module: cuda, module: cudnn, module: docs, module: edge cases, module: error checking, module: forward ad, module: functorch, module: initialization, module: memory usage, module: nn, module: norms and normalization, module: onnx, module: performance, module: tf32, module: vmap, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23854 Fan out calculation broken for group (depthwise) convolution, #65510 nn.conv1d padding='same', #77821 cannot convert to channels last format for conv2d conv3d hybrid model, #121261 Convolution and norm with vmap jacobian failing, #121715 Depthwise conv2d op produces nan values with non-nan input, #123533 Memory allocation for Conv2d",d9c8a634dbe503d577f50c73f9c634b361528f39d7d1322f7ec68f6864fb80ad
issue-collapse:630966c03862e848,module: convolution stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"21462, 46027, 53178, 112376, 119677, 142467, 144445, 150614, 152992, 153139, 157334, 161778, 165744, 165976",,,,"module: convolution, module: correctness (silent), module: cuda, module: cudnn, module: dependency bug, module: error checking, module: fft, module: floatx (formerly float8), module: functorch, module: mps, module: nans and infs, module: nn, module: nnpack, module: performance","enhancement, function request, low priority, module: NaNs and Infs, module: convolution, module: correctness (silent), module: cuda, module: cudnn, module: dependency bug, module: error checking, module: fft, module: floatx (formerly float8), module: functorch, module: mps, module: nn, module: nnpack, module: performance, topic: fuzzer, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21462 Slow convolution with large kernels, should be using FFT, #46027 Padding mode for ConvTransposeNd, #53178 Exceedingly different F.conv2d outputs on cuda/cpu, #112376 Contradictory Error Message for stride Argument in torch.conv_transpose3d(), #119677 Unexpected Gradient Results From conv1d on MacOS, #142467 Wrong error message",5b3ca09c3150acafbf6ff32450a8a9d45300a04ab1241e8bb783542bcedad0a3
issue-collapse:3a9ebfc69960b759,module: complex stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"35312, 36175, 44817, 45690, 48000, 49133, 51714, 52457, 54753, 56012, 57358, 60941, 62884, 64186",,,,"module: complex, module: dispatch, module: nans and infs, module: nn, module: numpy, module: tests, module: trigonometric functions, module: type promotion","complex_autograd, enhancement, low priority, module: NaNs and Infs, module: complex, module: dispatch, module: nn, module: numpy, module: tests, module: trigonometric functions, module: type promotion, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#35312 Method/constructor which takes as input angle and magnitude and returns a complex tensor, #36175 New dtype ComplexPolarFloat (phasor), #44817 ComplexHelper.h contains non-inline functions, #45690 Complex and real results do not agree when computing reciprocal or pow(-1) of 0, #48000 Inconsistent complex results with NumPy when computing non-positive power of 0, #49133 Unknown failure for test_reference_numerics_sinc_cpu_complex64",e4f13386bebf9ddb0e25d590d22826926dcab2ae988a186ec4ab04676dcd5234
issue-collapse:c3f6d55dafbed79c,module: complex stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50098, 61909, 65306, 67333, 69287, 71472, 71635, 71671, 74748, 76166, 77046, 78170, 81982, 86340",,,,"module: complex, module: half, module: jiterator, module: linear algebra, module: nans and infs, module: numpy, module: optimizer, module: primtorch, module: reductions, module: sparse, module: special, module: type promotion, module: viewing and reshaping","Stale, actionable, feature, module: NaNs and Infs, module: complex, module: half, module: jiterator, module: linear algebra, module: numpy, module: optimizer, module: primTorch, module: reductions, module: sparse, module: special, module: type promotion, module: viewing and reshaping, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50098 Errors when coercing complex numbers of various sizes, #61909 Support matrix operations between complex and real tensors, #65306 [numpy] Add `iscomplexobj` and `isrealobj`, #67333 Feature request: Complex Number Support for Special Functions, #69287 `out=` variant for `conj` unary operator, #71472 Feature request: Add complex support to `torch.nanmean`",1e8b4948ccb4aeba746cc445d3bb248fe8fbf632b0506dd76c43cb7829b58c99
issue-collapse:77c47566bfc67b82,module: complex stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38650, 48486, 51210, 53440, 72134, 83376, 90447, 92683, 92818, 94395, 96704, 104466, 108882, 115164",,,,"module: complex, module: correctness (silent), module: distributions, module: functorch, module: masked operators, module: nans and infs, module: numpy, module: python array api, module: random, module: ux, oncall: jit","complex_autograd, feature, function request, module: NaNs and Infs, module: complex, module: correctness (silent), module: distributions, module: functorch, module: masked operators, module: numpy, module: python array api, module: random, module: ux, oncall: jit, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38650 Implement torch.erf for complex dtypes, #48486 [complex] torch.abs: does not match numpy, #51210 Complex->Real cast is a warning, calling `real` or `imag` on non-complex tensors is an Error., #53440 Add complex support for torch.unique, #72134 Add support for `complex` `mean` to the `normal` operator, #83376 Complex-Valued Gaussian distributions",7b690f1a98863b854bda4688317fd5671e06fa67414436535b9dd66596f82a66
issue-collapse:475e1df08300020f,module: complex stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"43579, 92398, 104516, 116018, 117232, 117236, 117245, 121588, 125746, 135210, 136568, 136621, 151182, 154496",,,,"module: complex, module: correctness (silent), module: cpp, module: cpu, module: fft, module: inductor, module: linear algebra, module: mps, module: nans and infs, module: nn, module: norms and normalization, module: random, module: vectorization, oncall: cpu inductor, oncall: pt2","feature, module: NaNs and Infs, module: complex, module: correctness (silent), module: cpp, module: cpu, module: fft, module: inductor, module: linear algebra, module: mps, module: nn, module: norms and normalization, module: random, module: vectorization, oncall: cpu inductor, oncall: pt2, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43579 For the same complex dtype and same value, comparing a PyTorch tensor with a NumPy array results in False, #92398 [Inductor] support complex dtypes, #104516 vec_test_all_types_xxx with dtype c10::complex<float> and c10::complex<double> has failures on division, #116018 Inconsistent behavior of torch.einsum on float32 and complex128, #117232 ```torch.Tensor.renorm```: Internal assert failed, #117236 ```torch.linalg.lstsq```: Internal assert failed",0be87bbaaf686a3c68b8540acfd5b2a48d7144bd71c184ce991a8450087a05ee
issue-collapse:c82fe4faae2a4578,module: complex stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"63138, 80256, 116414, 125198, 125718, 131283, 139349, 148670, 154501, 154606, 157272, 159870, 160629, 163984",,,,"module: complex, module: cpu, module: cuda, module: dynamo, module: edge cases, module: jiterator, module: linear algebra, module: mps, module: nans and infs, module: nn, module: numpy, oncall: pt2","actionable, complex_autograd, enhancement, feature, has workaround, low priority, module: NaNs and Infs, module: complex, module: cpu, module: cuda, module: dynamo, module: edge cases, module: jiterator, module: linear algebra, module: mps, module: nn, module: numpy, oncall: pt2, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#63138 Sign in slogdet is set to requires_grad = False even when using complex numbers., #80256 [complex] dropout and it's variants should support complex tensors, #116414 Convolution NN for complex numbers and more special functions, #125198 [CUDA][Compex] `test_reference_numerics_large_jiterator_unary_cuda_complex64` broken after updating to `numpy >= 1.25.0`, #125718 `torch.compile` and complex numbers, #131283 `input` argument of `nan_to_num()` works with `complex` type but `nan`, `posinf` and `neginf` argument don't work with `complex` type",8cf5d9d3da201f7f880996b7173d9149030f0c2a33c40a324deea1606ab61536
issue-collapse:ad122d0c88c3cba1,module: codegen stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17257, 43114, 44738, 49416, 50987, 53023, 54001, 58087, 58124, 58765, 62591, 63356, 66341, 73425",,,,"module: codegen, module: error checking, module: internals, module: pybind, module: typing","better-engineering, enhancement, module: codegen, module: error checking, module: internals, module: pybind, module: typing, triaged",,Handle this module: codegen stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17257 Generated `__init__.pyi` contains invalid default values, #43114 out kwargs are sometimes inconsistent with returned named tuple field name (and in some cases, cannot be made consistent), #44738 [tools.codegen] Remove byte-for-byte compatibility code, #49416 A few years after #701 and PyTorch is still using implicit __all__ imports., #50987 Scalar/Tensor arg type for op schemas, #53023 [codegen] Make it easier to codegen call to API",47bdfd781689311352970a30e6fbb4f0ac15a692c20a92fb8e8ce104761dc86b
issue-collapse:2ccfa772fca70ef9,module: codegen stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"69471, 73505, 75904, 77251, 83214, 84265, 84616, 87178, 88598, 100062, 103160, 103370, 111157, 162398",,,,"module: codegen, module: cpp, module: dispatch, module: infra, module: optimizer, module: python frontend, module: structured kernels","Stale, better-engineering, module: codegen, module: cpp, module: dispatch, module: infra, module: optimizer, module: python frontend, module: structured kernels, topic: build, triaged",,Handle this module: codegen stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#69471 Include Declarations.yaml in Libtorch distributions, #73505 Figure out what to do with functions that take both Tensor and TensorOptions, #75904 Move _SKIP_PYTHON_BINDINGS to native_functions.yaml, #77251 The codegen unconditionaly generate code even when it is not going to be used, #83214 torchgen.model.FunctionSchema.parse fails with following ops' schema, #84265 torch.Tensor.to.dtype_layout overload is not available in Python",b29dedb1997a34f7ac69b333403b8d58d96366ff7b8ad1f1e1473a4546f2edb0
issue-collapse:b35f4e5c814af10e,module: ci stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72555, 72933, 73083, 73531, 73755, 74256, 74281, 76028, 76972, 77340, 78260, 78634, 79083, 79893",,,,"module: ci, module: docker, module: docs, module: ios","enhancement, module: ci, module: docker, module: docs, module: ios, topic: bug fixes, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72555 Enforce quotas on CI users, #72933 Why there are 8 flavors of iOS build jobs for every commit, #73083 Improve test_test_history.py, #73531 Mention docker build process in RELEASE.md and automate building those for release, #73755 CI: Bake as many dependencies as we can in the AMI (windows), #74256 Create secure credential storage for metrics credentials and associated documentation on how to regenerate them if needed",d21c246421dbbfc9dd2424c3ff217b6aef0cc9086f6703045ee66fedf8292362
issue-collapse:e404ad5978a1eef0,module: ci stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72749, 75703, 76838, 79787, 79895, 79977, 80118, 80238, 81110, 81768, 81945, 81955, 81983, 82033",,,,"module: ci, module: convolution, module: nn, module: xla","module: ci, module: convolution, module: nn, module: xla, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72749 pytorchmergebot doesn't react to comments left from ""files"" tab, #75703 Let's host NVIDIA dependencies in our own S3, #76838 Windows CUDA TTS tracking task, #79787 Improve clarity by making sharding a static nightly update, #79895 Modify update-viable-strict GHA to use internal version of checkout, #79977 [CI] Do we run all cpp tests on CI?",73da21a39e88b4074b117aba521859b3b079571e03485ca53a34892137505dfc
issue-collapse:665474299b18222f,module: ci stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14701, 65718, 82259, 83081, 83082, 84259, 86131, 86717, 88309, 88408, 92173, 95645, 96637, 103860",,,,"module: ci, module: cpp, module: cuda, module: devx, module: flaky-tests, module: nvfuser, module: testing, module: tests, module: windows","better-engineering, enhancement, module: ci, module: cpp, module: cuda, module: devx, module: flaky-tests, module: nvfuser, module: testing, module: tests, module: windows, topic: binaries, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14701 CI: Flaky download from download.pytorch.org, #65718 Migrate current Windows CI scripts off of batch, #82259 [Reproducibility] Make tests say when unusual environment variables are set that change behavior of the test, #83081 [CI] Split up periodic.yml into forward-fixable.yml and periodic.yml, #83082 [CI] Create periodic fuzzy testing for PyTorch build flags, #84259 would you like upload to the cpp libtorch to vcpkg package repo?",27f7c11945c4d9891ec5a03145616d138df9644c4708754074362bb3717360fc
issue-collapse:efe09527b9ac2827,module: ci stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72814, 85499, 87501, 104808, 106073, 106237, 107253, 115659, 117424, 117576, 123169, 123506, 124417, 125146",,,,"module: ci, module: cudnn, module: docs, module: testing, module: tests, oncall: r2p, oncall: releng","module: ci, module: cudnn, module: docs, module: testing, module: tests, oncall: r2p, oncall: releng, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72814 Standardize Naming for Workflows/Jobs, #85499 Execute smoke test for Better Transformer feature, #87501 Synchronize domain builds to be executed after core build have completed, #104808 PyTorch built with CuDNN-8.8.1 crashes if CuDNN-8.9.2 is installed on the system, #106073 Potential lack of CI testing on older NVIDIA GPU, #106237 ghstack + mergebot race condition",0966d06b445a2333414d5cbb845e8779dc70065c3ed9ac8bd12b04963e4298b2
issue-collapse:ca1a6fb94cceab55,module: ci stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"87782, 106544, 107780, 117357, 117705, 123532, 126323, 127950, 129386, 129784, 131213, 131695, 136647, 137941",,,,"module: ci, module: devx, module: docs, module: mkldnn, oncall: pt2, oncall: releng","actionable, enhancement, module: ci, module: devx, module: docs, module: mkldnn, oncall: pt2, oncall: releng, topic: binaries, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#87782 Libtorch windows binaries publishing, #106544 Branch name in double quotes """", #107780 Add caffe2 ideep/onednn tests to OSS CI, #117357 [RFC] CI test stats, #117705 Add GPU runners for Dynamo shards, #123532 Request: Restrict use of tags to strictly formal release versions",3a364428068fab42e161b297b1d0109284c7aa17b8768acab81644893bbe575a
issue-collapse:e66513953e6134a3,module: ci stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"130396, 137931, 138042, 138564, 139188, 139201, 140017, 140018, 140345, 141558, 143079, 143215, 144875, 145388",,,,"module: ci, module: correctness (silent), module: third_party","actionable, feature, module: ci, module: correctness (silent), module: third_party, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#130396 Add some way to automatically opt-out of CI features, like TD., #137931 TD skips relevant tests, #138042 `Calculate Docker Image` step times out if run on LF runners, #138564 [CI] Continuous spam of PyTorch CI build status notifications in fork, #139188 PCH build fail with sccache-v0.8.2, #139201 First run lint on just the changes in the PR before running it over the entire PR",06aa0216e431ce733c51588073cef506cba597ff2ab09aac2b532124fd60ff02
issue-collapse:fb5344ff6f37f5cb,module: ci stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"141204, 143041, 148762, 149110, 149169, 150078, 151322, 152212, 152489, 152566, 152699, 152884, 154560, 156607",,,,"module: ci, module: cuda, module: flaky-tests, module: infra, module: log classifier, module: third_party, module: xpu, oncall: pt2, oncall: releng","better-engineering, enhancement, feature, module: ci, module: cuda, module: flaky-tests, module: infra, module: log classifier, module: third_party, module: xpu, oncall: pt2, oncall: releng, security, triaged, unstable",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#141204 Stack trace from pytest is very far away and far too find on some tests, #143041 Use labels to create dedicated pools of non-scalable runners for trunk jobs, #148762 magma builds should be part of the docker image builds, #149110 [XPU] Enable Windows CI/CD test for XPU, #149169 [RelEng] wheel testing for new arch versions, #150078 [RFC] Remove periodic/unstable jobs that has been continuously broken for more than 30 days",643dedb15f54762977099832656cc7e0cfe050fc79a52755b29daf58e5c72bc1
issue-collapse:72901743decf3b76,module: ci stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"141984, 157379, 158127, 158782, 158876, 159866, 160240, 161281, 162182, 163302, 163577, 163657, 163902, 164951",,,,"module: ci, module: correctness (silent), module: docker, module: flaky-tests, module: lint, module: xla, oncall: pt2","better-engineering, enhancement, module: ci, module: correctness (silent), module: docker, module: flaky-tests, module: lint, module: xla, oncall: pt2, triaged, unstable",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#141984 CI feature mirroring or caching for Docker Hub images, #157379 [CI] Need better way to detect OOMs especially on pet instances, #158127 A ""Merge-when-ready"" label when ci finishes?, #158782 [BE] linter for base docker image configurations with docker_builds.yml, #158876 UNSTABLE pull / linux-jammy-py3_9-clang9-xla / test (xla), #159866 BE: enable auto reformat List to list and similar typing",d500233363ed0e0fbd012b0269365914a9e9aa4d926611652d78b36604920146
issue-collapse:14a0a6f661dc793c,module: ci stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14839, 46065, 46066, 47208, 53910, 54938, 55330, 64868, 64939, 66775, 66902, 68193, 71152, 71407",,,,"module: ci, module: dataloader, module: fft, module: lint, module: macos, module: tests, oncall: mobile","better-engineering, feature, module: ci, module: dataloader, module: fft, module: lint, module: macos, module: tests, oncall: mobile, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14839 Test OpenCV4 in CI, #46065 TestDataLoader.test_proper_exit takes 2.5min to finish, #46066 TestXNNPACKConv1dTransformPass.test_conv1d_with_relu_fc takes 2+ min to finsh, #47208 [RFC] Add test execution time analysis CI workflow, #53910 There should be a CI build with OpenBLAS in addition to MKL, #54938 [tests] `cumprod` OpInfo tests take long time to run (around 1min)",ca1c3eb4c879403b4a22ee190eed6a7258604848dc8c155a336c5a927277ff20
issue-collapse:a709c958e540a237,module: ci stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"120710, 154074, 164493, 164860, 165173, 165684, 166011, 166072, 167351, 167670, 167746, 169184, 170266, 170365",,,,"module: ci, module: cuda, module: devx, module: docker, module: infra, module: lint, module: sparse, module: xpu","enhancement, module: ci, module: cuda, module: devx, module: docker, module: infra, module: lint, module: sparse, module: xpu, triaged, unstable",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#120710 Reduce duplication in number of places where we specify dependency pin, #154074 UNSTABLE Build manywheel docker images for s390x / build-docker-cpu-s390x, #164493 [CI][CUDA][SymmMem] B200 Periodical Job Failure with test_triton_sum_reduce_int64, #164860 Driver update on H100 and A100 instances, #165173 Add CUDA 13.0 in CI, #165684 Python version in test CI job doesn't align with the python version in the docker build log",e4fe6758aa5bb7450bdbdee2e3edf256543e0955655f922efadf4d5ae521ac11
issue-collapse:bfcfd00f8f6018e3,module: c10d stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"53658, 54021, 56836, 57547, 59256, 63802, 67312, 68803, 71134, 73121, 73683, 73825, 80022, 80417",,,,"module: c10d, module: data parallel, module: ddp, module: nccl, oncall: distributed","Stale, better-engineering, module: c10d, module: data parallel, module: ddp, module: nccl, oncall: distributed, pt_distributed_rampup, triaged",,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#53658 torch.distributed with NCCL can hang if the first operation is a barrier, #54021 Support scatter/gather for NCCL backend, #56836 Remove single device constraint from ProcessGroupNCCL profiling, #57547 Add distributed testing for CUDA aware MPI, #59256 ProcessGroupGloo creation crashes when world_size > 150, #63802 [torch.distributed.launch|run] Hangs on SIGINT when using a TCPStore backed rdzv_backend",746b691616acff8392dc98668ceeed9b98c3b250427879d84be73f426398aab2
issue-collapse:e7b251b7a0af83eb,module: c10d stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"76288, 85621, 86225, 95572, 108900, 113543, 113546, 122931, 126008, 126032, 126381, 126544, 131755, 140301",,,,"module: c10d, module: elastic, module: nccl, oncall: distributed","feature, module: c10d, module: elastic, module: nccl, oncall: distributed, triaged",,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76288 Multi-GPU distributed training reports errors, #85621 Gloo DDP SocketTimeout error on Windows, #86225 [Feature] Dispatching PyTorch Distributed Collectives, #95572 Support datatype argument for torch.distributed.all_gather() (And the whole distributed module), #108900 [c10d] fix functional collective reduce op naming convention, #113543 [PTD] Logging Improvements Main Task",e2132fa95e76d0c39ccea176dc3737f4cdcfc707bbabf9c51205831b379b49c6
issue-collapse:fdcec0b957594ab3,module: c10d stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"75773, 80372, 82091, 85604, 97268, 137268, 146956, 147502, 147631, 147677, 149177, 151965, 152527, 154847",,,,"module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: pt2","better-engineering, module: c10d, module: ddp, module: nccl, oncall: distributed, oncall: pt2, triaged",,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#75773 Handle noncontiguous inputs in distributed backend layer, #80372 [BE] Update ProcessGroupWrapper tests to test other collective message, #82091 RuntimeError: [1] is setting up NCCL communicator and retreiving ncclUniqueId from [0] via c10d key-value store by key '0', but store->get('0') got error: Timeout waiting for key: default_pg/0/0 after 1800000 ms, #85604 Multiple GPUs get ""errno: 98 - Address already in use"", #97268 Desync debugger encounters traceMap error, #137268 DDP deadlock ProcessGroupNCCL's watchdog got stuck",341170a4e27ba7fbf2c31e1d2b0bf83cb08302234fcb10642853e4f3b351d822
issue-collapse:6f45b36b7ca3d51d,module: build stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"41119, 55865, 65313, 114175, 114388, 115020, 115257, 115672, 116012, 116478, 117099, 120000, 121215, 122027",,,,"module: build, module: cpp, module: cuda, module: ios, module: mkl, module: nccl, module: protobuf, module: rocm, module: sparse, module: third_party","enhancement, low priority, module: build, module: cpp, module: cuda, module: ios, module: mkl, module: nccl, module: protobuf, module: rocm, module: sparse, module: third_party, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#41119 TORCH_CUDA_ARCH_LIST deprecation warning, #55865 Don't define __assert_fail on systems using `musl-dev` headers, #65313 openSUSE Build: CMake Generate step failed. Build files cannot be regenerated correctly., #114175 Unable to build on CUDA 11.8 due to cutlass incompatibility, #114388 Respect user-specified USE_ROCM/USE_CUDA, #115020 [Libtorch/iOS] Build for iOS as dynamic library fails in linker phase.",b500f8466679531da40f9b09da52bb451ba6031868f38dd65251508dfeb780b6
issue-collapse:33642d39a1569927,module: build stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"58267, 72582, 102936, 103805, 112942, 119557, 119740, 122187, 122592, 123091, 124383, 125243, 125539, 128314",,,,"module: build, module: ci, module: cuda, module: docs, module: macos, module: mkl, module: mkldnn, module: power, module: third_party, module: windows","matrix multiplication, module: POWER, module: build, module: ci, module: cuda, module: docs, module: macos, module: mkl, module: mkldnn, module: third_party, module: windows, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#58267 Pytorch build issues on PowerPC, #72582 Avoid using thrust:: directly, use THRUST_NS_QUALIFIER:: instead, #102936 torch.cuda.is_available() returns False on GTX 1650 with cuda 11.7 and torch==2.0.0+cpu, #103805 Error when building with USE_TENSORRT=1, #112942 Can't build PyTorch 2.1 from source by GCC 13.2 on M1 MacOS, #119557 Python 3.12 CPU Build system cannot find MKL libraries",73db5ae336ccb1c349c2e497d9417130aeb640b3567b84f504ed282c32e7ffdf
issue-collapse:e6fa5449ff8920e2,module: build stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15167, 42791, 61244, 81856, 91566, 105728, 112089, 113948, 124270, 129310, 130863, 132013, 132599, 132600",,,,"module: build, module: cublas, module: cuda, module: cudnn, module: docker, module: mkl, module: mkldnn, module: openmp, module: static linking, module: third_party, module: windows","actionable, enhancement, module: build, module: cublas, module: cuda, module: cudnn, module: docker, module: mkl, module: mkldnn, module: openmp, module: static linking, module: third_party, module: windows, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15167 cudnn not found, #42791 Build fail on Ubuntu 18.04: caffe2/CMakeFiles/init_test.dir/build.make:106: recipe for target 'bin/init_test' failed, #61244 error: invalid conversion in vec256_bfloat16.h, #81856 Compatibility with newest MKL, #91566 Build Error: OpenMP library could not be found. Proceeding might lead to highly sub-optimal performance., #105728 Compile error PyTorch 2.0.1 / GCC 13.1.0",438d01c5dca000af3840fe1eb7c87e60b1fe666b7147c43ea680fb3b5bffc1cc
issue-collapse:d58cbf122f07ca34,module: build stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10684, 106971, 116416, 126394, 134399, 134439, 135604, 136201, 137261, 139170, 139874, 139958, 139962, 140088",,,,"module: build, module: magma, module: third_party, module: vulkan, module: windows, oncall: mobile","actionable, module: build, module: magma, module: third_party, module: vulkan, module: windows, oncall: mobile, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10684 make install error： [third_party/gloo/gloo/CMakeFiles/gloo.dir/all] Error 2, #106971 Found two conflicting CUDA installs, #116416 error: cast from 'PyObject *(*)(THPCppFunction *, void *)' (aka '_object *(*)(torch::autograd::THPCppFunction *, void *)') to 'getter' (aka '_object *(*)(_object *, void *)') converts to incompatible function type [-Werror,-Wcast-function-type-strict], #126394 Building libtorch for android fails, #134399 Build errors while building PyTorch with BLIS, #134439 configure ignores USE_QNNPACK=0 and still looks for third_party/QNNPACK",942b0b929f1b7468497a98e8f7526ab15453874fdf81da60b80f9f92155cbc9a
issue-collapse:bfe65d8496020451,module: build stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"26340, 51151, 75184, 77939, 103150, 125134, 126211, 127388, 134682, 143109, 144328, 144495, 146982, 148518",,,,"module: build, module: core aten, module: cpu, module: cuda, module: cudnn, module: mkldnn, module: openmp, module: power, module: wsl, oncall: distributed, oncall: quantization, oncall: releng","enhancement, feature, module: POWER, module: build, module: core aten, module: cpu, module: cuda, module: cudnn, module: mkldnn, module: openmp, module: wsl, oncall: distributed, oncall: quantization, oncall: releng, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#26340 CPU version of PyTorch on PyPI, #51151 -fno-omit-frame-pointer by default in our builds, #75184 PyTorch source code compile fail after ""Built target fbgemm_avx2"", #77939 Fails to compile with GCC 12.1.0, #103150 Build fails at linking torch_shm_manager on aarch64, #125134 binaries/dump_operator_names.cc missing iostream include",147ee723250ec5bbe7e647d60e6f765e3f0dc0d30f15595d4d5bbc81b2ceb976
issue-collapse:31f5266dd2be6a4d,module: build stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"36437, 68923, 76082, 134576, 139290, 148546, 149796, 149803, 149845, 150571, 150718, 150873, 151592, 153361",,,,"module: build, module: ci, module: cpp, module: cpp-extensions, module: cuda, module: docs, module: xpu","better-engineering, enhancement, module: build, module: ci, module: cpp, module: cpp-extensions, module: cuda, module: docs, module: xpu, topic: build, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#36437 Issue when linking C++ code with libtorch_cpu: cuda not detected, #68923 [cpp extension] making it possible to parallelize the building process, #76082 Eliminate uses of deprecated `FindCUDA.cmake`, #134576 Would it be possible to have a CMAKE option USE_SYSTEM_FMT, #139290 Using xpu module in the cuda version Pytorch, #148546 Pytorch nightly broken Flash Attention 3 compile with sycl commit - TypeError: _write_ninja_file() got an unexpected keyword argument 'sycl_cflags'",0edc9f8e0090dce9f3ad8d32ef0a969a7128fe5cb4ba810f7d9131d8a1bdfc24
issue-collapse:f31986b7a8180dbd,module: build stale/low urgency  batch 37,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"21318, 21818, 22495, 22868, 23512, 23676, 23768, 23896, 24167, 24417, 24419, 24456, 24500, 24798",,,,"module: build, module: ci, module: docs, module: internals, module: macos, module: third_party, oncall: distributed","better-engineering, enhancement, low priority, module: build, module: ci, module: docs, module: internals, module: macos, module: third_party, oncall: distributed, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#21318 Undefined symbols for architecture x86_64: ""testing::internal::UntypedFunctionMockerBase::UntypedInvokeWith(void const*)"" on Mac OS X, #21818 C++ ABI - Coupling different libraries issue, #22495 `binary_linux_libtorch_2.7m_cu100_devtoolset3_build` times out after running for 5 hours, #22868 Make it easier to bisect on PyTorch, #23512 Build reconfiguration should consistently honor env variables, #23676 Bogus ""Your compiler (clang++) is not compatible"" message",83efafcce62963fa6bbcaf747477780adc5fa1050a680ee58664e71e03423548
issue-collapse:8f203cc23e90bcb5,module: build stale/low urgency  batch 36,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14277, 16710, 21738, 22682, 24831, 25460, 26058, 26258, 26409, 26796, 26863, 26964, 26967, 27015",,,,"module: build, module: cpp, module: cuda, module: flaky-tests, module: lint, module: protobuf, module: static linking, module: third_party, module: vision","better-engineering, has workaround, module: build, module: cpp, module: cuda, module: flaky-tests, module: lint, module: protobuf, module: static linking, module: third_party, module: vision, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14277 Provide Protobuf library if libtorch was built with included version, #16710 culibos linker errors on binary_linux_conda_3.6_cu90_build, #21738 symbol lookup error: libmkl_intel_lp64.so: undefined symbol: mkl_blas_dsyrk (binaries built with static linking -DBUILD_SHARED_LIBS=OFF fail due to dynamic linker problem), #22682 CMAKE_PARSE_IMPLICIT_LINK_INFO Function invoked with incorrect arguments, #24831 subprocess.CalledProcessError: Compile source in NVIDIA TX2, #25460 Build link not right",572938bc8b9dc49b9ac2b3e05c9ed3c6f4010b09bd58941cd589f0d8646c36e7
issue-collapse:7ac35c22830f5240,module: build stale/low urgency  batch 35,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12535, 15291, 16405, 17666, 27409, 28515, 28655, 28745, 28756, 28780, 29235, 30463, 30744, 31300",,,,"module: build, module: cuda, module: macos, module: mkldnn, module: openmp","module: build, module: cuda, module: macos, module: mkldnn, module: openmp, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12535 [CMake] Linking against Intel OpenMP, #15291 linked error of Pytorch 1.0 release, #16405 error on cmake_version from tools/build_pytorch_libs.py, #17666 Building pytorch on ARM failed, #27409 Installer not setting rpath for MAGMA (OS X w/ GPU), #28515 Get rid of libc10.so",a6cdfc414c065353a4139927b08eaf0528d0446bbfcaaf082a84684e7c00167d
issue-collapse:d35de6d818673bd4,module: build stale/low urgency  batch 34,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5900, 31487, 31916, 32214, 32403, 32416, 32687, 33049, 33129, 34409, 34621, 35035, 35097, 35234",,,,"module: build, module: mkldnn","better-engineering, enhancement, low priority, module: build, module: mkldnn, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5900 Build Fails on Gentoo with CUDA 9.1, GCC 6.4, Python 3.5, #31487 Disable PSIMD? Why Pytorch is trying to download PSimd when PSIMD_SOURCE_DIR is defined?, #31916 I'm not able to build pytorch with tensorrt (current master), #32214 Pytorch compilation error, #32403 How to accelerate the compiling of pytorch, #32416 Is there a way to use SYSTEM_INSTALLED 3rdparty libraries?",1c4aafc5f9dd15ad35deba23726aec914be2f7c3f1ef9eb45495f51657804238
issue-collapse:21b61373535474f6,module: build stale/low urgency  batch 33,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"31264, 35414, 35832, 35942, 36356, 36407, 36713, 36792, 36804, 37282, 37510, 37629, 38212, 38354",,,,"module: build, module: ci, module: cuda, module: docs, module: macos, module: mkldnn, oncall: jit","module: build, module: ci, module: cuda, module: docs, module: macos, module: mkldnn, oncall: jit, small, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#31264 Failed to config caffe2_rocksdb in cmake, #35414 MacOS Error: subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--target', 'install', '--config', 'Release', '--', '-j', '12']' returned non-zero exit status 1., #35832 Linking error after marking an op `manual_kernel_registration: True`, #35942 Build PyTorch-1.4.0 from source failed, #36356 Memory leak issue still exists in CI, #36407 Unable to build wheel from RC3",274e8a346aa0faf75aa8d52006e62cdf4ae1e84bd3d06727dbfdf995f519120d
issue-collapse:36d20b7911aa059c,module: build stale/low urgency  batch 32,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14699, 21794, 33668, 36378, 36570, 36572, 38204, 38239, 38719, 38854, 38855, 39064, 39101, 39305",,,,"module: build, module: cpp, module: cuda, module: docs, module: mkldnn, module: nccl, oncall: distributed","enhancement, module: build, module: cpp, module: cuda, module: docs, module: mkldnn, module: nccl, oncall: distributed, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14699 [Feature Request] linux distribution friendly build system, #21794 Linker errors when building project with OpenCV, #33668 /usr/bin/x86_64-linux-gnu-ld: warning: libcusparse.so.10.0, needed by /pytorch_master/build/lib/libtorch_cuda.so, not found (try using -rpath or -rpath-link), #36378 CMake targets wrongly forward unknown options to NVCC (v1.5+), #36570 Libtorch build error when setting both `USE_GLOO` and `USE_SYSTEM_NCCL` to `ON`, #36572 [1.4.1] cmake3 not found",3bd2e4747807b0aa117c7d4be10bc134f4a77af8d8c6b83e11b2adcbbd7bbe66
issue-collapse:1a3af2dd00653540,module: build stale/low urgency  batch 31,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4758, 31883, 36125, 39990, 40563, 40854, 41135, 41178, 41673, 41915, 41920, 41921, 41938, 41999",,,,"module: build, module: cpp, module: cuda, module: mkldnn, module: power, module: third_party, oncall: distributed","feature, module: POWER, module: build, module: cpp, module: cuda, module: mkldnn, module: third_party, oncall: distributed, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4758 Clang color diagnostics don't work with ninja, #31883 The dependency target ""nccl_external"" of target ""gloo_cuda"" does not exist., #36125 build Pytorch-1.5.0-rc1 from source fail, #39990 libtorch 1.5 macos crash when loading on some mac, #40563 Compiling PyTorch with 11.0, V11.0.167, #40854 Build pytorch with cuda11",7f41e7d19643c094bc8a9ce2a3ab6df9c5f5ef96d98373920ec204d5315330b0
issue-collapse:066e8119b34b0c55,module: build stale/low urgency  batch 30,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14200, 15294, 23730, 41816, 42362, 42526, 42807, 42916, 43030, 43113, 43604, 44362, 44948, 45129",,,,"module: build, module: build warnings, module: cuda, module: dependency bug, module: pybind","better-engineering, module: build, module: build warnings, module: cuda, module: dependency bug, module: pybind, small, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14200 Failed to run 'bash ../tools/build_pytorch_libs.sh --use-nnpack --use-mkldnn --use-qnnpack caffe2', #15294 manylinux2014 compatible wheels, #23730 Failed to build pytorch ..., #41816 install path can set prefix install dir, #42362 libpytorch_jni is not provided in conda builds, #42526 Python3 segfaults or doesn't load torch after installation",aab48583b1ee12dc8ac72e0f637ca935557580b6d4037456eeeff11483389345
issue-collapse:dae4497d02dcb92b,module: build stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33890, 51511, 88802, 135126, 139108, 144767, 146221, 150328, 153606, 154096, 155350, 156725, 157465, 159156",,,,"module: build, module: cpp, module: cuda, module: dependency bug, module: docs, module: nccl, module: onnx, module: pybind, module: static linking, module: tbb, module: third_party, module: windows, oncall: pt2","has workaround, module: build, module: cpp, module: cuda, module: dependency bug, module: docs, module: nccl, module: onnx, module: pybind, module: static linking, module: tbb, module: third_party, module: windows, oncall: pt2, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33890 Libtorch segfault when used with libqpOASES, #51511 Python-3.9: Importing numba and torch triggers internal error, #88802 ImportError: libcupti.so.11.2: cannot open shared object file: No such file or directory, #135126 ""RuntimeError: CUDA error: operation not supported"" fixed by downgrading toolkit version, #139108 LibTorch build error on Windows for CUDA version (debug/release), #144767 Unable to build with ATEN_THREADING=TBB option",b12e27d2b8bd0bc0073999ee926eea1476ab1ba4e8eea61ea9cee533fa0dc75e
issue-collapse:115eb4bb2386291a,module: build stale/low urgency  batch 29,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5924, 6071, 6328, 31598, 34451, 41916, 41918, 45444, 45573, 46035, 46139, 46241, 46367, 46659",,,,"module: build, module: ci, module: macos, module: mkl, module: power, module: protobuf, module: xnnpack, oncall: mobile","better-engineering, enhancement, has workaround, module: POWER, module: build, module: ci, module: macos, module: mkl, module: protobuf, module: xnnpack, oncall: mobile, todo, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5924 Install doesn't work with spaces in directory, #6071 Repeated 'python setup.py install' with clang leads to -lcpuinfo not found, #6328 [feature request] Include libomp support (macOS), #31598 ATen not compiled with MKL support, #34451 Mac build from source error, #41916 ASAN build broken when using USE_ASAN=1",5e48c4d44898407e87a632f4d81ab2a7999ddfb4eb7565fa9f29100f021e418c
issue-collapse:a8b20f831c3ef40e,module: build stale/low urgency  batch 28,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7214, 7352, 7353, 25004, 25714, 38765, 40968, 46766, 46807, 47557, 47681, 48033, 48064, 48457",,,,"module: build, module: build warnings, module: cpp, module: cpp-extensions, module: cuda, module: openmp, module: vectorization","better-engineering, enhancement, module: build, module: build warnings, module: cpp, module: cpp-extensions, module: cuda, module: openmp, module: vectorization, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7214 Do not put system paths in RPATH, #7352 Check for F2C convention (for blas) at runtime, #7353 Windows MAGMA binary requires explicit linking against MKL LAPACK, or it will silently give incorrect results, #25004 [C++] Call find_package(Torch REQUIRED) more than one time in downstream project causes CMake configuration error, #25714 Backtrace prints many <unknown function>, #38765 LibTorch 1.5.0 not supporting GLIBC < 2.23",7ad0fb17152281f686a8b5a88d6355d53fb61dac285c3cbd83e668b81f35cfbd
issue-collapse:45113d8479247b8d,module: build stale/low urgency  batch 27,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8097, 8107, 8561, 8871, 17920, 32113, 45493, 46644, 48291, 49677, 49784, 49865, 50136, 50830",,,,"module: build, module: cpp, module: cuda, module: linear algebra","feature, module: build, module: cpp, module: cuda, module: linear algebra, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8097 manager.cpp:64: undefined reference to `shm_open (when building with GCC 5.x (sic)), #8107 Export CC is ignored when I build pytorch, #8561 cleanup BLAS detection, #8871 Use target_compile_options to set warning flags, #17920 Build compact libtorch from source with cmake, #32113 Inconsistent linking flags result in error when building lib/libc10_cuda.so",111b1ff021522474cf08a23afdf9b817640b1ac4016b17b0c92d5bb83b1085be
issue-collapse:2f6cfc10056cc859,module: build stale/low urgency  batch 26,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9893, 10691, 38193, 39634, 50206, 50888, 50944, 51015, 51074, 51084, 51231, 51302, 51868, 52096",,,,"module: build, module: cuda, module: macos, module: static linking","module: build, module: cuda, module: macos, module: static linking, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9893 interaction with FindCUDA causes spurious re-cmakes, #10691 Build system doesn't prevent ATen/core from including non-core files, #38193 Can't Build Pytorch0.4.1 From Source, #39634 Installation from source fails on macOS (No CUDA), #50206 Build Fail with vulkan on ARM64 and wayland drivers, #50888 Does pytorch support cuda 11.1",310d2c6bf2806189ef77b15afdc15374ce9d90287179a6a4122342c4da065262
issue-collapse:b525cc896d98df75,module: build stale/low urgency  batch 25,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4906, 11113, 11854, 12609, 52168, 52208, 52318, 52598, 52605, 53710, 54279, 54829, 55329, 55461",,,,"module: build, module: cpp, module: cpp-extensions, module: cuda","has workaround, low priority, module: build, module: cpp, module: cpp-extensions, module: cuda, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4906 Rebuild from no-CUDA to CUDA leads to: error: #error ""Expected GLOO_USE_CUDA to be defined"", #11113 use_system_nccl flag does not work?, #11854 error: ‘array_size’ is not a class template, #12609 Request for stripped down / inference only pytorch wheels, #52168 CMake errors building project that uses PyTorch, #52208 C++ standard library confusion",540bb5d9765f30244859ec6088307369e6149a2db2060bd516edae91805f2698
issue-collapse:340941807e62274b,module: build stale/low urgency  batch 24,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11213, 12828, 12830, 23482, 44172, 44814, 56465, 56710, 56777, 57108, 57123, 57370, 57440, 58672",,,,"module: build, module: build warnings, module: ci, module: tensorpipe, module: undefined reference","caffe2, feature, module: build, module: build warnings, module: ci, module: tensorpipe, module: undefined reference, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11213 Ubuntu 16.04 setup.py error - undefined reference to elfLink_Get_FatBinary_From_Object' /usr/lib/x86_64-linux-gnu/libcuda.so: undefined reference to elf32_section_header', #12828 BUILD_BINARY is a lie, #12830 Test that (cd build && ninja) immediately after build is no-op in CI, #23482 Build error due to unintended include path /usr/include, #44172 build from scratch failed because of the expansion of macro, #44814 LNK2019 error after static compilation of Libtorch",2e6eea6eefbcdcb321d90e2990ab32c066fce6bffd507901877a6a8b4e1bc5a8
issue-collapse:2c5aab6d42020e8e,module: build stale/low urgency  batch 23,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14477, 15303, 37113, 42689, 45448, 54905, 56161, 58596, 58944, 59457, 59531, 59542, 59935, 60448",,,,"module: build, module: codegen, module: cuda, module: cudnn, module: determinism, module: power","better-engineering, enhancement, feature, module: POWER, module: build, module: codegen, module: cuda, module: cudnn, module: determinism, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14477 Add a debug mode which is -O0 for framework code, but -O for kernels, #15303 CMake Error at CMakeLists.txt:10 (find_package) in C++, #37113 pip install torch==1.4.0 is broken when using CUDA 10.1, #42689 How can I specify NumPy while building PyTorch?, #45448 CuDNN version not found, #54905 Allow CUDA build without requiring a physical GPU device.",77cca5950adb5f504dfb7f087c560670133e0165594a515ccb93d0e14cff4119
issue-collapse:c9882eecf2a6efb2,module: build stale/low urgency  batch 22,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15812, 16090, 20206, 31943, 47281, 49810, 60674, 60848, 60950, 61510, 61695, 61718, 61827, 61930",,,,"module: build, module: macos, module: protobuf, module: xnnpack","better-engineering, caffe2, has workaround, module: build, module: macos, module: protobuf, module: xnnpack, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15812 Possible regression in incremental build, #16090 build failure with NNAPI enabled (Caffe2 path but under pytorch umbrella), #20206 Failed to build with system protobuf, #31943 error when building pytorch 1.1.0 from source, #47281 Error: invalid use of register, #49810 Undefined reference to `cv::imread`",88c310647e0a473e33a5d16c6ef9546aee892af5f8076e489f562adf3e14bad8
issue-collapse:990ab26d1c2a7aa3,module: build stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17644, 32083, 45073, 48483, 56038, 61468, 61730, 62656, 63574, 64308, 64527, 65062, 65446, 65475",,,,"module: build, module: ci, module: power, module: protobuf, module: pybind, module: tensorpipe, module: third_party, oncall: distributed, oncall: releng","module: POWER, module: build, module: ci, module: protobuf, module: pybind, module: tensorpipe, module: third_party, oncall: distributed, oncall: releng, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17644 Not depend on third_party submodules, but self-built libraries?, #32083 build pytorch from source fauled: undefined reference to `cusparseGetErrorString(cusparseStatus_t)', #45073 Compilation errors on power-pc, #48483 Vulkan Api Backend Build Error with GCC, #56038 [Meta] PyTorch features build/test matrix, #61468 Version is set to 1.8.0 in the torch tag v1.8.1",09692942e5871e6abdb8254293204a7c190fd55bb39484f3608a84d7e794b98d
issue-collapse:640fa9ca4542fab0,module: build stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1505, 9433, 16499, 31994, 42353, 56110, 65572, 66196, 67284, 67642, 67758, 67968, 68163, 70521",,,,"module: build, module: cuda, module: dependency bug, module: mkl, module: vulkan","module: build, module: cuda, module: dependency bug, module: mkl, module: vulkan, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1505 Cannot find Intel MKL, #9433 Build error: cannot find -lonnxifi_loader, #16499 TorchConfig.cmake always sets _GLIBCXX_USE_CXX11_ABI, #31994 Trying to disable cuda to run torch on OSX run it on CPU, #42353 Error while trying to install pytorch on pycharm, #56110 caffe2/utils/signal_handler.cc is failing to symbolize libtorch_cpu.so",88d9fc26a5fc73143fe5aaf928cf37c634678315ee63458f10773294c45b6068
issue-collapse:4fe4c5084d48373e,module: build stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"58054, 129358, 141235, 144931, 147001, 153933, 159838, 160167, 160225, 162280, 162283, 162352, 163950, 164786",,,,"module: build, module: cpp-extensions, module: cuda, module: custom-operators, module: debug-build, module: functorch, module: nativert, module: power, module: pt2-dispatcher, module: third_party, module: windows, oncall: pt2","actionable, enhancement, module: POWER, module: build, module: cpp-extensions, module: cuda, module: custom-operators, module: debug-build, module: functorch, module: nativert, module: pt2-dispatcher, module: third_party, module: windows, oncall: pt2, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#58054 PyTorch master failed to build with glog 0.5.0., #129358 Pytorch build fail with GCC 14.1.0 due to third_party/fbgemm/src/UtilsAvx512.cc:970:35: error: ‘r’ may be used uninitialized [-Werror=maybe-uninitialized], #141235 Enable CI and Build Support for PyTorch on PPC64LE Architecture, #144931 Error loading ""torch\lib\aoti_custom_ops.dll"" or one of its dependencies, when importing Torch, when building from Source on Windows 11 with cuDNN., #147001 Preload CUDA fails if CUDA libs in different PYTHONPATH, #153933 PyTorch build fail with GCC 15.1.0 due to third_party/gloo/gloo/types.h:66:11: error: ‘uint8_t’ does not name a type",d965dfe1687727c28ed22cf2293a35c1c0309d42a14277f4f18f4eeb6d181879
issue-collapse:544c79deeb32d672,module: build stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17525, 49795, 69086, 69286, 69360, 70138, 71228, 71396, 71471, 71660, 71682, 73062, 73355, 73394",,,,"module: build, module: ci, module: cublas, module: cuda, module: infra, module: ios, module: macos, module: undefined reference, oncall: releng","Stale, enhancement, module: build, module: ci, module: cublas, module: cuda, module: infra, module: ios, module: macos, module: undefined reference, oncall: releng, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17525 cmake fails with -DBUILD_PYTHON=OFF -DUSE_NNPACK=OFF, #49795 ninja: build stopped: subcommand failed., #69086 Confusing value of `TORCH_CUDA_ARCH_LIST`, #69286 subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--target', 'install', '--config', 'Release', '--', '-j', '32']' returned non-zero exit status 1., #69360 1.10.0 failed to build due to missing TensorBody.h, #70138 build_ios.sh prevents iOS.cmake from configuring ios deployment target correctly",60514a280529da8448a5e3d6355c150c899419cad6d352b90b1679f4acf066f2
issue-collapse:69f6132427691c33,module: build stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24810, 44726, 46157, 66271, 71518, 74398, 74421, 74896, 75180, 75565, 75895, 75911, 77411, 77546",,,,"module: build, module: cuda, module: linear algebra, module: lint, module: power, module: sparse, module: vectorization","module: POWER, module: build, module: cuda, module: linear algebra, module: lint, module: sparse, module: vectorization, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24810 Doesn't install the python module ""torch"", #44726 Pytorch Installation from source fails (GCC-8.4/CUDA-10.2/RHEL7) and (GCC-7.5/CUDA-10.2/RHEL7), #46157 Ability to disable cusolver, #66271 1.10.0-rc1 fails to compile with gcc 6.4.0, #71518 Compilation instructions are not exhaustive: <<parameter packs not expanded with ‘...’>> on Fedora 35/CUDA 11.6, #74398 _run_ninja_build failing with clang",6ae5617e61e8f166a38f23612e687d6882eae2d12235733b0195e46a0d619a94
issue-collapse:bc915d1390aa3331,module: build stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33385, 46803, 59592, 64692, 76206, 77614, 77738, 78530, 78882, 79021, 79202, 79387, 81543, 81544",,,,"module: build, module: codegen, module: cuda, module: dispatch, module: macos, module: nnpack, module: selective build, module: tbb, module: third_party, oncall: distributed","module: build, module: codegen, module: cuda, module: dispatch, module: macos, module: nnpack, module: selective build, module: tbb, module: third_party, oncall: distributed, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33385 Illegal instruction: 4 - OSX 10.13.6 install from source, #46803 About 2 minor bug fixes on CUDA macOSX 10.13.6, #59592 Build failed with TBB, #64692 Remove the dependency with CUDA_LIBRARIES in TorchConfig.cmake, #76206 Update NCCL to 2.12, #77614 PyTorch fails to build on gcc 12 due to gloo",0d7f880d6ffd445afe7fa21e9ef78c3da463141ce72a7d7b9f71854c9d8475a0
issue-collapse:54602dc2b4b6ce9e,module: build stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14367, 23233, 39917, 80742, 82276, 82710, 82871, 83157, 83714, 83977, 83979, 84050, 84805, 85590",,,,"module: build, module: cuda, module: docs, module: internals, module: macos, module: sparse, module: static linking","enhancement, has workaround, module: build, module: cuda, module: docs, module: internals, module: macos, module: sparse, module: static linking, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14367 Error: DeviceGuardImpl for cpu is not available (static linking PyTorch), #23233 Unofficial ARMv6, ARMv7, ARMv8 builds, #39917 build QT program use libtorch-cxx11-abi-shared-with-deps-1.5.0+cu101 ok with CPU but error with cuda GPU, #80742 Add TorchDynamo as a submodule to Pytorch?, #82276 DEBUG=1 env var doesn't actually set DEBUG preprocessor macro, #82710 build fail when using lto with gcc",622d6686c01240dca5da5541b9ac782500c12542ff6d155bdcaa9f77aed9b892
issue-collapse:94ce7d8bfeb0b64c,module: build stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32694, 47568, 85618, 86192, 86544, 86660, 87080, 87352, 87389, 87589, 87642, 88004, 88389, 88616",,,,"module: build, module: cpp, module: cpp-extensions, module: cuda, module: cudnn, module: docker, module: m1, module: third_party","actionable, module: build, module: cpp, module: cpp-extensions, module: cuda, module: cudnn, module: docker, module: m1, module: third_party, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32694 ninja: build stopped: subcommand failed., #47568 Failed to get generated_cpp list, #85618 Build from source failed with error of different gpu architecture (compiler shows sm_30-related error but I use sm_86 GPU), #86192 fmt/src/os.cc: error: unknown type name 'error_code'; did you mean 'std::error_code'?, #86544 Compile failed at allreduce without gloo, #86660 libtorch make failed",78ad17c3045541b0ea7c5c7963e69e6eaf53338a72f331a755856eb19e3a56a3
issue-collapse:175aa34ce30233b7,module: build stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38536, 50718, 55437, 72393, 73563, 76105, 87957, 88265, 89293, 90256, 90395, 90895, 91406, 91417",,,,"module: build, module: cuda, module: docker, module: docs, module: static linking, module: tbb","module: build, module: cuda, module: docker, module: docs, module: static linking, module: tbb, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38536 build libtorch problem: Configuring incomplete, errors occurred! _mm256_abs_epi16, #50718 dependency_links is deprecated, #55437 Error building docker image: No module named 'typing_extensions', #72393 Some system-installed headers are mistakenly used., #73563 Pytorch Installation from source fails, #76105 Failed to build on Ubuntu 18.04 due to bad MPI linker flags",c0e8e62eb69352cd6674d0a870b95c00bd8899558fcd0ac567378aaff25b32b1
issue-collapse:19c61e37decc5e68,module: build stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"57423, 62168, 62630, 69063, 78248, 83509, 91697, 93240, 93944, 94739, 94801, 95135, 97397, 97421",,,,"module: build, module: build warnings, module: cuda, module: docs, module: macos, module: tensorpipe, module: vectorization","Stale, enhancement, module: build, module: build warnings, module: cuda, module: docs, module: macos, module: tensorpipe, module: vectorization, shadow review, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#57423 internal compiler error compiling PyTorch master on Fedora 33, #62168 Error installing from source, #62630 Instructions for CMAKE_PREFIX_PATH seem to be broken, #69063 make: Makefile: No such file or directory, #78248 USE_NATIVE_ARCH flag causes nvcc build failure due to ""'arch=native': expected a number"", #83509 Build and Run QNNPACK on X86",d76fb6b8392c1ccc4566ea54dbe2375b4ca278ab9d4956f8c00897da78e43ef7
issue-collapse:781509c578f74665,module: build stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"18993, 36041, 73554, 93206, 94912, 95369, 98587, 98917, 100123, 100957, 102732, 102753, 102761, 103131",,,,"module: build, module: cpp-extensions, module: cuda, module: internals, module: macos, module: mkldnn, module: power, module: rocm, module: third_party, module: windows","actionable, feature, module: POWER, module: build, module: cpp-extensions, module: cuda, module: internals, module: macos, module: mkldnn, module: rocm, module: third_party, module: windows, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#18993 ninja: build stopped: subcommand failed., #36041 [DISCUSSION] Better user experience for debugging on Windows, #73554 Build failure using GCC 11.2.0, #93206 Build from Source Issues on MacOS Ventura 13.2, #94912 PR #88607 breaks build for POWER9 CPU, #95369 build failed when strictly following the guidelines",b3498ea1f0058bdc2d873d752511b302f6b3747da429fa5d697494955a347333
issue-collapse:bcdb82dc322da94b,module: build stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20686, 22541, 34611, 38090, 63188, 66573, 87449, 89490, 94804, 99035, 100054, 103212, 105060, 105904",,,,"module: build, module: cpp, module: cuda, module: cudnn, module: docs, module: mkldnn, module: rocm, module: windows","actionable, module: build, module: cpp, module: cuda, module: cudnn, module: docs, module: mkldnn, module: rocm, module: windows, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20686 In-source build causes repeating filename annotations (Windows doesn't support out-of-source build), #22541 make[2]: *** No rule to make target 'libtorch/lib/libc10.so', #34611 Eigen version for PyTorch ?, #38090 ImportError: libtorch_cpu.so: cannot open shared object file: No such file or directory, #63188 ROCm build documentation lacks needed dependencies, #66573 tools/amd_build/build_amd.py should fail if any file fails to write",18efb39f67e4d7f3ad138efe566f2611c8f29f990f3e2ae8747f30bb331ddcc3
issue-collapse:1f6f4a19f085925b,module: build stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"70243, 81065, 98459, 102966, 107389, 108851, 110238, 110795, 111410, 111566, 111654, 111907, 112330, 112857",,,,"module: build, module: ci, module: cpp, module: cuda, module: nccl, module: nvfuser, module: static linking, module: tbb, module: vision, module: vulkan, oncall: mobile","actionable, ciflow/periodic, enhancement, has workaround, module: build, module: ci, module: cpp, module: cuda, module: nccl, module: nvfuser, module: static linking, module: tbb, module: vision, module: vulkan, oncall: mobile, topic: build, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#70243 Unable to compile PyTorch when libcudart_static.so is not available, #81065 [Releng] Improve the tutorials release process, #98459 Fail to pass test HAVE_XXX_REGEX while building pytorch, #102966 how to workaround the error ""don't have an op for vulkan_prepack::create_linear_context"" ?, #107389 caffe does not respect CUDNN_LIB_DIR when building from source (cmake), #108851 about nccl not work",288469dfedd2b7bdad62be0e2697778d09b776f0d825c28d54e2882455a55c2c
issue-collapse:fc00c155a393e2cc,module: build stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"65161, 124018, 139065, 153863, 156434, 159232, 160248, 165100, 166064, 166167, 167970, 169072, 169675, 169816",,,,"module: build, module: cpp-extensions, module: cuda, module: onnx, module: sanitizers, module: third_party, module: vectorization, module: windows, oncall: quantization","actionable, better-engineering, enhancement, has workaround, module: build, module: cpp-extensions, module: cuda, module: onnx, module: sanitizers, module: third_party, module: vectorization, module: windows, oncall: quantization, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#65161 error: ‘_mm512_set_epi8’ was not declared in this scope when build from source on AMD 3600 CPU, #124018 Fix OOMs during parallel builds, #139065 Cannot compile with latest LLVM-19, #153863 cmake: add USE_SYSTEM_{KLEIDI,CUDNN_FRONTEND,CUTLASS,FMT} options to USE_SYSTEM_LIBS, #156434 Suggest to use the torch cmake target instead of ${TORCH_LIBRARIES} in the c++ docs, #159232 Support installing Python bindings in CMake",fd78903a6304ffb93313585ae43ebd305fb93d298cc1435c91caeca73c3a3529
issue-collapse:d05b4a4132c3075c,module: bootcamp stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15822, 25641, 31467, 32497, 37010, 37782, 38233, 38437, 38548, 40456, 40790, 43253, 44341, 44898",,,,"module: bootcamp, module: error checking, module: internals, module: memory format, module: pooling, module: rpc, module: tensorpipe, oncall: distributed, oncall: jit","better-engineering, enhancement, feature, module: bootcamp, module: error checking, module: internals, module: memory format, module: pooling, module: rpc, module: tensorpipe, oncall: distributed, oncall: jit, small, triaged",,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15822 AvgPool2d doesn't test if kernel is smaller than input size, #25641 [distributed] all_gather on a List of Tensors directly, #31467 Distributed hangs on process termination with world_size=1, #32497 Don't unnecesarily send cleanup dist autograd context RPCs to other nodes, #37010 Make it an error to def() an operator multiple times, #37782 Guard Gloo and TensorPipe related code in RPC with #ifdef",01dfe206c44e6dfcd6e4043ad131d21908f8a3ad26ca9ddda7810f06157df4e3
issue-collapse:f8e468fd662785b0,module: bootcamp stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40550, 46321, 48605, 49178, 49539, 52583, 52940, 54941, 54942, 54943, 55005, 56482, 60431, 62655",,,,"module: bootcamp, module: dispatch, module: internals, module: nn, module: performance, module: rpc, module: tests, oncall: distributed, oncall: jit","NNC, better-engineering, enhancement, module: bootcamp, module: dispatch, module: internals, module: nn, module: performance, module: rpc, module: tests, oncall: distributed, oncall: jit, pt_distributed_rampup, triaged",,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40550 RemoteModule enhancements, #46321 Add distributed examples into PyTorch CI tests, #48605 [NNC] Bugs Exposed in Binary Op Testing, #49178 [NNC] Some ops have type promotion logic which adds extra casts & does compute in different dtype than eager, #49539 Improve AssertionError for torch.nn.functional.pad 'replicate', #52583 [TensorExpr] Add IR Verifier",683f0a6bcbdc85e97c715d450d1122e64a9cca208822275491538f24f8907a9c
issue-collapse:87fe166b6ee3e424,module: bootcamp stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10375, 45760, 46076, 55191, 58015, 58139, 61122, 61835, 62141, 62790, 64744, 64745, 72257, 73517",,,,"module: bootcamp, module: c10d, module: codegen, module: complex, module: cuda, module: ddp, module: docs, module: multiprocessing, module: nans and infs, oncall: distributed","better-engineering, enhancement, module: NaNs and Infs, module: bootcamp, module: c10d, module: codegen, module: complex, module: cuda, module: ddp, module: docs, module: multiprocessing, oncall: distributed, pt_distributed_rampup, triaged",,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10375 Sending CUDA tensor to process, and then back, does not work, #45760 Complex Number support for distributed, #46076 NCCL watchdog thread should log warnings about long-running GPU operations instead of silently hanging, #55191 `torch.tensor(..., device='meta')` doesn't work, #58015 Support a dist.group_like API, #58139 [docs] Strange torch.unique function signature",d1d8f878043f9a2e0830e4da93eeae24335b5f1bd28ea5b7fcd97af747278223
issue-collapse:b5ca37b64bd9cf7a,module: bootcamp stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8374, 11936, 15821, 18229, 48845, 54899, 55198, 56191, 62554, 62917, 63291, 75926, 87514, 140596",,,,"module: bootcamp, module: cuda, module: fx, module: internals, module: linear algebra, module: memory format, module: multiprocessing, module: nn, module: pooling, module: python frontend, module: typing, oncall: export, oncall: jit, oncall: pt2","actionable, empathy-day, enhancement, module: bootcamp, module: cuda, module: fx, module: internals, module: linear algebra, module: memory format, module: multiprocessing, module: nn, module: pooling, module: python frontend, module: typing, oncall: export, oncall: jit, oncall: pt2, small, todo, triaged",,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8374 [Feature request] Add torch.multiprocessing.Pipe, #11936 torch.bmm doesn't support CUDA uint8 (byte) tensor, #15821 No test coverage for kwargs of AvgPool2d and AvgPool3d, #18229 Add support for tuple type deduction in C++ custom operators, #48845 memory_format kwarg doesn't work on most factory functions, #54899 [fx] forward references in annotations do not produce correct source code",2dccc18345590dda0bfecf97fef0e33ea8cd21b2f3c3a382f90ab9bdf2b5d9b6
issue-collapse:8b0382522f8637a4,module: boolean tensor stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"32210, 41323, 44969, 64041, 67684, 77962, 106579, 106662, 107041, 111925, 117246, 117249, 136578, 139609",,,,"module: boolean tensor, module: correctness (silent), module: edge cases, module: loss, module: nn, module: numpy, module: python frontend, module: safe resize, module: sorting and selection, module: sparse, module: type promotion","enhancement, module: boolean tensor, module: correctness (silent), module: edge cases, module: loss, module: nn, module: numpy, module: python frontend, module: safe resize, module: sorting and selection, module: sparse, module: type promotion, triaged",,Handle this module: boolean tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#32210 dist.send/recv ""IndexError: map::at"" error when bool tensors are used with mpi backend, #41323 [discussion] Comparison operator chaining, #44969 Tensordot does not support Bool, #64041 torch.unique acting up for a binary tensor, #67684 [feature request] torch.clamp on BoolTensors, #77962 Operating on boolean torch tensor and numpy array casts to `unit8`",bf0a61b35d2903e11c8eb448208a1b2525c8cc86d49bb0feee0bb5b295da6040
issue-collapse:4c036cebea88bb74,module: binaries stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"47109, 48485, 49405, 49445, 54332, 55120, 55611, 56926, 59263, 60288, 61516, 63555, 63849, 66359",,,,"module: binaries, module: build, module: cpu, module: cuda, module: docs, module: lts, module: macos, module: memory usage, module: xla, oncall: releng","module: binaries, module: build, module: cpu, module: cuda, module: docs, module: lts, module: macos, module: memory usage, module: xla, oncall: releng, tracker, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#47109 Pytorch 1.7/Cuda 11.1 binaries, #48485 Add wheels for all cuda versions on pypi, #49405 PyTorch 1.7.1 on (macOS/python 3.9/conda) links libtorch_global_deps.dylib with libomp.dylib instead of libiomp5.dylib, #49445 [doc] how to prevent pytorch-nightly from being replaced by a released version on pip install, #54332 Unit testing failures when porting PyTorch wheel, #55120 Why isn't pip finding the correct pytorch dependency?",9cb5bb78772513057fbc643fe3502238c702a135488fbe01b7f1e52e14de89ed
issue-collapse:0872663b69fa4b97,module: binaries stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24205, 37332, 47916, 49406, 56901, 69072, 70199, 71364, 71473, 73034, 77052, 79359, 79606, 81172",,,,"module: binaries, module: build, module: cuda, module: docker, module: ios, module: mkl, module: static linking","Stale, enhancement, low priority, module: binaries, module: build, module: cuda, module: docker, module: ios, module: mkl, module: static linking, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24205 deprecate cuda arch 3.5/3.7 in nightlies, #37332 Concerning default confiugration for distribution packages, #47916 Hash mismatch for METADATA file, #49406 Bitcode enable for iOS, #56901 Static prebuilt libraries contain shared libraries, #69072 Libtorch dll versioning",40e1b4242cc498e06c432361681342bf7f12605ad13dadbc2831cf64fb053f7f
issue-collapse:5cbd28f49240af0a,module: binaries stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8276, 38624, 55454, 68055, 69202, 69505, 70394, 71022, 83064, 85390, 86965, 87267, 91334, 92907",,,,"module: binaries, module: build, module: convolution, module: cuda, module: cudnn, module: dependency bug, module: intel, module: macos, module: mkl, module: openmp, module: power, module: rocm, module: wsl","Stale, enhancement, module: POWER, module: binaries, module: build, module: convolution, module: cuda, module: cudnn, module: dependency bug, module: intel, module: macos, module: mkl, module: openmp, module: rocm, module: wsl, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8276 [Installation]: Support conda/pip install with ppc64le(power8), #38624 Wheels not manylinux1 compliant, #55454 Include the cuda headers in the libtorch pre-built packages, #68055 Release official Python 3.9 images, #69202 CPU only mode with cudatoolkit 11.3, #69505 with the same environment ,pytorch 1.8 worked but 1.10 can't work",6b29f100475e6fc7591afe0eb554f220c64e16477f6477f549be8149d9e38797
issue-collapse:d16e0a655e5d9ac0,module: binaries stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"75158, 80857, 85098, 92273, 92292, 93884, 96608, 98361, 98863, 99637, 100366, 103959, 106382, 108244",,,,"module: binaries, module: cuda, module: docs, module: python frontend, oncall: releng","feature, module: binaries, module: cuda, module: docs, module: python frontend, oncall: releng, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#75158 Offical libtorch (pytorch c++ frontend) docker image, #80857 Compatibility List, #85098 install libtorch cxx11 ABI as default in PyTorch pip installation, #92273 Always install cpu version automatically, #92292 ""Get Started"" tells us to use the anaconda installer for PyTorch 3.x - but this should be python 3.x, #93884 Importing tensorflow (2.12) before torch (2.0) hangs at import torch",91ae15b8a9831dd6ab370fcb263908eb045b45ac04a145d50999a4d9b0b66d73
issue-collapse:2d3ea4d991b61f2f,module: binaries stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"43760, 96595, 103578, 109112, 110250, 110819, 113621, 113941, 114948, 114949, 115077, 116576, 120291, 120319",,,,"module: binaries, module: build, module: cuda, module: dependency bug, module: ios, module: sparse, module: windows","module: binaries, module: build, module: cuda, module: dependency bug, module: ios, module: sparse, module: windows, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#43760 Error : lib/libstdc++.so.6: version `CXXABI_1.3.11 not found, #96595 Why doesn't PyTorch install the REAL nvidia cuDNN pip package?, #103578 ImportError: undefined symbol: cublasSetWorkspace_v2, version libcublas.so.11, #109112 Unable to install the latest version of PyTorch using mamba., #110250 Accessing Particular Nightly Builds Don't Work, #110819 Segmentation fault on aarch64 (Rpi4) using Pytorch 2.1.0 & torchaudio",4bbe0b8859e768755119d02cdfc76bfb77374bd41e78d9e514e25f778442acc7
issue-collapse:1ed0e7ebd98fc522,module: binaries stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"80874, 89417, 95225, 110074, 115138, 121090, 121308, 122571, 123075, 124047, 124566, 125502, 126286, 128033",,,,"module: binaries, module: build, module: cpp, module: cpp-extensions, module: docker, module: macos, module: memory usage, module: protobuf","better-engineering, has workaround, module: binaries, module: build, module: cpp, module: cpp-extensions, module: docker, module: macos, module: memory usage, module: protobuf, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#80874 C++ extensions inject a bunch of compilation flags, #89417 OSError: libcublas.so.11: cannot open shared object file: No such file or directory, #95225 Remove conda virtualenv from the docker image, #110074 ImportError: libc10_cuda.so: cannot open shared object file: No such file or directory, #115138 ImportError: /lib64/libstdc++.so.6: version `GLIBCXX_3.4.20' not found, #121090 Pytorch nightly docker images set PYTORCH_VERSION variable that doesn't match the installed torch version",38f5a375ed048df2c9c6ea3105325eaebe921dd0adb3df72852f8b8e7b747887
issue-collapse:f05c30e8a0a2711f,module: binaries stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"70513, 99781, 112872, 119072, 119447, 125862, 126005, 129336, 129555, 130128, 132617, 137412, 141337, 141467",,,,"module: binaries, module: cpp, module: cuda, module: cudnn, module: docker, module: docs, module: intel, module: nccl, module: python frontend, module: third_party","module: binaries, module: cpp, module: cuda, module: cudnn, module: docker, module: docs, module: intel, module: nccl, module: python frontend, module: third_party, topic: binaries, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#70513 Conda Repodata.json file not found in Pytorch channel, #99781 2.0.0+cu118 package missing proper libnvrtc-builtins.so.11.8, #112872 libtorch exports miniz symbols, #119072 Importing PyTorch 2.2 fails with undefined symbol error: ncclCommRegister, #119447 Cannot install or import PyTorch Nightly, #125862 Pytorch nightly docker image invalidated layers",b18cee3ec55614fe056e4c534a2c3e83d6925a8a93db49917aae4e6921abe8a8
issue-collapse:9f368b276d2e9f1f,module: binaries stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17835, 38412, 52134, 98115, 118173, 121668, 134236, 139005, 141094, 141977, 143434, 152426, 155203, 156024",,,,"module: binaries, module: cpp, module: cuda, module: cudnn, module: mkl, module: performance, module: windows, oncall: releng","actionable, feature, module: binaries, module: cpp, module: cuda, module: cudnn, module: mkl, module: performance, module: windows, oncall: releng, topic: binaries, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17835 PyPy support, #38412 No MKL Compatible Conda installation for PyTorch 1.5, #52134 Build using Py_LIMITED_API and then build wheels with the stable ABI abi3 tag, #98115 Request to cherrypick a fix into v1.13.1 (v1.8 has a CVE), #118173 ModuleNotFoundError: pynvml does not seem to be installed or it can't be imported., #121668 Warning static library kineto_LIBRARY-NOTFOUND not found when trying build tutorial torch.compiler_aot_inductor.html",0072ec9fb5f7e4abb3316ddb5cbd1fdfec8bb8e18df26bab599ec0450514080b
issue-collapse:7adb92b7f4992d86,module: binaries stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"20963, 27752, 28863, 31697, 32419, 32972, 34064, 36062, 37588, 43140, 43758, 47291, 47739, 47751",,,,"module: binaries, module: build, module: cuda, module: dependency bug, module: error checking, module: protobuf, module: selective build, oncall: distributed, oncall: releng","better-engineering, module: binaries, module: build, module: cuda, module: dependency bug, module: error checking, module: protobuf, module: selective build, oncall: distributed, oncall: releng, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#20963 official libtorch static build zip file error, #27752 Version number is still duplicated in a bunch of places, #28863 Command for downloading torch 1.12.0 CUDA 10 linked to cu92 version, #31697 the cmake problem with build from source ?, #32419 MAGMA libraries, #32972 which pytorch do i install for running this project in my windows.link for this is :: https://github.com/xiaojunxu/SQLNet .IN this they have used python 2.7 ,but i am unable to install pytorch on python 2.7 environment .help me with this",0bb9b665f0240ad4d2fb7e7a2b8d08671ec48e26ce17e33c8d1aa5406bfdfb61
issue-collapse:0620a87f0274001c,module: binaries stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"76557, 139761, 147274, 150399, 156694, 156810, 157548, 159947, 162971, 164122, 165902, 166233, 167979, 177082",,,,"module: binaries, module: build, module: ci, module: cuda, module: docker, module: docs, module: linear algebra, module: nativert, module: openmp, module: sparse, module: third_party, module: windows, oncall: releng","actionable, better-engineering, bot-triaged, enhancement, has workaround, low priority, module: binaries, module: build, module: ci, module: cuda, module: docker, module: docs, module: linear algebra, module: nativert, module: openmp, module: sparse, module: third_party, module: windows, oncall: releng, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76557 Pip packaging and publishing improvements in pytorch wheels for better integration with poetry, #139761 [RFC] PyTorch - PyPi PEP-759 proposal (wheel-next), #147274 PaddleOCR PyTorch conflicts, #150399 torch wheels are unusable if CUDA RPMs are installed on the system (was Import error in nvidia/cuda:12.6.3-cudnn-devel-rockylinux9), #156694 Vendored wheels on PyTorch pip repository are outdated (e.g., `cmake`, `certifi`), #156810 Add the XPU item to pytorch.org/get-started",84e1008c8f4acf8be0c036fb81d74bb3a4ca2c05cd62d55dbb7b7c8493a45b41
issue-collapse:62fa51815916e14f,module: bfloat16 stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"15607, 57773, 62016, 67682, 67979, 75427, 75667, 80574, 94333, 101781, 112575, 116120, 120463, 121389",,,,"module: bfloat16, module: bootcamp, module: complex, module: cpu, module: cublas, module: cuda, module: docs, module: nn, module: numpy, module: performance, module: power, module: scatter & gather ops, module: sorting and selection, module: sparse","enhancement, matrix multiplication, module: POWER, module: bfloat16, module: bootcamp, module: complex, module: cpu, module: cublas, module: cuda, module: docs, module: nn, module: numpy, module: performance, module: scatter & gather ops, module: sorting and selection, module: sparse, triaged",,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#15607 documentation for adding a new type via C++ extensions, #57773 BFloat16 CUDA GEMM ops unsupported on Nvidia P100 (SM_60) on CUDA 11.3, #62016 ppc64le build fail: invalid conversion from Bfloat16 in functional_base.h:180, #67682 Feature request: bfloat16 support for CUDA matmuls, #67979 Feature Request: Support prelu_cuda for BFloat16, #75427 `torch.cuda.is_bf16_supported()` seem to not work properly",431fe4dd88556f437dc8c0b603ff4c731b81d7ee9f2154c17b060bbdab07470d
issue-collapse:ac83e676c3f251ef,module: bfloat16 stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"88136, 117961, 136186, 140348, 140892, 147134, 151934, 154086, 154351, 157808, 163425, 163574, 166853, 172085",,,,"module: bfloat16, module: correctness (silent), module: cpu, module: cublas, module: cuda, module: half, module: linear algebra, module: numerical-stability, module: performance, module: python frontend, module: random, module: rnn, module: sparse, module: special, module: tests","actionable, matrix multiplication, module: bfloat16, module: correctness (silent), module: cpu, module: cublas, module: cuda, module: half, module: linear algebra, module: numerical-stability, module: performance, module: python frontend, module: random, module: rnn, module: sparse, module: special, module: tests, triaged",,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#88136 Placing LSTM model on bfloat16 on GPU causes error, #117961 ReplicatePad2d cuda not support bfloat16, #136186 ""bmm_sparse_cuda"" not implemented for 'BFloat16', #140348 inconsistency in ```torch.special.i0e``` on CPU and GPU, #140892 inconsistency in ```torch.special.xlogy``` on CPU and GPU, #147134 `torch.mul` uses `OpMathType` for computation.",e6a48bea0617541bce3d604efe39bbe1c7e8c4664d4315dd715269c9e29156d2
issue-collapse:105afe070a531daa,module: bc-breaking stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8104, 33122, 39310, 41027, 46291, 50530, 61574, 68464, 72317, 79195, 109781, 141139, 151110, 157764",,,,"module: bc-breaking, module: c10d, module: cpp-extensions, module: cuda, module: deprecation, module: distributions, module: docs, module: embedding, module: functional ux, module: intel, module: meta tensors, module: mkldnn, module: nn, module: norms and normalization, module: numerical-stability, module: numpy, module: primtorch, module: python frontend, module: safe resize","ezyang's list, feature, module: bc-breaking, module: c10d, module: cpp-extensions, module: cuda, module: deprecation, module: distributions, module: docs, module: embedding, module: functional UX, module: intel, module: meta tensors, module: mkldnn, module: nn, module: norms and normalization, module: numerical-stability, module: numpy, module: primTorch, module: python frontend, module: safe resize, topic: bc breaking, topic: improvements, triaged",,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8104 load_state_dict unexpectedly does not load Tensor to buffers that currently have None value, #33122 [feature request] make torch.multinomial behaviour compliant with rnn output dimension, #39310 Make torch.cross `dim` parameter work intuitively, #41027 out= resizing (and restriding) behavior is confusing, #46291 ProcessGroup::Work API changes, #50530 Unable to compile CUDAExtension with Pytorch 1.5 or above",b908551ffd2d20bafdcb524d2de78fcc42def1b5d98a9dad575331890702a50b
issue-collapse:80decb4108424e75,module: backend stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55879, 55884, 91737, 93347, 95207, 98070, 98924, 100914, 112691, 113369, 116255, 123072, 151303, 165811",,,,"module: backend, module: cudacachingallocator, module: intel, module: memory usage, module: mps, module: privateuse1, module: python frontend, module: serialization, module: vulkan","enhancement, module: CUDACachingAllocator, module: PrivateUse1, module: backend, module: intel, module: memory usage, module: mps, module: python frontend, module: serialization, module: vulkan, triaged",,Handle this module: backend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55879 Vulkan backend on desktop platforms, #55884 [Bug] RuntimeError: could not create a primitive on Xeon, #91737 [MPS] Improve the performance of torch.linear(), #93347 when I want to use a new backend, how to deal with the op with 'device' argument?, #95207 new backend privateuseone with ""to"" op, #98070 Request custom backend device memory Allocator.",f49c5986fd508c4c73f90f03174a292e8846f743e8779e2b24c4622c4b84b691
issue-collapse:a85fe976a49dbc6b,module: autograd stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"67463, 68331, 69714, 71698, 72041, 72055, 72061, 73137, 74222, 74299, 74604, 76585, 76780, 76786",,,,"module: autograd, module: ci, module: complex, module: cpp, module: cuda, module: deadlock, module: double backwards, module: forward ad, module: linear algebra, module: nn, module: nn.utils.parametrize","Stale, actionable, complex_autograd, enhancement, feature, module: autograd, module: ci, module: complex, module: cpp, module: cuda, module: deadlock, module: double backwards, module: forward ad, module: linear algebra, module: nn, module: nn.utils.parametrize, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#67463 `test_forward_mode_AD` hangs for `nn.functional.cosine_embedding_loss`, #68331 Casting real parameter to complex during forward produces warning on backward, #69714 Possibly out of date error in autograd codegen, #71698 [libtorch]can not save a vector<int> to AutogradContex->saved_data., #72041 RuntimeError: Trying to backward through the graph a second time (or directly access saved tensors after they have already been freed). Saved intermediate values of the graph are freed when you call .backward() or autograd.grad(). Specify retain_graph=True if you need to backward through the graph a second time or if you need to access saved tensors after calling backward, #72055 index_add : Inconsistent between CPU and CUDA",9e6adf883b1b1a874ec402387f80f748f1c26ef8d1d0b661500f74bacc23058a
issue-collapse:2179694bc7121e12,module: autograd stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"63623, 76783, 77515, 77552, 77963, 78151, 78346, 78624, 79705, 80204, 80206, 80803, 80804, 81448",,,,"module: autograd, module: distributions, module: double backwards, module: edge cases, module: forward ad, module: linear algebra, module: loss, module: nn, module: pooling, module: tests","actionable, enhancement, feature, has workaround, module: autograd, module: distributions, module: double backwards, module: edge cases, module: forward ad, module: linear algebra, module: loss, module: nn, module: pooling, module: tests, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#63623 Expose a SavedTensorsHooks nn.Module for users to register saved tensors hooks, #76783 `gradcheck` fails for `torch.trace`, #77515 Inplace Bool API + `sum` will trigger INTERNAL ASSERT FAILED, #77552 Functional Jacobian does not work with Torchdiffeq, #77963 `logaddexp2` fails to backward, #78151 `ctc_loss` will backward crash",08fcae6e122e66b87116a850a5267dc5bc05a6166c589ab4951a59bd06d10a81
issue-collapse:d97ae3b01a82d8a2,module: autograd stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"56500, 68208, 74677, 78482, 80061, 80762, 81417, 82145, 82545, 83537, 86271, 86444, 88691, 89574",,,,"module: autograd, module: batching, module: cuda, module: dependency bug, module: edge cases, module: forward ad, module: functorch, module: ios, module: loss, module: molly-guard, module: norms and normalization, module: tests, module: viewing and reshaping, oncall: mobile","actionable, enhancement, feature, module: autograd, module: batching, module: cuda, module: dependency bug, module: edge cases, module: forward ad, module: functorch, module: ios, module: loss, module: molly-guard, module: norms and normalization, module: tests, module: viewing and reshaping, oncall: mobile, tensor subclass, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#56500 Marking variable not required in backward calculation if it is not needed, #68208 Make `torch.Tensor.view` support autograd for appropriate cases, #74677 `check_batched_forward_grad` fails for `torch.norm` and related ops, #78482 RuntimeError: Event device type CUDA does not match blocking stream’s device type CPU, #80061 pow CUDA tensor raised to CPU scalar tensor result can't backward properly, #80762 `atan2` will gradcheck fail when `other` is a tensor with `int8` dtype",bd98f3e7096e7bc3d5577ae398d1d34ee14d9872b838d8aa0926d05f954054c5
issue-collapse:409ee3df9b1aa604,module: autograd stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"49825, 70870, 76522, 85485, 87755, 87956, 90065, 94451, 96187, 96305, 96769, 96908, 98004, 98498",,,,"module: autograd, module: cuda graphs, module: forward ad, module: functorch, module: mps, module: numerical-reproducibility, module: numerical-stability, module: performance, module: sparse","ZeroTensor, actionable, better-engineering, enhancement, feature, has workaround, module: autograd, module: cuda graphs, module: forward ad, module: functorch, module: mps, module: numerical-reproducibility, module: numerical-stability, module: performance, module: sparse, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#49825 Ensure inplace views are properly handled, #70870 Add flag for functional.Jacobian to return output as well, #76522 `torch.clamp` does not distribute gradients as element-wise`min/max` do, #85485 `max_pool2d_with_indices(self, ...)` shouldn't need to save `self` for backward, #87755 Autograd precision for CONV + BN between pytorch version 1.11.0 and 1.12.1, #87956 Autograd doesn't stop executing backward graph early enough in situations involving set_",d4e79b24e3e5207050bb2e7c2099ab57c4e90f62a57b4bee22a5be0141e77caf
issue-collapse:45b9f0465bb72eba,module: autograd stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3625, 66357, 91996, 94111, 94990, 95132, 97827, 99147, 99806, 101069, 102556, 103518, 104040, 105220",,,,"module: autograd, module: complex, module: double backwards, module: edge cases, module: forward ad, module: functorch, module: linear algebra, module: sparse, module: tests, module: vmap","actionable, feature, has workaround, module: autograd, module: complex, module: double backwards, module: edge cases, module: forward ad, module: functorch, module: linear algebra, module: sparse, module: tests, module: vmap, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3625 Proposal: combine requires_grad and retain_grad(), #66357 OpInfos disabled for batched forward grad computation, #91996 autograd.functional.jacobian : Imaginary part is lost for functions with real input and complex output., #94111 `svd` triggers INTERNAL ASSERT FAILED when computing jacobian in forward mode, #94990 interactions between views + autograd.Function + AOTAutograd causes memory leak, #95132 Differentiate with regard a subset of the input",a9c583f264e9c89043311cec69f9ef1877b7ffa220f86b38bb9efd14f24e0d03
issue-collapse:96569b05301b09f2,module: autograd stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"54216, 61222, 68429, 87559, 92331, 95711, 104933, 105846, 106265, 115785, 116320, 118883, 121842, 123996",,,,"module: autograd, module: bc-breaking, module: cuda, module: derivatives, module: fft, module: forward ad, module: functorch, module: half, module: mps, module: nn, module: norms and normalization, module: optimizer, module: rocm, module: sparse, module: vmap, oncall: distributed","actionable, enhancement, function request, module: autograd, module: bc-breaking, module: cuda, module: derivatives, module: fft, module: forward ad, module: functorch, module: half, module: mps, module: nn, module: norms and normalization, module: optimizer, module: rocm, module: sparse, module: vmap, oncall: distributed, topic: bc breaking, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#54216 Add inplace variant for torch.minimum / torch.maximum, #61222 The syncbatchnorm can not be employed in the MODEL that the gradients including in the loss function, #68429 istft gradcheck fails on ROCm, #87559 diagonal of Jacobian matrix, #92331 backward(inputs= does not need to execute grad_fn of the inputs, #95711 `copy.deepcopy` does not copy gradients of nn.Parameter",82a73a35a73f229c4e2e9c733370922d5d50f8cf6e5ee5bda05aaff27ce594bd
issue-collapse:51c933f8478df664,module: autograd stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"67240, 80411, 94397, 101407, 127541, 128020, 128064, 128600, 128703, 134069, 134612, 134901, 135229, 135447",,,,"module: autograd, module: complex, module: correctness (silent), module: cudnn, module: docs, module: double backwards, module: error checking, module: functorch, module: memory usage, module: mps, module: nn, module: norms and normalization, module: rnn, module: sparse, module: vmap","actionable, better-engineering, complex_autograd, enhancement, feature, module: autograd, module: complex, module: correctness (silent), module: cudnn, module: docs, module: double backwards, module: error checking, module: functorch, module: memory usage, module: mps, module: nn, module: norms and normalization, module: rnn, module: sparse, module: vmap, topic: docs, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#67240 Feature: Add derivative for channel_shuffle, #80411 Tracker: Slow gradcheck failures possibly indicating incorrect gradients, #94397 jacfwd and jacrev are fundamentally broken for complex inputs, #101407 Delete old vmap prototype, #127541 RuntimeError: derivative for aten::_spdiags is not implemented, #128020 Make `torch.autograd.Function` support `vmap`",d6d63e5fc81730558d8500decbaa19bf0f59c3b3966ac3a97f72747fec2ad016
issue-collapse:4a621e62bb71f389,module: autograd stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"29734, 80827, 83510, 99556, 128971, 140516, 141845, 144639, 145356, 146036, 149315, 154332, 159667, 162907",,,,"module: autograd, module: compiled autograd, module: correctness (silent), module: cpu, module: determinism, module: flex attention, module: forward ad, module: functional ux, module: functorch, module: higher order operators, module: pt2-dispatcher, module: sparse, module: vmap, oncall: pt2","actionable, enhancement, fixathon, module: autograd, module: compiled autograd, module: correctness (silent), module: cpu, module: determinism, module: flex attention, module: forward ad, module: functional UX, module: functorch, module: higher order operators, module: pt2-dispatcher, module: sparse, module: vmap, oncall: pt2, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#29734 Tensor.nbytes() returns itemsize * numel for sparse tensors, #80827 Inconsistent computation of gradient in MaxUnPooling, #83510 Cdist backward dependent on compute_mode, #99556 torch.func.jacrev fails if model contains full_backward_hook, #128971 Trying to use forward AD with _scaled_dot_product_flash_attention that does not support it because it has not been implemented yet., #140516 [autograd engine && compiled autograd] support for partial backward compilation",418a9ffd454002b3c76b41b050f9a14b6325fafe6a8089acf0fc3adce903b92d
issue-collapse:e452c6b65a72306c,module: autograd stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9701, 18643, 22035, 22778, 23032, 25941, 26379, 27579, 27641, 27642, 28925, 31185, 32576, 32884",,,,"module: autograd, module: cpp, module: docs, module: macos, module: memory usage, module: optimizer, module: performance, module: rpc, module: sparse, module: tests","enhancement, feature, module: autograd, module: cpp, module: docs, module: macos, module: memory usage, module: optimizer, module: performance, module: rpc, module: sparse, module: tests, quansight-nack, todo, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9701 [request] speed-up multidim slicing backward, #18643 Memory not being deallocated in backward(), #22035 Have a different way to check if gradient was computed in the optimizer (not checking for None), #22778 In-place updating the original value tensor should also update version counter of sparse tensor's values_ tensor, #23032 Proposal: Optional AutogradMeta for Variable, #25941 TestAutograd.test_deep_reentrant fails with SIGBUS on macOS",e92d742bd9f2e080465a7eccb89f16b0649f3180b5cabfebf86cab7bae466425
issue-collapse:e1d6f86feed1fcb1,module: autograd stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2512, 2691, 4501, 4564, 7786, 28549, 31579, 34082, 34367, 34998, 35128, 37354, 37860, 38057",,,,"module: autograd, module: cuda, module: data parallel, module: docs, module: double backwards, module: mkldnn, module: nn, module: performance, oncall: distributed, oncall: jit","better-engineering, enhancement, feature, module: autograd, module: cuda, module: data parallel, module: docs, module: double backwards, module: mkldnn, module: nn, module: performance, oncall: distributed, oncall: jit, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2512 detach_() variant that affects all past uses too, #2691 Feature Request: Support grad of grad in fused RNNs, #4501 Bind in Python _backward ATen functions, #4564 Mixed Tensor/TensorList arguments in ATen functions with explicit derivatives, #7786 [feature request] Simple and Efficient way to get gradients of each element of a sum, #28549 Expose DifferentiableGraphBackward to python",25ce48fd3561d820c6ab9b818e39c53d1e6d6ebd0d734a31aa47894db38eac34
issue-collapse:2b44634e21b62700,module: autograd stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3818, 4952, 5159, 32531, 33373, 33855, 40736, 41926, 42156, 43347, 45308, 46643, 46820, 46918",,,,"module: autograd, module: batching, module: docs, module: linear algebra, module: memory usage, module: performance, module: power, module: sparse","actionable, feature, function request, module: POWER, module: autograd, module: batching, module: docs, module: linear algebra, module: memory usage, module: performance, module: sparse, quansight-nack, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3818 Memory leak when doing backward with grad as yourself, #4952 [Feature request] Optimize autograd/ATen when a gradient is clearly zero, #5159 TakeBackward taking a significant portion of backward time, #32531 Batched torch.eig() and gradient of torch.eig() for real eigenvalues, #33373 torch.clamp_ not inplace during backward, #33855 Unclear output for Pytorch Profiler",b603aa50aa3eff6deda1b9bab577f70f46d16e161f6cdc92c1853cba02543818
issue-collapse:af69f2df7200129e,module: autograd stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4731, 8874, 9282, 25814, 33628, 35797, 40480, 45459, 47824, 47901, 48227, 49673, 49724, 50613",,,,"module: autograd, module: bootcamp, module: cpp, module: ddp, module: nans and infs, module: nccl, module: numerical-stability, module: optimizer, module: sparse, module: tests, module: viewing and reshaping","actionable, better-engineering, enhancement, fixathon, module: NaNs and Infs, module: autograd, module: bootcamp, module: cpp, module: ddp, module: nccl, module: numerical-stability, module: optimizer, module: sparse, module: tests, module: viewing and reshaping, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4731 Consider disallowing Variables that require grad in NCCL/comm functions, #8874 Issues with dynamically created grad_fn for views, #9282 [gradcheck] warn about the case that mulitple inputs share storage, #25814 (PyTorch1.1 and 1.2) RuntimeError: Can't detach views in-place. Use detach() instead, #33628 Remove .data subset1 for fixathon, #35797 Backward function causes device error in C++ when changing module's device repeatly.",594d88c59f061f2a1d5f18ff1d745c05f14e7cb4e26178ce1932f0afb0dd3ba2
issue-collapse:ad6e309308bd8c44,module: autograd stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12248, 29378, 47502, 49409, 50219, 51117, 52390, 52762, 53407, 53649, 53666, 54375, 55557, 56177",,,,"module: autograd, module: correctness (silent), module: linear algebra, module: nn, module: tests, module: vmap","better-engineering, feature, low priority, module: autograd, module: correctness (silent), module: linear algebra, module: nn, module: tests, module: vmap, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12248 .cuda() changes a module's behavior when there are registered buffers with requires_grad=True, #29378 Support out= parameters with autograd, #47502 Hackable python-only autograd engine, #49409 Add global gradcheck setting, #50219 `requires_grad_` in `no_grad` context returns incorrect value with tensor subclasses, #51117 TestLinalgCPU.test_cholesky_solve_autograd_cpu_float64 fails on different random seed",c749c23a1cb36ec04af4a80156e4555139d2b8991038a63b40aedde72dd27f91
issue-collapse:c93623d58d0a16d3,module: autograd stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13818, 37751, 47028, 56370, 57157, 57502, 57503, 57507, 58037, 58442, 58479, 59493, 59896, 60159",,,,"module: autograd, module: complex, module: custom-operators, module: data parallel, module: docs, module: linear algebra, module: loss, module: nn, module: norms and normalization, module: pybind","actionable, complex_autograd, feature, module: autograd, module: complex, module: custom-operators, module: data parallel, module: docs, module: linear algebra, module: loss, module: nn, module: norms and normalization, module: pybind, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13818 test_spectral_norm: Backward is not reentrant, #37751 Documentation of _CtxMethodMixin: must be tensors?, #47028 torch.autograd.backward() fails to sync with other stream, #56370 Acquiring ""is_grad_enabled()"" inside an autograd function, #57157 Using register_full_backward_hook with target module for intermediate activation, #57502 [structured] Fail if CompositeExplicitAutograd is in dispatch table of structured_delegate function",73bc0e53cf70913c329a883d957f5d77b28d1589dd2a6a1a31a6a9175a4878db
issue-collapse:64f9fb285dd4233a,module: autograd stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7343, 54591, 55609, 60426, 60796, 61410, 61485, 61513, 61619, 61645, 62934, 63205, 63624, 63644",,,,"module: autograd, module: codegen, module: deprecation, module: docs, module: memory usage, module: nn, module: tensor creation","actionable, better-engineering, module: autograd, module: codegen, module: deprecation, module: docs, module: memory usage, module: nn, module: tensor creation, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7343 [memory leak] [PyTorch] .backward(create_graph=True), #54591 Various issues in derivatives.yaml, #55609 requires_grad does not get propagated properly when using the JIT compiler, #60426 Output of non-inplace, non-CompositeImplicitAutograd op has TensorImpl > 1 or StorageImpl use_count != 1, #60796 AccumulateGrad short circuits before calling variable hooks when the incoming grad is None, #61410 requires grad get lost during transform.",06b8d0319cf4dae731f04e161cb1dea93f580f1ef77592750ef29408d7512a36
issue-collapse:11d1fa51fa638969,module: autograd stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"23938, 35736, 52052, 54346, 56202, 56235, 56275, 57447, 62473, 62849, 64306, 64931, 65536, 65683",,,,"module: autograd, module: codegen, module: complex, module: cpp, module: cuda, module: determinism, module: distributions, module: linear algebra, module: magma, module: memory usage, module: nans and infs, module: nn, module: sparse, module: tests, module: viewing and reshaping, oncall: distributed, oncall: jit","actionable, days, module: NaNs and Infs, module: autograd, module: codegen, module: complex, module: cpp, module: cuda, module: determinism, module: distributions, module: linear algebra, module: magma, module: memory usage, module: nn, module: sparse, module: tests, module: viewing and reshaping, oncall: distributed, oncall: jit, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#23938 torch.nn.DataParallel causes incorrect gradients, #35736 Simple C++ custom autograd function code throws error ""CUDA error: driver shutting down"", #52052 common_jit.check_against_reference() should check gradients w.r.t. weights, #54346 Missing tests for gradcheck, #56202 Tolerance for non-determinism operators in gradcheck, #56235 `test_cholesky_solve` gradgradcheck fails sometimes",afb83c01f878a621716a8200b095888b392e1ef7b93c32f901f4bd1804e48422
issue-collapse:e8ed169497025fdb,module: autograd stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"30987, 59137, 62654, 66066, 66199, 66875, 67187, 67387, 67493, 67594, 67653, 67772, 68046, 70342",,,,"module: autograd, module: docs, module: nans and infs, module: nn, module: sanitizers, module: tests, module: xla","actionable, better-engineering, enhancement, feature, hackathon, low priority, module: NaNs and Infs, module: autograd, module: docs, module: nn, module: sanitizers, module: tests, module: xla, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#30987 Remove `.data`, #59137 [autograd] `fill_` gradgradcheck raises RuntimeError, #62654 Checkpointing without re-entrant autograd, #66066 Finishing OpInfos: test_autograd.py, #66199 Move the `queue_callback()` API out of `Variable._execution_engine` and into a public API, #66875 Make the evaluated value of function f(x) accessible from `torch.autograd.functional.jacobian(f,x)`",f6d42fa626afd96ff4ac1cb6d67c8d294123df7c852f0435a40eddfa1c966fce
issue-collapse:e38337b4eadb12d6,module: autograd stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11390, 68966, 69610, 75432, 77230, 117582, 125984, 140520, 154226, 164331, 164759, 166131, 167254, 169055",,,,"module: autograd, module: cuda, module: docs, module: forward ad, module: functorch, module: linear algebra, module: nestedtensor, module: nn, module: onnx, module: sdpa, oncall: export, oncall: profiler, oncall: pt2, oncall: visualization","actionable, feature, module: autograd, module: cuda, module: docs, module: forward ad, module: functorch, module: linear algebra, module: nestedtensor, module: nn, module: onnx, module: sdpa, oncall: export, oncall: profiler, oncall: pt2, oncall: visualization, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11390 Views created in no_grad block still have requires_grad=True, #68966 `torch.ldexp` generated tests fail on call to `torch.mul`, #69610 [Question] How to extract/expose the complete PyTorch computation graph (forward and backward)?, #75432 Rollup: Top forward-over-reverse formulas, #77230 fast `gradcheck` fails when outputs that do not require grad precede outputs that do, #117582 Add forward AD support for nn.functional.grid_sample",9bf2a2a7b7e9aaf9b4b65895a08e36bb230d0870fe8e4ffed0f77d2cd5f6c624
issue-collapse:930f7bc65b1678c8,module: assert failure stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"16589, 16695, 22122, 51855, 59743, 70388, 84510, 86644, 89275, 92594, 94304, 96789, 122467, 130073",,,,"module: assert failure, module: autograd, module: ci, module: cpp, module: cudacachingallocator, module: cudnn, module: docs, module: dynamic shapes, module: error checking, module: internals, module: molly-guard, module: multiprocessing, module: nvfuser, module: sparse, module: type promotion","Stale, module: CUDACachingAllocator, module: assert failure, module: autograd, module: ci, module: cpp, module: cudnn, module: docs, module: dynamic shapes, module: error checking, module: internals, module: molly-guard, module: multiprocessing, module: nvfuser, module: sparse, module: type promotion, triaged",,Handle this module: assert failure stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#16589 Make sure `data_ptr` for non-zero-size input tensors stays the same after the VariableType dispatch, #16695 Check whether _cudnn_rnn_flatten_weight can avoid changing the TensorImpl or Storage pointer of tensors in `weight_arr`, #22122 ASSERT FAILED at /pytorch/torch/csrc/autograd/generated/VariableType_0.cpp:12721, #51855 Attempted to read a PyTorch file with version 3, but the maximum supported version for reading is 2. Your PyTorch installation may be too old. (init at /pytorch/caffe2/serialize/inline_container.cc:132), #59743 m.fallback(torch::CppFunction::makeFromBoxedFunction<&my_fallback>) gives bad error message, #70388 RuntimeError: tensor has too many (>25) dims when permuting tensor with GPU backend",753bcc7dd1d5c4ad6afe7729dff4df029afb1caa2b4162d3645e7c615b31d811
issue-collapse:c35c1e20ba273d1d,module: arm stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"22550, 39680, 46522, 47476, 50650, 51535, 52206, 52478, 54104, 59262, 59437, 59936, 60834, 65673",,,,"module: arm, module: binaries, module: build, module: convolution, module: correctness (silent), module: cpu, module: cuda, module: docker, module: jetson, module: macos, module: performance, module: vision, module: vulkan","feature, function request, module: arm, module: binaries, module: build, module: convolution, module: correctness (silent), module: cpu, module: cuda, module: docker, module: jetson, module: macos, module: performance, module: vision, module: vulkan, triaged",,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#22550 Eigen Tensor library for convolutions on CPU, #39680 CUDA cannot be found, #46522 ModuleNotFoundError when installing PyTorch via pip on aarch64 environment, #47476 Reciprocal and reciprocal square root instructions are too inaccurate on ARM64, #50650 build from source fail in jeston tx2 Python3.8, #51535 Why isn't there a pip-installable PyTorch for Raspberry Pi?",a9f217c81f8b882b9a8f5fb2581ce197d9a1bc575d12ff1e87d8f244339dad6b
issue-collapse:786563afe60776ed,module: arm stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50471, 56333, 59243, 59356, 61110, 66410, 68701, 73051, 74503, 76689, 79421, 81317, 86676, 99544",,,,"module: arm, module: binaries, module: build, module: ci, module: cuda, module: jetson, module: macos, module: numerical-stability","Stale, enhancement, module: arm, module: binaries, module: build, module: ci, module: cuda, module: jetson, module: macos, module: numerical-stability, triaged",,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50471 linker error when trying to use Metal backend in PyTorch Mobile, #56333 Unable to pip install PyTorch on M1 Mac - Errors, #59243 Can't install pytorch =1.7 with python 3.8.5 on Raspberry Pi, #59356 pytorch build failure on arm64, #61110 Incorrect exponential calculation on Jetson devices with float32 dtype, #66410 CIFAR10 doesn't work on M1 MacBook",f1595a170d4dacbc6ab3d3dd5968a0c3f05aac05cd25e3f95a13f3e99a3307be
issue-collapse:71f8a2e96f8b852a,module: arm stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"59439, 63558, 87159, 88968, 89336, 90536, 91182, 93197, 97188, 104502, 105335, 107693, 126980, 130524",,,,"module: arm, module: binaries, module: build, module: ci, module: convolution, module: cpp, module: cuda, module: int overflow, module: jetson, module: macos, module: nn, module: sleef, module: torchbind, module: type promotion, oncall: quantization","feature, low priority, module: arm, module: binaries, module: build, module: ci, module: convolution, module: cpp, module: cuda, module: int overflow, module: jetson, module: macos, module: nn, module: sleef, module: torchbind, module: type promotion, oncall: quantization, topic: improvements, triaged",,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#59439 NaN values on torch.nn.functional.conv2d (aarch64), #63558 universal binaries for libtorch on mac (x86_64+arm), #87159 torch/csrc/utils/python_arg_parser.h:424:94: error: format ‘%ld’ expects argument of type ‘long int’, but argument 7 has type ‘int’, #88968 M1 runner i-090e1df32b6f48a20 run out of disk space, #89336 Conv2d error on M1 mac, RuntimeError: NNPACK SpatialConvolution_updateOutput failed, #90536 Illegal hardware instruction following Real Time Inference on Raspberry Pi 4 tutorial",2addc44bf82652f65302e5f093e27624f3c28d61b7d361b62d6c3ebf563abb21
issue-collapse:95b926c7de227371,module: arm stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"86989, 91368, 120897, 128414, 132132, 132412, 139210, 143571, 145716, 148078, 153471, 157139, 158041, 169294",,,,"module: arm, module: binaries, module: build, module: ci, module: correctness (silent), module: cpu, module: cuda, module: docker, module: jetson, module: macos, module: memory usage, module: mps, module: nans and infs, module: performance, module: tests, module: vectorization, module: windows, module: xnnpack, oncall: profiler, oncall: pt2, oncall: quantization","OSS contribution wanted, enhancement, feature, low priority, module: NaNs and Infs, module: arm, module: binaries, module: build, module: ci, module: correctness (silent), module: cpu, module: cuda, module: docker, module: jetson, module: macos, module: memory usage, module: mps, module: performance, module: tests, module: vectorization, module: windows, module: xnnpack, oncall: profiler, oncall: pt2, oncall: quantization, topic: build, topic: performance, triaged",,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#86989 Error: unknown architecture `armv7-a;' and Error: selected processor does not support `command' in ARM mode, #91368 PyTorch memory leak reference cycle in for loop, Mac M1, #120897 [memory profiling] about supporting the C++ Call Stack on the ARM Architecture, #128414 How to enable XNNPACK instead of NNPACK/MKLDNN in Windows?, #132132 Tests stopped working after merging #126905 on Arm64, #132412 [CI] PyTorch eager is slow in pytorch-linux-jammy-aarch64-py3.10-gcc11-inductor-benchmarks docker",9a87684ba51c5983502fbc9a6dc918413aba6d01b3b9722f6797fe73c1a8a21d
issue-collapse:fea1b93929513a43,module: android stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"50723, 51020, 52495, 58171, 58172, 58393, 58548, 63305, 66284, 74155, 78261, 83947, 101972, 107627",,,,"module: android, module: arm, module: binaries, module: build, module: ci, module: memory usage, module: undefined reference, module: vulkan, oncall: java, oncall: mobile","enhancement, mobile_perf, module: android, module: arm, module: binaries, module: build, module: ci, module: memory usage, module: undefined reference, module: vulkan, oncall: java, oncall: mobile, triaged",,Handle this module: android stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#50723 [PyTorch Mobile] Can't use Vulkan backend, #51020 arm64-v8a not compiling due to libpytorch_jni.so, #52495 link failed when using custom build pytorch on Android, #58171 Error while converting model for NNAPI ""Unsupported node kind ('quantized::batch_norm2d')"", #58172 Mobile Android: Could not run 'aten::quantize_per_tensor' with arguments from the 'Vulkan' backend, #58393 In Android, the memory used by the tensor or model cannot be recycled. Is there any way to solve it?",ad3ad9c787d9b86c6793227e1c5eb1b8b16a0be2887a62c6a61144c365ea855d
issue-collapse:3d35350d297c6484,module: amp (automated mixed precision) stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"42460, 43544, 43556, 46178, 47906, 49282, 49653, 53094, 53345, 56979, 60164, 60724, 63078, 63313",,,,"module: amp (automated mixed precision), module: cuda, module: custom-operators, module: dependency bug, module: interpolation, module: memory format, module: nn, module: performance, module: sparse","module: amp (automated mixed precision), module: cuda, module: custom-operators, module: dependency bug, module: interpolation, module: memory format, module: nn, module: performance, module: sparse, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#42460 Training with AMP gives overflow warning, #43544 AMP much worse performance with groupped Conv2d than fp32, #43556 FP16 gives NaN loss when using pre-trained model, #46178 torch.cuda.amp! when I use @autocast() on DCN(DeformConv), the error ""RuntimeError:expect scalar type Float but Half "", #47906 amp for custom op, #49282 Bug when combining AMP with channels_last memory format",447bfffb9e27e2a2c20a4c96dfe35aa6019d093f3c11a6736de4cf306e7afc11
issue-collapse:65b8c49a2755202f,module: amp (automated mixed precision) stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40427, 42838, 45412, 52279, 55374, 67879, 68315, 70346, 72421, 73924, 80167, 82212, 83733, 86270",,,,"module: amp (automated mixed precision), module: bfloat16, module: cuda, module: docs, module: half, module: linear algebra, module: performance, oncall: distributed","Stale, enhancement, feature, has workaround, module: amp (automated mixed precision), module: bfloat16, module: cuda, module: docs, module: half, module: linear algebra, module: performance, oncall: distributed, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40427 Can we make torch.inverse FP16? - RuntimeError: ""inverse_cuda"" not implemented for 'Half', #42838 RuntimeError: ""triangular_solve_cuda"" not implemented for 'Half', #45412 Autocast but there are no kernels registered for this dispatch key. The operator is p�8ߎU Aborted (core dumped), #52279 [docs] Official apex -> torch.cuda.amp migration guide, #55374 [RFC] Extend Autocast to CPU/CUDA with BF16 data type, #67879 AMP does not improve performance",0043d505dd140b190290783bb19e2e0a201c921fa5c1fde2d5a4c7bc3b8cf538
issue-collapse:0d9492e47639c2d9,module: amp (automated mixed precision) stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"38487, 49038, 70166, 82076, 86548, 87979, 88474, 90057, 92542, 93227, 99710, 100705, 100932, 105682",,,,"module: amp (automated mixed precision), module: bfloat16, module: cpu, module: cuda, module: data parallel, module: docs, module: half, module: linear algebra, module: memory format, module: nn","feature, module: amp (automated mixed precision), module: bfloat16, module: cpu, module: cuda, module: data parallel, module: docs, module: half, module: linear algebra, module: memory format, module: nn, shadow review, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#38487 expected scalar type Half but found Float with torch.cuda.amp and torch.nn.DataParallel, #49038 Make the forward function of a nn.Module and/or a certain function to work with only float32 when used with autocast, #70166 [feature request] Autocast module and function wrappers, #82076 Autocast documentation examples would break, #86548 About autocast, #87979 amp with `bf16`: backward happens in `f16` when using `@torch.cuda.amp.custom_bwd`",6b5def70cd3a492135ffe392fe2bb559fd7cc9ba8b8d3f02362d875365b834df
issue-collapse:10bbda0df0b2ee25,module: amp (automated mixed precision) stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"72897, 74739, 98814, 98860, 108627, 110966, 111739, 117832, 119473, 120994, 129027, 130847, 132388, 132602",,,,"module: amp (automated mixed precision), module: autograd, module: cuda, module: ddp, module: deadlock, module: docs, module: fsdp, module: numerical-stability, oncall: distributed","actionable, module: amp (automated mixed precision), module: autograd, module: cuda, module: ddp, module: deadlock, module: docs, module: fsdp, module: numerical-stability, oncall: distributed, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#72897 Hanging Validation, #74739 GradScaler support FP16 gradients?, #98814 autocast does not work properly on embedding module, #98860 Sharded Grad Scaler Issue Tracker, #108627 autocast not consistent across different GPUs (A100 and RTX A6000), #110966 `torch.is_autocast_enabled()` always False on CPU",7db22d4f479d2bf3e37509f1f24bb085021f87b3796225cb5304f65abc263553
issue-collapse:e69044270a8fe971,module: amp (automated mixed precision) stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40497, 45910, 61173, 67598, 81876, 107663, 115500, 130067, 137080, 160415, 160438, 161621, 165324, 165730",,,,"module: amp (automated mixed precision), module: autograd, module: cuda, module: docs, module: error checking, module: mps, module: nans and infs, module: nn, module: norms and normalization, module: numerical-stability, module: performance, module: privateuse1","actionable, module: NaNs and Infs, module: PrivateUse1, module: amp (automated mixed precision), module: autograd, module: cuda, module: docs, module: error checking, module: mps, module: nn, module: norms and normalization, module: numerical-stability, module: performance, release notes: torch.func, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40497 Mixed precision causes NaN loss, #45910 with torch.cuda.amp.autocast() get out of memory error when using with torch.no_grad() during validation, #61173 Increased memory usage with AMP, #67598 unnecessary warning with autocast, #81876 dtype mismatch when after using auto mixed precision, #107663 `RuntimeError: expected scalar type BFloat16 but found Float` with `torch.nn.TransformerEncoder`",a1deaafdd964ae73a13178ec31503584f042d820e449abb03b2994245240b373
issue-collapse:83ec081ebcc3bcbd,module: advanced indexing stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11635, 15284, 45125, 53257, 54062, 55143, 56896, 59140, 63921, 63935, 71331, 76347, 79997, 96111",,,,"module: advanced indexing, module: complex, module: docs, module: error checking, module: numpy, module: performance, module: sparse, module: tests, module: type promotion, module: viewing and reshaping","enhancement, module: advanced indexing, module: complex, module: docs, module: error checking, module: numpy, module: performance, module: sparse, module: tests, module: type promotion, module: viewing and reshaping, triaged",,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11635 [Feature request] Advanced indexing in functions like `expand`, #15284 index_add_ with scalar values instead of tensors, #45125 Boolean indexing of an ndarray with a torch.tensor mask breaks for size=1, #53257 Inconsistent `out=` behaviour on advanced indexing operations., #54062 Many advanced indexing operations have untested large tensor branches, #55143 Add suport to torch.gather for negative indices",0473714afeea16ee66c7ab8a7dba6550855ca0290c26b9bc78a631d239de2cf6
issue-collapse:a1edcd45308ae130,module: advanced indexing stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"73501, 74952, 99297, 107700, 111159, 119548, 121719, 130525, 135487, 136736, 150017, 161970, 167825, 170350",,,,"module: advanced indexing, module: boolean tensor, module: cpu, module: cuda, module: determinism, module: edge cases, module: embedding, module: error checking, module: nn, module: numpy, module: onnx, module: partial aliasing, module: performance, module: privateuse1, module: python frontend, module: type promotion","enhancement, feature, module: PrivateUse1, module: advanced indexing, module: boolean tensor, module: cpu, module: cuda, module: determinism, module: edge cases, module: embedding, module: error checking, module: nn, module: numpy, module: onnx, module: partial aliasing, module: performance, module: python frontend, module: type promotion, triaged",,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#73501 `index_copy` has different index behavior with `index_fill`, #74952 [feature request] Make `index_select` parallel., #99297 Slicing and indexing support negative steps, #107700 where.self_out doesn't fail gracefully when inputs have different dtypes, #111159 Wrong onnx model from `torch.onnx.export` when using `index_add_` function with duplicate `index` values., #119548 Inconsistent behavior when indexing a Tensor with a list of lists",7ed3f1c0c00469a8ff8f37aaf3faae13d688ce609b1fbdb857f706a0cf6387a7
issue-collapse:97943c53be971881,module: activation checkpointing stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7801, 13234, 23732, 68423, 71322, 79016, 82156, 82465, 86097, 90481, 102600, 125591, 129293, 130096",,,,"module: activation checkpointing, module: autograd, module: bc-breaking, module: cuda graphs, module: data parallel, module: forward ad, module: memory usage, module: nn, module: performance, oncall: distributed","actionable, enhancement, has workaround, module: activation checkpointing, module: autograd, module: bc-breaking, module: cuda graphs, module: data parallel, module: forward ad, module: memory usage, module: nn, module: performance, oncall: distributed, release notes: distributed (sharded), sharded_tensor, triaged",,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7801 Checkpointing is slow on nn.DataParallel models, #13234 torch.utils.checkpoint is not compatible with nn.DataParallel, #23732 nn.Module.forward signature with **kwargs, #68423 `torch.utils.checkpoint.checkpoint_sequential` is not optimal, #71322 Allow users to pass use_reentrant=False to checkpoint_sequential, #79016 Corner cases of ShardedTensor checkpoint when using TorchRec",20b475fd3efe227cc7553c3d2116f6e8965c551f108184c51a1029db55298cdf
issue-collapse:4f0871d860df8fc1,module: activation checkpointing stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"24005, 98668, 131062, 143176, 143774, 144470, 144928, 146707, 147449, 152752, 153334, 165880, 166646, 170284",,,,"module: activation checkpointing, module: autograd, module: compiled autograd, module: functorch, module: pt2-dispatcher, oncall: distributed, oncall: pt2","feature, module: activation checkpointing, module: autograd, module: compiled autograd, module: functorch, module: pt2-dispatcher, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, triaged",,Handle this module: activation checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#24005 Using `torch.utils.checkpoint.checkpoint_sequential` and `torch.autograd.grad` breaks when used in combination with `DistributedDataParallel`, #98668 Cannot use `checkpoint_sequential` with `torch.compile`, #131062 Add some more testing around checkpoint over compile, #143176 Compiled Autograd + Activation Checkpointing/Offloading, #143774 CUDA error when compiling loss function, #144470 Partitioner's auto-AC misbehaves with mixed dtypes",50563a2127d5e5cbee67a5853af80e7b7ea146977987c45c26362471d190b6f4
issue-collapse:3fb7ca402d266f95,module: __torch_function__ stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"55440, 61544, 64187, 64690, 65215, 65753, 73860, 74491, 75462, 75554, 75568, 76037, 76294, 76734",,,,"module: __torch_function__, module: lazy, module: nn, module: tests, module: typing, module: vmap","enhancement, module: __torch_function__, module: lazy, module: nn, module: tests, module: typing, module: vmap, triaged",,Handle this module: __torch_function__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#55440 Have the ability to disable `__torch_function__` dispatch for torch.nn.functional functions, #61544 `__torch_function__` always records default kwargs as kwargs, #64187 Uniformly test that defaulted Tensor arguments appropriately handle __torch_function__, #64690 _disabled_torch_function_impl is unimplementable (must be special cased), #65215 _disabled_torch_function_impl should not be necessary for __torch_dispatch__, #65753 [POC] Convolution-relu fusion via torch function and one-step lazy evaluation",c4477f79f18d99cdd0b019b1dff909cf97e50c59d95a99eac0d9e5b2b18e6211
issue-collapse:ddb2ce91c5783c8f,module: __torch_function__ stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"63911, 76751, 77668, 81323, 92701, 93362, 105644, 106302, 110249, 119194, 122814, 123245, 131947, 135660",,,,"module: __torch_function__, module: activation checkpointing, module: dynamic shapes, module: functorch, module: linear algebra, module: nestedtensor, module: performance, module: testing, oncall: export","enhancement, module: __torch_function__, module: activation checkpointing, module: dynamic shapes, module: functorch, module: linear algebra, module: nestedtensor, module: performance, module: testing, oncall: export, pre_dispatch tracing, tensor subclass, triaged",,Handle this module: __torch_function__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#63911 Add efficient symmetric tensor representations, #76751 Default repr of __get__ methods in __torch_function__ is bad, #77668 Extend BC test to test for __torch_function__ overridability, #81323 `torch.overrides.get_testing_overrides` does not function as intended for native tensor methods/operations, #92701 set_default_device/torch.device has performance impact for non-factory functions, #93362 Make torch.testing functions overrideable with torch_function?",78483caf3e1b9303625aefcc514f6743b2659e38c3e543677cfd4e70dd2d3461
issue-collapse:8657ace2bfabcf67,module: __torch_dispatch__ stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"65423, 68622, 69530, 72179, 73150, 73930, 74991, 76844, 77118, 77265, 77544, 78845, 79563, 81240",,,,"module: __torch_dispatch__, module: __torch_function__, module: autograd, module: bootcamp, module: functionalization, module: nn, module: python frontend","Stale, actionable, better-engineering, feature, module: __torch_dispatch__, module: __torch_function__, module: autograd, module: bootcamp, module: functionalization, module: nn, module: python frontend, tensor subclass, triaged",,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#65423 Enable is_contiguous slowpath for torch_dispatch, #68622 torch.{tensor, Tensor, LongTensor, ...} isn't captured under `enable_python_mode()`, #69530 Extend composite compliant testing to backward formulas, #72179 Mechanism for Tensor subclasses to ""disable autograd"", #73150 How do we handle metadata-modifying in-place operators (like `squeeze_`) with `__torch_dispatch__`?, #73930 User raised TypeError from __torch_dispatch__ gets turned into generic ""unsupported operand type"" message",0c39b6341404b984e33c26d964d86f9063e94009ee09f3b62ac57fc4462fcaaf
issue-collapse:8943a8f64a16130b,module: __torch_dispatch__ stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"75166, 81565, 81682, 82239, 82479, 85105, 86128, 91437, 108983, 114652, 121917, 131753, 131772, 163013",,,,"module: __torch_dispatch__, module: autograd, module: complex, module: dispatch, module: functionalization, module: functorch, module: higher order operators, module: lazy, module: library, module: primtorch, module: pt2-dispatcher, oncall: pt2","ZeroTensor, actionable, feature, module: __torch_dispatch__, module: autograd, module: complex, module: dispatch, module: functionalization, module: functorch, module: higher order operators, module: lazy, module: library, module: primTorch, module: pt2-dispatcher, oncall: pt2, tensor subclass, triaged",,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#75166 Disable Python mode (torch dispatch mode) inside of mode-induced __torch_dispatch__call, #81565 __getitem__ is returned as an OverloadPacket instead of an OpOverload in __torch_dispatch__, #81682 Python operator registration API for subclasses, #82239 primtorch refs should be composite compliant, #82479 Cloning conjugate tensor in torch_dispatch context produces non equality., #85105 Python dispatch for PyOps needs to respect tensor subclasses",7b833090aad97129166b98bd36da472c27ffbc8cb981514456bc3be78d64ca7c
issue-collapse:f284fb594084c593,general stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"40087, 97408, 97504, 97653, 97961, 99790, 100617, 100785, 100807, 101319, 101536, 102741, 102921, 103382",,,,,"better-on-discuss-forum, enhancement, inductor_pattern_match, intel, low priority, topic: build, topic: docs, topic: new features, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#40087 Make Scaling in BatchNorm optional, #97408 Further memcopy improvement at FX body level, #97504 Burn benchmark suites into CI docker image. Not only this saves test time, but also it will get rid of occasional model installation failures. (@weiwangmeta ), #97653 transposed 2d copy bfloat16 support, #97961 [CI/Infra] Record keeping: runner shutdown spike, #99790 Cannot compile torch 1.10 in CentOS 7.3",80b2ac6d6c799903b0f5e9fd1aafff7b08265ac8596b5366756b581dd055f6fe
issue-collapse:055842f6c9756094,general stale/low urgency  batch 8,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"27815, 33298, 84193, 92838, 99025, 103241, 103530, 103802, 104282, 108888, 110786, 111210, 112318, 114044",,,,,"ZeroTensor, enhancement, fx, hacktoberfest, release notes: devx, topic: devs, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#27815 Improve the error message when trying to install in a 32-bit Python environment, #33298 [feature request] torch.expand to not require unsqueeze and match -1 to existing dimensions if tensor.shape.count(-1) == tensor.ndim, #84193 Attach execution time to each node in an fx trace, #92838 Enable OnDemand for Open Source CI, #99025 Is there a way to get the full call stack of pytorch from python to C/C++?, #103241 Image Processing with Pytorch",52949f142b270cee2058a712bac24ad15f682c3c028728a8a40ffdbfc8aa35c1
issue-collapse:8f1a4a9db7a0a814,general stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"73665, 100419, 102360, 112608, 117967, 118970, 119377, 119387, 119417, 119418, 120002, 120037, 120104, 120541",,,,,"enhancement, export-triaged, lazy, tensor subclass, topic: binaries, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#73665 Nice way to override string representation of tensor subclasses, #100419 Subgraph rewriter: Unable to match constant args, #102360 PyTorch hangs at import when used together with TensorFlow, #112608 `grad_fn` is not defined in `__torch_dispatch__` arguments when lazy device is used, #117967 Manually modified parameter during trainning, gradients are not being correctly updated. Help, #118970 When concatenating empty variable with non-empty variable, I get CUDA error: illegal memory access encounter",fe163af533342dbaa2b3db3357ba59bde94eff5fbd992c4e0f8f79bccea43af4
issue-collapse:144e75d59da56b62,general stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"33520, 65140, 70540, 77537, 77880, 104045, 120534, 121737, 122552, 122756, 123120, 123466, 124998, 127555",,,,,"enhancement, function request, lazy, topic: docs, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#33520 Reverse Cumulative Sum, #65140 profiler_legacy `_parse_legacy_records` deduplication logic has false-positive., #70540 cumcount / cumulative count, #77537 batch Kronecker product, #77880 Make the appropriate backend `DimensionNode` visible to LTC core, #104045 ""TorchDynamo Deeper Dive"" doc is missing some part",06bd4c7519651a0f99c8376aca9fce290277229c769e05fd931097826bafbaff
issue-collapse:29950c0369955634,general stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"4073, 10396, 68258, 101191, 111806, 128960, 129447, 129625, 130238, 130636, 130792, 131274, 131962, 134201",,,,,"better-engineering, dynamo-torch-function, enhancement, release notes: python_frontend, security, topic: docs, topic: security, triaged, upstream triton",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#4073 Feature request: Correlation module, #10396 [feature request] Provide a way to redirect shared memory prefix ""/torch_"", #68258 Option to configure TorchElastic multiprocessing/api.py PContext.close() timeout, #101191 custom_op API follow-ups, #111806 Revisit security implications of #31875, #128960 [Dynamo x torch_function] magic methods and methods with regular Tensors don't seem to work",47daa528e098f3b56caa9466f35199dd454b2b736676cd4a755fffa268bba82c
issue-collapse:decbeffb84a4e9ad,general stale/low urgency  batch 46,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6552, 6846, 6901, 7227, 7480, 7493, 7729, 7897, 7945, 7986, 8128, 8186, 8328, 8546",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6552 [Caffe2] - Run on different Cuda streams, #6846 Jetson TX1 Caffe2 installation fails on latest JetPack release, #6901 [Caffe2] [feature request] Dilations in grouped convolutons, #7227 [Caffe2]Install problem, #7480 [caffe2] How to use multiple CPUs?, #7493 Build from source with anaconda [caffe2]",844cd3923adfe1e07d3cc3c8b8fa805d39914d77efd47c6a9f257caaf76cced7
issue-collapse:6de78e4e3fa32178,general stale/low urgency  batch 45,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7781, 8012, 8479, 8548, 8607, 8648, 8650, 8730, 8902, 8979, 8984, 9003, 9068, 9098",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7781 [Caffe2] Fail to build after upgrading to cuda 9.2, #8012 [Caffe2] VideoInput/LMDB Reader to Standalone Predictor Question, #8479 An error occurred while creating a new notebook., #8548 [caffe2] [feature request]Doese caffe2 support conv_nd with group?, #8607 [Caffe2] Wrong prediction with simple FF, #8648 Cannot allocate memory Error from operator",18772b185a940ce704827ebe746a4e2b9e3dabee07140921357f817b0e38bc0f
issue-collapse:d131f6e26fb4bad9,general stale/low urgency  batch 44,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7571, 9024, 9028, 9150, 9166, 9174, 9207, 9218, 9449, 9881, 9883, 9921, 9945, 10088",,,,,"awaiting response (this tag is deprecated), caffe2",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7571 the latest version of caffe2, Which file is the function loadToNCHW(), #9024 cmake error:Could NOT find IDEEP (missing: /usr/local/include), #9028 [caffe2]build problem, can not find caffe2_pybind11_state_hip, #9150 [caffe2] Drop connections, #9166 nvcc fatal : A single input file is required for a non-link phase when an outputfile is specified, #9174 [Caffe2] build_ios.sh => 'is only available on iOS 11 or newer'",f740841d4e6e66eecec75d6012cc92d4e17ae1b917db9e396c5e1e8bf91c05e4
issue-collapse:3fc5f73b54286c70,general stale/low urgency  batch 43,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8982, 9016, 9761, 10070, 10161, 10170, 10176, 10215, 10256, 10314, 10342, 10460, 10461, 10515",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8982 parallelize_bmuf_distributed_test intermittently hangs, #9016 SimpleMetaNetDefInitializer intermittently hangs, #9761 [caffe2] Questions about conv_transpose && ConvTranspose, #10070 UnicodeDecodeError while loading caffe2 model, #10161 error when building caffe2 from source with gcc 7 & cuda gcc 5.5, #10170 build error",49f1f213e82e99aa85765ceac4418c02a4b2e5a15eb1218148e805ff10691dc3
issue-collapse:1de25b9dbea483de,general stale/low urgency  batch 42,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7047, 7789, 7946, 10341, 10582, 10714, 10738, 10808, 10902, 10984, 10985, 10991, 11119, 11197",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7047 ""Undefined symbols for architecture arm64"" when linking libcaffe2.a for iOS, #7789 detectron net create error, #7946 how to use Softmax when do segmentation, #10341 [Caffe2] Which document should we follow to install caffe2?, #10582 [Caffe2] Unable to use MPI rendezvous in Caffe2, #10714 [feature request][caffe2] extend FC/FCTranspose op to handle 2d bias.",c8f29d5af2f13433a56bb4e28560bb2925ac852089c9017b78767f142e839715
issue-collapse:8269209c3df0b29c,general stale/low urgency  batch 41,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9708, 9975, 10875, 10990, 11058, 11180, 11267, 11268, 11363, 11433, 11678, 11885, 11890, 11976",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9708 [caffe2]mpirun multi-node multi GPU in Distributed mode ,run resnet50_trainer.py get RuntimeError, #9975 python caffe2/python/operator_test/activation_ops_test.py Segmentation fault (core dumped), #10875 how can i run two model in a simple project?, #10990 [Caffe2] [Question] How to disable the bias parameter?, #11058 A bug in roi_align.cc? [Caffe2], #11180 A Caffe2 Implementation of Pose Estimation",5b77969aafa11889835174b9824a2db1dbe4094d360fe2276dc96f2b9add837d
issue-collapse:de7ca3f799103454,general stale/low urgency  batch 40,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6660, 8944, 10675, 11850, 11996, 12010, 12045, 12086, 12134, 12136, 12181, 12257, 12280, 12297",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6660 [caffe2] GANs, #8944 [caffe2] UnicodeDecodeError when running LeNet, #10675 GenerateProposals CUDA implementation, #11850 [Caffe2/Bug] Cannot enable MKL-DNN, #11996 Test the Caffe2 Installation with GPU erro, #12010 Caffe2 compiled with MKLDNN doesn't have device_type = MKLDNN",d3c67871c866d9378dab0a30cbe632ec10caf134bd8c2b2b11ceeb181c97104c
issue-collapse:6ca12fe37b25667e,general stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8937, 9407, 37443, 50258, 55228, 57117, 57230, 78954, 96265, 129228, 138197, 139520, 140640, 140711",,,,,"actionable, better-engineering, caffe2, enhancement, function request, security, topic: docs, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8937 [caffe2] Is there any method to implement learning rate scheduler?, #9407 [feature request] Make `torch.gather` broadcastable., #37443 torch.clamp_max clamp_min shouldn't be there, #50258 More general MultiHeadAttention and Transformer modules, #55228 Requesting a feature to calculate attention weights., #57117 MultiheadAttention.out_proj.weight is not explicitly initialized",6b36616e47e493d22b93c7eba8d321b89c2c19891be8244847d8f8638700555c
issue-collapse:340477368a4ac4e9,general stale/low urgency  batch 39,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6618, 7912, 12501, 12576, 12640, 12642, 12646, 12658, 12755, 12763, 12808, 12879, 13041, 13120",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6618 [Caffe2] CUDNN_STATUS_BAD_PARAM Error with the LRN layer while trying to run the code using CUDA. The training works fine on CPU, #7912 [Caffe2] Operators of Detectron module not registered/compiled when built on windows, #12501 C++: Calling Workspace::RunNet for a prediction on a different thread each time causes a GPU memory leak, #12576 [feature request] `ignore_label` argument in Caffe2 `SoftmaxWithLoss`, #12640 how to store a bounding box in Tensor?, #12642 fail to visualize caffe2 model",88e6712a0f7eac02c53d602abf66139a7f4a1fe7a5a317a752897aaf6cb28a26
issue-collapse:4fd31e74a2eed422,general stale/low urgency  batch 38,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"11636, 11791, 12931, 13208, 13297, 13373, 13405, 13498, 13655, 13667, 13671, 13684, 13715, 14030",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#11636 undefined reference to caffe2, #11791 [caffe2]Can i build caffe2 library only for cpu inference purpose and reduce the binary size?, #12931 USE_OPENMP=OFF is ignored [Caffe2], #13208 [caffe2] How to use the operators that are not included in brew through the python API?, #13297 [Caffe2]How to accelerate group convolution in Caffe2?, #13373 cannot run mobilenet_v2_quantized on pytorch/caffe2",7f3bf59bf5125d23e9cfac46b0eea6ef4bd393b3d2fef051928bf5f6816f17bb
issue-collapse:4c96513a318cbfd5,general stale/low urgency  batch 37,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8471, 10940, 12515, 13226, 14185, 14222, 14262, 14312, 14320, 14328, 14337, 14348, 14366, 14380",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8471 Does the caffe2 have the depthwise separable convolution oprater ?, #10940 [Caffe2] Model work Alright in Python but Failed in C++, #12515 how to use mask-rcnn in caffe2 c++ gpu, #13226 [Caffe2] cmake3 detection error?, #14185 [Caffe2] Non-spatial batchnorm optimization, #14222 Caffe2->ONNX conversion issue due to spatial-bn",c20d4c5361b34d18a0e0495e322bd9a04d43698651b7574982e1b86dc21ace56
issue-collapse:eb7e4d22005cac58,general stale/low urgency  batch 36,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7078, 11861, 12913, 13493, 14408, 14431, 14508, 14538, 14652, 14685, 14726, 14788, 14796, 14970",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7078 Fail to import Caffe2Backend, #11861 [Caffe2 installation problem], #12913 ubuntu16.04 build from source caffe2, #13493 How to do inference with float16 (or half) in caffe2?, #14408 DeviceOption::set_device_type() doesn't accept caffe2::CPU anymore, #14431 BUILD_CAFFE2_OPS=OFF is not tested in CI",ea22224877bf94de9478377e1cbf5accccb70c74bb2501357c1ce02ad148b68e
issue-collapse:c3f487ba66a4120e,general stale/low urgency  batch 35,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9531, 11735, 11865, 12007, 12322, 12530, 13772, 14799, 14925, 14963, 15009, 15249, 15313, 15608",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9531 [caffe2]train on GPU and test on cpu failed, #11735 Caffe2 installation: libcaffe2_gpu.so: undefined reference to `caffe2::ClipTransformRGB, #11865 [Caffe2] Errors occured when running Cpp Predictor, #12007 caffe2's installation is still problematic..., #12322 [Caffe2] Segmentation fault (core dumped) while import caffe2.python.core, #12530 [caffe2] Memory usage",1631ded314b443de3b84649f766025d4c459bafdf1497656767e18913d2295b2
issue-collapse:72aa9f0ffe265cd4,general stale/low urgency  batch 34,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6868, 9676, 13598, 14459, 14939, 14940, 14989, 15431, 15432, 15465, 15771, 15869, 16138, 16195",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6868 [Caffe2] Android NNApi integration bugs., #9676 [caffe2]running problem, No module named caffe2_pybind11_state_hip, #13598 caffe2: RuntimeError: [enforce fail at reshape_op.h:110] with Alexnet onnx test with cuda, #14459 [Caffe2] Error protos.protos_size() == OutputSize() when loading dataset created by regular Caffe (datum), #14939 caffe2/predictor missing from libtorch, #14940 caffe2::Predictor: Cannot find operator schema for GivenTensorFill. Will skip schema checking.",6cd42d92cf89e3fc6baaf312b93e2fbf16d619ba2bffd776bf354c91cf2ca21b
issue-collapse:43514a19c4d0d165,general stale/low urgency  batch 33,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"9704, 11354, 15290, 16254, 16280, 16385, 16436, 16494, 16590, 16666, 17437, 17632, 17634, 17712",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#9704 A newer protobuf should be used, so that cmake install will work when building for mobile, #11354 caffe2 failed to build from source, #15290 [caffe2] Installation problem on OSX, #16254 I hope Caffe2's python interface add DataLoader Module Similar to pytorch's DataSet and DataLoader, #16280 caffe2: cudaHostRegister() and mbind()-related test failures on ppc64le, #16385 Adding new module to caffe2",d42913428d828de69ee601c66e0740ca64068398cf747b7a5759edf63a671a81
issue-collapse:17f6803ee786080d,general stale/low urgency  batch 32,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6408, 8886, 14486, 14731, 15120, 16299, 17800, 17850, 17872, 17887, 17907, 17936, 18273, 18944",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6408 [caffe2] Prefetching blobs for memory optimization, #8886 [Caffe2] Dose caffe2 support the function similar to itersize of caffe, #14486 building bundled nccl fails in (caffe2, cuda 8, cudnn 7) CI environment, #14731 Caffe2 C++ runs single threaded, #15120 Different implementations of upsampleBilinear between pytorch and caffe2, #16299 caffe2::CudnnConvOp::RunOnDevice() fails on Squeezenet",531a44b2d6e7a9da1f716ebd2bf9038f4ee5d0f0f70119592e3d679a43f44b25
issue-collapse:5e5e6a933ed26e6e,general stale/low urgency  batch 31,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10232, 12401, 12509, 17495, 18000, 19107, 19244, 19416, 19708, 19930, 19960, 20268, 20776, 20777",,,,,"awaiting response (this tag is deprecated), caffe2, todo, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10232 [Caffe2] AVX not enabled for pytorch/caffe2, #12401 AttributeError: Method VideoInput is not a registered operator, #12509 The text design (color, type) makes it hard to read, #17495 RuntimeError: storage_.IsType<T>() ASSERT FAILED, #18000 Unsupported type of tensor c10::half when running resnet50_trainer.py in Caffe2, #19107 pytorch/caffe2/onnx/backend.cc:1668:57: error: invalid conversion from ‘google::protobuf::int32 {aka int}’ to ‘onnx_torch::TensorProto::DataType {aka onnx_torch::TensorProto_DataType}’ [-fpermissive]",0d9369312b34f14a6a68dfa800f9c2969b00fe72ceaf78487886e9bbca10d112
issue-collapse:549b969b4cc72d60,general stale/low urgency  batch 30,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13664, 13836, 14131, 18011, 18677, 20273, 20975, 21012, 21277, 21684, 21766, 22040, 22358, 22375",,,,,"caffe2, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13664 Caffe2 install, #13836 Pytorch-Caffe2 export: ""Arrays are not almost equal to 3 decimals"", #14131 running caffe2 float16 tensors results in aten runtime error, #18011 caffe2 Segmentation fault (core dumped), #18677 How to compile/install caffe2 with cuda 9.0?, #20273 Linker errors when linking statically (avx perfkernels) [Caffe2]",e0e76532bfb86d84079445541c23239a6d1320d1a0f5a573af4132d5680d5c0b
issue-collapse:3c4b5c44ead32dec,general stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"93454, 93455, 93456, 93463, 146889, 148044, 148113, 150531, 150861, 153041, 153967, 155703, 157721, 159228",,,,,"enhancement, tensor subclass, triaged, upstream triton",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#93454 MMDet 3.x cannot run successfully in inductor mode, #93455 Performance regression on interpolation in Kornia, #93456 Way to run accuracy minifier on only one particular subgraph, #93463 Traceable tensor subclasses cannot actually be used with AOTAutograd, #146889 How to customize a torch.Tensor() method to access the underlying data structure of a PyTorch tensor., #148044 Sweep through Potentially BC Breaking Commits in Triton",fda083bf35853a6a17182981d41d6ac4cfd19c09436d7576c50f93b65e3eaf9c
issue-collapse:ff77c07c92189cfb,general stale/low urgency  batch 29,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6402, 6794, 12855, 13224, 13304, 17458, 22609, 22859, 23364, 23429, 24079, 24353, 25112, 25880",,,,,"caffe2, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6402 [caffe2] How to freeze a layer?, #6794 [Caffe2] TensorProtosDBInput AttributeError, #12855 Model with Caffe2 runs much slower than it with pytorch in GPU mode !!!!, #13224 Caffe2: Causes error when using flag remove_legacy_pad while converting from caffe to caffe2, #13304 ASSERT FAILED at /opt/conda/conda-bld/pytorch-nightly_1539602533843/work/aten/src/ATen/core/blob.h:79, #17458 Overflow in fbgemm",b967d9ac217408c45074c7a76a7e71d8f9c5c2a04eedf270ba9adb5f4808621a
issue-collapse:a3d8be928bca59f0,general stale/low urgency  batch 28,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"8028, 9029, 9208, 9699, 11451, 11978, 12869, 14370, 14487, 17796, 27034, 28249, 28859, 30155",,,,,"better-engineering, caffe2, caffe2-op, enhancement, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#8028 Build error : mpi/mpi_gpu_test.cc.o: undefined reference to symbol '_ZN3MPI8Datatype4FreeEv', #9029 DataWorkersTest::testRNNInput timeout, #9208 [Caffe2] compiling error with gcc-6, #9699 [Build error] libcudnn.so: error adding symbols: File in wrong format, #11451 [caffe2] caffe2 openmp linking error with xcode 9.0(AppleClang 9.0) on mac, #11978 [Caffe2] Attempting to install Caffe2 in Google Colab",6629a650461ab10be5ed6c5494fa2cadd16bf4f5fd96a2ae0054a049730dbeaf
issue-collapse:703f6fd35b18d247,general stale/low urgency  batch 27,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6613, 24835, 26802, 28093, 30405, 30418, 31548, 31606, 32033, 32105, 32783, 33180, 33341, 33611",,,,,"better-engineering, caffe2, enhancement, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6613 [caffe2] benchmark performance for different operators, #24835 NetworkX's Version, #26802 Google Summer of Code, #28093 Move options logic in _like functions from parsing layer to implementations, #30405 Codegen refactoring master task, #30418 More detailed information about TensorType in error messages",dd973194786757adffe6b8e8fe876066827d088ec59e33a5d4286d6324571f62
issue-collapse:86aa3e14420be132,general stale/low urgency  batch 26,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2739, 6760, 32916, 33653, 33787, 34207, 34715, 35681, 35755, 35944, 35964, 36436, 36733, 36753",,,,,"better-engineering, caffe2, enhancement, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2739 Proposal: simplify overloaded Tensor function signatures, #6760 [Feature request] LayerNormLSTMCell and LayerNormLSTM, #32916 from torch._C import * (ImportError: DLL load failed), #33653 master build error, #33787 MSE Loss Implementation, #34207 Support vars/__dict__ on torch.return_types tuples e.g. topk",a6830059bccd94b3ec8027f8a7454ee349fc4f2e6082ffc32018c2394533d835
issue-collapse:c78240b0f9e15a8d,general stale/low urgency  batch 25,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3176, 8550, 27840, 35609, 36754, 37065, 37204, 37434, 37684, 38170, 38640, 39298, 39603, 39665",,,,,"awaiting response (this tag is deprecated), better-engineering, caffe2, enhancement, internals, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3176 Add safety checks in `index_add`/`scatter_add`, #8550 caffe2: Does caffe support the model running on different device (CPU and GPU) same time ?, #27840 caffe2 install VS2019 CUDA 10.1 lib\torch.lib : fatal error LNK1248: image size (10028FA9F) exceeds maximum allowable size (FFFFFFFF), #35609 Randomly error reports, #36754 Clearer guidance on when to define an operator as a method on torchbind'ed class versus standalone function, #37065 Deprecated mask fill mask type can causes pages and pages of repeated messages",a7943af7f751f4fb4e8abd1d89dc5a4ca8122d7aa0e5f43059ecc2199ba7e9b7
issue-collapse:bf115e73fd08a810,general stale/low urgency  batch 24,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5266, 13746, 16322, 18101, 33085, 39972, 40030, 40269, 41660, 41801, 41823, 41904, 43040, 43486",,,,,"better-engineering, caffe2, caffe2-op, enhancement, jit-backlog, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5266 [feature request] warnings for functions with unspecified dim arguments, #13746 [caffe2] test depthwise3x3_conv_op_test fails to run, #16322 We're binding a bunch of crap to 'torch' namespace which shouldn't be there, #18101 Android NNAPI support of Caffe2, #33085 not able to import * from fastai.vision in Google collab, #39972 Port old registration API to new one",2329118e0277a9e671cb934832a084d74594d39be16ea7a444d194da8939a24c
issue-collapse:0eb8f93ad82e5ee0,general stale/low urgency  batch 23,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"3823, 7546, 16402, 22447, 24313, 31679, 42173, 42597, 43815, 43964, 44185, 46400, 46787, 46869",,,,,"Merged, better-engineering, caffe2, enhancement, fb-exported, function request, open source, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#3823 Fuse bias to CuDNN convolution, #7546 [Caffe2] Build broken on macOS High Sierra: can't find sys headers in /usr/local/include., #16402 error with nccl when distributed training on caffe2, #22447 Weak Symbols Resolution Causes Segmentation Fault in External Libraries, #24313 Value_select to perform region-wise selection, #31679 broadcast randperm with dim specification",d5f73e6275329ac0e628ec5f733763e247af2eec3f55abb30faf859f90369d14
issue-collapse:c5ad0071178e0ee7,general stale/low urgency  batch 22,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"7745, 9133, 9375, 14864, 15378, 28218, 41232, 47954, 48929, 48998, 49166, 51282, 51410, 51451",,,,,"OSS contribution wanted, actionable, better-engineering, caffe2, enhancement, shadow review, small, todo, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#7745 [Caffe2] Align Element-Wise Ops Broadcasting to Numpy, #9133 [Caffe2] Error running net train when running resnet50, #9375 [Caffe2] Not handle attribute 'spatial' of ONNX operator 'BatchNormalization', #14864 [discussion] Recommend a different file extension for models (.PTH is a special extension for Python), #15378 Projective Transformation grid generator, #28218 index_sub, index_mul and index_div",25e6cf7a53ceee83c8603cb143081958cda2a4daf9ac204d85667f895d340206
issue-collapse:46410431ae495e15,general stale/low urgency  batch 21,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"10913, 11516, 11537, 40034, 52745, 52767, 53072, 53193, 53386, 53442, 53463, 54192, 54362, 54364",,,,,"NNC, better-engineering, caffe2, enhancement, jit-backlog, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#10913 Cannot find operator schema for 'ATen' Caffe2 Ios, #11516 at::Device makes it very easy to write buggy code, #11537 [CLEANUP] Context functions should return TypeExtendedInterface, not Type, #40034 Doing optimizing compilations in a separate thread, #52745 Sort torch.topk() output following input index order, #52767 [nnc] Grouped convolutions",3b5499f319259e67fbec49d74275db6ed302246a628d99a987a3998b1ea00c06
issue-collapse:44611781f2e3b140,general stale/low urgency  batch 20,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"12690, 27336, 52694, 54340, 54366, 54566, 54757, 54800, 54802, 54897, 55696, 55933, 55941, 56285",,,,,"NNC, better-engineering, caffe2, complex_autograd, enhancement, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#12690 How can I build caffe2_gtest_main under pytorch/caffe2/test/ folder?, #27336 Allow one inferred axis in torch.split, #52694 [TensorExpr] Loop Transformations: `compute_inline` should use `Let` for inlined expressions, #54340 RuntimeError: polar does not support automatic differentiation for outputs with complex dtype., #54366 [package] Heuristics for extern-ing common dependencies, #54566 Support docstring definition in torch/library.h operator definition",8e5de148547ed50ca4df55e5505f4b35d914fd2e5932d59cd8c99499319a1b41
issue-collapse:c46412eecaf70efa,general stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"93464, 93465, 93486, 93488, 93580, 93592, 93596, 93602, 93604, 93609, 99999, 106885, 146945, 153745",,,,,"actionable, low priority, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#93464 wav2vec2 model: error trying to do inference, #93465 Graph breaks with HuggingFace Stable Diffusion, #93486 A Simple Function Causing Graph Break, #93488 If minifier test fails, stderr/stdout of subprocess calls is not printed, #93580 AssertionError: Unknown expression s2, #93592 Turning on minifier causes bug to go away (on DebertaForMaskedLM)",4b4c6a1db6aa50429eaab676ed8797f087e5f8e4e41aee53c154b26c01f99e62
issue-collapse:b390185cdf62ae35,general stale/low urgency  batch 19,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"13034, 13508, 56100, 56357, 56634, 56809, 57005, 57359, 57380, 57415, 57495, 57691, 57742, 58091",,,,,"NNC, better-engineering, caffe2, enhancement, inference mode, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#13034 [caffe2]How can I export init_net.pb and predict_net.pb files on my own?, #13508 OpenCL support for smartphones, #56100 [package] implicit externing can miss cases where the stdlib depends on further stdlib modules, #56357 [NNC] All NNC APIs should return bool flag, #56634 [package] Module name reported in error message does not always match what is needed to extern/mock it, #56809 Autograd engine worker thread initialization fails with python 3.9 debug build",a3c6e7dce6e87befcd095a2fb2680a9edffae2c79a9bd6e32c82f32ab7a85e89
issue-collapse:cba2b1d1da3b2ca5,general stale/low urgency  batch 18,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"14702, 16556, 44489, 53952, 58016, 58167, 58617, 58791, 59575, 60083, 60102, 60802, 60991, 61843",,,,,"NNC, TSRootCause:TypeRefinement, better-engineering, caffe2, ezyang's list, op-bench, pipeline parallelism, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#14702 pytorch_doc_push is racing with itself, #16556 KeyError: 'No translator registered for layer: name: ""res3b_relu_norm""\ntype: ""Normalize""\nbottom: ""res3b""\ntop: ""res3b_relu_norm""\nnorm_param, #44489 for_each doesn't support integer input to float output type promotion, #53952 [Pipe] supporting None and non-Tensors in forward's input/output, #58016 Eliminate potential double device check, #58167 Allow `Union` as a `Dict` key",62311cd77304daf68a481afc4ce1d0ecb8782d81509b35d474133b6be025b08a
issue-collapse:1b5b5c7e50701b65,general stale/low urgency  batch 17,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"6505, 7020, 53462, 54630, 56707, 57404, 57987, 58930, 60924, 62699, 62742, 64845, 65494, 67377",,,,,"better-engineering, caffe2, enhancement, low priority, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#6505 [Caffe2] Onnx to Caffe2 Blob error, #7020 [Caffe2] LT and GT cannot be exported to ONNX, #53462 Second order derivatives for F.grid_sample(), #54630 torch-tensor-repr gdb macro doesn't seem to work in opt mode, #56707 [TorchScript JIT] `dtype` propagation pass (meta-issue), #57404 [lint] Make sure that all CMake options have a corresponding output in the summary",75addddff17464f097d7678c1cd5cb47b81e11c55dd0602f81c928549870bfcc
issue-collapse:94c6f0dea155ad2d,general stale/low urgency  batch 16,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"17538, 42919, 60999, 61384, 62673, 64124, 67751, 68474, 68475, 68498, 69676, 69893, 69932, 70413",,,,,"OSS contribution wanted, Stale, better-engineering, enhancement, jit-backlog, lazy, shadow review, tensor subclass, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#17538 ONNX->TensorRT parser library duplicated, #42919 Legacy Python2 and early Python3 leftovers, #60999 Merge seemethere/upload-artifact-s3 and driazati/upload-artifact-s3, #61384 Merge fork of upload-artifact-s3 to a single commit, #62673 Make printing to stdout / stderr in tests in CI an error, #64124 ciflow tracking meta issues",5a336dc3afe1342ec6c2f5980ffefc2cb324c6a801e4f8474ae62add86a3ca2f
issue-collapse:1c5dd24988c0bc5d,general stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"62011, 62973, 72448, 74248, 74300, 74485, 74514, 75798, 75925, 75986, 76040, 76047, 76323, 76326",,,,,"NNC, better-engineering, enhancement, release notes: python_frontend, sharded_tensor, tensor subclass, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#62011 Enable clang-tidy on all of master, #62973 Audit use of `C10_UNUSED`, #72448 Use Module `__getattr__` in `torch.ops` and friends, #74248 dir() on torch.ops.aten doesn't work, #74300 repeat_interleave is not meta friendly; not transform friendly either, #74485 dir(torch._VF) doesn't work",ee361f3d27dc642ffbf3e9170bf2bd2958963109e384ffb115c3707dce7ca39a
issue-collapse:8b1a7027ca8f6912,general stale/low urgency  batch 14,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"76327, 76331, 76441, 76451, 77894, 77899, 77946, 77967, 77981, 78507, 79013, 79030, 79031, 79050",,,,,"enhancement, inference mode, lazy, release notes: python_frontend, sharded_tensor, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#76327 [checkpoint] Add extension points to avoid the default serialization behavior, #76331 [checkpoint] Avoid loading whole tensor when resharding, #76441 [checkpoint] Switch away from pickle-base serialization, #76451 [checkpoint] Extension hooks to support logging and telemetry, #77894 Floating point exception in channel_shuffle, #77899 Floating point exception in native_channel_shuffle",ac193628b94827fc80ed92ba483f58a73e62d05a74721fc510975c33ad074371
issue-collapse:673358102cec4e5a,general stale/low urgency  batch 13,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"25091, 79032, 79052, 79053, 79118, 79137, 79336, 79349, 79395, 79407, 79425, 79453, 79469, 79592",,,,,"enhancement, lazy, topic: docs, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#25091 JavaScript (Web Assembly) target for trained models, #79032 Add a test that shows that lazy_ir reuse breaks SizeNodes, #79052 Mirror and implement `SymbolicIntNode` API for `SymInt` so we can trace in C++, #79053 [MetaIssue] Propagating SymInts through Autograd, #79118 About the source code of, #79137 [AUTOGRAD] codegen to use sym_sizes for ops w/ symint overloads in derivative formulas",423ca0639f46584ced2e2b0b373b56161ed5b144a659ab0d105382d3f16b8ce8
issue-collapse:fc79e0a29b36d281,general stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"39820, 53438, 76656, 79138, 80985, 81932, 81935, 81937, 81938, 81943, 82197, 82357, 82751, 82764",,,,,"better-engineering, has workaround, lazy, pipeline parallelism, release notes: nested tensor, tensor subclass, topic: not user facing, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#39820 I would like to install pytorch-nightly from requirements.txt, #53438 Pipe should automatically build appropriate partitions., #76656 Clarify dependency on NumPy (related to maskedtensor?), #79138 [AUTOGRAD] support implicit reductions with SymInts in autograd., #80985 Nested tensor: Support Noncontiguous Buffer, #81932 Make sure we always redispatch through a dispatcher for all SymInt ops",6db060c9e80cd528c87d0c6aec719b0025356c95f7f774f2e7dd15b8ff04d6d8
issue-collapse:4156c7fc9d9f85cd,general stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"83024, 83764, 83817, 83851, 84346, 84597, 84691, 85058, 85082, 85625, 90905, 91624, 91678, 91759",,,,,"NNC, core issue, enhancement, fx, topic: build, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#83024 New PR template suggests a pattern that does not close PR, #83764 Missing header file, #83817 [feature request] Add new device type works on CPU, #83851 torch fx cannot trace assert for some cases, #84346 fix ATen tests that do not compile, #84597 Accept SymInts and SymFloats For Scalar Inputs",eb9633675e21151188e1367feff7981540f5d57be6c5aa49e96e4f8a93049470
issue-collapse:b715a890e5f64eb0,general stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"52668, 90998, 91716, 92075, 92781, 92811, 92828, 92967, 93444, 93446, 93458, 93579, 95779, 95917",,,,,"complex_autograd, enhancement, triaged",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#52668 Independent axis support for torch.trace, #90998 Crash in `index_select` with singleton `self`, non-singleton `index`, #91716 linspace (and arange) behaves differently on GPU and CPU, #92075 Hessian produces wrong results, but works if I add a perturbation, #92781 Floating point exception when running torch.nn.AdaptiveMaxPool3d, #92811 [complex] Jacobian of a non-holomorphic complex valued function",4454b653661b16df47c56374882df5558ed280a79dd6ed921863a142b6486a51
issue-collapse:100b4e6b4e9fd449,general stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"2569, 15829, 109098, 152134, 156745, 159932, 160122, 164089, 165042, 165992, 169318, 169955, 170674, 171100",,,,,"ZeroTensor, better-engineering, function request, release tracker, security, tensor subclass, topic: docs, triaged, upstream triton",,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#2569 Add a [strong] hash function for tensor data, #15829 computing entropy of a tensor, #109098 Can't initializa NVML, #152134 [RFC] Proposed Changes to Feature Tracking & Classification for PyTorch Releases starting Release 2.8, #156745 [v.2.8.0] Release Tracker, #159932 Cannot view dtype of tensor which is contiguous due to numel=1 but has non-standard striding",9391063f863b3196cef8c5a417341c9270d698c30cc9eb71c65b633685d6dd40
issue-collapse:cd655552fd3a971d,feature stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1641, 1861, 7250, 11612, 12764, 20613, 26109, 32088, 32219, 32452, 33390, 34086, 37300, 38737",,,,,"feature, low priority, shadow review, todo, triaged",,Handle this feature stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1641 [Feature request] In-place 'max' method for Tensor, #1861 Factorized Output Layer, #7250 feature request: support for new/future hardware accelerators, #11612 [feature request] Triangular Matrix Representation, #12764 [feature request] Operator Overloading, #20613 groupby function",37882fe1106845c0c2798a50e9c6fb50d1579252f87b0a2bbdf293b25147ffa7
issue-collapse:1cc854226cbead9f,feature stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"5385, 9417, 38550, 38684, 42440, 45540, 48366, 49076, 49110, 51440, 52553, 52673, 55274, 59386",,,,,"feature, function request, has workaround, shadow review, todo, triaged",,Handle this feature stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#5385 Handle python_arg_parser dtype constants better, #9417 [feature request] allow forward_pre_hook to ""preprocess"" input, #38550 Returning a tensor instead of a list in split and chunk, #38684 Add Plackett-Luce distribution, #42440 [feature request].Time Series Algorithms, #45540 Why does NLLLoss and CrossEntropyLoss require type long?",aa4e582ad8eee3662aef6261479cbf86fa4eeac99d4e1f0fdd6560cdb279bbbf
issue-collapse:6743f7940f5fc76a,feature stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"630, 711, 1442, 9624, 9898, 17268, 51456, 65002, 65950, 68967, 69654, 74539, 77736, 84721",,,,,"Stale, better-engineering, feature, triaged",,Handle this feature stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#630 Add Peephole connections for LSTMs?, #711 Feature Request: Easier to extend base RNN implementation, #1442 Optimizers can't be moved to a different GPU, #9624 Request: Pycuda interoperability, #9898 [Feature Request] Checkpoint manager, #17268 Generic object to tensor dispatching",620e59a8b308f71a4e6a4a62a037ff4382b33911306b67a23322c344a366b41d
issue-collapse:f752f02f2ae38dd3,feature stale/low urgency  batch 1,batch action queue,stale/low urgency,2,medium,bucket_component_queue,14,0,0,0,"1642, 33859, 39395, 41164, 44255, 64932, 70529, 72737, 73646, 74669, 87350, 88880, 93719, 129330",,,,,"feature, function request, low priority, matrix multiplication, triaged",,Handle this feature stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 14 open issues; raw issue rows remain available for audit.,"#1642 feature request: On-the-fly Operation Batching in Dynamic Computation Graphs, #33859 Possibility to support int4 data type, #39395 Morphological operations, #41164 Pytorch.js would be nice!, #44255 Allow to use system zstd, #64932 [RFC] TorchStore - A Shared-Memory Tensor Store",d52b487f17c7c0191bb7daed10252f77fc6333be809823649670db23b609c2d1
issue-collapse:df576fc0578f15e0,module: sorting and selection stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"11072, 19150, 29398, 32305, 57301, 60825, 67597, 68610, 80296, 88142, 88227, 108642, 140396",,,,"module: sorting and selection, module: tests","enhancement, feature, function request, has workaround, module: sorting and selection, module: tests, todo, triaged",,Handle this module: sorting and selection stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#11072 [...] operator for masked select does not broadcast anymore, #19150 Improve unit test coverage of torch.unique, #29398 [feature request] torch.kthvalue to support a new argument largest, #32305 [FR] histc should (optionally) return a long tensor, #57301 Avoid code repeat in create sample inputs for sort/msort, #60825 Where OpInfo doesn't handle cases where one of the inputs is a scalar",8396294e46aa347df1bc90c0d5ff0a152ef7ee5113bf2f213ef367738a82a97a
issue-collapse:e1c1e4d026393613,module: pooling stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"1468, 9570, 14907, 17169, 33086, 39757, 39959, 46929, 51038, 51622, 54499, 73841, 107412",,,,module: pooling,"actionable, better-engineering, enhancement, feature, function request, hackamonth, module: pooling, todo, triaged",,Handle this module: pooling stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#1468 avg_pool functions hold input for backward, #9570 [feature request] More options for Fractional Max Pooling, #14907 Missing dilation from several pooling modules (AvgPool), #17169 Global Second Order Pooling, #33086 F.max_pool*d/F.min_pool*d should support integer dtypes and bool tensors, #39757 max_unpool2d and max_unpool3d cpp signature should be similar",598603f88c57acf232486b2f6e0bbf2f6bb91674a6730794096ce51f7902950b
issue-collapse:d4ad196ee79165dc,module: onnx stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"39245, 52230, 61167, 64477, 65657, 65698, 74034, 75625, 89634, 89675, 94675, 99770, 111186",,,,module: onnx,"module: onnx, onnx-needs-info, onnx-triaged, triaged",,Handle this module: onnx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#39245 pytorch-crf model to onnx conversion, #52230 Removing mutation on block input should print any meaning message., #61167 Torch .pth model export to ONNX returned empty graph, #64477 torch.onnx.export crash while export quantized model, #65657 Transpose channels when exporting to ONNX, #65698 torch.onnx.export error",98907b69ab02fe75060382e4fcd53ec9f35dfe2e5e12eb490622f7e681239dea
issue-collapse:c2e89d6ad31bb6dc,module: mta stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"106427, 106931, 108445, 111351, 117233, 117884, 117981, 124726, 128395, 129329, 133586, 139795, 162911",,,,"module: mta, module: nn, module: optimizer, module: performance, module: pt2 optimizer, module: python frontend, module: tests, oncall: pt2","actionable, better-engineering, enhancement, feature, module: mta, module: nn, module: optimizer, module: performance, module: pt2 optimizer, module: python frontend, module: tests, oncall: pt2, triaged",,Handle this module: mta stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#106427 `torch.nn.utils.clip_grad_norm_()` causes H2D sync with foreach ops., #106931 Other overloads of `_foreach_clamp`, #108445 _foreach_copy_ with scalar second arg, #111351 _foreach_copy_ supports fast copy between cpu and cuda devices., #117233 Submit a bug when trying to assign arg 'out=' in api torch._foreach_add_, #117884 Foreach Where support",16ea55fbb786cce62ce0bf44131ed1c000a8173d090fc615bff92d1b7420f274
issue-collapse:79b494a4b698f247,module: mkldnn stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"4186, 21120, 22415, 23657, 27497, 32826, 32909, 60459, 62788, 62933, 96140, 103272, 114188",,,,"module: mkldnn, module: onnx, module: performance, module: serialization, module: tests, oncall: jit, oncall: quantization","feature, module: mkldnn, module: onnx, module: performance, module: serialization, module: tests, oncall: jit, oncall: quantization, onnx-needs-info, onnx-triaged, triaged",,Handle this module: mkldnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#4186 Feature Request: CPU performance optimization with MKL-DNN, #21120 [RFC] Adding MKL-DNN Int8 functions to PyTorch/Aten/JIT backend, #22415 Add support for serializing Mkldnn Tensor, #23657 Accelerate PyTorch just-in-time compilation using MKL-DNN, #27497 TSAN failure related to mkldnn, #32826 MKLDNN doesnt work and is slower than normal cpu mode",4bca7d70ea2909c946d92b04f952cac86a958e15585487ec7bed0ec3c7741a46
issue-collapse:f43824a179e7ce44,module: masked operators stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"74301, 81985, 87076, 89416, 92350, 109009, 115422, 115624, 117287, 132608, 150186, 166401, 169191",,,,"module: masked operators, module: nestedtensor, module: nn, module: norms and normalization, module: padding, module: performance, module: sorting and selection, module: sparse","module: masked operators, module: nestedtensor, module: nn, module: norms and normalization, module: padding, module: performance, module: sorting and selection, module: sparse, triaged",,Handle this module: masked operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#74301 Error in lobpcg when using largest=False, #81985 nn.InstanceNorm and nn.GroupNorm are affected by padding, so they need to masking, #87076 Nandense layer for missing values, #89416 When the torch.masked_select operator passes in the same parameters, it behaves differently on CPU and GPU., #92350 TypeError: no implementation found for 'torch._ops.aten.max.default' on types that implement __torch_dispatch__: [<class 'torch.masked.maskedtensor.core.MaskedTensor'>], #109009 TypeError: mask must have dtype bool",cc3a6061720ec66e8304c3b3483bf95c86a9db709a99fc2719a2588a9dd58cae
issue-collapse:38d8944886cbfb97,module: logging stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"39503, 40687, 52913, 82789, 86265, 98872, 103173, 109850, 120079, 135135, 142493, 143274, 146913",,,,"module: logging, module: tests, oncall: export, oncall: pt2","better-engineering, enhancement, export-triaged, feature, low priority, module: logging, module: tests, no-scrub, oncall: export, oncall: pt2, triaged",,Handle this module: logging stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#39503 pytest suppresses stderr from Python startup by default, #40687 libtorch: macros in logging_is_not_google_glog.h have very common names like CHECK or LOG, #52913 Create a standalone python execution viewer like livepython, #82789 [LibTorch] the C++ api needs detailed error reports like pytorch, #86265 TORCH_WARN is executed just once per set of parameters, #98872 Expand component configurable logging system to C++",b001aae1cdd318ae978c3eefda5126dcc7e1e5b8eee8a7864ec519eb4a575b5b
issue-collapse:2aff493344f28266,module: int overflow stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"107539, 107541, 107543, 107544, 107545, 107546, 107548, 107550, 107552, 107554, 107555, 107556, 107702",,,,module: int overflow,"module: int overflow, triaged",,Handle this module: int overflow stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#107539 Integer overflow when running torch.nn.functional.max_unpool2d, #107541 Integer overflow when running torch.nn.MaxUnpool2d, #107543 Integer overflow when running torch.nn.AdaptiveAvgPool2d, #107544 Integer overflow when running torch.nn.ReplicationPad3d, #107545 Integer overflow when running torch.nn.functional.upsample, #107546 Integer overflow when running torch.nn.functional.upsample_bilinear",6289640bc581d898d158a5439cf65ba4999d4ac86c6d43dc79752a0780c2e25b
issue-collapse:f45c0b5d60284515,module: fx stale/low urgency  batch 9,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"43659, 43679, 44049, 44851, 45220, 45534, 45683, 46210, 47481, 48209, 48813, 49265, 49618",,,,"module: fx, oncall: jit","module: fx, oncall: jit, triaged",,Handle this module: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#43659 Exploring supporting some cases of `__bool__` on Proxy, #43679 [FX] List unpacking is broken, #44049 [FX] Can't symbolically trace callsites to ScriptFunction, #44851 fx: Python range function used on tensor shape is not symbolically traceable, #45220 FX should preserve type annotations and not break TorchScript-ability, #45534 [FX] Make tracer return a Graph, not a GraphModule",fef64b78ed4984c39f2ca442bcc71e5b722061967acab0f4311b182ddf226524
issue-collapse:7f4d06a8a6625c11,module: doc infra stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"19263, 20183, 27708, 27843, 38012, 44719, 48465, 49873, 49874, 54677, 55751, 57759, 62446",,,,"module: doc infra, module: docs, oncall: distributed","better-engineering, enhancement, feature, module: doc infra, module: docs, oncall: distributed, triaged",,Handle this module: doc infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#19263 GCP Base Image Wrong CUDA Version, #20183 Latex Errors when Compiling documentation to latexpdf, #27708 Documentation makefile should include torchvision, #27843 Be able to build torch.distributed documentation easier, #38012 Formatting of topic and sub-topic pages, #44719 404 Page not found - inference_api.md & management_api.md",e303ed0d3ae649753b2aebc91efcf25e305830537f72640c78a71f834fff9e9e
issue-collapse:a22b145ff150a4e2,module: cublas stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"21459, 21587, 30596, 34576, 38553, 42800, 43325, 53473, 58670, 75549, 83272, 91653, 91958",,,,"module: cublas, module: cuda, module: cudnn, module: performance","feature, matrix multiplication, module: cublas, module: cuda, module: cudnn, module: performance, triaged",,Handle this module: cublas stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#21459 RuntimeError: cublas runtime error, #21587 RuntimeError: cublas runtime error, #30596 Overlapping strides not supported by cublas, #34576 Significant speed difference between P100 and V100, #38553 When I use cuda(), wg = th.matmul(extra_obs, extra_obs.transpose(-2, -1)) take a mistake, #42800 RuntimeError: cublas runtime error : the GPU program failed to execute at /pytorch/aten/src/THC/THCBlas.cu:27 for spectral norm",88c29d5a93964b45ed2f077359fa82996dd7d3c319b36fa7c06189b9ff10d3a3
issue-collapse:692da56b4983b784,module: bazel stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"46631, 63972, 66565, 72880, 79351, 79355, 79851, 79853, 81167, 86194, 91751, 101470, 119279",,,,"module: bazel, module: bootcamp, module: build, module: build warnings, module: cuda, module: lazy, module: tests, module: third_party","lazy, low priority, module: bazel, module: bootcamp, module: build, module: build warnings, module: cuda, module: lazy, module: tests, module: third_party, triaged",,Handle this module: bazel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#46631 error in bazel build //..., #63972 bazel build warning: Artifact 'torch/csrc/api/include/torch/version.h' is duplicated, #66565 Bazel target all_tests improperly reports failures on CPU-only (non-CUDA) build, #72880 Bazel fails in an obscure way if submodules are not initialized, #79351 [bazel] ability to run gpu tests on gpu machines in RBE, #79355 [bazel] build spams warnings",de4cf58aa00193a125a8c9f0d7d427815d0b0cdf7acc8a3fcd8f5104e4e6b488
issue-collapse:607d0992bcae2678,module: batching stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,13,0,0,0,"10172, 15253, 16432, 21856, 22169, 33035, 40375, 40441, 54147, 54743, 75909, 101564, 146697",,,,"module: batching, module: cuda, module: data parallel, module: distance functions, module: error checking, module: functorch, module: linear algebra, module: nn, module: norms and normalization, module: numpy, module: vmap","enhancement, feature, function request, module: batching, module: cuda, module: data parallel, module: distance functions, module: error checking, module: functorch, module: linear algebra, module: nn, module: norms and normalization, module: numpy, module: vmap, triaged",,Handle this module: batching stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 13 open issues; raw issue rows remain available for audit.,"#10172 [Feature request] Batch eig/symeig functions (for small matrices, with CUDA), #15253 [Feature Request] cdist: pairwise distances between two sets of tensors with batch mode, #16432 A bug in parallel.data_parallel when module_kwargs is not None, #21856 Batch Normalization axis, #22169 [RFC] NestedTensor - 0.0.1, #33035 Support batch linear transformation",6a93b00f6fe40f77950f2b7c8f66898c1c7e803d4fb6969ed3feaad665d19fe8
issue-collapse:87c18ce6a9c1b822,module: viewing and reshaping stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"10978, 27276, 57095, 59617, 64709, 69433, 76891, 78759, 106173, 107026, 153542, 169870",,,,"module: viewing and reshaping, oncall: pt2","better-engineering, enhancement, function request, module: viewing and reshaping, oncall: pt2, triaged",,Handle this module: viewing and reshaping stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#10978 [feature request] padding for torch.cat, #27276 Add in-place view (view_), #57095 [FR] unflatten support empty shape if unflattened dim has size 1, #59617 torch.split() infer -1 entry from the other split sizes, #64709 Add out= resize warning to cat and maybe_native_stack, #69433 `torch.transpose` should raise an error when indexing 0 for 0 dimensional tensor.",023b3d11dc3251f8389acb47fb106cebf11da745df0cbecddd1321c13f48585e
issue-collapse:07955e329a4b2353,module: tensor creation stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"33040, 38197, 38470, 40419, 51112, 68703, 73832, 82230, 82242, 87861, 92920, 98673",,,,module: tensor creation,"enhancement, feature, function request, has workaround, module: tensor creation, triaged",,Handle this module: tensor creation stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#33040 [debatable] Better infer dtype in torch.as_tensor, #38197 Why no `torch.randperm_like`?, #38470 [Feature] Option to have zeros/ones/full output tensor with zero strides, #40419 Add a `like` argument to creation ops, #51112 torch.tensor(x) fails inconsistently, assumes 0 index exists in 'Series', #68703 Feature request: Extend `torch.eye` for creating a batch of eye matrices",d70d7f8add4b457549e1e86df945fbdcc21636c118d32131fbb7097d35b49782
issue-collapse:64126ac739162cff,module: power stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"2466, 2703, 12980, 39671, 40114, 41531, 41844, 54407, 57533, 74614, 109870, 115820",,,,"module: power, module: random, module: tests, module: third_party, module: xnnpack, oncall: distributed, oncall: jit","NNC, module: POWER, module: random, module: tests, module: third_party, module: xnnpack, oncall: distributed, oncall: jit, triaged",,Handle this module: power stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#2466 Autograd test failure on ppc64le, #2703 Have ppc64le docker images?, #12980 Installing pytorch from source on labs.cognitiveclass.ai, #39671 TestTorchDeviceTypeCPU.test_float_to_int_conversion_finite_cpu_uint8 is broken on PowerPC, #40114 ppc64le: test cpp_extensions/rng_extension.cpp failure (without altivec override), #41531 Test failure in test_shared_allgather_nccl: NCCL error in: ../torch/lib/c10d/ProcessGroupNCCL.cpp:537",890e8ebda8d05a21f616e7f062a4aa187786cf4148dba37bdd99e928f88de3f1
issue-collapse:4573aaf88992282d,module: norms and normalization stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"8035, 12189, 12760, 12812, 20551, 49431, 50090, 73161, 78656, 98075, 104853, 137172",,,,"module: norms and normalization, module: numerical-reproducibility, module: numerical-stability, module: performance","enhancement, module: norms and normalization, module: numerical-reproducibility, module: numerical-stability, module: performance, triaged",,Handle this module: norms and normalization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#8035 [PyTorch] EmbeddingBag comparison vs Embedding fails w/ small max_norm on CUDA, #12189 [Feature] Support Adaptive Max Gradient Norm / Clipping, #12760 [feature request] Spectral norm support in torch.norm or factor out Power Iteration from spectralnormalization in some other place and orthogonalization from PowerSGD hook, #12812 Pytorch BatchNorm2D Unstable, #20551 torch.norm produces incorrect results, #49431 clip_grad_norm_ performance regression",c3eab6e1d94ef38096ffc234630f797244617aaf39d90fc5d35176d9a463a8d5
issue-collapse:9a4cb249d147e93c,module: nn stale/low urgency  batch 27,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"15994, 18206, 19682, 20022, 20165, 20204, 22277, 22879, 25034, 26551, 28962, 29429",,,,"module: nn, module: numerical-reproducibility, oncall: distributed","enhancement, low priority, module: nn, module: numerical-reproducibility, oncall: distributed, small, triaged",,Handle this module: nn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#15994 More data type support for gather_map, #18206 Rename ignore_index to ignore_target in CrossEntropyLoss, #19682 View in Sequential - reasoned case for it, #20022 Precision of sparse float embeddings differs from dense embeddings on CPU, #20165 torch.nn.threshold cannot accept tensor as a threshold, #20204 SyncBatchNorm should support 2D input (B, C)",ae735a78acc296fef4b244877ae9d3b95d258a8303285dc207665795d7f854cc
issue-collapse:c5a6f3eb563ef423,module: interpolation stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"10482, 10604, 52845, 61182, 63016, 63217, 66366, 76487, 80929, 84630, 86204, 94816",,,,"module: interpolation, module: memory usage, module: nn, module: padding","actionable, enhancement, feature, function request, module: interpolation, module: memory usage, module: nn, module: padding, triaged",,Handle this module: interpolation stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#10482 Reduce code duplication in interpolate and make it more generic, #10604 Bilinear interpolation behavior inconsistent with TF, CoreML and Caffe, #52845 [feature request] F.interpolate_as, #61182 Feature Request: torch.functional.interpolate to quietly ignore ""align_corners"" when mode is set to ""nearest"", #63016 grid_sample should not require grid to be the same dimension as input., #63217 Torch.nn.functional.interpolate() function memory release problem",b3f6f1b8902661090a2657368799790c30e3bf009cff5d101fa5b04ab9622c7d
issue-collapse:940d7405a616c6d8,module: inductor stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"92259, 93482, 93639, 95021, 95423, 103652, 115966, 117251, 117976, 118570, 124145, 126708",,,,"module: inductor, oncall: cpu inductor, oncall: pt2","enhancement, feature, internal ramp-up task, low priority, module: inductor, oncall: cpu inductor, oncall: pt2, triaged",,Handle this module: inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#92259 [discussion] Fused MLPs, #93482 [inductor] Add more matmul configurations to `TORCHINDUCTOR_MAX_AUTOTUNE=1` mode, #93639 Explore TorchInductor optimization pass to reorder kernel bodies, #95021 Custom operations in inductor, #95423 PT2 Computes Multi Device Backward in a Single Thread, #103652 inductor: support horizontal reduction with vec_transpose to improve TIMM swin_base_patch4_window7_224 dynamic shape performance",9d7aa9c1b3dc85d1871117bc130db4d942d7f3686f5f9854311ea866becc1bcb
issue-collapse:e36110bf795c410d,module: forward ad stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"75503, 75652, 78961, 91280, 92072, 93982, 94085, 94208, 98164, 113368, 129176, 146901",,,,"module: forward ad, module: functorch, module: sdpa","enhancement, module: forward ad, module: functorch, module: sdpa, triaged",,Handle this module: forward ad stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#75503 make_dual errors out when primal is a Tensor and tangent is a subclass Tensor, #75652 Op segfaults with ForwardAD and Subclassed Tensor as Tangent, #78961 [forwardAD] torch.no_grad has no effect under forward_ad, #91280 custom Function that supports functorch jvp doesn't work with in-place, #92072 Implement forward AD with grid_sampler_2d, #93982 `linalg.lstsq` fails the gradient computation in forward mode",7876679df2d1f703076e3a66105ae1233ddf422ffe722a2940f20cf206e2f08a
issue-collapse:3c19e3feeb045204,module: floatx (formerly float8) stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"107087, 113831, 123761, 128257, 129803, 130359, 131196, 132624, 141749, 154028, 160651, 162458",,,,"module: floatx (formerly float8), module: mps, oncall: quantization","enhancement, feature, matrix multiplication, module: floatx (formerly float8), module: mps, oncall: quantization, triaged",,Handle this module: floatx (formerly float8) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#107087 RuntimeError with operations on torch.float8_e5m2 and torch.float_e4m3fn data types, #113831 Update _scaled_mm to support addmm with bias.size == out.size, #123761 Support for no-frills FP8 matmuls, #128257 can not create custom gradient when fp8 type is used, #129803 3D matrix support in _scaled_mm, #130359 FP8 Rowwise Scaling does not work on L4 GPU",795047a42f1519724aceec0e36282ba0c1c20c1c3998fab9c038bbc20c97ca22
issue-collapse:46f217a1f1dbe12a,module: deprecation stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"7834, 7900, 9381, 32580, 37863, 46187, 50332, 50352, 53124, 54588, 55951, 55964",,,,"module: deprecation, module: docs, module: loss, module: nn, module: numpy, module: optimizer, module: random, module: tensor creation, module: ux","module: deprecation, module: docs, module: loss, module: nn, module: numpy, module: optimizer, module: random, module: tensor creation, module: ux, triaged",,Handle this module: deprecation stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#7834 torch.Tensor.new() disappeared in 0.4 doc, #7900 Deprecate torch.Tensor, #9381 [doc] many losses still mention size_average in formula, #32580 SGD optimizer with deprecation warning, #37863 Deprecate type() and type_as() call, #46187 Better support for operators that return (named) tuples of tensors",1bb972456d7e991f9a6e654a075d9b8f851f7e77f0c65518dc774f2d1575b926
issue-collapse:cddddd36dc11f3a9,module: decompositions stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"94591, 106375, 113007, 113024, 113329, 117348, 118227, 118642, 134821, 137128, 145094, 147287",,,,"module: decompositions, module: dynamic shapes, module: dynamo, module: faketensor, module: fx, module: primtorch, module: pt2-dispatcher, oncall: pt2","better-engineering, module: decompositions, module: dynamic shapes, module: dynamo, module: fakeTensor, module: fx, module: primTorch, module: pt2-dispatcher, oncall: pt2, topic: fuzzer, triaged",,Handle this module: decompositions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#94591 Issue with `upsample_nearest2d` decomposition, #106375 nll_loss reference shouldn't be registered as a decomposition., #113007 `make_fx` failing for `_scaled_dot_product_flash_attention` decomposition, #113024 Upsample bilinear 2d decomposition does not match native implementation for uint8, #113329 Mark aten.normal as a core aten op, #117348 The bitwise_XX.Scalar_Tensor should be dispatched to core IR but is currently not",a78ded3ed347c02319768c4bac1eb55bb0c3ca78d430b41dd58788ab165b0509
issue-collapse:bb8aa27dd9837704,module: dataloader stale/low urgency  batch 10,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"11633, 16901, 20416, 21554, 22136, 22585, 23720, 29804, 30225, 31657, 31867, 32150",,,,"module: dataloader, module: error checking, module: multiprocessing, oncall: distributed","better-engineering, feature, module: dataloader, module: error checking, module: multiprocessing, oncall: distributed, triaged",,Handle this module: dataloader stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#11633 DataLoader: Could not wrapper a exception in threads, #16901 Issue with dataloader using pin_memory = True, #20416 Class based Sampler for Class Incremental/Continual Learning research, #21554 Pytorch hangs when dataloader multiprocessing workers are killed, #22136 [dataloader] SIGCHLD handler should poll the queue for exception first, #22585 [dataloader] Mysterious error when using spawn start_method",f94005fb4a2703fc1854576fbeb26393747430d4993023ec60094b1699711caf
issue-collapse:ee6c8b190af4a33f,module: data stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"11980, 13053, 46711, 49180, 49398, 51777, 64766, 65067, 66116, 68925, 71543, 74014",,,,"module: data, module: dataloader, module: distributions, module: docs, module: memory usage","enhancement, feature, module: data, module: dataloader, module: distributions, module: docs, module: memory usage, triaged",,Handle this module: data stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#11980 [Enhancement] Increase user-friendliness of dataset.random_split, #13053 Generalized Data Class, #46711 A random split function that return the datasets following specific target (label) distribution., #49180 torch.utils.data.DistributedSampler allow uneven inputs, #49398 Can we add try-except for list/slice indices even when auto_collation is True?, #51777 Analytical metric package",7bed4a7029cf323a649f39fb26bc274cb649b546419e891289fce4b8bb980912
issue-collapse:3c583854e2459624,module: cpp stale/low urgency  batch 12,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"12795, 15491, 19499, 20531, 21050, 21405, 21828, 21899, 21902, 22536, 23301, 25520",,,,"module: cpp, module: custom-operators, module: docs, module: nn, module: performance","enhancement, low priority, module: cpp, module: custom-operators, module: docs, module: nn, module: performance, triaged",,Handle this module: cpp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#12795 C++ frontend: how to debug nan gradients, #15491 LibTorch: include cmake files for all distributed headers, #19499 C++ API 'nn::Sequential' has inconsistent behavior with python conterpart, #20531 c++ torch::nn::Sequential increments count on name errors, #21050 output values not same and much slower than Python API, #21405 how libtorch can work with tensor data as same as pytorch",aea6e9fd9f6ca6e6564d2b3306282786c29de6b19513972e47dbb2de183d567d
issue-collapse:ef6cab37335fdf96,module: build stale/low urgency  batch 38,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"7035, 13118, 15309, 16843, 20576, 20753, 21135, 21399, 21777, 22370, 22406, 22580",,,,"module: build, module: build warnings, module: collect_env.py, module: cpp, module: cuda, module: dependency bug, module: docs, module: third_party","low priority, module: build, module: build warnings, module: collect_env.py, module: cpp, module: cuda, module: dependency bug, module: docs, module: third_party, triaged",,Handle this module: build stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#7035 [Compilation] how to disable caffe2?, #13118 warning: attribute namespace ""clang"" is unrecognized; High Sierra / Fedora compilation with clang results in spurious clang errors in nvcc, #15309 libtorch without cmake, #16843 Allow to build pytorch for a *specific* architecture, #20576 NVRTC_ERROR unknown when using self-built libtorch, #20753 cmake for Torch unusable in archlinux",ed347350323b3a8e9ae2ed7a5cca1bb7ec3a5ae1ea6fea63f254a0f9660e471e
issue-collapse:4f57efe9e4b97f19,module: abi stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,12,0,0,0,"17658, 28754, 41797, 51781, 62650, 64645, 69013, 73829, 102142, 102272, 163290, 166101",,,,"module: abi, module: binaries, module: build, module: cpp, module: cudnn, module: protobuf, module: rocm, module: third_party","Stale, module: abi, module: binaries, module: build, module: cpp, module: cudnn, module: protobuf, module: rocm, module: third_party, triaged",,Handle this module: abi stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 12 open issues; raw issue rows remain available for audit.,"#17658 C++ API: Crash in cudnnDestroy() when deconstructing, #28754 ABI backwards compatibility, #41797 [C++] Libtorch error in Release build only, #51781 Libtorch lock thread, #62650 Error while importing torch: libtorch_python.so: undefined symbol: PyThread_tss_alloc, #64645 Compilation from source fails",1651d79478910a62e535519dde90fd6ff02eb270895d969b45305501afb00dea
issue-collapse:af0ebc740b0fb6a0,module: windows stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"21176, 39226, 42095, 55869, 68405, 89688, 101968, 114565, 146944, 155667, 162786",,,,"module: windows, oncall: distributed","module: windows, oncall: distributed, triaged, windows-triaged",,Handle this module: windows stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#21176 Adding a method called `T` in native_functions causes undefined behavior on Windows, #39226 Try Address Sanitizer in MSVC builds, #42095 [RFC] Add Windows support to torch.distributed package, #55869 We should perhaps prevent linking against LibTorch DLLs with the wrong configuration (debug vs release), #68405 Windows multi machine multi card training, card initialization in GLOO communication., #89688 Enable NCCL for PyTorch on Windows",a0795e3b7924e29dc8792f1159652d714c8da92680f338d599301da12526876a
issue-collapse:cf3c61e4db487220,module: tensorboard stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"23749, 23771, 24236, 26097, 26790, 38564, 38980, 39537, 45481, 46202, 47627",,,,"module: tensorboard, oncall: visualization","enhancement, module: tensorboard, oncall: visualization, triaged",,Handle this module: tensorboard stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#23749 Wrong device in graph - Tensorboard SummaryWriter, #23771 SummaryWriter doesn't read comment if log_dir precised, #24236 Tensorboard: Add disable flag for debugging, #26097 [RFC] TensorBoard extensions and improvements for PyTorch, #26790 Importing tensorboard jams CUDA device selection, #38564 tensorboard projector mode with custom metadata_header with only one label name",24a0fc380a3fdfeefcec499da0ac7f6a25df38a549b978532cd0d47e2ea46852
issue-collapse:60e5c85d51ba152a,module: serialization stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"5106, 8430, 10577, 21745, 24432, 25892, 26567, 31615, 32165, 37985, 38643",,,,"module: serialization, oncall: jit","enhancement, module: serialization, oncall: jit, triaged",,Handle this module: serialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#5106 Saving model with runtime code changes, #8430 batchnorm2d track_running_stats, #10577 [BUG]: unstable happend in saving model., #21745 torch.save also saves docstrings into pickle for some reason, #24432 Use a ScriptModule on GPU that was saved from CPU, #25892 load_state_dict on CPU first",a34d6a19f5f73cf7531e4799c0264a7bf63d124b221437a46218387fa937239a
issue-collapse:aaa7059fc37f6a40,module: rpc stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"24946, 26119, 26446, 27671, 28882, 29402, 31842, 31937, 31979, 32269, 32363",,,,"module: rpc, oncall: distributed, oncall: jit","better-engineering, feature, module: rpc, oncall: distributed, oncall: jit, todo, triaged",,Handle this module: rpc stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#24946 [RPC] Make ProcessGroupAgent send task non-blocking, #26119 Make GloballyUniqueId a common type for both rpc and dist autograd, #26446 Avoid sending zero grads over the wire in distributed autograd backward pass, #27671 Avoid RTTI in DistEngine, #28882 Support RRef[T].__call__(*args) which invokes T.__call__(*args) on owner, #29402 [RFC] RPC timeout",9e27d03e9ed7ff47b6eebf7b58bef371d102ee6d8811922ba7529d1dadd3257e
issue-collapse:2625638aef915a87,module: error checking stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"4622, 18933, 22513, 24031, 24354, 24463, 26354, 37457, 38202, 40237, 44704",,,,"module: error checking, module: internals, module: numpy, module: partial aliasing, module: pooling, module: pybind, module: scatter & gather ops, module: tensor creation","better-engineering, enhancement, low priority, module: error checking, module: internals, module: numpy, module: partial aliasing, module: pooling, module: pybind, module: scatter & gather ops, module: tensor creation, triaged",,Handle this module: error checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#4622 Protect user from No module named _C import error, #18933 Completion of error handling, #22513 No assertion when using scatter_ on a non-contiguous tensor, #24031 fractional_max_pool2d_with_indices silently ignores output_ratio if output_size is provided, #24354 Default warning handler in C++ doesn't seem to unique warnings, #24463 Crashes on torch.cuda.memory_allocated(device)",315a4de35024ed8b37b884ee19aaee6e83536ab345dec74492544a0580bca727
issue-collapse:d411e223ff9ecfde,module: dynamo stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"89158, 96396, 101117, 102878, 114353, 118058, 118390, 121333, 126360, 130464, 134249",,,,"module: dynamo, module: inductor, module: logging, oncall: export, oncall: pt2","dynamo-dataclasses, dynamo-dicts, dynamo-must-fix, enhancement, feature, inference mode, internal ramp-up task, low priority, module: dynamo, module: inductor, module: logging, oncall: export, oncall: pt2, release notes: dynamo, triaged",,Handle this module: dynamo stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#89158 Support disallowing calls to certain instance methods in TorchDynamo, #96396 [dynamo] add hook to modify instructions before/after instructions be generated, #101117 Barriers to using torch.compile directly in PyTorch library code, #102878 [dynamo] Diffusers - Graph break on OrderedDict, #114353 add support for linalg lstsq in dynamo, #118058 torch.export doesn't handle functools.partial correctly.",c8085d02288815266f703619233f9683c3fe6f13c2d2e9f53888b6b95342b6b9
issue-collapse:584feb0cb29e4436,module: distance functions stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,11,0,0,0,"15918, 25156, 37363, 37408, 37734, 48306, 49426, 52005, 61509, 69524, 70040",,,,"module: distance functions, module: docs, module: nn, module: numerical-stability, module: performance","enhancement, function request, module: distance functions, module: docs, module: nn, module: numerical-stability, module: performance, small, triaged",,Handle this module: distance functions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 11 open issues; raw issue rows remain available for audit.,"#15918 Broadcasting and Additional Dimensions for pairwise_distance, #25156 euclidean distance between two tensors, #37363 cdist docs imply only one batch dimension is possible, but actually arbitrarily many are allowed, #37408 torch.cdist() implementation without using contiguous() calls, #37734 torch.cdist returns inconsistent result, #48306 [doc] example for pairwise distance matrix",755442c191d28f52f6ce43eaf69b71922fcfee5bc5161603255b23831c856699
issue-collapse:4b9389f5a0f64497,module: special stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"69451, 69452, 69453, 71071, 74940, 86326, 108948, 137888, 140701, 140796",,,,module: special,"feature, module: special, topic: new features, triaged",,Handle this module: special stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#69451 Exponentially scalable modified Bessel function of the second kind, #69452 Complete elliptic integral of the first kind, #69453 Jacobian elliptic functions, #71071 Feature Request: torch.special.ellipe, #74940 Feature Request: Add J0 J1 J2 H0 H1 H2 Bessel functions, #86326 `torch.special.round` doesn't support the same dtypes as `torch.round`",3608538ad60683f6031c193ecc454e3bd07d4d473d546398675c14f5072ef692
issue-collapse:c805fc88128ca938,module: pybind stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"7330, 23540, 34172, 43646, 49370, 49476, 55784, 91251, 94474, 162685",,,,module: pybind,"enhancement, feature, module: pybind, shadow review, triaged",,Handle this module: pybind stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#7330 Issue when importing both retro (from OpenAI) and torch, #23540 Error from PyTorch when finalizing Python embedded in C++, #34172 END_HANDLE_TH_ERRORS_PYBIND prevents pybind11 Exception translation, #43646 c10::string_view pybind11 custom type caster., #49370 PyEval_SaveThread: the function must be called with the GIL held, but the GIL is released, #49476 Use METH_FASTCALL protocol in Python bindings",468dc42d212930d4ddc35d98ee10863bf6ef77a14eb172701c1c30016c80b0da
issue-collapse:e897a6b6c4cf0ace,module: primtorch stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"76804, 76871, 76944, 77216, 77675, 78053, 78070, 79089, 79177, 80154",,,,"module: primtorch, module: tests, module: type promotion, module: viewing and reshaping","module: primTorch, module: tests, module: type promotion, module: viewing and reshaping, triaged",,Handle this module: primtorch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#76804 torch.add bool x bool allows integer alpha, inconsistent with other dtype type checking, #76871 PrimTorch binary refs do not handle CUDA + CPU scalar tensors correctly, #76944 primTorch references don't handle scalar x scalar inputs correctly, #77216 [primTorch] Reduction references don't return views consistent with their original operators, #77675 PrimTorch refs do not match argument naming with their PyTorch counterparts, #78053 toleranceOverride should override atol and rtol even when explicitly specified in a test",6eedbebb7f7156265fb743742ae2f6f143935cf16ffc15937ca99e472c7d6ff2
issue-collapse:ba4dbaa871553247,module: named tensor stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"27175, 28142, 28146, 29010, 31759, 33006, 33007, 38457, 43692, 46254",,,,module: named tensor,"enhancement, feature, module: named tensor, triaged",,Handle this module: named tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#27175 Easier way to create tensors with names, #28142 Named Tensor: Support -1 in `unflatten`., #28146 Named tensor: align_to align_as error messages, #29010 torch.stack does not yet support named tensors, #31759 Allow using digits in names of named tensors, #33006 matmul: no warning when contracting differently named dimensions",9791810611ebc5f2b9818dc4b322f6f66ed655a91e310f21a5f275d0eb265093
issue-collapse:5cd69aa617bb1182,module: multiprocessing stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"8913, 29172, 30461, 31596, 33799, 41639, 41694, 42049, 44545, 46093",,,,module: multiprocessing,"better-engineering, enhancement, module: multiprocessing, todo, triaged",,Handle this module: multiprocessing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#8913 Multiprocessing Self Test Error, #29172 Modules without copying in multiprocess, #30461 `torch.multiprocessing.spawn` fails when `join=False`, #31596 pytorch forward hangs in multiprocess environment, #33799 Add `start_process_in_context` to `torch.multiprocessing`, #41639 Error with minimal hogwild test (multiprocessing shared memory)",0bd5db1b61d1684c7443b55afbd49c124ad6fa9d2fce5272d7cdf5f64ba4cd68
issue-collapse:2301e28ee6093318,module: linear algebra stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"23434, 25978, 36314, 37556, 41622, 48587, 49177, 50068, 52712, 56483",,,,"module: linear algebra, module: numpy, module: performance, module: type promotion","enhancement, function request, module: linear algebra, module: numpy, module: performance, module: type promotion, triaged",,Handle this module: linear algebra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#23434 einsum equation with conditional mask works in numpy but not in PyTorch, #25978 Provide a way to select SVD algorithm in PyTorch?, #36314 [discussion] Refactor spectral_norm to use the newly merged lowrank solvers and proposal for Linear Algebra Cookbook page, #37556 torch.cartesian_prod(*tensors) error when you have tensors with [x,y], #41622 New Feature : A very fast algorithm for computing matrix rank, #48587 [feature request] Give `torch.cholesky` an optional fallback to test whether the tensor is positive definite",40117566ab2fd947ef16913a9c3432d7a148d1a8985d48e08334a4469b4bd491
issue-collapse:0bdeb378f0ef40da,module: ios stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"35535, 44313, 47904, 49456, 77538, 90551, 95946, 102833, 107394, 107581",,,,"module: ios, module: memory usage, oncall: mobile","module: ios, module: memory usage, oncall: mobile, triaged",,Handle this module: ios stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#35535 Apple Review Rejected. ITMS-90338: Non-public API usage In Pytorch For iOS, #44313 `Undefined symbol: torch::kNearest` when building App iOS, #47904 Memory leak in nn.MaxPool2d layer when run on iOS, #49456 Impossible to run tests target with LibTorch as dependency inside a cocoapods., #77538 Failed to run on iOS - Couldn't find an operator for `aten::conv1d`, #90551 Could not run 'aten::as_strided' with arguments from the 'Metal' backend.",ebe1e12ce7ee093cc316ea60fbb1ae121cbbeccfa9930918b6d66dac8e251e0d
issue-collapse:1cc3fde4d63109bc,module: intel stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"65448, 94021, 95146, 96738, 97757, 106602, 107294, 116322, 142484, 161163",,,,"module: intel, module: linear algebra, module: magma, module: memory format, module: mkl, module: mkldnn, module: mps, module: nn, module: performance, module: random","actionable, enhancement, medium, module: intel, module: linear algebra, module: magma, module: memory format, module: mkl, module: mkldnn, module: mps, module: nn, module: performance, module: random, triaged",,Handle this module: intel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#65448 Magma : Intel MKL Errors, #94021 Set AVX2 is minimum supported instruction set for Linux X86, #95146 Whether to consider native support for intel gpu？, #96738 Many padding Module fail memory_format tests, #97757 oneDNN 3.0+ support, #106602 onednn ops supported in pytorch",480d9197ef2d3489e53dfd85451e290671be14dc0637ac8492deabf76930689b
issue-collapse:1c9e97ab266df53b,module: ddp stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"46248, 49891, 54232, 54351, 54445, 55156, 55876, 56460, 58522, 60343",,,,"module: ddp, module: nccl, module: rpc, module: serialization, module: tests, oncall: distributed","module: ddp, module: nccl, module: rpc, module: serialization, module: tests, oncall: distributed, triaged",,Handle this module: ddp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#46248 ""distributed"" NCCL tests fail when having more than 3 GPUs, #49891 Support DDP find_unused_parameters=True mode when combined with Pipe, #54232 DistributedDataParallel: `DDP(model)` hangs on non-master node, #54351 Numpy.float64 vs native python float breaks DDP, #54445 add tests to cover DDP on mixed dense-sparse models, #55156 multiprocessing function cannot pass cuda objects use when calling inside from a DDP process",0ced3567da7a9a3c29d164ce758a835a222504482a39ea556b9bfa70b627bf8f
issue-collapse:9f33e9b9cc8e069a,module: complex stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"36469, 37314, 39123, 39128, 39270, 40590, 41244, 41246, 44611, 44741",,,,"module: complex, module: cpu, module: cuda, module: macos, module: nans and infs, module: numpy","module: NaNs and Infs, module: complex, module: cpu, module: cuda, module: macos, module: numpy, triaged",,Handle this module: complex stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#36469 Wrong results for multiplication of non-finite complex numbers with real numbers, #37314 equal_nan keyword not implemented for complex torch.isclose, #39123 The MacOS compiler is generating illegal instruction for the division of c10::complex, #39128 Problems implementing complex support for acosh, asinh, tanh, #39270 Current implementation of c10::complex does not support being used in shared memory, #40590 Inconsistent behavior between numpy.exp and torch.exp on CPU for complex numbers",c7cc020e4024fc210094a1d7d0feba8091a2ad49366adfc13cc6477c015c3c45
issue-collapse:2e9a1c7ee3eb5492,module: compile-time stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"93716, 97155, 109074, 119128, 128153, 130119, 135489, 149957, 154456, 156520",,,,"module: compile-time, module: cuda graphs, module: dynamic shapes, module: dynamo, module: guards, module: inductor, module: python dispatcher, oncall: pt2","enhancement, ezyang's list, low priority, module: compile-time, module: cuda graphs, module: dynamic shapes, module: dynamo, module: guards, module: inductor, module: python dispatcher, no-scrub, oncall: pt2, triaged, upstream triton",,Handle this module: compile-time stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#93716 End-to-End AMP training with GradScaler, #97155 Custom recurrent network takes very long to compile for long sequences, #109074 torch.compile/triton holding GIL during compilation and CompiledKernel call results in deadlocks during distributed training, #119128 [dynamo] Refactor switch statements to improve compile times, #128153 torch.compile Jamba: long compilation time with backend=""eager"", #130119 Compilation time for packing function is quite long",5599c6832a18698ac6cd2a4338e5f55033e81a148760dbcc14e58c63c4c80fcf
issue-collapse:e008a10eed34fcc3,module: benchmark stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,10,0,0,0,"63971, 66992, 77801, 106801, 113854, 134079, 149209, 153826, 159676, 159680",,,,"module: benchmark, module: error checking, module: performance, module: rocm, module: sorting and selection, module: tests","module: benchmark, module: error checking, module: performance, module: rocm, module: sorting and selection, module: tests, op-bench, topic: new features, triaged",,Handle this module: benchmark stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 10 open issues; raw issue rows remain available for audit.,"#63971 benchmark.Compare raises: TypeError: object of type 'NoneType' has no len(), #66992 Bug? :Run torch.unique twice get different running time?, #77801 `torch.utils.benchmark.examples.blas_compare` can not be parsed by Python-3.7 runtime, #106801 Timer benchmark stores only one time value, and therefore has broken mean/median/etc metrics, #113854 Timer: Add Standard Deviation for Single Measurement, #134079 operator benchmark diag test cuda problem",b1f252ec3ccb8a5d85be877fea1c0e0934d121f2176dedb13e434e545e59005e
issue-collapse:5e8c5bc119fe4724,oncall: mobile stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"37343, 38319, 47609, 53365, 73160, 92086, 104606, 112716, 116515",,,,"oncall: mobile, oncall: quantization","oncall: mobile, oncall: quantization, triaged",,Handle this oncall: mobile stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#37343 qnnpack's quantized-add gives wrong result, #38319 c10/macros/cmake_macros.h not exists, #47609 Error while trying to load a pretrained ResNet34 VTN (Video Transformer Network) model in Android, #53365 Pytorch mobile Vulkan API usage, #73160 Preprocessing function for backend coreml is not registered, #92086 Error while building pytorch mobile binaries from source",ef8511210bbc04a0a7e5d58265757d6281d57ff2e8eb6b6645391f135e9edf01
issue-collapse:062a667d53350200,module: vulkan stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"72046, 77733, 82316, 90920, 92285, 95956, 106549, 123702, 142331",,,,module: vulkan,"ciflow/periodic, module: vulkan, triaged",,Handle this module: vulkan stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#72046 Building PyTorch with Vulkan backend don't work, #77733 Adding Vulkan Support, #82316 Add more Vulkan operations, #90920 Support for Transformer Models on Android with Vulkan Backend, #92285 InstanceNorm operator support for Vulkan devices, #95956 [vulkan] missing aten::reflection_pad1d.out operator",dfa00ccbf94e6952578a8aac93d18a4e506a58517890f617d0f00f23d91b363b
issue-collapse:b998321a9b5d4f1e,module: ux stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"27147, 28223, 39260, 58820, 60956, 67745, 68575, 69288, 127737",,,,"module: ux, module: viewing and reshaping","Stale, enhancement, feature, module: ux, module: viewing and reshaping, triaged",,Handle this module: ux stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#27147 Stop binding in-place methods to `torch.*`, #28223 Convert manually bound `cuda` `cpu` `byte` `float` operators to native_functions, #39260 Naming inconsistency: padding_mode vs pad_mode + F.conv* docs ops miss padding_mode arg at all, #58820 [docs] Tensor.bernoulli_ formatting is hard to read and UX inconsistent with the function variant, #60956 Naming convention for variants of operations that never short circuit, #67745 A `sieve` operation for separating values of a tensor into two tensors based on a condition.",00472dc568e8d1091777d86fa517db6273ad41a12e664bd35b33d6d510d19750
issue-collapse:61c77adf68b82910,module: sdpa stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"103836, 112015, 113524, 116237, 119188, 119653, 121465, 124718, 132700",,,,"module: sdpa, module: sparse","module: sdpa, module: sparse, topic: new features, triaged",,Handle this module: sdpa stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#103836 Please consider the SCFA/dynamic flash attention for your implementation of scaled dot product attention, #112015 op scaled_dot_product_attention case different results, #113524 _fused_sdp_choice returns wrong backend in pt2, #116237 support Tensor type for scale in scaled_dot_product_attention, #119188 scaled_dot_product_attention: different output between the reference implementation and the one from torch.nn.functional, #119653 I hope PyTorch will open the sliding window local attention feature as soon as possible.",b921af7806eb8112d527710619099b9f19676f12ab96919562c47c88060c97bf
issue-collapse:f8a484fd511f9f08,module: python frontend stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"76106, 78513, 80377, 81186, 84353, 85302, 91039, 91809, 93121",,,,"module: python frontend, module: shape checking","actionable, better-engineering, enhancement, feature, module: python frontend, module: shape checking, triaged",,Handle this module: python frontend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#76106 Allow any operation that takes a Storage to also take a contiguous Tensor instead, #78513 clear input shape declaration on pytorch model inputs and outputs, #80377 Modify _add_docstr to also set the correct module for the APIs, #81186 grad not preserved during copying or pickling, #84353 torch.Size should convert all elements to ints, #85302 Remove `TypedStorage` and use only `UntypedStorage`",53468dc09b4425e219c0ffe4ad7f076d101f9a02eefd0ed347fdf7ef9c4c7e93
issue-collapse:3628ec9fa5ab78c0,module: pipelining stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"127964, 128665, 128910, 129732, 138726, 144530, 147348, 151269, 164756",,,,"module: pipelining, oncall: distributed","module: pipelining, oncall: distributed, triaged",,Handle this module: pipelining stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#127964 [pipelining] Supporting skip connections in 1F1B schedules, #128665 [RFC][pipelining] PipelineStage should let user control send/recv endpoints, #128910 [pipelining] Free memory in stage after use, #129732 Torch Distributed Pipelinging API datatype mismatch, #138726 [PP] Add unit tests to check for memory regressions, #144530 [Pipelining] PP+DDP does not work for Zero Bubble",71a48e24a4bb325f57e79b9e2ab15c30383f413062fe76a12703a8c37bedb314
issue-collapse:023017d224125801,module: nestedtensor stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"1591, 75940, 80577, 82687, 83749, 83910, 84009, 84247, 86888",,,,"module: nestedtensor, module: tensor creation, module: testing","enhancement, hackamonth, module: nestedtensor, module: tensor creation, module: testing, release notes: nested tensor, triaged",,Handle this module: nestedtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#1591 Pad PackedSequences to original batch length, #75940 Add nesting of nested Tensor, #80577 2-dimensional arange, #82687 Move nested-tensor tutorial from prototype, #83749 [Nested Tensor] Update TestCase.AssertEqual, #83910 Tracking nested tensor functions with backward kernels registered in derivatives.yaml",9ed6c0e5865c85c270988df611b9048ea4dbb31e4581d63ca00fce6a4470013a
issue-collapse:5b343f6a2454cc8b,module: dependency bug stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"21956, 22815, 23721, 26344, 31289, 37004, 53891, 70170, 113905",,,,"module: dependency bug, module: initialization, module: internals, module: multiprocessing, module: nccl, module: nn, oncall: distributed","enhancement, module: dependency bug, module: initialization, module: internals, module: multiprocessing, module: nccl, module: nn, oncall: distributed, quansight-nack, triaged",,Handle this module: dependency bug stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#21956 nn.init.orthogonal_ doesn't work with multiprocessing, #22815 TensorImpl de-virtualization, #23721 Using PyTorch on AWS EFA network, #26344 How to get rid of zombie processes using torch.multiprocessing.Pool?, #31289 Intel OMP multiprocessing assertion failure: Assertion failure at z_Linux_util.cpp(2338), #37004 RuntimeError: NCCL error in ProcessGroupNCCL.cpp:290, unhandled system error",637bc7e7718af6edeada87fbef227ce96f51f4badd7a8d3910882048861a7f0a
issue-collapse:cb0b183d0d9a47b7,module: cudacachingallocator stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,9,0,0,0,"111366, 112377, 119452, 130162, 154603, 159149, 159164, 160405, 166116",,,,module: cudacachingallocator,"feature, module: CUDACachingAllocator, triaged",,Handle this module: cudacachingallocator stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 9 open issues; raw issue rows remain available for audit.,"#111366 RuntimeError: CUDAPluggableAllocator does not yet support cacheInfo, #112377 RuntimeError: NVML_SUCCESS == r INTERNAL ASSERT FAILED at ""/opt/conda/conda-bld/pytorch_1695392020201/work/c10/cuda/CUDACachingAllocator.cpp"":1154, please report a bug to PyTorch., #119452 RuntimeError: !block->expandable_segment_ INTERNAL ASSERT FAILED at ""../c10/cuda/CUDACachingAllocator.cpp"":2814, please report a bug to PyTorch., #130162 caching_allocator_delete does not free up the memory allocated by caching_allocator_alloc, #154603 RuntimeError: devptr INTERNAL ASSERT FAILED at ""/build/source/c10/cuda/CUDACachingAllocator.cpp"":3696, please report a bug to PyTorch. entry in cache has missing shared_ptr, #159149 ❌ Error generating speech: NVML_SUCCESS == r INTERNAL ASSERT FAILED at ""../c10/cuda/CUDACachingAllocator.cpp"":995, please report a bug to PyTorch.",b86394077d12cbb35f0d0ea6db64a060427acca34202bd4d2803f153aae0642a
issue-collapse:fea37ae34ffd6656,module: vision stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"15864, 18998, 21477, 27740, 31471, 31685, 32402, 32751",,,,"module: vision, oncall: jit","feature, module: vision, oncall: jit, triaged",,Handle this module: vision stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#15864 Deformable Convolution, #18998 torch.from_PIL() Request ?, #21477 Not obvious how to install torchvision with PyTorch source build, #27740 Scripting torchvision.models.detection.maskrcnn_resnet50_fpn, #31471 Mnasnet0_5 first layer shape incorrect, #31685 What is the significance of torchvision._is_tracing()?",fc5496921ac1515770224fe3fe7157082b790b1eb3b8a44d7050eae6828ad3cb
issue-collapse:644eba3539cb5aec,module: structured kernels stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"52274, 53141, 56691, 57949, 63057, 64644, 69831, 74788",,,,"module: structured kernels, module: tests, module: type promotion","enhancement, module: structured kernels, module: tests, module: type promotion, triaged",,Handle this module: structured kernels stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#52274 Port unary elementwise ops to structured kernels, #53141 Structured kernels for operators that don't have out= variants, #56691 [structured] at::cpu_unchecked functions, #57949 [structured] Clarify if NoNamesGuard is actually needed in impls or not, #63057 Find out why TestSparseCPU.test_softmax_cpu_float64 grad and input dtypes are different, #64644 Improve meta tensor testing",2e49bca0ae1c6341f946558eb7180af3e28e1fa59d150b6ab42e25796c1fa826
issue-collapse:c8ae27df7893143f,module: rnn stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"7032, 9193, 12482, 12715, 26110, 30150, 37731, 43072",,,,"module: rnn, module: typing","enhancement, module: rnn, module: typing, todo, triaged",,Handle this module: rnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#7032 [feature request] norm argument for RNNCells, #9193 [Feature Request] Additional torch.nn.LSTM functionality, #12482 [Feature request]: add `LayerNormLSTMCell`, #12715 pack_padded_sequence throws IndexError when only kwargs are specified, #26110 Access data_ptr in RNN.cpp, #30150 I have reconstructed LSTM model and tested by mnist data but the loss is not changed (loss=2.3)",4a863492ae1e68ef215f2c865a107f037c4e5c00786576aa1d7bbdcc9ace281b
issue-collapse:eee9976fcbffa8ca,module: random stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"22557, 25070, 25071, 29692, 32047, 33664, 37373, 38662",,,,module: random,"enhancement, module: random, triaged",,Handle this module: random stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#22557 CPU torch.exponential_ function may generate 0 which can cause downstream NaN, #25070 [FR] torch.(Generator|random).seed allows specifying the seed value, #25071 Generator objects should not always use the same seed, #29692 [FR] add generator= kwarg support for torch.randn and torch.rand, #32047 Sobol point implementation, #33664 torch.rand() not having same values on using torch.manual_seed(0)",8c9d56498c3c3993673a033871c1e1742550404ca9bcbd8a39119fdaa1f1e676
issue-collapse:32deecefabf0fe9b,module: performance stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"3076, 22578, 23655, 23865, 30248, 32754, 36140, 39864",,,,"module: performance, module: rpc, oncall: distributed, oncall: jit","better-engineering, feature, module: performance, module: rpc, oncall: distributed, oncall: jit, triaged",,Handle this module: performance stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#3076 BN slows down double-backprop enormously, #22578 torch.gels runs 100 time slower on gpu than on cpu, #23655 TorchScript GPU Fuser Doesn't Handle In-Place Operations, #23865 model->to(device) costs over a millisecond when doing nothing, #30248 Add scripts for comprehensive benchmark TensorIterator, #32754 JIT performance discrepancies",f7486e2016497a9f4a285c5b1f28917d3f2fd4d100d09ac0b13bc1b3359f5712
issue-collapse:0c34f78875c0bb52,module: numerical-reproducibility stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"8242, 51426, 54947, 58758, 68102, 73475, 80808, 121595",,,,"module: numerical-reproducibility, module: numerical-stability, module: numpy, module: random, module: rnn, module: tensor creation","actionable, low priority, module: numerical-reproducibility, module: numerical-stability, module: numpy, module: random, module: rnn, module: tensor creation, triaged",,Handle this module: numerical-reproducibility stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#8242 Different behavior of LSTM and LSTMCell implementation, #51426 Different results on RTX 2060s vs GTX 1060 with gpt model, #54947 Results surprisingly differ on different backends, #58758 precision/consistency issue in `linspace`, #68102 Can not guarantee reproducible for `nn.ReflectionPad2d`, #73475 different results on each batch_size in torch==1.10.2+cu113 on RTX 3080",500694d63a1a2ba0df2a99ed0e3c424fb6cff75f58bd567873b41ce4ca9955f8
issue-collapse:88211d6702cbe0e4,module: lazy stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"63113, 63970, 65386, 65646, 65837, 66208, 67535, 67540",,,,module: lazy,"LazyTensor_nvfuser_integration, lazy, module: lazy, triaged",,Handle this module: lazy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#63113 Implement half_to_float flag for both _log_softmax and _softmax operators, #63970 [LTC] Investigate a testing utility to make sure correct lowering has happened, #65386 [LTC] Setup a subset of TorchBench for the current implementation, #65646 [LTC] Building unit tests on Debian results in double defined _GLIBCXX_USE_CXX11_ABI flag, #65837 NVFuser Examples to Help with Integration with Lazy Tensor Core, #66208 Add LTC-TS support for ops used in SGD optimizer",3bc40874c493f4fff673c1736f53a24ec8dac4eaad2359d86431a2ac6a658581
issue-collapse:3dac0cf760d2ddbf,module: internals stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"17249, 19408, 20342, 20363, 20412, 21824, 21847, 22766",,,,"module: internals, module: memory usage, module: molly-guard, module: numpy","feature, module: internals, module: memory usage, module: molly-guard, module: numpy, triaged",,Handle this module: internals stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#17249 Proposal: Add __tensor_wrap__ method similar to numpy __array_wrap__, #19408 Massive memory overhead over NumPy, #20342 TensorIterator resizes output to a scalar if there are no inputs, #20363 Define portable M_PI replacement, use it instead of non-standard M_PI in math.h, #20412 String in tensor, #21824 How about add torch::end for slicing in c++ frontend",543747c0cb83487bd196b80e8c40aada436251b792fea4dc73f0560bf27198d8
issue-collapse:7d1880ca657fc502,module: infra stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"40319, 83932, 95244, 108862, 124953, 128735, 135362, 169458",,,,module: infra,"module: infra, triaged",,Handle this module: infra stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#40319 Too many labels in the repo, #83932 Hitting rate limits for pytorchbot token, #95244 Make this ridiculously long error message more user friendly, #108862 A100 runners down: apt-get install nvidia-docker2, Could not get lock /var/lib/dpkg/lock-frontend, #124953 PRs with unbalanced quotes can not be merged, #128735 Flaky test page should include retry runs",7946e6729f58c338045325b3c6e5661a80803a3b6b4a3355045039d4b521eee9
issue-collapse:763e35aaa86097a2,module: embedding stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"51003, 52851, 54861, 55549, 55981, 56064, 123119, 127422",,,,"module: embedding, module: mps, module: nn, module: sparse, module: vulkan","feature, module: embedding, module: mps, module: nn, module: sparse, module: vulkan, new-layer, shadow review, triaged",,Handle this module: embedding stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#51003 Proposing new features: Developing Echo State Network layers, #52851 `embedding_bag(..., include_last_offset=True)` should always error if `offsets[-1] != input.size()`, #54861 Occured error in loss.backward() when using sparse=True in Embedding layer, #55549 `embedding_bag` has unexpected behavior when given `offsets` that are not monotonically increasing, #55981 LazyEmbedding, an embedding layer with a dynamically sized vocabulary, #56064 Optionally include padding_idx items in the EmbeddingBag reduction",0aee4a2eb3ad0f7f66faf98d7c9755ec914c3a2cc11a5280ef4a250320eff5ff
issue-collapse:b9c861062b987de0,module: edge cases stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"73180, 73181, 73182, 73504, 73840, 78122, 78125, 78126",,,,"module: edge cases, module: error checking, module: pooling, module: random, module: type promotion","module: edge cases, module: error checking, module: pooling, module: random, module: type promotion, triaged",,Handle this module: edge cases stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#73180 Segmentation fault in _sobol_engine_ff_, #73181 Segmentation fault in _sobol_engine_initialize_state_, #73182 Segmentation fault in _sobol_engine_scramble_, #73504 `ge` and `div` behaves differently when converting an overflow number, #73840 max_pool1d(): argument 'dilation' must be tuple of ints, but found element of type int at pos 1, #78122 Segfault in bincount",6fddd80e3fc1a994e65197c2ddb3049c34264d4f42f73d50e0ce88b10f425683
issue-collapse:fa02d166182b8510,module: docker stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"5070, 12773, 17286, 39444, 48083, 58386, 60932, 61030",,,,"module: docker, module: memory usage, module: multiprocessing, module: vision, module: xla, oncall: releng","enhancement, module: docker, module: memory usage, module: multiprocessing, module: vision, module: xla, oncall: releng, triaged",,Handle this module: docker stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#5070 NVIDIA_DRIVER_CAPABILITIES env variable is missing in pytorch docker images, #12773 Build the docker image from source, but torch.cuda.is_available()==false, #17286 Use standard docker image for XLA build, #39444 Torch hub: object has no attribute nms, #48083 CPU Tensor with Python MP Freezing in Docker Container, #58386 Better shared memory allocation under Docker",e4f9f98b3370ce6b1dbedc8b8bfc66a7a3825f3f4106a2aee3ffcbabd4b44ef1
issue-collapse:26222c61e1289df4,module: devx stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"96585, 103482, 105954, 110130, 116775, 116896, 117647, 118015",,,,"module: devx, module: flaky-tests, module: infra, module: testing","module: devx, module: flaky-tests, module: infra, module: testing, triaged",,Handle this module: devx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#96585 Proposal: Disable GC in test suite; GC after every test case, #103482 Document CI retry rules, #105954 Ensure PRs are rebased on top of a recent commit (CI check), #110130 Create a new heuristic TD rule for failures coming from base commit of the pull requests, #116775 CI on the main branch should run with keep-going, #116896 keep-going UX is not good",1520c4bf93620d213a8b1ab0846a735c1c825908f2b8f82c278115adb9f3d71d
issue-collapse:fec7871fb054028a,module: deploy stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"56489, 58238, 58876, 73268, 74384, 81568, 82382, 82419",,,,"module: deploy, module: rpc, oncall: distributed","module: deploy, module: rpc, oncall: distributed, triaged",,Handle this module: deploy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#56489 [torchdeploy] ar: _bz2module.o: No such file or directory, #58238 torchdeploy doesn't prevent Obj from being used on wrong interpreter, #58876 [deploy] Enable `torch.distributed.rpc` Python bindings in deploy, #73268 Add detection of interned submodule of externed module in PackageExporter, #74384 Package and Deploy: Generate External Python Registration files at runtime [WIP], #81568 Improve interaction of PyTorch downstream libraries and torchdeploy",5faf10a059ae58c3888e2e6876dff8bada1ebb6612664a0f3b6e9af0d0b5c5da
issue-collapse:675f2aef3f264e4d,module: cuda graphs stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"66858, 71896, 85201, 87163, 91950, 92217, 102830, 119873",,,,"module: cuda graphs, module: numerical-reproducibility, module: random","module: cuda graphs, module: numerical-reproducibility, module: random, triaged",,Handle this module: cuda graphs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#66858 CUDAgraph error while capturing, #71896 CUDA Graph API Improvement, #85201 performance between manually created graph and CUDAGraph.replay, #87163 New APIs for cuda graph inspection and manipulation, #91950 RuntimeError: philox_cuda_state for an unexpected CUDA generator used during capture, #92217 false INTERNAL ASSERT FAILED at ""../c10/cuda/CUDAGraphsC10Utils.h"":73, please report a bug to PyTorch. Unknown CUDA graph CaptureStatus32680",9f6467a6045ce8861d7cca9c65860576a9677669488932b2857b25993e94b47e
issue-collapse:be623044ac047061,module: build warnings stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"22252, 27021, 39682, 54149, 55952, 62588, 69290, 81140",,,,"module: build warnings, module: lint","Stale, better-engineering, module: build warnings, module: lint, triaged",,Handle this module: build warnings stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#22252 [libtorch] header warning suppression, #27021 Cmake warnings during build, #39682 Run `clang-tidy` on the `aten` folder?, #54149 OSS build doesn't hard fail on nodiscard, #55952 Compiler warnings tracking, #62588 static library kineto_LIBRARY-NOTFOUND not found.",e9f9fc2aaae38e8292b1c875bcec3bf61cf5df7fdd393426f500023285f3bdfc
issue-collapse:5b8bc244563409a7,module: 64-bit stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"25747, 133751, 137368, 147118, 147613, 149301, 154322, 169163",,,,"module: 64-bit, module: correctness (silent), module: cuda, module: cudnn, module: edge cases, module: mps, module: nn, module: norms and normalization, module: numerical-stability, module: optimizer, module: padding, module: pooling, module: rnn","module: 64-bit, module: correctness (silent), module: cuda, module: cudnn, module: edge cases, module: mps, module: nn, module: norms and normalization, module: numerical-stability, module: optimizer, module: padding, module: pooling, module: rnn, topic: fuzzer, triaged",,Handle this module: 64-bit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#25747 Int32 overflow in bincount indexing, #133751 `torch.nn.LSTM` using cuDNN fails with sequences longer than 65,535, #137368 Fused AdamW causing Illegal memory access on H100 for large models, #147118 padding fails on view from large tensor, #147613 torch.nn.AvgPool2d fails with stride >= 2^31 on CUDA, #149301 Unexpected results w/ LayerNorm -- suspecting possible memory issue?",0706c278ce0cf3cb388983912429c68da544dd33ad7a9ede6ade07d6ef2e4d4c
issue-collapse:85589ab644dd9b2c,general stale/low urgency  batch 47,batch action queue,stale/low urgency,2,medium,bucket_component_queue,8,0,0,0,"6542, 6783, 6845, 6857, 6875, 6902, 7133, 7179",,,,,caffe2,,Handle this uncategorized stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 8 open issues; raw issue rows remain available for audit.,"#6542 【Train issue with caffe2 detectron】Aborted at 1523501813 (unix time) try ""date -d @1523501813"" if you are using GNU date *** PC: @ 0x7fa35103733a (unknown), #6783 [caffe2] IfOp, #6845 [caffe2] global average pool in caffe2, #6857 [caffe2] Understanding WorkFlow for Training and Testing, #6875 [Bug] VS reports unresolved external symbol from caffe2 observer, #6902 [Caffe2] Flatten Layer in caffe2",faa50840350584c02bf8745a011b0e7be630176028fbeafa0410824d9271cbb1
issue-collapse:17066772e67a5df2,oncall: quantization stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"25314, 29657, 39495, 42336, 47311, 55464, 55912",,,,oncall: quantization,"feature, low priority, oncall: quantization, triaged",,Handle this oncall: quantization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#25314 ConstQuantizerPtr is misnamed, #29657 Improved detection of repeated observers, #39495 Misannotation of layer_norm parameters causes internal assert failure, #42336 Audit quantization functions to ensure proper argument sizes, #47311 Quantized modules should properly implement __getstate__ and __setstate__ (copy.deepcopy doesn't work on quantized model), #55464 Support quantized functional linear/conv operators with 32b output",8b1f1c98a4e4477c4c9dd35945164c8e083be5a7e7ea1e833ce26ad905e4d142
issue-collapse:ce3d2f707dd5f92a,oncall: cpu inductor stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"93477, 101273, 119181, 126691, 130948, 133458, 158664",,,,"oncall: cpu inductor, oncall: pt2","intel, oncall: cpu inductor, oncall: pt2, skipped, triaged",,Handle this oncall: cpu inductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#93477 [Inductor] [CPU] optimize thread parallel and loop collapse, #101273 IPEX as TorchDynamo Backend Performance Dashboard, #119181 [inductor][cpu]Background_Matting fP32 static default wrapper multiple threads performance regression, #126691 TORCH_CHECK used within torch.compile does not throw legible errors, #130948 [inductor][cpu]opacus_cifar10 performance regression in 2024-07-15 nightly release, #133458 [CPU] jx_nest_base AMP both inductor and eager performance regression in 2024-08-10 nightly release",cf0cb1a44836fdbc3d57d9eec51354ce30d6acd1ca1d7d1f479716500052e9ed
issue-collapse:07aa59be60e7942e,module: scatter & gather ops stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"6850, 56464, 71300, 77231, 108569, 124783, 151978",,,,"module: scatter & gather ops, module: sparse","feature, module: scatter & gather ops, module: sparse, triaged",,Handle this module: scatter & gather ops stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#6850 [feature request] More `index_*_` functionality and/or lambda functionality, #56464 Scatter tests missing when passing a reduction operation., #71300 Kernel fusion for Gather, Apply, Scatter (GAS) model, #77231 `torch.scatter_add` will succeed when the `index` is a complex tensor, #108569 Call for a deterministic implementation of scatter_add_cuda_kernel, #124783 Scatter_add limitation when accumulate beyond 2^24 under float32 precision",66a7b52f503c49b0786735dc584eba5d7be7ac30511457ff4be1310b545bad9f
issue-collapse:d90557bac080cf10,module: sanitizers stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"88185, 88944, 88945, 88947, 88949, 88950, 90526",,,,module: sanitizers,"module: sanitizers, triaged",,Handle this module: sanitizers stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#88185 `torch.mm` Trigger RuntimeError with UndefinedBehaviorSanitizer, #88944 `torch.topk` Trigger RuntimError under UndefinedBehaviorSanitizer, #88945 `torch.linalg.eigvals` Trigger RuntimeError under UndefinedBehaviorSanitizer, #88947 `torch.Tensor.msort` Trigger RuntimeError under UndefinedBehaviorSanitizer, #88949 `torch.set_rng_state` Trigger RuntimeError under UndefinedBehaviorSanitizer, #88950 `torch.nn.functional.embedding_bag` Trigger RuntimeError under UndefinedBehaviorSanitizer",08681179b908659a0c60d87fa7acb21931a82e73fc188f8a622356111500206f
issue-collapse:5fb015df7fd51b32,module: privateuse1 stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"125895, 133183, 134293, 135153, 156444, 161954, 171956",,,,"module: privateuse1, module: tests, oncall: distributed","module: PrivateUse1, module: tests, oncall: distributed, triaged",,Handle this module: privateuse1 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#125895 Support for third-party use of pipeline, #133183 convolution_overrideable not implemented when executing VAEEncode in ComfyUI, #134293 [PrivateUse1 Backend] Unable to register backend to PrivateUse1 dispatch key: ModuleNotFoundError: No module named 'torch.privateuseone', #135153 TestSDPAPrivateUse1Only tests causes irreversible changes to PyTorch internal state, #156444 Tensor.is_pinned() raises error after renaming privateuseone backend., #161954 [RFC] CUDA code decoupling and directory refactoring",5e519347d37723ac394cc87662d910315fbd5989feeb04362d3b74c937952734
issue-collapse:6046a603bdca4ac4,module: printing stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"8821, 12797, 21028, 35071, 63855, 74138, 84710",,,,"module: printing, module: ux","enhancement, module: printing, module: ux, triaged",,Handle this module: printing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#8821 Storages still use legacy printing, #12797 Port dragon4_scientific for pretty float tensor print., #21028 Zero-dim Tensors (scalars) should be printed at full precision, #35071 Options for printing the shape with print(tensor), #63855 Include tensor shape in its default print formatting, #74138 `torch.set_printoptions` overwrites settings of its own previous calls",fcdfd7cc5b297d854cd8439ae5a87c344df9c5474774faa3f6e3c8fa600246b1
issue-collapse:6b243a1e9bca18d3,module: nvfuser stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"75371, 75474, 76007, 76070, 76491, 76933, 79246",,,,module: nvfuser,"module: nvfuser, triaged",,Handle this module: nvfuser stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#75371 NVFuser bad ""reshape"" performance, #75474 [NVFuser] call kernels with informative names (e.g. pow_mul_add), #76007 Build a default NVFuser comparison callback, e.g. for use with torchbench, #76070 NVFuser microbenchmark classifier - hash on memory formats, #76491 [NVFuser] Automated generation of microbenchmarks, #76933 NVFuser opinfos - check for CudaFusionGroup in the graph",5a46caaf19c17cde226ed26f37c0f4bfe41ecabbc6586098530943d2a12a78ba
issue-collapse:807b5e61923a4c07,module: loss stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"22604, 25057, 26228, 29754, 34879, 36553, 39435",,,,"module: loss, module: nn, module: type promotion","enhancement, feature, module: loss, module: nn, module: type promotion, triaged",,Handle this module: loss stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#22604 ""CrossEntropyLoss"" should mention in its name that it takes softmax for target, #25057 ctc_loss computes different losses and gradients on batched utterances vs. individual utterances, #26228 Higher dimension support for `MultiLableSoftMarginLoss`, #29754 [FR] general nll_loss and cross_entropy along arbitrary dimension, #34879 Addition of Siamese loss or contrastive loss function, #36553 Typecasting issue in MSELoss",0c3b380a1aed099321a23bf08f4b29b485ea1ea8ca71a2822d26279a495c5a79
issue-collapse:5bbd351db7ff8e7c,module: library stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"83583, 99681, 107387, 107695, 125183, 125234, 127615",,,,"module: library, module: pybind, module: type promotion, module: typing","module: library, module: pybind, module: type promotion, module: typing, triaged",,Handle this module: library stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#83583 Silent promotion of bool to int in the dispatcher, #99681 `torch.ops.aten.empty` is not discoverable from `dir(torch.ops.aten)` until explicitly calling getattr, #107387 Incorrect type hint for `torch.library.Library.define`, #107695 New variables in torch._ops.py pollute the torch.ops namespace, #125183 torch/_refs/__init__.py is not autoreloadable in bento / jupyter notebook, #125234 torch.Library can easily cause segfault on loading/unloading",d2c9a434711686e722dadc54bfe619cf811ca49d588008b017b50ac5d8b14988
issue-collapse:6c919b9d7d9cf586,module: half stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"26218, 54774, 54983, 55629, 55755, 56231, 58841",,,,"module: half, module: linear algebra, module: numerical-stability, module: numpy, module: optimizer, module: sorting and selection","enhancement, feature, module: half, module: linear algebra, module: numerical-stability, module: numpy, module: optimizer, module: sorting and selection, triaged",,Handle this module: half stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#26218 Default adam epsilon to 1e-7 when on fp16, #54774 RuntimeError: ""log2"" ""_vml_cpu"" not implemented for 'Half', #54983 Accept objects with `__float__` wherever regular `float`s are accepted, #55629 torch.arctanh not implemented for torch.float16, #55755 Some float16 inputs to CPU matmul are supported, others aren't, #56231 Half precision support for torch.sparse.mm",ff2815e21f00e7a13fd590f6bd0fb3ae073020e9f1dc62f2898665d4b962d8e6
issue-collapse:6e779f247b316ac0,module: deadlock stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"17745, 26447, 39706, 63441, 68507, 134067, 162154",,,,"module: deadlock, module: linear algebra, module: multiprocessing, module: multithreading, module: nccl, oncall: distributed","module: deadlock, module: linear algebra, module: multiprocessing, module: multithreading, module: nccl, oncall: distributed, triaged",,Handle this module: deadlock stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#17745 distributed data parallel, gloo backend works, but nccl deadlock, #26447 parallel_for may hang when called in main process and then on daemon process, #39706 [distributed] calling nccl reduce with inconsistent dst hangs, #63441 OpenCV causes backpropagation to get stuck, #68507 Multiprocessing hang and queue short circuiting, #134067 subprocess tensor creation gets stuck",b9958652fc37d58ff57a46641b3e81afb44a4c59d214e048aa1f087ae548d9d6
issue-collapse:cb97397ccac73c7c,module: data parallel stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"25514, 27046, 27218, 28275, 31360, 31553, 33081",,,,"module: data parallel, module: determinism, module: memory usage, module: multiprocessing, module: performance, oncall: distributed","feature, module: data parallel, module: determinism, module: memory usage, module: multiprocessing, module: performance, oncall: distributed, triaged",,Handle this module: data parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#25514 Multithreaded backpropagation with custom autograd.Functions, #27046 The Gather problem in DataParallel: dimension are not matched., #27218 pytorch data_parallel oom on gpu:0, #28275 nn.parallel.replicate in v1.1+ is much slower than v1.0, #31360 DDP/MP not yielding nontrivial speedup, #31553 allow setting different batch size splits for data_parallel.py and distributed.py",d814ebd38517aa52dac2d24ecb58293b3b5cd6af014aa6d536b6fe1f38e53a38
issue-collapse:8fbd52ded185367e,module: cpu stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"3468, 19437, 19825, 21602, 22687, 26534, 28390",,,,"module: cpu, module: internals, module: mkldnn, module: multithreading, module: performance, module: random","feature, module: cpu, module: internals, module: mkldnn, module: multithreading, module: performance, module: random, triaged",,Handle this module: cpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#3468 improve performance of common CPU clone / contiguous calls with HPTT, #19437 The `unary_kernel` call should be pushed into CopyKernel.cpp and it should completely replace the current `copy_kernel`., #19825 Performance issue master (a25b79531), #21602 Strange latency overhead of F.conv2d, #22687 DispatchStub should report what operator it failed to find kernel for, #26534 MKLDNN+AMD BLIS path for PyTorch",76a20e2668ed60ed6b8a303145daa009696431b41b18c7b9d07af9b1a6433ef2
issue-collapse:c2b04685121e39a8,module: correctness (silent) stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"5272, 24168, 48281, 55355, 59515, 61538, 70397",,,,"module: correctness (silent), module: cublas, module: cuda, module: distance functions, module: linear algebra, module: memory format, module: nn, module: numpy, module: safe resize, module: tensor creation, module: tf32","module: correctness (silent), module: cublas, module: cuda, module: distance functions, module: linear algebra, module: memory format, module: nn, module: numpy, module: safe resize, module: tensor creation, module: tf32, triaged",,Handle this module: correctness (silent) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#5272 Unsafe out= keyword argument with tensors sharing storage, #24168 torch.unique is inconsistent with NumPy's unique, #48281 Batched inplace mm changes stride when out size is correct, #55355 simple matrix multiplication yields wrong result on Ampere (3080), #59515 Conv1d with large batch size and half precision in cuda returns incorrect result, #61538 The use of epsilon in torch.nn.functional.pairwise_distance leads to incorrect minima of the distances",2a1e29ffffc30ddcc35b7e0140fe322c89e38e4a07c3fe82cbc9e9c7e6b36c91
issue-collapse:d268c19fa454e34f,module: compiled autograd stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"124866, 138215, 138901, 142104, 143759, 144982, 155105",,,,"module: compiled autograd, module: dynamo, module: higher order operators, module: pt2-dispatcher, module: tests, oncall: pt2","actionable, dynamo-autograd-function, enhancement, feature, module: compiled autograd, module: dynamo, module: higher order operators, module: pt2-dispatcher, module: tests, oncall: pt2, triaged",,Handle this module: compiled autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#124866 support side effects in HOPs?, #138215 Dynamo x autograd.Function: leverage compiled autograd on graph break in backwards, #138901 Runtime error: retains_grad_hooks not implemented for compiled autograd, #142104 [compiled autograd] Does compiled autograd under torch 2.4 support using selective checkpoint?, #143759 compiled autograd tests should use expecttest, #144982 [compiled autograd] It would be nice if the compiiled autograd graph was actually runnable",225bfc3acefc4afc4dcd8382b5bd54580e677067ac5c4edc83bba2c50b250436
issue-collapse:dbfacb461b98cee5,module: ci stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"18692, 20111, 33477, 37512, 37847, 37880, 42655",,,,"module: ci, module: tests, oncall: distributed","better-engineering, module: ci, module: tests, oncall: distributed, todo, triaged",,Handle this module: ci stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#18692 Add build tests for feature environment vars, #20111 Split libtorch binary build CI job into separate variants, #33477 Add 32-bit CI (e.g., Raspberry PI CI), #37512 Improve visibility in test suite timings, #37847 llvmlite version issue when upgrading CI docker image to python 3.8, #37880 [RFC] Don't install CI dependencies in build scripts, install them in underlying docker images",7035a9d9c9b1926382091192d5601a80774b83271aedc6a7249a87887ae1ef92
issue-collapse:f3a7dd67db12e458,module: bc-breaking stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"3697, 10683, 12962, 22707, 30263, 31776, 38205",,,,"module: bc-breaking, module: cuda, module: deprecation, module: docs, module: molly-guard, module: nn, module: numerical-stability, module: scatter & gather ops, module: ux","enhancement, feature, module: bc-breaking, module: cuda, module: deprecation, module: docs, module: molly-guard, module: nn, module: numerical-stability, module: scatter & gather ops, module: ux, triaged",,Handle this module: bc-breaking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#3697 Deprecate inplace argument in torch.nn.functional, #10683 Tensor.register_hook is not passing the tensor object to the hook function, #12962 Come up with a better strategy for noticing BC-breaking attribute additions to serializable classes, #22707 torch.fill_() exists and modifies the input tensor: Expected or bug?, #30263 pairwise_dist eps argument is confusing, #31776 Documentation for `scatter` incorrectly states that index values must be unique",2b107e1a447b5699b560f65736651b5188330e2c98fdba874e255eb86723cd08
issue-collapse:68d021cced617644,module: backend stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"7944, 26691, 47546, 54546, 54646, 63477, 83657",,,,"module: backend, module: codegen, module: dispatch, module: internals","module: backend, module: codegen, module: dispatch, module: internals, triaged",,Handle this module: backend stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#7944 Better error message in DataChannelTCP::_receive, #26691 Support FPGA Xilinx, #47546 Whats Pytorch's policy on adding support for a wider range of hardwares for training and inference?, #54546 Let backends specify a schema version when registering kernels, #54646 Optional enhanced logging for operator calls for backend implementor debugging, #63477 Supporting Pytorch for Custom Compiler Backend",d384a71f7722c9dd6e7bb59812532e74f67a142ce664314fffa518f968c57761
issue-collapse:fd1daacd168cfe56,module: aotinductor stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,7,0,0,0,"127928, 147475, 150364, 161108, 163624, 169154, 169469",,,,"module: aotinductor, module: torchbind, oncall: export, oncall: pt2","export-triaged, module: aotinductor, module: torchbind, oncall: export, oncall: pt2, topic: new features, triaged",,Handle this module: aotinductor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 7 open issues; raw issue rows remain available for audit.,"#127928 Ability to do aot/inductor compilation from a jit model (or torch.exported model), #147475 [Export AOTI] dynamic_shapes export and compile degraded output, #150364 AOTI doesn't error if dynamic shape gets specialized during lowering, #161108 AoTInductor gives segmentation fault with torchbind registration, #163624 [aoti] [xpu] [null-pointer-deference] potential npt issue in `sycl_runtime_wrappers.h`, #169154 AOTIModelPackageLoader occupies too much disk space when launching multiple instances",f879c8981ffce090e87dd74c5b3eca63752c98241fbc33bcbe98690b84fcad6f
issue-collapse:4d9f6a93fe787380,oncall: jit stale/low urgency  batch 15,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"19504, 20734, 21165, 21454, 24336, 24343",,,,oncall: jit,"oncall: jit, triaged",,Handle this oncall: jit stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#19504 [jit] Bad error when calling `ScriptModule`s with attributes/parameters, #20734 [jit] set up views for Autodiff and autograd hooks, #21165 Getting Access to Blob/Tensor reference in jit::script::Module, #21454 [JIT] Memory Leak during tracing?, #24336 Auto tuner takes too much time in serialized model, #24343 Unified representation for enum types",a77d61c6ce62d32a5369cdc31fbd48e84eccc00aeca637173cca9a3dd27fa921
issue-collapse:df3236858b9fa882,module: xpu stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"144766, 151308, 157714, 160787, 162837, 167170",,,,module: xpu,"module: xpu, triaged",,Handle this module: xpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#144766 Matmul with int32 parameters on Intel GPU leads to errors, #151308 [Intel GPU][UT failure] Depthwise conv related UT failures, #157714 AssertHandler::printMessage On Intel GPU, #160787 [Intel Arc XPU] NaN Values in Gradients after SDPA with is_causal, #162837 XPU backend is falling back to run on the CPU., #167170 [XPU][Tracker] Enable `torch._scaled_matmul`",f201c956ac0f1db5503835862ed5f599263e517dc13fbd476bcce62d2bc11d50
issue-collapse:edb059c8fbbba489,module: xla stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"77808, 92909, 97676, 111033, 113382, 130985",,,,"module: xla, oncall: fx","module: xla, oncall: fx, triaged",,Handle this module: xla stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#77808 `Could not start gRPC server` flakiness in XLA tests, #92909 [RFC] XLA Lazy Backend Support In DistributedTensor API, #97676 Pytorch 2 compile + fsdp + transformers crash, #111033 `CapabilityBasedPartitioner` returns invalid partitions., #113382 TorchXLA - owner @JackCaoG, #130985 Runtime error when running huggingface pretrained model with torch-xla",212e6a3039fa1f5327e8e0711d9110571388988dee60f2fb44535c1f50c48837
issue-collapse:59ed58e6bda5237d,module: vmap stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"47378, 52295, 94593, 130768, 141890, 151286",,,,module: vmap,"module: vmap, triaged",,Handle this module: vmap stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#47378 torch.vmap giving INTERNAL ASSERT FAILED error, #52295 Printing a Tensor that is being vmap'ed over in C++ raises an error, #94593 Memory corruptions can be triggered in torch._remove_batch_dim, #130768 Migrate gradcheck with check_batched_grad=True to use the new torch.vmap, #141890 Does vmap accept None out_dims?, #151286 vmap- out of memory",12bd3841fad9d038b68e34424661e8da4354a0944a2967aad52bd881e386a3ef
issue-collapse:5f0d107af50bd333,module: vllm stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"148983, 150190, 166733, 169133, 170495, 170563",,,,"module: vllm, oncall: pt2","compile-cache, dynamo-triage-jan2025, module: vllm, oncall: pt2, triaged, vllm-compile",,Handle this module: vllm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#148983 inference_mode Tensors do not always need to be guarded on, #150190 auto_functionalize HOPs are cacheable, also we should check that we allow for manual overrides for cacheability, #166733 benchmark and advertise torch.compile's ability to reduce triton kernel launch overhead, #169133 vLLM tlparse sometimes missing compilation_metrics, #170495 vLLM Blackwell Fusion and Compile Tests failing PyTorch 2.10, #170563 vLLM v1/distributed/test_eagle_dp.py Tests failing on PyTorch 2.10",b18d1ca8801cf26575705bab30095a2394a22ebf4ea6ce0b2071559a3562af70
issue-collapse:e11778b834ea92c2,module: type promotion stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"32098, 37895, 40891, 50213, 71058, 71918",,,,module: type promotion,"enhancement, feature, module: type promotion, triaged",,Handle this module: type promotion stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#32098 torch.gt ge lt le, #37895 torch.addmv can't take as input tensors with different dtypes, #40891 [FR] torch.floor/ceil should support output int dtype, #50213 torch.float_power out= and inplace variant errors on non-matching output dtype instead of casting, #71058 `torch.Tensor.where` cannot work when `y` is float, #71918 Support mixed python scalar/tensor types for torch.clamp's min/max args",db843110a0b5e2cc886e8017702fb4ccb9797a2af0a1e4a80db6fa21201f2a3b
issue-collapse:66195f6e254f8596,module: reductions stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"67457, 71629, 101334, 105790, 143561, 153360",,,,"module: reductions, module: sparse, module: unsigned int","enhancement, feature, module: reductions, module: sparse, module: unsigned int, triaged",,Handle this module: reductions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#67457 Reductions on zero-size dims: 1) by accepting a custom default value, 2) if tensor has another non-reduced zero-size dim, #71629 `torch.{max,min}` have strange error message when `input.numel()==0`, #101334 Functions for Calculating Skewness and Kurtosis, #105790 torch.sparse.mm() with reduce operator for GPU support and COO matrices, #143561 ""aten::any"" and ""aten::all"" behavior for unsigned tensors different than signed, #153360 torch.aminmax reducing over all dims should not accept keepdim",2bf598ad5ecc46d56323cd1982d673f571c14936aa98570a0bd79b20c165b789
issue-collapse:38465c6b3e086d5e,module: pickle stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"3473, 6107, 18436, 19529, 51836, 145665",,,,"module: pickle, module: serialization, oncall: distributed","better-engineering, feature, module: pickle, module: serialization, oncall: distributed, onnx-triaged, topic: security, triaged",,Handle this module: pickle stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#3473 Feature Request: Distributed send arbitrary objects, #6107 Well documented, safe method to deserialize model parameters from untrusted sources, #18436 UnpicklingError when trying to load multiple objects from a file, #19529 failed to load model which is saved as text format(pickle_protocol=0) instead of binary format, #51836 TypeError: can't pickle _thread.lock objects when using spawn / forkserver, #145665 Pickling duplicates Storage",8497a43371076f0cf062fd855435c761f5de504b76196ba1986abd8880b8d541
issue-collapse:d2690df4614e9715,module: memory usage stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"3931, 5161, 5528, 28055, 36874, 38177",,,,"module: memory usage, module: tests","enhancement, module: memory usage, module: tests, triaged",,Handle this module: memory usage stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#3931 Suppress hidden state output of RNNs?, #5161 ASAN detected leaks on python -c 'import torch', #5528 TestNN.test_data_parallel takes 10G of memory, #28055 Traced resnet101 leaks memory during `forward`, #36874 Memory leak upon exception for interactive consoles, #38177 valgrind says libtorch has memory leak",d001da87680f5b86c23a422ba0dc21771553bf0ad46b222206106f9cbcd31925
issue-collapse:e51452a2b3f090fe,module: initialization stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"43411, 54030, 54055, 65153, 124719, 136410",,,,"module: initialization, module: nn, module: sparse","enhancement, module: initialization, module: nn, module: sparse, triaged",,Handle this module: initialization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#43411 Add support of random state generator objects to nn.init module, #54030 [FR] Configurable gain value for kaiming_normal_ / kaiming_uniform_ in torch.nn.init, #54055 Wrong initialisation gain for SNNs/SELU, #65153 Unification of model initialization methods / naming across domain libraries + support of skip_init, #124719 Questions about parameter initialization, especially with torch.bfloat16 precision, #136410 Implement kaiming_uniform_ for sparse tensors.",6c904b1a5576be39902b3b7a20b6a8c26ef24b2f38454c498f28df5fb1b9a51b
issue-collapse:406e8a7f2fd59cc9,module: fsdp stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"98484, 98808, 101999, 104252, 104479, 108190",,,,"module: fsdp, oncall: distributed","module: fsdp, oncall: distributed, triaged",,Handle this module: fsdp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#98484 Improvements to FSDP debugability, #98808 [FSDP] move up the first all gather, #101999 [FSDP] Ensure full precision checkpoints, #104252 [RFC] Make `_HYBRID_SHARD_ZERO2` public as `HYBRID_SHARD_GRAD_OP`, #104479 FSDP Optimizer Overlap - follow ups, #108190 [FSDP] incorrect backward prefetch order when using BackwardPrefetch.BACKWARD_POST",f8700bde039f72746a2d165907f30c54773442576cee312f0cf56f33ccbc9313
issue-collapse:876076dbe8c3d7c9,module: dynamic shapes stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"117681, 123765, 128053, 128548, 130432, 135255",,,,"module: dynamic shapes, oncall: pt2","feature, module: dynamic shapes, oncall: pt2, triaged",,Handle this module: dynamic shapes stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#117681 Improved workflow for tracing dynamism through complicated model, #123765 Accurately typed ValueRanges / convenient OpsHandler dtype inference, #128053 Handling reasoning about rationals in symbolic shapes, #128548 [torch.compile] Llama2 failure using dynamic shapes with Torch 2.4 nightly, #130432 Deferred runtime asserts triggered during backwards should not trigger assert not self.runtime_asserts_frozen, #135255 Log missed optimization opportunities due to dynamic shapes in Inductor",c56e25f052284e1c31838b681b3001bb26062d5649888e475c102c62e5680329
issue-collapse:cf2d379a8719e403,module: docs stale/low urgency  batch 28,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"8153, 16804, 19652, 20375, 20481, 20641",,,,"module: docs, module: rnn, module: tests","enhancement, module: docs, module: rnn, module: tests, triaged",,Handle this module: docs stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#8153 GRU is implementation of GRU v1 draft rather than final GRU paper algo, #16804 Wrong description of positive class weight in BCEWithLogitsLoss, #19652 [docs] torch.set_flush_denormal(...) to mention default mode, #20375 Better documentation / molly-guards around use of multiprocessing with spawn in Jupyter/ipython notebooks, #20481 [docs] Automatically detect docs missing in rst, #20641 Main page broadcasting (?) example image bug",959e00be9d1dc00b37a0e49d0aaae4d4b32e107e275601037dffb9d59c94396d
issue-collapse:82c08dfb6aa0b86b,module: cudnn stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"4542, 4551, 5280, 16655, 19911, 42545",,,,"module: cudnn, module: dependency bug, module: deprecation, module: multi-gpu, module: rnn","module: cudnn, module: dependency bug, module: deprecation, module: multi-gpu, module: rnn, small, triaged",,Handle this module: cudnn stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#4542 Carefully audit contiguity requirements of code, #4551 Met 'cudnnDestroyDropoutDescriptor' while run multiply gpu-based models in multiply processes, #5280 BatchNorm1d raises RuntimeError (CUDNN_STATUS_BAD_PARAM) on 3D input., #16655 CUDA cache doubles on the second batch and causes OOM, `empty_cache` doesn't empty it, #19911 BatchNorm1d does not support batchsize>65535 in eval mode with 3 dimension (NxCxL), raise CUDNN_STATUS_NOT_SUPPORTED, #42545 CuDNN RNN bindings are basically all deprecated in cudnn 8",d309578aadf9a0fc98c64cc5b4ec12d9c991fbc0b56b7c70a6c3c7c0f562e8f8
issue-collapse:e892a7b27a89fa32,module: cuda stale/low urgency  batch 36,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"12819, 17350, 19229, 19335, 19599, 20562",,,,"module: cuda, module: cudnn, module: error checking","module: cuda, module: cudnn, module: error checking, triaged",,Handle this module: cuda stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#12819 Initialization error when moving data to the GPU, #17350 torch.nn.CrossEntropyLoss with ""reduction"" sum/mean is not deterministic on segmentation outputs / labels, #19229 Different behavior of torch.nn.MultiMarginLoss on CPU/GPU Tensors, #19335 CUDA optimization: using `__restrict__` whenever possible, #19599 cudnn conv doesn't check batch_size > 0, #20562 Performance issue when accessing an extremely large (10GB) longtensor",d5501d916b9b45c9c78ec10c34452cccef6ca983bfbac69ec617959cdd6e9cb1
issue-collapse:fae7d800a90096d1,module: c10d stale/low urgency  batch 4,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"10800, 16318, 46428, 48305, 49217, 53866",,,,"module: c10d, module: ddp, module: docs, module: rpc, oncall: distributed","distributed-backlog, module: c10d, module: ddp, module: docs, module: rpc, oncall: distributed, triaged",,Handle this module: c10d stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#10800 [distributed] Synchronization on CUDA side with MPI backend, #16318 Pytorch with CUDA aware OpenMPI for Infiniband not working with HCOLL and MXM, #46428 send_object/recv_object APIs for c10d, #48305 [RFC] CUDA-aware future for distributed, #49217 Bug in profiler on CI machines when profiling NCCL distributed calls, #53866 Add comprehensive subgroup testing to torch.distributed",04f40efde262b52a710bc620b34f2adcb08bcc30f602dbe11f73c95fe9fbfb80
issue-collapse:1ae287f1cc65b3a2,module: bootcamp stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"1736, 23299, 24273, 24836, 31107, 31422",,,,"module: bootcamp, module: data parallel, module: nccl, module: optimizer, oncall: distributed","enhancement, feature, module: bootcamp, module: data parallel, module: nccl, module: optimizer, oncall: distributed, triaged",,Handle this module: bootcamp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#1736 Expose optimizer options as attributes when there's a single param group, #23299 Add gatherv/allgatherv primitives to support non-equal contribution, #24273 Check PyTorch version when initializing process groups, #24836 Gloo scatter gives wrong result for stride != 1, #31107 Add torch.version.nccl, #31422 Support DataParallel with PackedSequence",260a76d4e2ff9cc55be5f1bd4833924068309099716268778dabb52e6bde9c5e
issue-collapse:fc50d6bc7bc064a4,module: binaries stale/low urgency  batch 11,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"3990, 8098, 23094, 25548, 29057, 29722",,,,"module: binaries, module: build, module: cuda, module: docker, module: mkl, module: performance, module: rnn","better-engineering, module: binaries, module: build, module: cuda, module: docker, module: mkl, module: performance, module: rnn, triaged",,Handle this module: binaries stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#3990 Raise an error when using magma built against wrong version of cuda, #8098 Please provide wheel package for windows on PyPI, #23094 Versioning for libtorch nightlies, #25548 The inference speed of the torch c++ dynamic library compiled manually is slower than the torch library officially provided, #29057 Docker issue for Pytorch 1.3, #29722 Slow (20-50x) RNN tutorial/example when torch is installed using pip comp. to conda installation",eca53a43418711bb55229f873eb4cb4d8b6682c6d0cbedc3e1554c0c1d9fd7b0
issue-collapse:156f68b46660da9f,module: aotdispatch stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"111950, 146693, 148146, 151884, 154083, 163048",,,,"module: aotdispatch, module: dynamic shapes, module: pt2-dispatcher, oncall: distributed, oncall: pt2","low priority, module: aotdispatch, module: dynamic shapes, module: pt2-dispatcher, oncall: distributed, oncall: pt2, triaged",,Handle this module: aotdispatch stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#111950 Operators that return dynamic-shape outputs that require_grad choke in AOTAutograd / unbacked SymInts are reallocated, #146693 Partitioner moves useless and memory-heavy op from bwd to fwd, #148146 Unify OpOverload._get_dispatch and HigherOrderOperator.dispatch, #151884 The ""eager"" and ""aot_eager"" backends have different behavior for the expected gradient tensor of the torch.expend_as operator, #154083 torch.compile joint trace materializes constant tensors, #163048 [compile/aot-autograd] `torch.autograd.grad` w.r.t. an intermediate tensor works in eager but fails under `torch.compile` (“tensor not used in the graph”)",109872ef472c236c17d24bf1b8cfc181c98ae5d94584804834ce79c46b1ca403
issue-collapse:0c2e27d636c77f16,module: accelerator stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"150663, 166054, 166696, 167210, 167447, 169371",,,,"module: accelerator, module: cpp, module: mps","enhancement, module: accelerator, module: cpp, module: mps, triaged",,Handle this module: accelerator stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#150663 Proposing torch.empty_cache (device_type) as generalization of torch.cuda.empty_cache() ., #166054 Add XLA to `at::accelerator::getAccelerator` and `at::accelerator::isAccelerator`, #166696 `torch.accelerator.is_available()` fails when privateuseone is used, #167210 [RFC] Introduce the Unified Memory Pool and Pluggable Allocator in torch.accelerator, #167447 Memory related APIs in torch.accelerator are not supported on MPS, #169371 C++ Generator API is platform dependent",59d9d4f81d77aa739af89cb26edd2645e70bd8dbd13be558090dc2bb35c8d6df
issue-collapse:ea64bfb77d7f7319,module: __torch_dispatch__ stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,6,0,0,0,"62888, 64548, 64874, 65339, 65619, 66026",,,,"module: __torch_dispatch__, module: deploy, module: meta tensors, module: pybind","module: __torch_dispatch__, module: deploy, module: meta tensors, module: pybind, triaged",,Handle this module: __torch_dispatch__ stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,6,Represents 6 open issues; raw issue rows remain available for audit.,"#62888 Add a mechanism to not universally dispatch to python for all for all operations when using __torch_dispatch__, #64548 Interpreter initialized state changed by torch dispatch, #64874 I am unable to detect torch.Tensor._version change during __torch_dispatch__ call, #65339 Make it easier to accurately reflect storage/view relationships in meta/wrapper tensors, #65619 Integration of __torch_dispatch__ with masked tensor semantics, #66026 `__torch_dispatch__` can result in returning `None` for an op that should return Tensors.",5fca3c77944cdfde3ddd3ccd57a080216305de554f4ca8be1ae5b4bb3fd75dc9
issue-collapse:cc2aaaf441fc58c2,oncall: r2p stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"68553, 69197, 69520, 77737, 78061",,,,oncall: r2p,"Stale, oncall: r2p, triaged",,Handle this oncall: r2p stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#68553 [distributed] elastic/agent/server/api.py could react faster on exit / Ctrl+D / Ctrl+C, #69197 RendezvousConnectionError when use C10d on multi nodes, #69520 race condition of agent's TCP store on exit_barrier(), #77737 torchrun leads to `ModuleNotFoundError: No module named 'tensorboard'`, but python -m torch.distributed.launch is ok, #78061 Python memory allocator called without holding the GIL when running torchrun under Python debug version",6e3f1b0d3a59c3a83a8e0e13cd9e3909eea8a2250c872287aa227f7da71c910c
issue-collapse:a859b43821264c37,oncall: pt2 stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"100241, 100637, 104531, 105092, 126190",,,,oncall: pt2,"NNC, compile-docs, enhancement, feature, oncall: pt2, topic: docs, triaged",,Handle this oncall: pt2 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#100241 Interaction of torch.no_grad and torch.autocast context managers with torch.compile, #100637 Optimal Batch Size Selection in Torchdynamo Benchmarks for Different GPU Memory Sizes, #104531 [RFC] Optional Modular Representation for FX Graph from `torch.compile()`, #105092 [RFC] Proposal to upgrade LLVM version, #126190 Centralized Mechanism To Pytorch Query Global Flags",9b73171c5c7b56a15ebb62d01537277505cb6d523ddb97e86e165e21adeed052
issue-collapse:35414a2f5bc7961f,oncall: profiler stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"58447, 60146, 78871, 155001, 158364",,,,oncall: profiler,"enhancement, oncall: profiler, triaged",,Handle this oncall: profiler stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#58447 TSAN issues in at::RecordFunction, #60146 [torch.profiler] enhancements + corrections, #78871 [RecordFunction] Hold a durable schema reference, #155001 The profiler does not seem to be able to record cuda runtime nodes, #158364 [feature request] [memory_viz] Flexible time/range memory event filtering controls to skip loading all memory events from pickle in the browser's memory",b72e7958933b36df4ec09d0a22048198d61ea9b80ec5e1066d361e4734dfac3a
issue-collapse:49e84820cc953bd0,oncall: distributed checkpointing stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"119070, 126047, 126881, 137327, 146566",,,,oncall: distributed checkpointing,"enhancement, oncall: distributed checkpointing, triaged",,Handle this oncall: distributed checkpointing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#119070 [DCP] Review and update examples and docstring to reflect DCP save/load API updates, #126047 `torch_save_to_dcp`'s produced dcp checkpoint doesn't fit the load interface for FSDP, #126881 [DCP] DCP does not support objects which are lazy initialized., #137327 [torch2.4][Distributed Checkpoint] new flatten logic is error-prone, #146566 [DCP] Allow for rank-specific tensors with duplicate keys",1400c222e0ae6a48ebb96d68c67511d4f52934d32048b6194e3aeca416f71b78
issue-collapse:b821ca7202a4f5df,module: torchbind stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"42949, 42950, 85392, 133689, 170458",,,,"module: torchbind, module: vllm","enhancement, module: torchbind, module: vllm, topic: performance, triaged, vllm-compile",,Handle this module: torchbind stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#42949 Verify TorchBind works with nested class type, #42950 TorchBind C++ Enum Class, #85392 please report a bug to PyTorch. Expected Object but got PyObject, #133689 [torchbind x PT2] Handle vllm's ScalarType, #170458 must_in_dispatch takes a long time in torch registered ops",154c72ac8e7b5b514011cdce17ace2e8e6233877898a29b1463c637bd9ac10f0
issue-collapse:f18c66c217461b01,module: pytree stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"109460, 111962, 113378, 116987, 159848",,,,module: pytree,"module: pytree, triaged",,Handle this module: pytree stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#109460 [BUG][pytree] treespec serialization for locally defined classes and namedtuple types, #111962 [pytree] `pytree.tree_map` does not respect type of `torch.Size`, #113378 [pytree] implement key path API, #116987 [pytree] class based registration api, #159848 `tree_flatten(some_pytree, lambda x: isinstance(x, Tensor))` yields `[None], _`",d1865f20f0fe5f728a48c01b6f802c8435ce3bd4eefc2fd25a453a6a8ee09156
issue-collapse:2d095069b97fbae2,module: pt2-dispatcher stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"121967, 158401, 161937, 168977, 169152",,,,"module: pt2-dispatcher, oncall: distributed, oncall: pt2","module: pt2-dispatcher, oncall: distributed, oncall: pt2, triaged",,Handle this module: pt2-dispatcher stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#121967 [PT2.2] PyTorch dispacthing non core ATen ops in .compile mode for torch ops., #158401 torch.compile hits internal tuple index out of range inside subclass_meta.unwrapped_idx, #161937 AOTAutograd runtime alias reconstruction applies view on non‑contiguous base under torch.compile (inductor + DDP): “Use .reshape(...) instead”. pre_grad_custom_pass and joint_custom_post_pass can’t override, #168977 Easier way to correlate DebugMode with FX graph, #169152 [debug mode] support pipeline parallelism",d8096e138149627c8db416f21c52cb75bcd5bf5cc1a9ac83aaf9c6fc5738fbb4
issue-collapse:e7d665a7439d3244,module: nccl stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"32162, 37598, 40095, 40633, 44169",,,,"module: nccl, module: rpc, module: tensorpipe, oncall: distributed","module: nccl, module: rpc, module: tensorpipe, oncall: distributed, triaged",,Handle this module: nccl stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#32162 Using Tensor.to(device) after distributed all_reduce intermittently causes deadlock with NCCL, #37598 NCCL fails to find cuda include dir, #40095 Bump up NCCL to 2.7.3, #40633 RuntimeError: broken pipe from NCCL, #44169 Deadlock with RPC and dist.barrier() for TensorPipeAgent and NCCL.",482f8ed2a5dfd3ebcecb10684d5d2b7a91f42d5c3bd4697558c2c4d8c00be0eb
issue-collapse:f6a8e46fcadfa53d,module: mpi stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"22791, 25570, 51207, 134700, 139178",,,,"module: mpi, module: nccl, oncall: distributed","enhancement, feature, module: mpi, module: nccl, oncall: distributed, triaged",,Handle this module: mpi stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#22791 How to use mpi backend without CUDA_aware, #25570 Remote memory access similar to MPI one-sided in pytorch, #51207 Initialize NCCL backend with MPI, #134700 [c10d][MPI] Attempting to create a new group after MPI causes ""RuntimeError: Underlying Non-PrefixStore shouldn't be null"", #139178 the mpi backend not support to transfer float16/bfloat16 tensor and would post ""IndexError: map::at"" when transferring float16/bfloat16 tensor",f3432291095cb2548aec48dbae19a037c037ec515c2a9ebf8cbe806a98880020
issue-collapse:a29cb222796f8f77,module: higher order operators stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"152870, 155230, 157919, 158566, 165999",,,,"module: higher order operators, module: pt2-dispatcher, module: user triton, oncall: pt2","dependencies, module: higher order operators, module: pt2-dispatcher, module: user triton, oncall: pt2, triaged",,Handle this module: higher order operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#152870 Docs Update `wrap_triton`, #155230 [scan] scan is broken in nightly, #157919 tighten pt2 FX graph invariants, #158566 torch.compile fails with UncapturedHigherOrderOpError when using torch.ops.higher_order.scan, #165999 [RFC] Improve the performance of `associative_scan()` with `cuda.compute`",1273bd70e36ded6946c7d8a08850729389119bf2b115dc4c8c2e30551d1730e3
issue-collapse:5496fb7ff3417ad8,module: flex attention stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"139131, 144810, 151695, 157015, 163428",,,,"module: flex attention, module: functorch, module: higher order operators, module: inductor, module: pt2-dispatcher, module: vllm, oncall: pt2","module: flex attention, module: functorch, module: higher order operators, module: inductor, module: pt2-dispatcher, module: vllm, oncall: pt2, triaged, vllm-compile",,Handle this module: flex attention stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#139131 [FlexAttention] Update the way we generate kernel options, #144810 Support torch.func.grad for Flex Attention, #151695 FP8 Support for FlexAttention, #157015 [WIP][RFC] Compilable flex_attention + Context Parallel, #163428 [vLLM] Crash during inferece in torch dynamo on 2xRTX2080Ti",c6b4d4569af9827c8c733a80224b8ab172bcc2aaa6fdcc16da6796bc1ecceb09
issue-collapse:4de61e830bfe32f5,module: dlpack stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"82610, 82823, 96318, 136250, 148306",,,,"module: dlpack, module: functorch, module: vmap","feature, module: dlpack, module: functorch, module: vmap, triaged",,Handle this module: dlpack stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#82610 contiguous() not work for rank 1 length 1 tensor., #82823 getDLContext in DLConvertor.h cannot be found, #96318 Support managed memory backed dlpack with torch.from_dlpack, #136250 A question about `getDLDevice`, #148306 torch.vmap incompatibility with DLPack functions",b7a0cc25b9ed578266a17865d6a0cb69747f48a9e204002d358c95db244909e9
issue-collapse:3d68b37a16650dd1,module: devicemesh stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"114629, 114632, 125358, 142469, 160337",,,,"module: devicemesh, oncall: distributed","module: DeviceMesh, oncall: distributed, triaged",,Handle this module: devicemesh stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#114629 [DeviceMesh] `init_device_mesh` does not support CPU-only, #114632 [DeviceMesh] back DeviceMesh initialization by custom_pg, #125358 [DeviceMesh] Add support for `group: Tuple[ProcessGroup, ...]` in `from_group()`, #142469 Using DeviceMesh Slicing with torch.compile(), #160337 [RFC] Simplify bookkeeping of DeviceMesh slicing, (un)flattening, ...",411f8545559ef6bd8e57f8fd8cbd8d93eb468be93a235c24609886b10f142371
issue-collapse:e30801e92fed01a1,module: custom-operators stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"46409, 57392, 97329, 110461, 132534",,,,"module: custom-operators, module: dataloader, module: deadlock, module: dispatch, module: internals, module: library","enhancement, module: custom-operators, module: dataloader, module: deadlock, module: dispatch, module: internals, module: library, triaged, upstream triton",,Handle this module: custom-operators stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#46409 Custom ops get stuck in multiprocess data loader under certain environments, #57392 Simplified API for custom inplace & view kernel, #97329 torch.ops.aten.pow(2.0, 3) return unexpected value with complex type, #110461 Custom tensor attributes not preserved with registered functions, #132534 triton import inside function doesn't work",b9e949b4b821e19f5bec9eb9d9dd7368219dcb1cf77ce16ae3792bb0643593a9
issue-collapse:2b4311583da0cb83,module: cpp-extensions stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"16873, 17849, 18805, 54077, 78708",,,,"module: cpp-extensions, module: cuda, module: docs, module: tests","enhancement, module: cpp-extensions, module: cuda, module: docs, module: tests, triaged",,Handle this module: cpp-extensions stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#16873 Allow building C++ custom ops that imports another custom ops, #17849 Tensor::options() returns false for requires_grad when it is true, #18805 Running custom operator tests manually is too difficult, #54077 [cuda extension] tutorial should mention device guard, #78708 BuildExtension does not choose correct CUDA installation",30c04e822a307f4a63cbcca26af7451ed397ee32f216e69f49c4805f89a0d96b
issue-collapse:4d299adcc404d859,module: compile ux stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"131263, 133944, 147428, 150021, 158873",,,,"module: compile ux, module: dynamo, oncall: pt2","dynamo-triage-jan2025, empathy-day, feature, module: compile ux, module: dynamo, oncall: pt2, triaged",,Handle this module: compile ux stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#131263 An API to re-enable dynamo-disabled functions, #133944 torch.compile ""run at compile time"" decorator., #147428 Add a config to allow print and generate all recompile reasons and not stop at first., #150021 [ued] I want to see the full sizes/strides in TORCH_LOGS=recompiles, #158873 Erroneous Recompile Warnings with Multiple Models",d85e5956c0317dec610c20722eb1371d74b1e5b2f40f719365a1170dd25f82a7
issue-collapse:e930ec4b873a5ea2,module: collect_env.py stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"18658, 72768, 96447, 124025, 144615",,,,"module: collect_env.py, module: devx, module: docs, module: mpi, module: nccl, module: numerical-stability","enhancement, module: collect_env.py, module: devx, module: docs, module: mpi, module: nccl, module: numerical-stability, triaged",,Handle this module: collect_env.py stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#18658 [FR] add CPU information in collect_env.py, #72768 Add NCCL and MPI version printing to torch.utils.collect_env, #96447 Information about CPU in `collect_env` is too verbose, #124025 Different results for single sample vs. batched, #144615 `collect_env.py` fails with `'NoneType' object has no attribute 'splitlines'` if pytorch is installed without pip",756b9efc3a0c1d0bb8349d258fed8972efd17719e16479ce84a28c05bc71da5e
issue-collapse:c5c5a3f14d0c773c,module: android stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,5,0,0,0,"32919, 34799, 45560, 63082, 64311",,,,"module: android, oncall: jit, oncall: mobile","feature, module: android, oncall: jit, oncall: mobile, triaged",,Handle this module: android stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,5,Represents 5 open issues; raw issue rows remain available for audit.,"#32919 Torchscript used to work, but now it fails with VariableTensorId error, #34799 Is it possible to run an object detection android app by using Pytorch Mobile?, #45560 Android: allow preallocation of output buffers, #63082 script/build_android nn::Module support, #64311 issues flutter in android studio",292f46f2ff7cbe15d094304f5570b3d5d6c2811b68e3abd74437fd706e683a65
issue-collapse:97af25c2acdf1088,oncall: releng stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"61261, 69009, 73985, 81654",,,,oncall: releng,"Stale, oncall: releng, triaged",,Handle this oncall: releng stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#61261 libtorch_cpu.so is getting too large., #69009 Mamba does not respect `cpuonly` when creating a conda environment from YAML file, #73985 Release note bot is not sending messages on PRs anymore, #81654 [packaging] Conda install missing python local version label (+cu123 or +cpu)",72a9d9af5b4e96d9275334a3eedb79069dcd9191b8f593e610b2fe85fce57a96
issue-collapse:97925fbc151f8465,module: typing stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"38876, 42971, 43091, 43617",,,,module: typing,"module: typing, triaged",,Handle this module: typing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#38876 Override `__call__` instead of `forward`, #42971 Enable torch.quantization.fuse_modules typechecks during CI, #43091 Enable torch.optim typechecks during CI, #43617 Type hints from _VariableFunctions and elsewhere clash",5d08d5bbb01d97d641d23c2b966287cb72a691b43c0f6c58fb21a26b164b9e76
issue-collapse:375463d7139b6639,module: tests stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"6135, 19016, 20245, 23103",,,,"module: tests, oncall: distributed","enhancement, module: tests, oncall: distributed, triaged",,Handle this module: tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#6135 Test suite should test implementations, #19016 Clean up and consolidate DDP tests, #20245 ProcessGroupMPI tests in test_c10d.py, #23103 Test utility for non-contiguous tensors",8e8932bb3c13fe8d32d5f84f1ed1b7434f04ed56be773656b473ff349ddcbcb4
issue-collapse:ecf71d1d4c881f43,module: static linking stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"46464, 48167, 48450, 57065",,,,module: static linking,"enhancement, module: static linking, triaged",,Handle this module: static linking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#46464 [META] Assert that expected libraries appear in libtorch, #48167 [libtorch] Build shared library with libtorch and compile a static library, #48450 Qt 5.14.2, libtorch1.7.0 cuda10.2，Error after using header file, #57065 Static Linking Pytorch",4670c396c182b5f8a8a48ba6ff56ea04b71f57bdea146c37bfedbed791d9baad
issue-collapse:b9fd41f95d47bc6d,module: opcheck stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"111924, 111930, 111931, 112251",,,,module: opcheck,"module: opcheck, triaged",,Handle this module: opcheck stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#111924 [opcheck] Cannot share failures_dict between multiple tests with differing sets of tests they run, #111930 [opcheck] Way to reduce Hypothesis sampling when running opcheck, #111931 [opcheck] Faster gradcheck execution, #112251 Optest is unable to xfail on parametrized tests",7b405be3da332066cf132ef064f98c3be9239fb0b538f71cc3421c5406f79a69
issue-collapse:6d7811082528f51d,module: jetson stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"72581, 78067, 94614, 144033",,,,module: jetson,"feature, module: jetson, triaged",,Handle this module: jetson stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#72581 Zero-copy on shared memory of NVIDIA Jetson devices, #78067 Installation on Jetson target board, #94614 Jetson CI needs Updates, #144033 Training fails with Torch 2.1.0 on Nvidia Jetpack 5.1.2",bab1309d04147cd5855e5270ef06f9aa5d232eee33855b5f6a2ad10b0025fd6f
issue-collapse:ea2e8bda12100136,module: flop counter stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"144001, 144564, 145947, 165570",,,,"module: flop counter, oncall: distributed","dependency issue, module: flop counter, oncall: distributed, triaged",,Handle this module: flop counter stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#144001 torch.utils.flop_counter.FlopCounterMode, #144564 Tabulate not official dependency of PyTorch but needed by features like FlopCounterMode, #145947 flops counter error with PyTorch2.5 and 2.6, #165570 Add support for element-wise operations to flop counter",b6ed5801ba6ee2da6a5556f5fa86f2927492991c363eea2e24ba483a072b29d3
issue-collapse:75597d623f24b9cc,module: faketensor stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"106130, 113456, 115362, 164732",,,,"module: faketensor, module: functorch, module: onnx, module: pt2-dispatcher, module: sdpa, oncall: pt2","module: fakeTensor, module: functorch, module: onnx, module: pt2-dispatcher, module: sdpa, oncall: pt2, triaged",,Handle this module: faketensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#106130 torch._subclasses.fake_tensor.DynamicOutputShapeException: aten.nonzero.default, #113456 Expose FakeTensor and FakeTensorMode as public APIs, #115362 torch._fused_sdp_choice for fake and real tensors returns different results, #164732 Fake tensor leakage in aot_autograd by itself",7974dcbb1a4581a5c428c43f489adb313bcdf384f925acea49e53ab08b33f605
issue-collapse:82f9334a11f133f9,module: convolution stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"22532, 23068, 25821, 26396",,,,"module: convolution, module: cpp, module: cuda, module: cudnn, module: docs, module: nn, module: nnpack, module: performance","enhancement, low priority, module: convolution, module: cpp, module: cuda, module: cudnn, module: docs, module: nn, module: nnpack, module: performance, triaged",,Handle this module: convolution stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#22532 Know which function is used by conv and force to use a function, #23068 [c++] torch::conv2d() expected output_padding to be a single integer value or a list of 3 values, #25821 Conv2D 2x~20x slower than Tensorflow when channel count is small, #26396 NNPACK condition should be changed (ARM processors)",8bfca565e6c6bf2f4c809d117d3be6525550db81c986d6c4ba31676a78b67e9c
issue-collapse:fe9a03aa1985f47e,module: codegen stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"4102, 40250, 42863, 46061",,,,module: codegen,"module: codegen, triaged",,Handle this module: codegen stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#4102 Make the generator tools data model more explicit, #40250 Include expanded TensorOptions version of op in at:: namespace, #42863 TensorOptions extensibility has rusted shut, #46061 Support undispatched ops in codegen",98fa043a98a0374dec5628f02eec491079577e404164dad75edc0cb26f61813b
issue-collapse:846393eb63d8b1d1,module: bfloat16 stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"22669, 26136, 28845, 54841",,,,"module: bfloat16, module: performance, module: tests","enhancement, module: bfloat16, module: performance, module: tests, triaged",,Handle this module: bfloat16 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#22669 BFloat16 numeric limits should contain more info, #26136 There should be gating around BFloat16, #28845 Implement operator<< for bfloat16, #54841 [testing] test_reference_numerics_extremal_clamp_cpu_bfloat16 fails on ci build with GCC5.4 and Python 3.6",5ed307930f2e0e057bd28aa633641d320faab53b38633ef893041a16c349ac5d
issue-collapse:3338b788f7e35787,module: assert failure stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"5857, 29026, 32036, 32227",,,,"module: assert failure, module: cuda, module: data parallel, module: error checking, module: sparse","module: assert failure, module: cuda, module: data parallel, module: error checking, module: sparse, small, triaged",,Handle this module: assert failure stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#5857 Change THCudaCheck to suggest that device-side asserts likely mean that you have out of bound indices, #29026 RuntimeError: !t.is_cuda() INTERNAL ASSERT FAILED at /pytorch/aten/src/ATen/native/sparse/SparseTensorMath.cpp:591, #32036 Internal assert failed using multiple GPUs with DataParallel, #32227 0 INTERNAL ASSERT FAILED at /pytorch/c10/util/intrusive_ptr.h:348, please report a bug to PyTorch.",84353390c613485e2c58baab841898cb33f700b22ab6bffa1d12d9c91d28a0f1
issue-collapse:9685727ac7305eca,feature stale/low urgency  batch 5,batch action queue,stale/low urgency,2,medium,bucket_component_queue,4,0,0,0,"1487, 6357, 25381, 31085",,,,,"caffe2, feature, low priority, todo, triaged",,Handle this feature stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,4,Represents 4 open issues; raw issue rows remain available for audit.,"#1487 [feature request/proposal] Relax scale_factor for nearest neighbor upsampling, #6357 [feature request] Unpooling layer in Caffe2, #25381 regarding builtin_function_or_method, #31085 Common lookup of generic types across full script and mobile parsers",6a9f0c742f97eca6df2134121df4fb17d86473761ad64561e11dfdc45b5e568e
issue-collapse:e4ac37d0462637fb,oncall: package/deploy stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"60977, 65261, 71317",,,,oncall: package/deploy,"days, imported, oncall: package/deploy, triaged",,Handle this oncall: package/deploy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#60977 [package] pattern usage can potentially create broken packages, #65261 Add an `all_paths` method to PackageExporter, #71317 [torch.deploy] Using zipped torch modules in torch.deploy interpreter",90b7622dd39ace16ff931266fca6cd76135b8782c2b76ee95c99e26c95033c0e
issue-collapse:797e09b48700e255,module: vectorization stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"10956, 43033, 56187",,,,module: vectorization,"better-engineering, feature, module: vectorization, triaged",,Handle this module: vectorization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#10956 Better dev docs for writing native CPU kernels with Vec256, #43033 No vectorization for int8 and uint8, #56187 AVX512 and Vec512",dace31e01c9d0ff8dcaf2b6acfeae1c3ece8edb046958ea5312b3328297dc1dd
issue-collapse:83f2954fdaba110e,module: user triton stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"141686, 152608, 155519",,,,"module: user triton, oncall: pt2, oncall: releng","module: user triton, oncall: pt2, oncall: releng, triaged",,Handle this module: user triton stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#141686 [PT2] Fine-grained custom triton kernels support, #152608 [triton pin update] Run Inductor CI on pin updates for Triton and the PyTorch nightly branch, #155519 [Tracking] Triton 3.2 deprecation",4aba5f4b351a73ef3c963592283acacf3aa281a9a53ba843ed63bc50b40889fb
issue-collapse:33595c6f8ef32379,module: third_party stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"70392, 149324, 150118",,,,"module: third_party, oncall: pt2, oncall: releng","module: third_party, oncall: pt2, oncall: releng, triaged",,Handle this module: third_party stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#70392 third_party/breakpad/ compilation failure, #149324 Unguarded Usage of Facebook Internal Code?, #150118 PyTorch 2.6 License Issues",bcf7de16f9d7bde2cdf8705eece0d3cc35bb74b74d16ea16dec2c3c675c0144e
issue-collapse:ded0371d794d6ba6,module: rocm stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"64426, 69233, 95921",,,,module: rocm,"Stale, module: rocm, triaged",,Handle this module: rocm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#64426 Add support for ROCm MIOpen miopenConvolutionForwardBias and miopenConvolutionBackwardBias, #69233 [ROCM] elementwise kernel launch, #95921 ROCm distributed flaky on test_distributed_spawn",d8b9e9a5501a499a981cb53b613e4ba23d22da1d99f4c95d14445aaa25dabc24
issue-collapse:7e37a43d556bf054,module: python array api stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"62255, 116610, 150368",,,,module: python array api,"module: python array api, triaged",,Handle this module: python array api stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#62255 Support `torch.linalg.trace`, #116610 Align `torch.sign` with Array API, #150368 Support `copy` kwarg in `torch.reshape()` following Python array API standard",0a0fba9dd43891d1223f6a7c4c5e26289b664942deda167767edf7546ef6dfea
issue-collapse:298f1f41c2146067,module: protobuf stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"24037, 69199, 127270",,,,"module: protobuf, module: tensorboard","caffe2, module: protobuf, module: tensorboard, triaged",,Handle this module: protobuf stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#24037 Tests do not pass with the latest protobuf, #69199 Protocol buffers library mismatch., #127270 Not supporting higher protobuf versions",11a3a95c5ee93f3021c32688f48aa2a7d552274641c863e843ac3d964d4a7566
issue-collapse:3f5a4b62dbb03c50,module: numpy stale/low urgency  batch 7,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"19172, 36058, 38272",,,,module: numpy,"enhancement, module: numpy, triaged",,Handle this module: numpy stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#19172 [Feature Request] Common constants in the torch.* namespace, #36058 Half type promotion with Numpy arrays is incorrect, #38272 PyTorch's rot90 returns a new tensor, inconsistent with NumPy's returning a view",b5e0899571f792f9416968724c8184c1724d759fd0649b5c412755c99b084038
issue-collapse:fb90c5898b50aa47,module: meta tensors stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"55192, 65577, 71305",,,,"module: meta tensors, module: structured kernels","enhancement, module: meta tensors, module: structured kernels, triaged",,Handle this module: meta tensors stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#55192 Consider adding meta:1, meta:2 devices, #65577 Support cat() for meta tensors, #71305 Convert a tensor with type caffe2::TypeMeta to std::vector",fb8bda5ac09c1df386b2ee058a919f949950bd14cfc543ccfe65a9ac18aeaa1e
issue-collapse:10b102c285d638b5,module: lts stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"71340, 73424, 77463",,,,module: lts,"module: lts, triaged",,Handle this module: lts stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#71340 locally installed PyTorch upgraded or superseded on windows/macos conda tests, #73424 ""munmap_chunk(): invalid pointer"" interaction error with pytorch (< 1.10), pybind, and cv_bridge, #77463 conda CPU installation for LTS fails with UnsatisfiableError",bba5aadc5f8d2c0bb039d508572983b45da4e4905193fe7e545fd33aef86893f
issue-collapse:1c3f34713b14b2ca,module: language binding stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"28233, 32698, 54035",,,,"module: language binding, module: pybind","feature, low priority, module: language binding, module: pybind, triaged",,Handle this module: language binding stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#28233 torch native functions cannot be used with inspect.signature, #32698 Add Julia Bindings to Torch backend, #54035 Web support?",c86c95aebe377bbd403035fcf763435fee7993e4ee48faaf8e7501fdd7f48cd9
issue-collapse:01708544e5ca1c56,module: functionalization stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"79867, 81385, 83923",,,,"module: functionalization, module: primtorch, module: viewing and reshaping","module: functionalization, module: primTorch, module: viewing and reshaping, triaged",,Handle this module: functionalization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#79867 All {view}_scatter variants should support all (or most) dtypes, #81385 ""Attempted to resize a view tensor to a larger size. This is not allowed in the functionalization pass"" reported on non view tensor, #83923 Support primtorch view ops in functionalization",b66e1bb74f051f090e35bd008355944efcc118a449f4cf7159f4c407e0d3f026
issue-collapse:27800f2cf704f9df,module: dtensor stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"105213, 106566, 116101",,,,"module: dtensor, oncall: distributed","module: dtensor, oncall: distributed, triaged",,Handle this module: dtensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#105213 [DTensor] Module parallelized with ColwiseParallel should return a sharded tensor, #106566 DTensor Sharding prop cache stats, #116101 [DTensor] Support API to shard to parent mesh",48d6e06d80fece9c4b6dff65c8a54b36898595ee7eb78d55c3360b1e63c47223
issue-collapse:70006651eb392aab,module: copy on write stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"97856, 109833, 132485",,,,"module: copy on write, module: cuda, module: internals, module: viewing and reshaping","module: copy on write, module: cuda, module: internals, module: viewing and reshaping, triaged",,Handle this module: copy on write stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#97856 make tensor data const correct, #109833 Implement Copy-on-write (COW) tensors, #132485 lazy_clone does not work for pinned memory",ffb52eeb9616f9e5bfaae8afa3611b9e272fa001bca022a7e82b9a1bd36d7e5b
issue-collapse:aeae13b6876ae3b9,module: amp (automated mixed precision) stale/low urgency  batch 6,batch action queue,stale/low urgency,2,medium,bucket_component_queue,3,0,0,0,"38788, 42311, 44724",,,,"module: amp (automated mixed precision), module: cudnn, module: memory format","module: amp (automated mixed precision), module: cudnn, module: memory format, triaged",,Handle this module: amp (automated mixed precision) stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,3,Represents 3 open issues; raw issue rows remain available for audit.,"#38788 Convergence issues when using pytorch's native AMP, #42311 Inconsistent results when trying to enable Tensor Cores on NVIDIA T4, #44724 16-bit input + AMP breaks AdaptiveLogSoftmaxWithLoss",d3c6ddd5691d83fa324e473260ab75a0d44271870772c35f1742022338e2cb35
issue-collapse:cdb83c0dd50456b0,oncall: visualization stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"34953, 36028",,,,oncall: visualization,"oncall: visualization, triaged",,Handle this oncall: visualization stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#34953 Graphic tool to view the backward(Gradient Graph) and forward graph in Pytorch, #36028 About tensorboard in pytorch record graph in different state",64ac6267e7dcb0aed83d467951194daab895a5d502cf584586308401619aca1c
issue-collapse:96cbf4e37301850f,oncall: java stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"45399, 58973",,,,oncall: java,"enhancement, feature, oncall: java, triaged",,Handle this oncall: java stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#45399 Load ExtraFilesMap with Java library, #58973 First-class Java API",d8b3d5090e670b66de089f0a45beedafbcf4c7a817c5c90fef25970b6728d996
issue-collapse:4a429cef38e92723,module: wsl stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"97503, 111187",,,,"module: wsl, oncall: distributed","module: wsl, oncall: distributed, triaged",,Handle this module: wsl stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#97503 torch.cppExtension won't work with wsl2, #111187 torch.distributed.DistBackendError: NCCL error in: ../torch/csrc/distributed/c10d/ProcessGroupNCCL.cpp:1331, unhandled cuda error (run with NCCL_DEBUG=INFO for details)",31dc2ba6f4a7a9d666ad251f76b686bae039d4bb0cb19d4fdb498a4f281c44e2
issue-collapse:c4bc0039f9ec2ed1,module: undefined reference stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"56144, 58591",,,,module: undefined reference,"module: undefined reference, triaged",,Handle this module: undefined reference stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#56144 libtorch static linking results in undefined references to onnx_torch and caffe2::EmbeddingLookup, #58591 link error on BatchNormImplBase<D, Derived>::pretty_print when using icpc",7a1d4b29c2a87522f2e8bf61e4d72e20d62d1c5ca118ac72788672ffadf82442
issue-collapse:ea43fa796b75de23,module: tf32 stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"69921, 76025",,,,module: tf32,"module: tf32, triaged",,Handle this module: tf32 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#69921 Context manager to enable/disabled TensorFloat32 on demand, #76025 Numerical instability: matrix multiplication got different results on cpu and gpu",3026f0c7a66bff9ccd7c9a3731d4752301d2a0d432f1bd5e000b15a5282d1263
issue-collapse:a0cb7f2a787b4587,module: tensoriterator stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"33166, 70238",,,,"module: tensoriterator, module: vectorization","enhancement, module: TensorIterator, module: vectorization, triaged",,Handle this module: tensoriterator stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#33166 TensorIterator does not work with different input/output types, #70238 `TensorIterator`: refactor `build_ternary_op` to match binary versions",cb06c5c4c8c36fbd2b4cbc33b41d653cb1d07c7a80dd79abd1faab8eb509b687
issue-collapse:1ec54e5f971d8188,module: pruning stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"85397, 157218",,,,module: pruning,"module: pruning, triaged",,Handle this module: pruning stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#85397 torch.nn.utils.prune.remove reorders the parameters of a module unexpectedly, #157218 [Feature] Taylor expansion pruning",7120c6dcf0b92d76e07123b2afcac64dfde3ccd3ae524fd421b39f7570d4f109
issue-collapse:209a138853874a68,module: proxytensor stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"89060, 132759",,,,"module: proxytensor, module: tests","module: ProxyTensor, module: tests, triaged",,Handle this module: proxytensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#89060 Extend test_proxy_tensor tests to support ops test non floating point types, #132759 Delete ProxyTorchDispatchMode.enable_tracing",885bd4c90aafee0a05a50346ba865703a0d2f9f06f1bdb64c615ab1ce7e8dab8
issue-collapse:f592d9a8dac5d657,module: porting stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"11439, 18351",,,,"module: porting, module: tensor creation","better-engineering, module: porting, module: tensor creation, triaged",,Handle this module: porting stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#11439 High leverage TH operations to port to ATen, #18351 Port SpatialConvolutionMM and VolumetricConvolutionMM to ATen",e35822fe144f4a5485a155743ea9ec0625e979e4ea86367d034fd8980639996c
issue-collapse:a701538d0684973d,module: partial aliasing stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"82303, 144861",,,,module: partial aliasing,"module: partial aliasing, triaged",,Handle this module: partial aliasing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#82303 Race condition between torch.tensor's view and /= (/= returns incorrect result), #144861 Region check for in-place read and write does not always work",4a86d5ff4189904c345a4e095f4cc506a75db66f72d7bc9caca727b85fc8746d
issue-collapse:39fb1d43358b4918,module: padding stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"32912, 170536",,,,module: padding,"function request, low priority, module: padding, triaged",,Handle this module: padding stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#32912 Optional seq_len argument to torch.nn.utils.rnn.pad_sequence, #170536 accept int padding value for F.pad",b7a04b7e53b317caaa966dbc87d8dde5245f4d6503e2bbe55a96422b64fb643a
issue-collapse:0e89814407e3c645,module: nnpack stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"84045, 107534",,,,module: nnpack,"module: nnpack, triaged",,Handle this module: nnpack stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#84045 COREMLTOOLs/NNPACK Python Issue, #107534 NNPACK slow down M1/M2 Mac CPU",8073bd235562d60c6a2e1cd17ba0efd9cc6d7b2887b16e49213acce72cfe9ad6
issue-collapse:adc07e5610338a1c,module: nn.utils.parametrize stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"102740, 125064",,,,"module: nn.utils.parametrize, module: rnn","module: nn.utils.parametrize, module: rnn, triaged",,Handle this module: nn.utils.parametrize stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#102740 parameterizations.orthogonal does not work as intended with nn.GRU or nn.LSTM, #125064 ValueError: weight_norm of 'weight' not found in ParametrizedConvTranspose1d",99a2f379115c65d0102fdb75bb7368900524b9a7f6ff067cf820f4e383156dc5
issue-collapse:8e6b9fa9a9342882,module: multi-gpu stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"78185, 87794",,,,module: multi-gpu,"actionable, module: multi-gpu, triaged",,Handle this module: multi-gpu stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#78185 test_to (__main__.TestTorch) fails with multiple gpus, #87794 CUDA error: operation not permitted when stream is capturing (2 GPUs)",229d543947f7d372a68075aed29b6b80109bef499cc38e34ab85776856a90662
issue-collapse:3bbb30485dee7af7,module: molly-guard stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"76383, 116533",,,,"module: molly-guard, module: nn, module: python frontend","actionable, module: molly-guard, module: nn, module: python frontend, triaged",,Handle this module: molly-guard stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#76383 Suggestion to throw a UserWarning when a user forgot .eval() mode during inference, #116533 `backward` implicit conversion from tuple to `torch.Tensor` results in an indexing error message",e0c35d1490e69fd7ed70d2fc9bbb837f2d5c70bc57e81679f79673be35291190
issue-collapse:5aec6b667335882f,module: mkl stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"74167, 93275",,,,"module: mkl, module: sparse","module: mkl, module: sparse, triaged",,Handle this module: mkl stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#74167 Slower performance of `torch.mm` method with sparse CSR tensor, #93275 Bug in torch.linalg.svd",9a7d531e6360bb34b659fda343ab322543f8ea32b50671ef53cc4a297ab29f73
issue-collapse:1b051bcdfbfc80d6,module: memory format stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"28221, 34812",,,,"module: memory format, oncall: jit","module: memory format, oncall: jit, triaged",,Handle this module: memory format stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#28221 TorchScript doesn't support torch.channels_last or any other memory format constants, #34812 Semantic differences between forward and backward of upsampling2d/3d in channels last (NHWC, NDHWC) format.",648d69a78398023414015745cbd34c9c24a1d5eabf4b3ae7e9368d8368826afd
issue-collapse:083e4f84ad2139b2,module: macos stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"19672, 31308",,,,module: macos,"module: macos, triaged",,Handle this module: macos stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#19672 torch.set_flush_denormal not working on some (old) OSX machines, #31308 [mac] Failure to import torch",65435015b8991c59927cdc0777c3bc9e449dcc0fc5345d3a6876d81a35adf112
issue-collapse:e23589ee8aec90f6,module: hub stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"43638, 45569",,,,module: hub,"enhancement, module: hub, triaged",,Handle this module: hub stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#43638 [FR] hub uses default github branch, #45569 Distutils Error in torch.hub Load()",35ed28c11139f838cec51969b52ce6346837760b4751fa0634b411f2687b8062
issue-collapse:7cf6ba913f10995d,module: first class dims stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"111525, 120572",,,,"module: first class dims, module: functorch","module: first class dims, module: functorch, tensor subclass, triaged",,Handle this module: first class dims stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#111525 Functorch FCD breaks with tensor subclasses, #120572 FCD and tensor subclasses",f069e8b054f36faece0782e1d86661aac4fbb8bf567c5b34c33718b41f4bf632
issue-collapse:fcfeb57671464447,module: fft stale/low urgency  batch 2,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"50552, 66305",,,,"module: fft, module: sanitizers, module: tests","module: fft, module: sanitizers, module: tests, triaged",,Handle this module: fft stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#50552 test_fn_grad_fft_fftn_cpu_complex128 and test_fn_grad_fft_rfftn_cpu_float64 are failing under TSAN, #66305 Modify convolution kernels for ops",4ba907b4aeee1f84ecc7fae4f8d2c72cd434c24cc161b84173788162654c3022
issue-collapse:2227bf50d27f37e2,module: empty tensor stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"131148, 153327",,,,"module: empty tensor, module: error checking","actionable, module: empty tensor, module: error checking, triaged",,Handle this module: empty tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#131148 The empty 1D or more D tensor for `median()/nanmedian()` with the deepest `dim` gets errors, not getting NaNs, #153327 FPE when calling `torch.pixel_shuffle()` with empty tensors and large upscale_factor",f5afcf884dc4530e34ea3054eec33a9080bf8c6a732ad65ff766d04ff25ed419
issue-collapse:48b5304607970ef4,module: determinism stale/low urgency  batch 3,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"38541, 39134",,,,"module: determinism, module: docs, module: nn","module: determinism, module: docs, module: nn, triaged",,Handle this module: determinism stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#38541 Doc update regarding predictability of experiments using Seeds and Workers, #39134 Crossentropy inconsistent results depending on tensor order",6542a31c8d3520f2c9783a85116fddff513771b2c06b10e015d8da6e1576d1f1
issue-collapse:43cc83ca78d5399b,module: derivatives stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"22274, 77973",,,,"module: derivatives, module: forward ad","enhancement, module: derivatives, module: forward ad, triaged",,Handle this module: derivatives stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#22274 second derivatives of unfold, #77973 forward-mode support for ""logically composite"" operators",c81e80c7c92ee3989812c5c8daff48405f757d2fad837cc2a719dd70a67406c2
issue-collapse:ace1a69ff0e1d7aa,module: context parallel stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"147793, 151878",,,,"module: context parallel, module: dtensor, module: sdpa, oncall: distributed","module: context parallel, module: dtensor, module: sdpa, oncall: distributed, triaged",,Handle this module: context parallel stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#147793 context_parallel fails with plain sdpa kernel SDPBackend.MATH, #151878 Potential error in calling scaled_dot_product_attention in pytorch/torch/testing/_internal/distributed/_tensor /common_dtensor.py ?",fc8b81ad64000bd215186972ab2eb04d51ef761fea4955c9e78b3135f818926e
issue-collapse:7e02796b8126544e,module: capabilitybasedpartitioner stale/low urgency,batch action queue,stale/low urgency,2,medium,bucket_component_queue,2,0,0,0,"81244, 81539",,,,"module: capabilitybasedpartitioner, module: fx, module: fx.passes","module: CapabilityBasedPartitioner, module: fx, module: fx.passes, triaged",,Handle this module: capabilitybasedpartitioner stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,2,Represents 2 open issues; raw issue rows remain available for audit.,"#81244 CapabilityBasedPartitioner does not work correctly with mutating operations, #81539 CapabilityBasedPartitioner treats non-compute ops inconsistently",aa85866e9ca90ac915bfc8103b75a0240a5a2fd4f829b217f4ad81b458cc591e
issue-collapse:990dd93250561e94,module: vllm needs triage,single issue review,needs triage,115,low,bucket_component_queue,1,0,0,0,174288,,,,"module: vllm, oncall: distributed, oncall: distributed infra","bot-triaged, high priority, module: vllm, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triage review, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#174288 [distributed] Batched isend/irecv with NCCL backend hangs on high load,3695c693e5069c1c01e9ed9dbb7e31f3cae4bcaa3b31ba6f87e5497732cfe007
issue-collapse:9faec3753d4f2d43,module: flex attention needs triage,single issue review,needs triage,115,low,bucket_component_queue,1,0,0,0,189144,,,,"module: flex attention, module: inductor, module: performance, module: regression, module: xpu, oncall: pt2","bot-triaged, has workaround, module: flex attention, module: inductor, module: performance, module: regression, module: xpu, oncall: pt2, triage review",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189144 [XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs,71b0a12d2db2aa189eb6703a3478bd003e07c86bef09ab47e6b7d884f4441324
issue-collapse:a5db04bcf82569ee,module: internals needs triage,single issue review,needs triage,97,low,bucket_component_queue,1,0,0,0,28786,,,,"module: internals, module: rpc, oncall: distributed","better-engineering, high priority, module: internals, module: rpc, oncall: distributed, triage review, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#28786 RPC couldn't match torch.ones with requires_grad=True,11a1f2d6e002bd623e73acd98189b5146c3dcf388afd6e57ffaa3aca6d908b3b
issue-collapse:4d9c42e08fc1d544,module: regression has linked PR,single issue review,has linked PR,94,low,bucket_component_queue,1,0,0,0,148317,,,,"module: regression, module: windows, oncall: pt2","module: regression, module: windows, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#148317 compile SageAttention faing error C2872: “std” for latest torch nightly,dc7fbf704d4777a487e922b139c4261d7a7b6a3a4850d2d3079f02ef2cdc2558
issue-collapse:80cae0750a06bc30,module: pytree has linked PR,single issue review,has linked PR,94,low,bucket_component_queue,1,0,0,0,65761,,,,module: pytree,"better-engineering, feature, high priority, module: pytree, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#65761 torch.utils._pytree -> stable,a7f2906f4cd61a786940e403fb267718654b69f71258e713ae35de08c45ce237
issue-collapse:a348d5d534eca7a5,module: pt2 accuracy high-priority/blocker,single issue review,high-priority/blocker,82,low,bucket_component_queue,1,0,0,0,113180,,,,"module: pt2 accuracy, oncall: pt2","high priority, module: pt2 accuracy, needs reproduction, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#113180 Higher train loss and worse evaluation metrics when using `torch.compile()`,d2b259a8da8878dab2096f5cc4849b11adef9c3d725f6f6c596feecf16efbec2
issue-collapse:dfe161eadd970275,module: guards has linked PR,single issue review,has linked PR,76,low,bucket_component_queue,1,0,0,0,139789,,,,"module: guards, oncall: pt2","high priority, module: guards, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#139789 compiling ""input mutation"" followed by ""input mutation + aliasing"" does not properly recompile",e7d4beceaf7456f26756f93ee3fcf49314ea31207d9517841c1bdf49cf75b0f2
issue-collapse:8687365b9a2dd863,module: pt2-dispatcher high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,124484,,,,"module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, better-engineering, high priority, module: pt2-dispatcher, months, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124484 HOP dispatch isn't faithful,811e0e4de1107121c3973f929bdbc0cc18e29587451e1fa95fb58f18213b93d0
issue-collapse:afc2516ffcc13cb2,module: intel high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,120601,,,,"module: intel, module: regression","module: intel, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120601 Torchrun arguments shadowing script arguments,2b60fc1a583792efe3d65d99db2cc0a04b79c1954bf8257399b3656ee6cda575
issue-collapse:d0b5e8f7b7d62a2d,module: faketensor high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,156819,,,,"module: faketensor, module: inductor, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, high priority, module: fakeTensor, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156819 FakeTensorUpdater doesn't support HOPs,5b11bad5b914374265abadbfd0c04bdcd09a1fff90dca247e41ada2f2d43abd1
issue-collapse:6888fec9815e308f,module: dtensor high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,168058,,,,"module: dtensor, oncall: distributed","high priority, module: dtensor, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#168058 [DTensor] inconsistency of distributed RNG (Shard(0) != Shard(1) or 8 GPU != 4 GPU != 1 GPU),8b455b1928f9a9652992886505d10b4c9c9c98b3c45629a4492bc0e8b9e0adea
issue-collapse:30a7e2681af40fbd,module: dispatch high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,189018,,,,"module: dispatch, module: meta tensors, module: regression","bot-triaged, matrix multiplication, module: dispatch, module: meta tensors, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189018 `_scaled_mm_v2` meta implementation checks for device-specific constraints,b0ac65d7b8f4361c8bfa37801db025778794b56d42ddf46cab2bf207caa62bac
issue-collapse:84bb1925d7d86ef5,module: custom-operators high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,184549,,,,"module: custom-operators, module: inductor, module: pt2-dispatcher, module: regression, module: vllm, oncall: pt2","bot-triaged, module: custom-operators, module: inductor, module: pt2-dispatcher, module: regression, module: vllm, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#184549 [vllm] [2.12 regression][Inductor] assert_size_stride (8192 vs 512) in BERT-with-rope AOT-compiled path for Nomic v2 MoE pooling model,cdbdbdff2107f707a44996b32f3bfd63bf43831e5f18ff8ee5f8ba653df0b06a
issue-collapse:dafb23ff1b86eee8,module: cuda graphs high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,167373,,,,"module: cuda graphs, module: dlpack, module: regression","PT2-Bug-Bash, module: cuda graphs, module: dlpack, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167373 Torch Tensor -> Nanobind ndarray breaks cuda graphs on torch 2.9,4bddcba5e5c8fba6785f39d16f73f21afc3e404d53b83ab6bb333057fd5eb31b
issue-collapse:57fa5e30b9928359,module: cublas high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,44428,,,,"module: cublas, module: cuda, module: linear algebra","function request, high priority, module: cublas, module: cuda, module: linear algebra, needs design, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#44428 Support for CUDA matrix multiplication on long (and other integer) tensors,0d95711b8b5015df4a3f289150b7cd3832cf2b1a42b23fa18e3c6f848a442db6
issue-collapse:22edae64ef760acf,module: aotinductor high-priority/blocker,single issue review,high-priority/blocker,70,low,bucket_component_queue,1,0,0,0,157128,,,,"module: aotinductor, module: build, module: regression, module: windows","PT2-Bug-Bash, module: aotinductor, module: build, module: regression, module: windows, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#157128 Windows Source Build Fails with OSError: [WinError 126] on aoti_custom_ops.dll for RTX 5080 (sm_120), Pre-built PyTorch Works",5f0e211621f8ac64a66c709ff6d305e32f1318c66f5fbaa327c58ab0aa31a91f
issue-collapse:21396b2a3c3b3113,module: xpu needs triage,single issue review,needs triage,65,low,bucket_component_queue,1,0,0,0,171689,,,,"module: xpu, oncall: profiler","module: xpu, oncall: profiler",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#171689 After enabling XPU Graph, I cannot collect profile data during the replay phase.",1f96da20a0d341d4b04f3d6efa8755a6142d07e0a06fa756c70dac2dabf0a301
issue-collapse:114cd45d83e54e4c,module: pt2-dispatcher needs triage,single issue review,needs triage,65,low,bucket_component_queue,1,0,0,0,124328,,,,"module: pt2-dispatcher, oncall: export","PT2-Bug-Bash, module: pt2-dispatcher, oncall: export",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#124328 PreDispatch tracing doesn't produce correct ""val"" for Sparse Fake tensors.",42112631f034b43f2503c9b9264fdb746571c460cc7f2ac175b25b174cf031cb
issue-collapse:5bec53438421da3f,module: deprecation needs triage,single issue review,needs triage,65,low,bucket_component_queue,1,0,0,0,184982,,,,"module: deprecation, oncall: quantization","module: deprecation, oncall: quantization, topic: deprecation",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#184982 deprecation of derived quantized dtypes (quint8, qint8, qint32)",117a1083edc89d273e83167c42560c3298eb1829128fc137b4c18c703c0d1030
issue-collapse:b470f8de50a9ffe8,module: aotdispatch needs triage,single issue review,needs triage,65,low,bucket_component_queue,1,0,0,0,189293,,,,"module: aotdispatch, module: pickle, module: pt2-dispatcher, module: python version, oncall: pt2","bot-triaged, module: aotdispatch, module: pickle, module: pt2-dispatcher, module: python version, oncall: pt2, tensor subclass, triage review",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189293 AttributeError: Can't get local object 'WeakValueDictionary.__init__.<locals>.remove' during AOTAutograd graph caching on Python 3.13,2611ea5c5060fd33f5c83cb6864bcad05030ac42b742ffee0ea718b2a98b8eea
issue-collapse:762e5fe111f1b4bb,module: activation checkpointing needs triage,single issue review,needs triage,65,low,bucket_component_queue,1,0,0,0,155924,,,,"module: activation checkpointing, oncall: distributed, oncall: distributed checkpointing","bot-triaged, module: activation checkpointing, oncall: distributed, oncall: distributed checkpointing, ptd-bot-triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#155924 Activation Checkpointing breaks ""torch.distributed.checkpoint.state_dict._get_fqns""",13cc8ff1a74b4074ab1cff75cfc5ce20279d820512e435c5eef527897283f4b0
issue-collapse:ae8eabdd43d7282b,module: nans and infs high-priority/blocker,single issue review,high-priority/blocker,64,low,bucket_component_queue,1,0,0,0,160876,,,,"module: nans and infs, module: regression","module: NaNs and Infs, module: regression, needs reproduction, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160876 F.silu(torch.tensor(float('-inf'))) returns nan instead of 0 in PyTorch 2.7.0+cu128,26d65ccda358f41a12879e5469c8dba33448d92d9f5a9b6542025e98087d189a
issue-collapse:815b4db72b51e58e,module: lint has linked PR,single issue review,has linked PR,62,low,bucket_component_queue,1,0,0,0,115016,,,,module: lint,"good first issue, module: lint, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#115016 Enable ruff rule PLW1510 codebase wide,eed7eb0700bc9c23df8450b9028e4b1083f0f770b5ce92d2d96ee559c85d641b
issue-collapse:4452d5917d00e1b6,module: nans and infs needs triage,single issue review,needs triage,59,low,bucket_component_queue,1,0,0,0,154708,,,,"module: nans and infs, oncall: quantization","module: NaNs and Infs, needs reproduction, oncall: quantization",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154708 torch.quantized_batch_norm API return different output when input is nan/inf or undefined mathematical calculation method,5b215c3bed0005002b99e15131166d60d7b5acfb79e3eefc3604b44a5b47e30e
issue-collapse:e4ec4987aa261c71,module: jiterator has linked PR,single issue review,has linked PR,56,low,bucket_component_queue,1,0,0,0,132756,,,,"module: jiterator, oncall: pt2","module: jiterator, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#132756 JITerator: Torch.prod produces RuntimeError: CUDA driver error: invalid argument,b0a78f711494591d454696b88da81d6c43e62fb4f210f2df62412e22cbe64c6f
issue-collapse:f2efb8046b9069e8,oncall: quantization high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,106432,,,,oncall: quantization,"high priority, oncall: quantization, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#106432 inconsistent dtype of scale and zero_point in observers,c6032be2734c72548578e7a8decfcdd9fc40c3475eb1882c8e518b70b1f5c5ff
issue-collapse:bf3ceeb1abf006ac,oncall: distributed high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,42793,,,,oncall: distributed,"high priority, oncall: distributed, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#42793 distributed training with c10d does not work with layerdrop in pytorch > 1.4,bd13281b98ea5de68b359caf6c36445aa69992ec50bb2077c77ee8f987574402
issue-collapse:ca79986f8e0cfa07,module: ux high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,31558,,,,module: ux,"high priority, module: ux, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#31558 More dynamic PyTorch APIs,4c58ca30272e9388ded679a2c0af39950414cfa543fa07192e6ecb234060dd86
issue-collapse:92f40d6349d4d39b,module: typing high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,58370,,,,module: typing,"feature, high priority, module: typing, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#58370 Add better python operators annotations for IDE type checking,92e94ee278ac7c57cb1c199ca93d6905bad7da4f7f8440f2c01ea54a5e385704
issue-collapse:351dc13c264fcf02,module: tests high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,39061,,,,module: tests,"high priority, module: tests, quansight-nack, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#39061 Some @slowtests are never run in CI,241aeef6fe47f24a6f0644fcd2bc25552873cf0428c8ed6eb918a23b0a15c4fc
issue-collapse:465db45efc22f133,module: tensor creation high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,48558,,,,module: tensor creation,"feature, high priority, module: tensor creation, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#48558 [FR] tensor ctors should support directly creating in shared memory,33aa11eca67cda42ebe5dc7d4e35edda212af21dd5193ddbda8cd819ddfff6ba
issue-collapse:047e55bc34a2cb19,module: rocm high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,85909,,,,module: rocm,"high priority, module: rocm, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#85909 Steam Deck Core Dump,7c1cc9083bfeb0e383513e487f856fd747109ecbda0eee1985d35b6b44126f84
issue-collapse:c78ccd7a42112714,module: pytree high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,152535,,,,module: pytree,"dependency issue, high priority, module: pytree, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#152535 optree package status in PyTorch,bb5fdc5132dd4abb41e799f32671d928888507ea76ec3b846c1aafa0a01f40f8
issue-collapse:dfc69019f6b9ce47,module: pybind high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,87283,,,,module: pybind,"high priority, module: pybind, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#87283 Missing string parsing for some parameter types in python arg parsing logic,28296d3e56b83f7b7c86fc2ee8252acda37e7cb4bfe058c28c3872b03071e414
issue-collapse:13031b3406184eb3,module: optimizer high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,110205,,,,"module: optimizer, module: regression","actionable, module: optimizer, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#110205 RuntimeError: Expected packed scalar Tensor to be of dimension 1. Got 0 instead.,6d2b3a7bfb2b27604b89d0d2facbace89e30288120deb3f739623dab2b94762a
issue-collapse:93cd38b6e2d48796,module: norms and normalization high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,100085,,,,"module: norms and normalization, module: regression","module: norms and normalization, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#100085 [regression] torch.norm with out dtype bfloat16 cause runtime error,3fec60df4b83c42a5099ac8b629544480ce144463bb58acbfa673f1e185ee7ed
issue-collapse:cde791b9d0a238b6,module: nestedtensor high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,122617,,,,"module: nestedtensor, module: nn","high priority, module: nestedtensor, module: nn, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122617 [Regression PyTorch 2.1 & 2.2] `torch._transformer_encoder_layer_fwd` outputs on CUDA/NestedTensorCUDA backends significantly diverge from truth for torch>=2.1 (SDP math)),c3c6e389ac5ccf47067004b009d1247a56ffe30eefce3639d27709f6e21d57d1
issue-collapse:6827751b107d5869,module: mkldnn high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,121150,,,,module: mkldnn,"high priority, module: mkldnn, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#121150 ☂️ oneDNN update functional/regression,b65f3dd55a27cf6591c6af2788af5af57706d7f9d15e7b24e7bdf7f9cee135bb
issue-collapse:8c4dbe9c09a92527,module: fsdp high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,102904,,,,"module: fsdp, oncall: distributed, oncall: distributed checkpointing","high priority, module: fsdp, oncall: distributed, oncall: distributed checkpointing, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#102904 Unable to checkpoint model and optimizer state when using Hybrid Sharding Strategy,4d51a921305a96dfa9fccd939a823895383e22ba1dc33596e9c848878ad022dd
issue-collapse:5d106b614244c727,module: flop counter high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,134804,,,,"module: flop counter, module: regression, module: sdpa","high priority, module: flop counter, module: regression, module: sdpa, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#134804 FlopsProfiler double counts sdpa flops,b5e1caab478bc8a2ec6a1ed0cc75b613f4d496e26d0818fa743b74e832b28b8a
issue-collapse:2fbda461ba87fa78,module: floatx (formerly float8) high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,113663,,,,module: floatx (formerly float8),"high priority, module: floatx (formerly float8), triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#113663 FP8 types should not participate in type promotion and should have no math ops defined on them,cb4324dd797d8c5e527b71551cdf9637d361b77724e5a27909daffcd6ae96913
issue-collapse:a3bca3d7de1834f0,module: error checking high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,81195,,,,"module: error checking, module: nn, module: norms and normalization, module: regression","module: error checking, module: nn, module: norms and normalization, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#81195 torch._weight_norm with specified dim returns wrong output,7c24461037ae363614f9b5bbd53ad2fde26e1a1d8344784d67d6271fd5a5422c
issue-collapse:10bb13fb8bfbe06d,module: docker high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,80431,,,,"module: docker, module: regression","module: docker, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#80431 CMake Error: File /opt/pytorch/build_variables.bzl does not exist.,46977623afd1a7d57a4e252c56b2303179804db7dba423bb452b19603659ffd9
issue-collapse:4ef4846a1c854088,module: distance functions high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,69512,,,,"module: distance functions, module: half, module: numerical-stability","high priority, module: distance functions, module: half, module: numerical-stability, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#69512 Inconsistent behavior of cosine_similarity between fp16 and fp32 inputs,cc495c33b1d7b6e58992d1f279824e1f8817261aa58163385e6528bdf89b12a6
issue-collapse:540d4e10d0737c37,module: deprecation high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,50013,,,,"module: deprecation, module: numpy","high priority, module: deprecation, module: numpy, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#50013 torch.Tensor.repeat is divergent from np.repeat,54af9aad887f2d05a4299cd9d75c085ba652167d243d7c6804c00231c014bc9d
issue-collapse:f4377d9865d8e208,module: deploy high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,71227,,,,module: deploy,"high priority, module: deploy, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#71227 Move torch::deploy tests to their own workflow job,dea5ee4cb1401e0c63bfea9e7cc99bb1066bea445ab8df1ca5ea7620a12bde10
issue-collapse:1bfca92eae52d686,module: data high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,103117,,,,"module: data, module: performance, module: regression","high priority, module: data, module: performance, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#103117 Observed regress in DataLoader spawn from PyTorch1.13 to PyTorch2.0,0ecd8e38f56bdda1f9d60fe4fe638ae6edeaf036c3f6ff180422243f04b9ff6d
issue-collapse:46d51820163f6b74,module: cuda high-priority/blocker  batch 2,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,49473,,,,"module: cuda, module: regression","module: cuda, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#49473 Pytorch 1.5+: Kernel dies / Segmentation Fault with torch.cuda.mermory_allocated() and torch.cuda.mermory_reserved(),b1ff4efa07c85a3a7091954c82fc09109bcbe2bd285faa28fd081963e77397d1
issue-collapse:e2fe6760a2f4796b,module: compile ux high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,146798,,,,"module: compile ux, module: graph breaks, module: regression, oncall: pt2","module: compile ux, module: graph breaks, module: regression, oncall: pt2, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146798 Torch 2.6 Unexpected Graph Break with SubConfigProxy,b89fdae463e6ffd2076d829222f10d0f9e3d56f985efca41e53dadef0bfa2efd
issue-collapse:6e70d5601fb235a5,module: c10d high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,59745,,,,"module: c10d, oncall: distributed","better-engineering, high priority, module: c10d, oncall: distributed, pt_distributed_rampup, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#59745 [c10d] Work objects should have a general operator<<,aa6ec7eaaf14b799ee84c5ce3d2036e578ac7c274ae7cbc8df549e06a38adb89
issue-collapse:4720f1a823c055da,module: bc-breaking high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,157461,,,,"module: bc-breaking, module: regression, module: vllm","module: bc-breaking, module: regression, module: vllm, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#157461 vLLM tests failing in torch 2.8rc but passing with torch 2.7,ed636482a69e332c4a8dcfa158fde96d91a698b0809d7d59ed9cda6be7fd98ae
issue-collapse:73ec7d180a3a2574,module: assert failure high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,51083,,,,"module: assert failure, module: dependency bug, module: regression","module: assert failure, module: dependency bug, module: regression, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#51083 apex internal assert failed,a7842693fb23ea0152efe6db2383cf9a6732bb94109ff2e7110793640010119a
issue-collapse:3dcd4fbe94579504,module: abi high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,20748,,,,"module: abi, module: cpp, module: crash","module: abi, module: cpp, module: crash, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#20748 Segmentation fault when use torch::from_blob,813b9e61afa542f0534a901332388417e2764984d72f46939bb8954580a1c289
issue-collapse:7dbcbac52c73e933,feature high-priority/blocker,single issue review,high-priority/blocker,52,low,bucket_component_queue,1,0,0,0,48365,,,,,"better-engineering, feature, high priority, triaged",,Keep these high-signal issues together so release/regression risk is reviewed early and consistently.,Review early as a priority/risk queue.,1,Represents 1 open issue; raw issue rows remain available for audit.,#48365 Custom exception for out of memory,ef4a9b98aa06c461ab1de476e99419b92925cdb108a2822549860abf55a0662f
issue-collapse:50529f24ecee42ef,oncall: releng needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,160005,,,,oncall: releng,"better-engineering, oncall: releng",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160005 [BE][download.pytorch.org] Index fix user navigation,2caabd5eab6f226bc9a83a730845dfceaff99d34c01e6a4d81738617f8209e31
issue-collapse:c773849a29f02111,oncall: mobile needs triage  batch 6,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,33124,,,,oncall: mobile,oncall: mobile,,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#33124 QNNPACK: GNU aarch64 assembler does not support 4s on neon mov,49d07c3f8442bf624834b4162bed23d0ff8e22056aa057132a975ed120df7297
issue-collapse:d1e25f2dc1fbdfc2,oncall: distributed checkpointing needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,162948,,,,oncall: distributed checkpointing,oncall: distributed checkpointing,,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#162948 Metadata object loses arbitrary attributes added dynamically,5b0440cf75212cad9e841926db800291633865ef9677f2dc29aca0d0218e9c60
issue-collapse:df0071ec1c76a365,module: xnnpack needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,92246,,,,"module: xnnpack, oncall: mobile","module: xnnpack, oncall: mobile",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#92246 test_qnnpack_add fails,4d88977a54062208fc99d779929ddd06559b2776e70bc0d99a8d9ba9dda91430
issue-collapse:d2b7562d5f13c658,module: xla needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,96448,,,,"module: xla, oncall: jit","module: xla, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#96448 Unable to move torch.jit.load-ed models to XLA devices,33e493db82ebed4f9bedc3e2a620af755099224bb08cbc8768f8b3b7fa3da3bd
issue-collapse:3e89b32f2a3a60eb,module: typing needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,123464,,,,"module: typing, oncall: jit","module: typing, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#123464 torch._C.JitType annotation is busted,79dacffc28cc0f902e2c5f316b75510ed916b4dbcc1076b776cc5cd6dd400800
issue-collapse:5b7ca938304f9739,module: testing needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,146776,,,,"module: testing, oncall: jit","module: testing, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#146776 INTERNAL ASSERT FAILED at ""/pytorch/torch/csrc/jit/testing/file_check.cpp"":607, please report a bug to PyTorch",50aef8e334a939397293105538a36c8a67523074bbb2d729e0971497567044cf
issue-collapse:1235490324b5d979,module: tensorpipe needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,60640,,,,"module: tensorpipe, oncall: distributed","module: tensorpipe, oncall: distributed",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60640 Building fails on a node with libibverbs installed with no Infiniband hardware present on RHEL8.,36ee4d7f4bb0455714f680acfe056ada6514470d53f69d5c269f681dd2f9032e
issue-collapse:8a80d988fc8ff4de,module: static linking needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,9797,,,,module: static linking,"caffe2, module: static linking",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#9797 Build torch as a submodule with static linking doesn't work (CAFFE2_PERF_WITH_AVX2 is not defined),7fda7dd156c03f5d9df529bbabec48d004bd9402b7bce2cc17d2a19457b7f042
issue-collapse:b13ba8d7f952aa85,module: sparse needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,160631,,,,"module: sparse, oncall: quantization","module: sparse, oncall: quantization",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#160631 torch.quantize_per_tensor not implemented for SparseCPU tensors,fe047011f6d164b96881e265122389d984915bb41f4a3c8d35846b9dc08c2b16
issue-collapse:60c33b06efde6ff5,module: sanitizers needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,71849,,,,module: sanitizers,"NNC, module: sanitizers",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#71849 addcmul, byte_channels_last fail test_nnc_correctness opinfo tests under UBSAN",d13b77ece6f670d726881478f34244e68e9c10ee72f67d35c87992a202f9eba9
issue-collapse:c57dfbe07ac640b8,module: optimizer needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,60440,,,,"module: optimizer, oncall: distributed","module: optimizer, oncall: distributed",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60440 DDP with cuda rpc failed with DistributedOptimizer Adagrad,c8d74a1f46494fb32dcd4dfa3e02cb314da3f5a3e1d09462b965bcec77049db8
issue-collapse:eed3696391715277,module: op-unification needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,14750,,,,module: op-unification,"caffe2, module: op-unification",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#14750 [caffe2] ConvTranspose with group attribute,1e5fddeb47f760117402ffffb92bcab1a7693c7486bf0bdd96ad263c4ac416e6
issue-collapse:39dc6dfb3082f50a,module: nn.utils.parametrize needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,60971,,,,"module: nn.utils.parametrize, oncall: jit","module: nn.utils.parametrize, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60971 jit scripting for the parametrizations,b64cab725a838a4cc1cc5919f8963e37fe5d4c439309ff1d02a70814e54e8ce6
issue-collapse:8232bb8f80ee4063,module: nn needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,115511,,,,"module: nn, oncall: export","module: nn, oncall: export",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#115511 [Export] aten::_transformer_encoder_layer_fwd is not core aten and has no decomposition.,2965a1f3c102e3b6ffbaa7692acf03010b392118eb09b207f6a237e324004176
issue-collapse:a1d949a5e70c4f74,module: nestedtensor needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,156544,,,,"module: nestedtensor, oncall: jit","module: nestedtensor, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156544 `TorchScript` does not allow accessing methods of nested tensors,85fad6a83b11468311da0dc73b37abc90d5399e1eb843d07ed3c4754f9246e6a
issue-collapse:43fea6babb973d05,module: memory usage needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,146698,,,,"module: memory usage, oncall: export, oncall: pt2","module: memory usage, oncall: export, oncall: pt2",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#146698 Possible memory leak when running test_model_exports_to_core_aten.py test,d9ecdf467b18501dde231c95aef24ffbb42f327d8a972ee4e99f09b21bb56ce8
issue-collapse:93262a850a3adb63,module: memory format needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,72110,,,,"module: memory format, oncall: mobile","module: memory format, oncall: mobile",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#72110 Conversion Error in pytorch mobile with metal,e4567d5509dd5565cacf580fd018470f6c18d032f3dbea2c84950768b98fb994
issue-collapse:9b13185993af0dbe,module: intel needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,61221,,,,"module: intel, module: mkldnn, oncall: jit","module: intel, module: mkldnn, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#61221 Unable to add a scalar to an MKLDNN tensor,f48cb28e9d4343f4d025537f296bfdd6f392b1be0638f8d5bbc129a7a48dc6cd
issue-collapse:e5e6420859af82b8,module: functionalization needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,84495,,,,"module: functionalization, oncall: jit","module: functionalization, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#84495 functionalize: Does not compose cleanly with torch.jit.script/torch.jit.trace,8024625ccfe2453abd470f09f77840b527356dcbc504990ce48c7a7de1bc03c5
issue-collapse:e29df6063862976e,module: error checking needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,149737,,,,"module: error checking, oncall: quantization","module: error checking, oncall: quantization",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149737 `INTERNAL ASSERT FAILED` in `interpolate` and `torch.import_ir_module`,ec04573e959dea6f6a48b39ae8c220792063ac5991e9d6ed7611d733220d6a82
issue-collapse:28d09feb5696d6dc,module: edge cases needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,156955,,,,"module: edge cases, module: error checking, oncall: jit","actionable, module: edge cases, module: error checking, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#156955 `torch.jit.load()` crashes with C++ `c10::Error` if file system is full (no graceful handling),2820e67db915449c3dc4bc413dbfe62325b56348341c6a580ce3a2c4bd560705
issue-collapse:5c214ad2a90a7584,module: dispatch needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,54728,,,,"module: dispatch, oncall: jit, oncall: mobile","module: dispatch, oncall: jit, oncall: mobile",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#54728 JIT prim ops v.s. dispatcher,f8df5e05566a86bbb3bc68a6f099c15d7370c51b6d4f30cdbb6db14f379500ff
issue-collapse:612e9e29c5aadaa5,module: decompositions needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,125266,,,,"module: decompositions, oncall: export","module: decompositions, oncall: export",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#125266 [export] linalg ops after core aten decomposition,c510c7246e2cdcd2138430e5eb8e55105d5376bd055b4d0404bd2d79ed0d55fc
issue-collapse:35a0b9ff6c55a5ea,module: data needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,92294,,,,"module: data, module: dataloader, oncall: jit","module: data, module: dataloader, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#92294 jit.fork stalls multiprocessing dataloader,5085fbfc5a38c99bae9d3c6fcf8e0acaeeee8716e113a5c1eb8f6d14b45824e2
issue-collapse:bf3693f4bdbd47d3,module: cudnn needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,6898,,,,module: cudnn,"caffe2, module: cudnn",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#6898 [Caffe2] cudnn versions compatibility issue.,eed47c246aa666ffa2c2d98743f5978f5f370bb88ff14eb4a26bbf95ab5f1a0c
issue-collapse:409bd700b1e4cf9d,module: cuda graphs needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,109719,,,,"module: cuda graphs, oncall: jit","module: cuda graphs, oncall: jit",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#109719 Make torch.cuda.graphs.is_current_stream_capturing() available in TorchScript,289870598952550bd791e353eb621a753f0df33d605243a69230368d85dc6638
issue-collapse:a7cf90d1183e88f4,module: assert failure needs triage,single issue review,needs triage,47,low,bucket_component_queue,1,0,0,0,21255,,,,module: assert failure,"caffe2, module: assert failure",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#21255 IsType<T>() ASSERT FAILED [Detectron e2e_mask_rcnn_R-50-C4_1x.yaml],c6c2b5e8a06f99201963011fe7029beb0ce6d6e9d8f8415c1f06b2c71670b17c
issue-collapse:91df7e8023c9d78b,oncall: distributed has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,140497,,,,"oncall: distributed, oncall: distributed infra","bot-triaged, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140497 C++ standard library functions not found when compiling c10d debug handlers,eb01b97c13b00391069f58f224283f793bf3f1b1dfb37e8cf01964e73e2d1048
issue-collapse:90bef6ac76b84311,module: vllm has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,138795,,,,"module: vllm, oncall: distributed, oncall: pt2","PT2-Bug-Bash, module: vllm, oncall: distributed, oncall: pt2, pt2d-triage-nov2024, triaged, vllm-compile",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138795 compile + allgather with group will fail for stack-style allgather,e38843473000d4f80023bbb65d913b300ccfc8dbd4f1d3c37abf35e3c433e737
issue-collapse:9497292eca4ad858,module: tests has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,174177,,,,module: tests,"actionable, better-engineering, bot-triaged, function request, module: tests, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#174177 [test] Add error_inputs for nn.Linear module,76ec65a23e054769231c7f310a1bf11bcdfb0215066cf220758533a6af268ea4
issue-collapse:e9df999cf22f536e,module: serialization has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,184219,,,,module: serialization,"actionable, module: serialization, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#184219 serialization_id is not deterministic across machines,941b62abc3ad2ae0f48cfa9f43d90df62f30cedd9a018440ea062ba7b5af0681
issue-collapse:9e6f3807b0bc20ea,module: pipelining has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,153485,,,,"module: pipelining, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: pipelining, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153485 Double backward error in Pytorch PP,0b0a4debacfb57f37195a9d505ecc3ec541adb7df65eeaa1e99312b8c2bbe0fe
issue-collapse:ac09916f836d5b92,module: performance has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,177127,,,,module: performance,"bot-triaged, function request, has workaround, module: performance, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#177127 torch.multinomial: add flag to skip input validation (66% of GPU time),127f57228f35a0fe39e1d05dbf1499082e992372ce8fa38b4c92bf5dfbe79c34
issue-collapse:981f9467a674bd36,module: meta tensors has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,96409,,,,module: meta tensors,"actionable, module: meta tensors, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#96409 Initialization on `meta` device failing for models containing `nn.utils.weight_norm`, with `NotImplementedError: Could not run 'aten::_weight_norm_interface' with arguments from the 'Meta' backend.`",4212853bfaf6217eefd1aa722bb3aade910a43dd56eeef306cf4a7c7ccc80947
issue-collapse:30daace6375740d1,module: memory format has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,184642,,,,"module: memory format, module: nn, module: typing","bot-triaged, enhancement, module: memory format, module: nn, module: typing, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#184642 `Module.to()` has no overload for `.to(memory_format=torch.channels_last)`,5384252b18166e44d431685af0c6acd977654ad2a95a769e8711372cf7a992cb
issue-collapse:e240f1be1b77d2ee,module: internals has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,129020,,,,module: internals,"good first issue, module: internals, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#129020 Expand Tag Set: views & reductions,cfff01fb430d88d67e1251c762e2fcfb0710c8cba8a49aa44ef732cf2060eee0
issue-collapse:5a1d0b8dc0c093de,module: functionalization has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,133512,,,,"module: functionalization, module: pt2-dispatcher, oncall: pt2","module: functionalization, module: pt2-dispatcher, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#133512 functionalization doesn't faithfully reproduce strides of intermediate values,6140613724f0f3a85b83e03864429e2b7c2b296c6dfc939a4e3acb4948349e84
issue-collapse:bff1ca018ac68b0d,module: dispatch has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,188052,,,,"module: dispatch, module: privateuse1","bot-triaged, feature, module: PrivateUse1, module: dispatch, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188052 Allow PrivateUse1 backends to opt out of CompositeExplicitAutograd decomposition,d2764ff7f045ba7bfeb74b60a88c006fed1485c810411563aadc9fe025853ff1
issue-collapse:66f05360297ae79e,module: derivatives has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,89459,,,,module: derivatives,"actionable, module: derivatives, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#89459 Higher order derivatives of sinc explode,089d92cef8aca77369c8793eef7f9431410f29278944165286e75a7ca8f9f22c
issue-collapse:fc1fe754e0ef2175,module: ddp has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,187298,,,,"module: ddp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, enhancement, module: ddp, oncall: distributed, oncall: distributed parallelisms, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187298 convert_sync_batchnorm silently drops the activation / extra behavior of `_BatchNorm` subclasses (e.g. timm `BatchNormAct2d`) - please emit a warning,abd66d6a9ef313923e3cbc387f2fdf9c12aa98fecd190a96ed519f25c1e97987
issue-collapse:5bc4744791b6e491,module: cuda graphs has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,175001,,,,"module: cuda graphs, module: higher order operators, module: pt2-dispatcher, oncall: pt2","PT2-Bug-Bash, bot-triaged, module: cuda graphs, module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#175001 torch.cond() and cuda graph mega issue,972542edfa6b7de1f9b38b2faf47cb0ffd84101fb6114abef61a826329a7381c
issue-collapse:9f45b1cdba187274,module: compiled autograd has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,173232,,,,"module: compiled autograd, module: windows, oncall: pt2","module: compiled autograd, module: windows, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173232 error C2872: 'std': ambiguous symbol when compiling `csrc/dynamo/compiled_autograd.h`,8ce90a907aafa077ac6a39a2e797efe1f477c9d9521970289ee6ace0f82d784b
issue-collapse:e6d34a189c03cbea,module: advanced indexing has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,180855,,,,"module: advanced indexing, module: cpu, module: floatx (formerly float8)","bot-triaged, enhancement, module: advanced indexing, module: cpu, module: floatx (formerly float8), triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#180855 `""index_copy_cpu""` not implemented for `'Float8_e5m2'`",79fd676f982dd73cc340016c60696e65850f0959eaf67c217b83e76d921327b5
issue-collapse:60288105302fbf5d,module: accelerator has linked PR,single issue review,has linked PR,44,low,bucket_component_queue,1,0,0,0,185975,,,,"module: accelerator, module: privateuse1","bot-triaged, feature, module: PrivateUse1, module: accelerator, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#185975 torch.accelerator.default_stream API is missing,229814c6c8fa2bd9045990230e41d737df6992204a1961ecc02c874ac482e0b1
issue-collapse:c6e09bd058c478a6,module: random needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,169803,,,,module: random,"enhancement, module: random, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#169803 [RFC] Add RGE256ctr: A Fast Counter-Mode ARX RNG for PyTorch GeneratorImpl,402c01a3d058ef791572ad6e28919f87d1df73e3698c82f6c9187e16303e428a
issue-collapse:e2cf1cb31f2956be,module: pt2-dispatcher needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,149258,,,,"module: pt2-dispatcher, oncall: pt2","module: pt2-dispatcher, needs design, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149258 Auto-selective activation checkpointing is not optimal for speed (issue with min_cut_rematerialization_partition),868c856eed0b5c6a24db0a01eafb483ddb380534861536ae6cf7b4781e3c2522
issue-collapse:f31b0d3b4ebb7995,module: lrscheduler needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,184517,,,,module: lrscheduler,"bot-mislabeled, bot-triaged, module: LrScheduler, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#184517 LRScheduler doesn't use _get_closed_form_lr when initalizing scheduler,0327febc107c05fbceaabde79defb1268d5905faffedf5b683377caa94e74018
issue-collapse:59ed933f8cb6438a,module: library needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,146804,,,,"module: library, module: pt2-dispatcher, oncall: pt2","module: library, module: pt2-dispatcher, needs design, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#146804 `torch.library.register_fake` respects only positional order, but not kwargs order",da69d7f83cb57469bf94c06e4ddad9d10deecfb17b9f207d987453c08becf840
issue-collapse:024496af63ae2ffc,module: faketensor needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,115516,,,,"module: faketensor, module: pt2-dispatcher, module: random, oncall: pt2","module: fakeTensor, module: pt2-dispatcher, module: random, needs design, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#115516 [FakeTensor] FakeTensorMode dispatch won't change generator state,6d98243ee7daa9fc684b6e0fee0adb423f0465b87a0972a58bd103831ec5bacc
issue-collapse:b59a31f67d095b0a,module: edge cases needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,171676,,,,"module: edge cases, module: error checking, module: molly-guard","module: edge cases, module: error checking, module: molly-guard, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#171676 PyTorch allows `tensor.storage().resize_(0)`, leading to segmentation fault on subsequent tensor operations",6969dced0f405ca3136b36f390a9b8bde72892bb3c8950607d3c1558d5875246
issue-collapse:ce7b19211d2a6cc9,module: docs good volunteer slice,single issue review,good volunteer slice,38,low,bucket_component_queue,1,0,0,0,9886,,,,module: docs,"good first issue, module: docs, triaged",,Turn this module: docs slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#9886 [doc] functionalities not documented,fcddd1f61b71ca5fa4d2952924c680b8fff000bfee792e1cf9c612bd8c1f2c18
issue-collapse:a8f4ca9983a91b81,module: cuda graphs needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,122192,,,,"module: cuda graphs, module: inductor, oncall: pt2","agents-banned, module: cuda graphs, module: inductor, needs design, oncall: pt2, repro:agent:h100:repro_success:reproduced, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122192 extra checks for reduce-overhead inconsistent when warmup/recording,b0834b50a7d4b9a095badeb2ea00a57d7793c0a108fe102afd17b4b80944f67a
issue-collapse:650e0473ae072879,module: compile ux needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,180845,,,,"module: compile ux, module: dynamo, oncall: pt2","bot-triaged, enhancement, module: compile ux, module: dynamo, needs design, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#180845 [Dynamo] Unsupported Tensor.item() call in list comprehension when capture_scalar_outputs=False,4ba4f20cae125199518aff3b64c370a1b3026860e5dd2c506510a076e3b1ecdc
issue-collapse:675de8a1b25d4d23,module: compile ux has linked PR,single issue review,has linked PR,38,low,bucket_component_queue,1,0,0,0,113800,,,,"module: compile ux, module: dynamo, module: logging, oncall: pt2","module: compile ux, module: dynamo, module: logging, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#113800 InternalTorchDynamoError rewrapping loses exception chaining,3062b287cab8b2f0824c5bfc2d50ab58769dbf967d31939ab8fe93e2ea076615
issue-collapse:c7e35d21aaa06d5b,module: bfloat16 needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,120376,,,,module: bfloat16,"feature, module: bfloat16, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120376 Stochastic rounding in bfloat16,c14042a149e151967dd36f35fb070408bbc2a0471e84655d13e03b5c4a3994f3
issue-collapse:e445ab5906348fa7,module: aotinductor needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,180107,,,,"module: aotinductor, module: cpp, oncall: pt2","bot-triaged, function request, has workaround, module: aotinductor, module: cpp, needs research, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#180107 Allow getting `AtenTensorHandle` from the tensor object via AOTI stable C API,2187fb4c5c612c66379a650eb22ea07d5f0c78d9eabc9b31908f882d37b38259
issue-collapse:91d267128891ddfa,module: accelerator needs design decision,single issue review,needs design decision,38,low,bucket_component_queue,1,0,0,0,189014,,,,module: accelerator,"bot-triaged, feature, module: accelerator, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189014 [RFC]: Generalize memory snapshot APIs under `torch.accelerator.memory`,e0856ac63ac4f682a0dcb2e6083fb8cdd46fdce4a7af62cd794520f822886b24
issue-collapse:c908d855be3b91e6,module: rpc needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,158724,,,,"module: rpc, oncall: distributed, oncall: distributed infra","bot-triaged, module: rpc, needs reproduction, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158724 Problem with RPC on dual stack networks,d56f2c0d2ded98086522cd0b168ed102214d811a55a36edddd2e8162dd1701f6
issue-collapse:1bd137ba2612d4b1,module: pybind needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,108041,,,,module: pybind,"module: pybind, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#108041 Undefined Symobl: pybind11::detail::type_caster<at::Tensor, void>::load(pybind11::handle, bool)",c61db64819827daa05a993eb6587a0a6af346b9f6e713f77bb3e95b6d099a6a1
issue-collapse:6fa28a88ecddc360,module: privateuse1 needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,136344,,,,module: privateuse1,"module: PrivateUse1, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#136344 Composite Compliance Tensor ut check fails in privateuse1 device,8d58df0d8ab5f2bfeba3d6558b552bb6181af8259494faf8d761405a9dc378da
issue-collapse:305f6a97271cceb6,module: intel needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,125954,,,,"module: intel, module: macos, module: mps","module: intel, module: macos, module: mps, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#125954 Error: command buffer exited with error status.,4a75855e684177e348ab4f36d898e3a6b4851eb968527e059474456834a20ac5
issue-collapse:adafce6dcc1d30aa,module: inductor needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,111182,,,,"module: inductor, module: performance, oncall: pt2","module: inductor, module: performance, needs reproduction, needs research, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111182 Training iresnet with torch.compile is slower than eager mode for torch 2.1.0,5be5a3495bb1c0f816301152b3f4d754f8553cff2dbb8864eabab2c83041f0ac
issue-collapse:6da3707436b51d05,module: fx needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,135196,,,,module: fx,"module: fx, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#135196 ModuleNotFoundError: No module named 'torch.fx.experimental.shape_inference',cc5b544ebea41698164da891c2b7b6522ceded09eb4c58f8cbea15bb1aa06fa0
issue-collapse:5fe43596f6255fdd,module: fsdp needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,120383,,,,"module: fsdp, oncall: distributed, oncall: distributed parallelisms","bot-triaged, module: fsdp, needs reproduction, oncall: distributed, oncall: distributed parallelisms, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120383 Numeric stability issue with full parameter fine tuning of Llama 2 7B model with FSDP,1822b50d4f9267bec15274ebc3b814048d787565dbb88d2d03bd786656826b77
issue-collapse:ae63e4f69813f102,module: flex attention needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,145782,,,,"module: flex attention, module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: flex attention, module: higher order operators, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#145782 Behavior Difference with Flex Attention + Sequence Packing,c0c0512d45f04d8c9e2b79e02afae189b99c414e3311f20a632766cf8928fdf8
issue-collapse:1440b026f819a22d,module: faketensor needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,141545,,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","module: fakeTensor, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#141545 Missing Fake/Meta/Abstract Kernels for Operators,00b6d323bf779683b537e9fabcbc73c7126e86f0cba63f1263f50886b15abffd
issue-collapse:bbecfdf6df9588d8,module: edge cases needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,184030,,,,"module: edge cases, module: nn","bot-mislabeled, bot-triaged, module: edge cases, module: nn, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#184030 torch.nn.PoissonNLLLoss returns NaN for negative input when log_input=False,dbf3b00844bb458f7450eba4e370c182d60e02abca9b9c0b71df591addeaa108
issue-collapse:6b89e7a5c697a5b5,module: dispatch needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,172944,,,,module: dispatch,"module: dispatch, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#172944 Tried to access the schema for which doesn't have a schema registered yet,c941b748fc70582f1a336cb8ed0b9eab75007095c2597aa2f1ed3224528600cc
issue-collapse:1d72077e3e5806b7,module: custom-operators needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,143579,,,,"module: custom-operators, module: pt2-dispatcher, oncall: pt2","module: custom-operators, module: pt2-dispatcher, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143579 `assert_size_stride` failing in `_inductor/utils.py` `return model(new_inputs)`,efe7ad99eda3fcac5fb5d4b669cb898a708e983b16f9d240ea08002c4043ef3c
issue-collapse:ae3ee7c9a4759f12,module: cuda graphs needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,171672,,,,"module: cuda graphs, oncall: pt2","PT2-Bug-Bash, module: cuda graphs, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171672 torch.compile max-autotune and reduce-overhead much slower than default due to repeated CUDA graph instantiation,9e05310cb834bf79ca285bc169f2faa5492c91e1e2ebebeb1eca4347be2a7231
issue-collapse:6d21dcbdb8339df1,module: collect_env.py needs reproduction,single issue review,needs reproduction,32,low,bucket_component_queue,1,0,0,0,177819,,,,"module: collect_env.py, module: macos","bot-mislabeled, bot-triaged, module: collect_env.py, module: macos, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#177819 MPS unavailable on macOS 26.3.1 arm64, and `device=""mps""` reports incorrect macOS requirement",f2cffbf2eb1a5f7e13e68bfa1b13d8a8d25fc6441abb2092650eea182fae0343
issue-collapse:865493371011bf3f,oncall: cpu inductor needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,93531,,,,"oncall: cpu inductor, oncall: pt2","oncall: cpu inductor, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#93531 TorchInductor CPU Performance Dashboard,6d5f05e50a2efe9d68851d997f191937f50dcb34975d7bab44d991f8b6aa09e3
issue-collapse:97b356b8e8535841,module: xnnpack needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,158343,,,,module: xnnpack,"module: xnnpack, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158343 Upgrade to latest XNNPACK,5333349539014e8a310b17ab1dd3905327985563a08e592473208272384349fb
issue-collapse:fda7d7a52af89734,module: vmap needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,116697,,,,"module: vmap, oncall: profiler","module: vmap, oncall: profiler, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#116697 Tracing GPU usage of `torch.vmap` vectorized operation using `torch.profiler` cause CUDA OOM,6dfff3004e74426057a77323370ddabf85aae49665e639b1bf0336b96f3a6704
issue-collapse:72c7c3851ca61412,module: viewing and reshaping needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,97892,,,,module: viewing and reshaping,"enhancement, module: viewing and reshaping, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#97892 A parameterized fill value for triu and tril functions,283454d4096f07c28892fbd8cdee180b53032698622c924b726cdf547d68aa9f
issue-collapse:a34e99f381adf90d,module: special needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,68110,,,,module: special,"module: special, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#68110 zeta(1/2) returns nan,d6db966663edeb5da545f23efc1fd404c0342e20515afab85b162f1e0905b137
issue-collapse:4ef53399add57213,module: scientific computing needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,74839,,,,module: scientific computing,"feature, module: scientific computing, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#74839 Feature request: Root-finding functionality (like scipy.optimize.root_scalar),2e0fc9244c76ff7218935921af505e80eb00dfe6b9782b356d658ec4cdaa9627
issue-collapse:b96f93531e210640,module: reductions needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,172854,,,,"module: reductions, module: tests","module: reductions, module: tests, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#172854 [test] Add unittest for nanmean output dtype validation (follow-up to #131043),b1b79851c0283b865347570e7631a1e5cb67bd39ae741c439a540f8d6f796e88
issue-collapse:f61ec9ba9f20c2b5,module: printing needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,51002,,,,module: printing,"enhancement, module: printing, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#51002 Consider adding context manager support for torch.set_printoptions,71cccec2e9250d57a7a05034f4d5ee37fb72bd55ffce2cb799ab2847750ffc7d
issue-collapse:cd76ab111bf9f665,module: power needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,109592,,,,"module: power, oncall: profiler","OSS contribution wanted, module: POWER, oncall: profiler, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#109592 test_memory_timeline fails on PPC due to extra temopraries,6841aa4d848d56f82ce890bdbb06f7425c38198385465839cef34da8516ff91d
issue-collapse:e82d14756bf074c2,module: numpy needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,106692,,,,"module: numpy, module: special","module: numpy, module: special, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#106692 torch.polygamma inconsistent with scipy.special.polygamma for n >= 1,3048855f802d703b62afca7d443809083e7058e8b1076cdde2cf2e34c61f99cf
issue-collapse:aec434c1e0c34658,module: multiprocessing needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,171796,,,,"module: multiprocessing, oncall: distributed","module: multiprocessing, oncall: distributed, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#171796 Exception in subprocess causes infinite hang and zombie processes,6d5b2941d724a5828071ad8bf68f6577ad4ee0f35affbeea03049ae4cccdf180
issue-collapse:e0bb37e2a07ec4ca,module: mpi needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,142095,,,,"module: mpi, oncall: distributed, oncall: distributed infra","bot-triaged, module: mpi, oncall: distributed, oncall: distributed infra, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#142095 torch.distributed.new_group got stuck when use backend=mpi,528a06f9a6a115f1178c0fe7126bbdd86e6f8cc4952310ee26b6ab5a4fc7b816
issue-collapse:c3f02657027388e2,module: molly-guard needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,182105,,,,"module: molly-guard, module: pickle, module: serialization, module: ux","bot-triaged, feature, module: molly-guard, module: pickle, module: serialization, module: ux, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#182105 Adding a UserWarning when pickling / deepcopying tensor views,b236145c2b471affc638173cc3b2515acd2de2e6a648aff77c33d322224c367d
issue-collapse:31dfa330fce522b5,module: meta tensors needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,188692,,,,"module: meta tensors, module: mkldnn, module: rnn, module: tests","bot-triaged, module: meta tensors, module: mkldnn, module: rnn, module: tests, skipped, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188692 DISABLED test_mkldnn_rnn_backward_gru_bias_shape (__main__.TestMetaKernelRegistrations),1672856e068d8829a10e452ffaf81267e0dd789632870003672e1ac8cabead9b
issue-collapse:0fa2f62cea5a557a,module: memory usage needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,154363,,,,"module: memory usage, module: sdpa","module: memory usage, module: sdpa, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154363 `scaled_dot_product_attention` broadcasting (GQA) is a memory footgun,a111b6f4713926a6246b65bd21cbdb412939868ad2bc877659d4a5d479058126
issue-collapse:e902a68d9e5a6244,module: memory format needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,187963,,,,"module: memory format, module: pallas, module: pooling, oncall: pt2","bot-triaged, module: memory format, module: pallas, module: pooling, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187963 Pallas CPU max_unpool2d does not preserve channels-last output layout,fcd97c23126f8c8c4d8b7276f73a659adc60ba78389d598281672874137dca3d
issue-collapse:9eb80a8d38b992cf,module: logging needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,150974,,,,"module: logging, oncall: pt2","module: logging, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#150974 feature: tlparse summary,20f9c7e17d3ae853889329b3876791a8cf087354d9a95e3d9ff37f94fccc9c5d
issue-collapse:1d8c62def7102e79,module: lint needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,149639,,,,"module: lint, module: typing","actionable, module: lint, module: typing, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#149639 Pylint error: ` torch.linalg.vector_norm is not callable`,fc814a2e2c1efd0e8dd8cad5b34d6c0c213a63c2524ec70499d8ea389030bbde
issue-collapse:a5be4d792acd8da5,module: jiterator needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,74415,,,,module: jiterator,"module: jiterator, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#74415 Warning: ""Specified kernel cache directory could not be created""",5ca3c921f50c84e8f12820ce30352c10771d5340c469efcc4cd3d3b1cbff53de
issue-collapse:f1a6534785247a39,module: intel needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,169670,,,,"module: intel, module: linear algebra, module: mkl","module: intel, module: linear algebra, module: mkl, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#169670 Energy normalization error on torch.fft.fft2 and torch.fft.ifft2 when the matrix size = [2048, 2048]",8adb049e67ed1274c0eb4f53656eebe6cfd3166af2f0e49fa6970a68b9345cee
issue-collapse:0c407f4a089c36d8,module: initialization needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,180968,,,,"module: initialization, module: nn, module: nn.utils.parametrize","bot-triaged, has workaround, module: initialization, module: nn, module: nn.utils.parametrize, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#180968 Initializing weights after parametrization silently fails,c6225be90a7bceca1a6bdccb4255c8aec5febeb92c3db8303d10b098c6792cd6
issue-collapse:15ffe127646aa9fb,module: fx.passes needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,188669,,,,"module: fx.passes, oncall: pt2","bot-triaged, module: fx.passes, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#188669 fx_graph_cse skipping duplicate element-wise CUDA optimization.,219943daaf0a77ca6e8638d6d133780fe3f63af3ba3566be4e985c90648611d8
issue-collapse:4fff9b7b9b0df282,module: fx needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,187649,,,,"module: fx, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: pt2","bot-triaged, feature, module: fx, module: higher order operators, module: inductor, module: pt2-dispatcher, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#187649 [Inductor] Support fx wrapper with non-triton fused kernel.,264186324d17da1b83c26816722cd6e0483f470f6408de73ade5c324039ffc87
issue-collapse:b018ef1687a3478a,module: first class dims needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,133559,,,,"module: first class dims, module: functorch","module: first class dims, module: functorch, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#133559 Support for variable number of dimensions in functorch.dim dims() and autonaming induced inconsistencies,ff97c5aea64bde1a70aae8a0cc66bccb03faacf013b9c7fb0d31740237a60462
issue-collapse:9b65e71ac27fd793,module: faketensor needs maintainer decision  batch 2,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,106732,,,,"module: faketensor, module: pt2-dispatcher, oncall: pt2","module: fakeTensor, module: pt2-dispatcher, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#106732 Hugging Face safetensor does not work with FakeTensorMode,1bc55ef611300c67260b7c19d7b4a84d33143ff7d270699a3e86baf218d32c50
issue-collapse:1c9e2a9b748766a3,module: double backwards needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,183603,,,,"module: double backwards, oncall: pt2","bot-triaged, module: double backwards, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#183603 torch.compile fails on second-order gradient after autograd.grad(create_graph=True),85e542102bc4d670455aec21443c207754d8a40c729904c4fb6ff0d1a2d91920
issue-collapse:4b0e6adfa4d23117,module: distance functions needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,59194,,,,"module: distance functions, module: numerical-stability","enhancement, module: distance functions, module: numerical-stability, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#59194 Add numerically-stable function for angle between vectors,4192bd24c2894bdadfa05921d5bb301558d084116cafdb73b33963687ac31af3
issue-collapse:a2783ab64522ce37,module: dependency bug needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,17023,,,,module: dependency bug,"module: dependency bug, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#17023 Download speed issues with the pytorch conda channel,25db30b9957e9fc65aa8022012d7642a4fe1574620fb87c4663ebd84ce313524
issue-collapse:865187b1c3435512,module: cudnn needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,170762,,,,"module: cudnn, module: inductor, oncall: pt2","module: cudnn, module: inductor, oncall: pt2, repro:agent:h100:repro_fail:not_reproduced, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#170762 [BUG][Inconsistency] aten.cudnn_batch_norm_backward produces numerical mismatch between eager and inductor,70573a1ac7b5490936b233b4ea0a7a046815c7f7b9610eb979fc1d9bafac6272
issue-collapse:c11a2712ac317519,module: cudacachingallocator needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,173104,,,,module: cudacachingallocator,"module: CUDACachingAllocator, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173104 Crash with `CUDACachingAllocator.cpp` when running Nvidia `nemotron-parse` at load,ce23141fe902c34657793a254bdb7a3940b020c4ba8e8cf6e593bfa2c8e1519f
issue-collapse:bd5c486d61522cb8,module: compiled streams needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,178132,,,,"module: compiled streams, module: inductor, oncall: pt2","bot-triaged, feature, module: compiled streams, module: inductor, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#178132 torch.compile streams support in AOTI,c2e88525c91d09dd05c2aabc02f4ae08f89c4c9920340e869310f99c49af20dd
issue-collapse:ebdc583592589e13,module: codegen needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,182450,,,,"module: codegen, module: lint, module: spin","bot-mislabeled, bot-triaged, enhancement, module: codegen, module: lint, module: spin, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#182450 Spin codegen run doesn't use self-contained environement,16f0094ee142eb4d87d21d998bf90544f1790109fb4b1924706d08c5bd35f68b
issue-collapse:ef416a4bf8a0a4a1,module: bootcamp needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,78842,,,,"module: bootcamp, module: c10d, module: distributed_tool, oncall: distributed, oncall: distributed infra","bot-triaged, enhancement, module: bootcamp, module: c10d, module: distributed_tool, oncall: distributed, oncall: distributed infra, pt_distributed_rampup, ptd-bot-triaged, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#78842 Add TORCH_SHOW_CPP_STACKTRACES when TORCH_DISTRIBUTED_DEBUG = detail,b44f4fee21713dfce7ec4d2aba3ef991ccd1d6b864378a25fe4a9ad738432b79
issue-collapse:d1eb8bdda61c43be,module: bazel needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,112903,,,,"module: bazel, module: build","module: bazel, module: build, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#112903 Support for Bazel workspace function or Bazel module,2eb32c90cb96f2806cafece804dfe66d95b3082614462fbf21a643136a90155e
issue-collapse:41a360e4318a0e58,module: batching needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,173802,,,,module: batching,"feature, module: batching, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#173802 Batching rule of aten::_weight_norm,7150464239dc1935b6b6cce6e272e21a5cb8b5515e06a06cfa565deb92807211
issue-collapse:a4bc368846f666fc,module: android needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,154449,,,,module: android,"executorch-needs-help, module: android, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154449 16KB pagination support for PyTorch Torch Vision and PyTorch Lite,021db4ffb2ee9c4ad2f2bec271135850b5eab05d2083e968b3bf002f3a51af6f
issue-collapse:aac78150909a8619,module: advanced indexing needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,189046,,,,"module: advanced indexing, module: autograd, module: backend, module: dispatch, module: privateuse1, module: reductions","bot-triaged, feature, module: PrivateUse1, module: advanced indexing, module: autograd, module: backend, module: dispatch, module: reductions, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#189046 Treat PrivateUse1 as CUDA-like under env override to reuse CUDA optimization paths in shared `at::native` operators,d0faf66b04a947062bc10af1e422de10c3387b1e8575b41f832c40be0eef8cfc
issue-collapse:6f180d4f0982a206,feature needs maintainer decision,single issue review,needs maintainer decision,28,low,bucket_component_queue,1,0,0,0,48972,,,,,"feature, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#48972 non-negative least squares solver feature request,008cf53ad2606a4f096821776e1714d1706910922ead9ff9462da0f222e8fe28
issue-collapse:e90a79ce0d87b069,oncall: quantization has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,111471,,,,oncall: quantization,"oncall: quantization, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111471 test_learnable_forward_per_channel fails due to integer overflow,ec73e354f9ab46bc4f1b4a1238b41bd848675d9986b71ddbd2d743d4bf2b95be
issue-collapse:7106b3e26c2d0692,module: windows has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,92600,,,,"module: windows, oncall: quantization","low priority, module: windows, oncall: quantization, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#92600 Update quantization to make source files complient with /Zc:lambda,5eb0bbdbf988bef332f03733cb8f45c09e4fc067478e1001f643252a91de8af3
issue-collapse:b2d2434a7c77ba88,module: vmap has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,122729,,,,module: vmap,"module: vmap, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#122729 torch.func transforms don't yet support saved tensor hooks.,a7bfedcff4e4581ee4c1840827603f8ec25fe0f6e1f865ff5b1ba968c431d52d
issue-collapse:f3e76d947f79b1df,module: sparse has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,136258,,,,module: sparse,"module: sparse, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#136258 `to()` Method Does Not Move Internal Components of Sparse Tensors,d6df05d9207f1a4a02731d0bf6d5b82f50f9999818999669fd95542120dce9ac
issue-collapse:ff04385e9aca3892,module: sdpa has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,117016,,,,"module: sdpa, module: vmap","actionable, module: sdpa, module: vmap, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#117016 There is a performance drop because we have not yet implemented the batching rule for aten::_scaled_dot_product_efficient_attention_backward,442358f44080ddc548f94d134fd593602b96baaee64726d3ddc19dc2837f0dfe
issue-collapse:d85f24d67cc0edae,module: pickle has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,39383,,,,module: pickle,"module: pickle, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#39383 Pickling a Tensor or a Storage is not deterministic,e0eae3e3d7c2f936409ec31e148106e74c2c6b33a9f7a6a5dec853815f794a2c
issue-collapse:e1abbbd6d9fdcabe,module: intel has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,132368,,,,"module: intel, module: mkldnn","module: intel, module: mkldnn, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#132368 Switch to new name `oneDNN` instead of `MKLDNN`,10d7b189bd5764141963eb7576ff43c02f7803a8c48f977a0408ad703c8c50c4
issue-collapse:c4cbdfdf7a0adb8d,module: higher order operators has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,158975,,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158975 Handle matmuls gracefully in foreach_map,13dda20558068824c3da26eaced3af1b683cff54be3feebaa5f9228953745b04
issue-collapse:a87814e579e3bd9d,module: fsdp has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,157832,,,,"module: fsdp, oncall: distributed","module: fsdp, oncall: distributed, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#157832 [FSDP2] should pass args as is instead of creating new ones,a29cf73a834be1afb5278daccfcad116a522ae4189128cc3f2b3fd8e3a3332ab
issue-collapse:af13f34f8c29c507,module: empty tensor has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,169208,,,,"module: empty tensor, module: error checking, module: norms and normalization","actionable, module: empty tensor, module: error checking, module: norms and normalization, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#169208 Segmentation fault in torch.native_batch_norm with empty running_var tensor,0b6512b4e69e1b5416b67e597819a4b6d9627a3bf92a6f4e773c594a9b6cd351
issue-collapse:8965747c19a8c948,module: distributions has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,152095,,,,"module: distributions, module: docs","actionable, module: distributions, module: docs, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#152095 To fix inconsistency between signature and doc on `torch.bernoulli()`,c322f80e05b20a20a927caac7364eeef971498d7854d8f974020a5ee324b7526
issue-collapse:8530db3874ae8fe4,module: core aten has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,137440,,,,"module: core aten, module: cuda","actionable, low priority, module: core aten, module: cuda, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#137440 `torch.pow` throws with `RuntimeError: ""reciprocal_cuda"" not implemented for 'Char'`",8362a6d69c113279b1242823385bd631a9b83389f86df306e726ac21e192445e
issue-collapse:0ce556062fd68474,module: compile-time has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,133734,,,,"module: compile-time, module: dynamic shapes, oncall: pt2","module: compile-time, module: dynamic shapes, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#133734 PyTorch `torch.compile` inductor backend shows enormous compile-time regressions.,8e5a52818d78c92c6ad155356c5a27fda5624bd0e7fb34466edbab15a6a03280
issue-collapse:6bac924e0045ba5c,module: __torch_dispatch__ has linked PR,single issue review,has linked PR,26,low,bucket_component_queue,1,0,0,0,140909,,,,"module: __torch_dispatch__, module: dynamo, oncall: pt2","module: __torch_dispatch__, module: dynamo, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#140909 [2.5.0+] Running a compiled model under FlopCounterMode regresses its performance,4a7eadedde2006dc26a567affd4f56bb1a7e0262dbc12e306a15787ebe75f63e
issue-collapse:2bf21fb4afc3c2ad,oncall: jit good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,16663,,,,oncall: jit,"good first issue, oncall: jit, triaged",,Turn this oncall: jit slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#16663 new_zeros is not traced correctly,069f34047da3de353e2e43b1819ead906ceaa3c3ad8dd91fc923fb605bd1ba4e
issue-collapse:3426ef94b903d0c9,module: vulkan needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,155986,,,,module: vulkan,"enhancement, module: vulkan, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#155986 Vulkan interoperability,7ebaad62b98ee0a60d719aa2cafe9080633b276bbe7e9368786cb1c9d1cfec63
issue-collapse:48859eb26770c306,module: typing needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,103109,,,,module: typing,"enhancement, module: typing, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#103109 Increased / more verbose type aliases for improved readability of user defined content,31b97e71bdbc4e44b3ab3eaa7645c4a611f99a1b81144dd857b15c772bbf5cce
issue-collapse:19beaa3cceefd752,module: tests needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,60223,,,,module: tests,"module: tests, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60223 [testing] SkipInfo should error if `cls_name` and `test_name` combination is not valid,1e47746faab6af514158d7a08b71c7e633fbcb4e5d8a1476f336e9e35f9739a3
issue-collapse:06d8870d709e7e97,module: tensor creation needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,121933,,,,module: tensor creation,"feature, module: tensor creation, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#121933 Please provide `total_repeat_length` field to PyTorch `repeat` / `repeat_interleave` API,0a176fe1814546584c9a1fe1f6f08e346bfe05cd442a6347da01732e55b25dbd
issue-collapse:04104e069b794545,module: sdpa needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,120589,,,,module: sdpa,"feature, module: sdpa, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120589 Softmax plus one for Flash attention,92467a6f654d31ec31676050de6d53bb8739af9dfeaacc93956c89de353a9572
issue-collapse:55445ce7e9432ce3,module: scatter & gather ops needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,138502,,,,module: scatter & gather ops,"feature, module: scatter & gather ops, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138502 Contribution Proposal: ScatterND Implementation for PyTorch,18c830db4e94e858f57873817257f991d8c16400aa89ae635dc785800d1aa3de
issue-collapse:601b3bc2c6224d2a,module: scatter & gather ops good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,63265,,,,module: scatter & gather ops,"enhancement, good first issue, module: scatter & gather ops, triaged",,Turn this module: scatter & gather ops slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#63265 torch.scatter_: support index.size(d) > src.size(d),d0e229162c8dda4b7c1ce4d7259ee96f2666560a34b67db92944abc8a5683efc
issue-collapse:791ed4bc2ffe58b6,module: rpc needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,44827,,,,"module: rpc, oncall: distributed","module: rpc, needs research, oncall: distributed, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#44827 [RFC] Pipeline Parallelism in PyTorch,4784ff81745672002e45798b327534312a296e4d26d75d1a4c81eb26c464038f
issue-collapse:df4388aa0d8fd3ec,module: rocm needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,153566,,,,module: rocm,"enhancement, module: rocm, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153566 Improve PyTorch Logging to Distinguish ROCm from CUDA,8f6134fb594fa6ef19abc296904f2929c04bb31c1c83ce345a963e462ab83921
issue-collapse:49d264c30e7f0782,module: reductions needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,131927,,,,module: reductions,"feature, module: reductions, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#131927 exclude_dim=[...] as a keyword argument to reduce type operations,a1b9dcd02a3c824be61efb174ede5aa214626be9e7eb5e6726fe1399bd2ec8aa
issue-collapse:b7780d485a9baeb8,module: random good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,67321,,,,module: random,"feature, good first issue, module: random, triaged",,Turn this module: random slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#67321 alias to generate tensor with random uniform distribution.,435cf59628b85645b2ccac5bbd6098f14da27207f744861ce7bd2d1094dc4701
issue-collapse:b969d3bc31df7ac7,module: pruning needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,39765,,,,module: pruning,"feature, module: pruning, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#39765 Support mainstream pruning techniques,935dbc668c30826c79416ae0ed1ebfe08cf3b89c36bf4f2a1fff8123d79fd7ed
issue-collapse:225236db14c4d66c,module: printing needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,76532,,,,"module: printing, module: python frontend","enhancement, module: printing, module: python frontend, needs design, tensor subclass, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#76532 Add ability to add custom suffixes to tensor repr,f2b49221b376c24eea845587f0b5f0ed032c76b38a78793fa42eea996d130ef2
issue-collapse:51cc605984f56a64,module: performance needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,151722,,,,"module: performance, module: python frontend","enhancement, module: performance, module: python frontend, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#151722 Add maxcount Parameter to torch.unique and torch.unique_consecutive,d446e781238f4a892149995f085d4a6ca0a7d6ccf81f7eee93f500a4e5345599
issue-collapse:66b00e56f8537bd3,module: numpy good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,93684,,,,module: numpy,"good first issue, module: numpy, triaged",,Turn this module: numpy slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#93684 Can we rewrite numpy operators to pytorch operators?,2e0cdec8a92903a9debabdb9f0e37dcfa4774507c7f307aa3f093293ccd56c2b
issue-collapse:a95e8a23b7ba39fd,module: norms and normalization needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,46516,,,,module: norms and normalization,"enhancement, module: norms and normalization, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#46516 Batchnorm support for tracking buffer statistics when using gradient accumulation,7fadd176702829c58c0b5fbabc0db19a483e463e26eb488cc992a8e60eac8fed
issue-collapse:fa3624804431a93a,module: nn good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,66813,,,,"module: nn, module: performance","actionable, better-engineering, good first issue, module: nn, module: performance, triaged",,Turn this module: nn slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#66813 Use `__slots__` for the `nn.Module` class,82fe2bff39641effe44ac68337fec284567f0e3c2f057598a084fe11b27ae475
issue-collapse:1f2404ad0a11b7c4,module: multiprocessing needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,124902,,,,"module: multiprocessing, module: privateuse1","feature, module: PrivateUse1, module: multiprocessing, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124902 Add support for IPC features for PrivateUse devices,72f0f2f3f0da01fe103ae2ebc3d4b0b1e9553726d856cfa5fb25ceb59d138c41
issue-collapse:89eb0502cfb1897b,module: infra needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,63038,,,,module: infra,"module: infra, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#63038 Use pytorch-probot for PyTorch-specific stuff + MOTD in Dr. CI comment,a818efae87230e476a684604b43fe9302e4ecb776c2f53bff04384bf6cd3eadd
issue-collapse:7b001d6212147cf5,module: higher order operators needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,120189,,,,"module: higher order operators, module: pt2-dispatcher, oncall: pt2","module: higher order operators, module: pt2-dispatcher, needs research, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#120189 Making Mamba first-class citizen in PyTorch,261f22f082c79beb1ba78ca46ea7227364b8d7553209befc696d818e3f6032c3
issue-collapse:76856b831554a1b5,module: flaky-tests needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,62902,,,,"module: flaky-tests, module: rocm, oncall: r2p","module: flaky-tests, module: rocm, needs research, oncall: r2p, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#62902 distributed/elastic/multiprocessing/api_test is flaky on ROCm,35a9b1e53f17f23b8ca86332473ebcbf4163f207395036b74c036f8009eaf272
issue-collapse:d3881d27703cd533,module: dynamic shapes needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,93419,,,,"module: dynamic shapes, oncall: pt2","low priority, module: dynamic shapes, needs design, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#93419 torchdynamo is not properly setting up input tracking (e.g., for symbolic shape guards) for view bases",a617f1140567b69afa224d8dab5aa5437483d71a32626e6448abceca371eb4cc
issue-collapse:df688d5d50593045,module: cpp-extensions needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,84782,,,,"module: cpp-extensions, module: m1, module: third_party","module: cpp-extensions, module: m1, module: third_party, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#84782 macOS Pyinstaller: libc++abi: terminating with uncaught exception of type c10::Error: Type c10::intrusive_ptr<ConvPackedParamsBase<2>> could not be converted to any of the known types,cf93f58b765de140b2914431d570d804b8d701a52f7da4aa7a1503e150376e7a
issue-collapse:ddc25b19d850a53e,module: compile ux good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,153669,,,,"module: compile ux, module: dynamo, oncall: pt2","dynamo-triage-jan2025, good first issue, module: compile ux, module: dynamo, oncall: pt2, triaged",,Turn this module: compile ux slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#153669 Make tlparse able to show a summary of distinct graph breaks,65b62ca3b5991c44f44df1351d4db4b92e836b941667d575c86a174df8d9839b
issue-collapse:2f5c251015925fac,module: collect_env.py needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,144536,,,,module: collect_env.py,"module: collect_env.py, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#144536 [bug report template format] Simplify version information with HTML tags,f8f391138f7da4619f02a40d4f9dae90a023aeb8c774422e92e5ab28f08e6454
issue-collapse:fa7751e1734cae6a,module: codegen needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,61664,,,,"module: codegen, module: internals","module: codegen, module: internals, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#61664 Support adding new keyword-only parameters without breaking FC,319b4e807752f9e33306a0e9a46e6f65f52dc02be86e8317bcb54c9a63190cca
issue-collapse:4153234bc8abc431,module: build good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,99278,,,,module: build,"good first issue, module: build, triaged",,Turn this module: build slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#99278 Build error on libstc++ header stl_alogbase.h on riscv,7bf21b13e45b1c223e5fa3c179c2337adbe8c10d85170438786d40daa99377dd
issue-collapse:f58e318efaaac6d8,module: bc-breaking needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,49227,,,,"module: bc-breaking, module: numpy","module: bc-breaking, module: numpy, needs design, topic: bc breaking, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#49227 Looking for a more convenient way to getter value of the upper/lower triangular matrix into 1D,c32bbb9730734400a6144fced80d9870ca297b3a24025007d1b9297e4c27291f
issue-collapse:ce38c9eb5c99815c,module: backend needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,133288,,,,"module: backend, module: privateuse1","enhancement, module: PrivateUse1, module: backend, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#133288 Give PrivateUse2 and PrivateUse3 the same treatment as PrivateUse1,b3e973ef73fea50c7b1dd2112ee7a3a430f7cf3ef4ff236711a7bcf5919c884f
issue-collapse:5a49c820b733700b,module: amp (automated mixed precision) needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,58228,,,,"module: amp (automated mixed precision), module: complex, module: cuda","module: amp (automated mixed precision), module: complex, module: cuda, needs design, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#58228 ""_amp_foreach_non_finite_check_and_unscale_cuda"" not implemented for 'ComplexFloat'",5a03753004bd945bd52aaee688ec77c2f619b116732122145cfa52b2d38a6fe8
issue-collapse:0d221c1b6efad51d,module: advanced indexing needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,116555,,,,"module: advanced indexing, module: autograd, module: cpu, module: performance","module: advanced indexing, module: autograd, module: cpu, module: performance, needs research, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#116555 The performance of `torch.index_select` and regular indexing differs dramatically based on the size of the tensor it is indexing,109f7ce253ae1b9b3dfd7b9cab489233e2fbb514e9968cad7f78874bed6e2785
issue-collapse:58a83c25227cc89e,module: __torch_function__ needs design decision,single issue review,needs design decision,20,low,bucket_component_queue,1,0,0,0,96579,,,,"module: __torch_function__, module: autograd","module: __torch_function__, module: autograd, needs research, tensor subclass, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#96579 Wrong return type from operation on custom tensor inside registered hook,d12f50f4611839de112b62d8f30f129f33631ee7843fe5e0dd52950a99f50db1
issue-collapse:bf0960a0e5e9b654,module: __torch_function__ good volunteer slice,single issue review,good volunteer slice,20,low,bucket_component_queue,1,0,0,0,117806,,,,"module: __torch_function__, module: dynamo","enhancement, good first issue, module: __torch_function__, module: dynamo, triaged",,Turn this module: __torch_function__ slice into one contributor-ready queue with shared scope and testing expectations.,Batch as volunteer-ready work; keep maintainer review focused on scope and tests.,1,Represents 1 open issue; raw issue rows remain available for audit.,#117806 Supporting custom attributes with `__torch_function__` tensor subclasses,a2c367ecea02156895b1f227a5aaa961036b44c91e94970ec10938f70e8b8afb
issue-collapse:ecefc424af54dea6,oncall: mobile needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,54577,,,,oncall: mobile,"needs reproduction, oncall: mobile, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#54577 Question when I use Pytorch Mobile,2c24eba0494eb904be16dc3683e853dcc413f6b698bd8a39d93d861fbe14928a
issue-collapse:ca9950efbdff5444,module: vmap needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,77893,,,,module: vmap,"module: vmap, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#77893 Segmentation fault in _remove_batch_dim,85468e044f1bd25ac43bb27d7e75c46022d6111efc9e387966a4c07b7eb85b82
issue-collapse:ed3b9aac0f903e60,module: vllm needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,167820,,,,"module: vllm, oncall: pt2","module: vllm, needs reproduction, oncall: pt2, triaged, vllm-compile",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167820 Why torch==2.9 compile qwen3 model with block ptr will crash?,c61feb2945afa875bd72dc342479b06632b328a159ccea379446e42487cec9c8
issue-collapse:5e5032ec40542794,module: third_party needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,143288,,,,"module: third_party, oncall: pt2","module: third_party, needs reproduction, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#143288 EXCEPTION : /python3.11/distutils/core.py,c755417c0b467e3b5220fc40f1d2163daf36d868f5799040d02e57767ec1fc09
issue-collapse:778f533693144139,module: testing needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,138455,,,,module: testing,"module: testing, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#138455 Tried to instantiate dummy base class,fa6ebb5a8ebd1a736a5fd66e789c5c3f8452fb60aeafc3e395c2ac2bc5fb322b
issue-collapse:649fb1ed2afa4c5a,module: special needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,94542,,,,module: special,"module: special, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#94542 torch.lgamma CUDA driver error,03677eb574b07dc2b8970205dec984a8c3ca4fc79588369a0c1c6e36f1eec29b
issue-collapse:a3e6b6f8170e58ad,module: sparse needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,111359,,,,module: sparse,"module: sparse, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#111359 I have a trouble with to_symmetric,a6d16b86207fd1ab8a23dcaccb530551483245180183713b9b8571c30df04741
issue-collapse:6115572c43d04196,module: rocm needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,60477,,,,module: rocm,"module: rocm, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60477 ROCm miopenStatusInternalError /MIOpen/src/sqlite_db.cpp:109: open memvfs: unable to open database file,a2b00d10c69b3df4d4571f20663f335dbd8c43eed54f1c3e1a78624077115605
issue-collapse:c7597db6d8f14d27,module: reductions needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,129303,,,,module: reductions,"module: reductions, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#129303 Incorrect index from torch.mode,2f7b2e5c680e8293fbedc5abe11009c3191af38e74cb1818004e141b8b12a9c1
issue-collapse:b01400c97086adef,module: random needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,63726,,,,module: random,"module: random, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#63726 a problem happened in torch.randperm,abd40875380081b612ba574266306b3ac0201289f7030e3c0d6f6bc6a99f5403
issue-collapse:5b402b7017a69097,module: pt2-dispatcher needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,161494,,,,module: pt2-dispatcher,"module: pt2-dispatcher, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#161494 OOM,f7ede4f842f163266eb2b639ab1587ae7e42bb4c4d507fa77808c3f56ccd1e84
issue-collapse:2c5be349f46fe306,module: performance needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,61243,,,,module: performance,"module: performance, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#61243 Pytorch 1.5+ is slower than pytorch 1.3,d0389fa38c3d82da5c0b7b15e02e4e876d3f409052ecc95c2b847d29284815b7
issue-collapse:4e85d183d380f4ac,module: numpy needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,84705,,,,module: numpy,"module: numpy, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#84705 Set dtype if tensor converted to numpy,d7129ba5ed5877de8234c887d995d537084d1f037770415536336f453bab29b8
issue-collapse:f0aea4ecd92a37a0,module: numerical-stability needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,5953,,,,"module: numerical-stability, module: optimizer","module: numerical-stability, module: optimizer, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#5953 LBFGS always give nan results, why",5eb54bdeeb4ef319ececee9d0b5f912ab39546c606d69ab10b5856a4bf2231bf
issue-collapse:f19c2673c576a8d6,module: numerical-reproducibility needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,107115,,,,module: numerical-reproducibility,"module: numerical-reproducibility, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#107115 H100 works differently than rtx4090 on same model,9c51e163ae5f3718ba78e00a09d124ea51dc8944b02d8289d841be8b173d5f9d
issue-collapse:8ea3eb4abdc1938a,module: nnpack needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,143646,,,,module: nnpack,"module: nnpack, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#143646 Getting ""Could not initialize NNPACK! Reason: Unsupported hardware."" warning even though NNPACK is enabled",f0e3b199038d933e0574b8276e0320e2d5a16ee223b11ed910b097557a945510
issue-collapse:c3538688a6e5303d,module: multi-gpu needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,59626,,,,module: multi-gpu,"module: multi-gpu, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#59626 multi scripted model on the same gpu on multi gpu machine,2f98f53343bcdc2a436d35889abd23dcd7dc42dc3d2f13729077d503f41e34eb
issue-collapse:a2fe4e77c56a4f5f,module: loss needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,129643,,,,"module: loss, module: nn","module: loss, module: nn, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#129643 There is a bug, only when self",c2dacd9aa9798ac6b2ec4b402982aa07bb8b942e7d518417a138e050f5109032
issue-collapse:6fdbbc801090adb7,module: lint needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,158906,,,,module: lint,"module: lint, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#158906 The make lint command got stuck.,88f863373c704a52bd4d2cafc53a333e4f5c20a3f4482a805afc8b93f60830f1
issue-collapse:0759a64692625e61,module: embedding needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,99268,,,,module: embedding,"module: embedding, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#99268 Embedding layer tensor shape,8add036e61113371b15f8a4775f087a629e35a7092ff980758cb054292c04c57
issue-collapse:092ffecdf4f5a52e,module: dtensor needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,154266,,,,"module: dtensor, oncall: distributed, oncall: pt2","module: dtensor, needs reproduction, oncall: distributed, oncall: pt2, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#154266 torch dynamo fails on GH200 with world size 5,40bcfd670523e3f104dd6030fbdf4ddbb4e4017b2c39edc76d59b4d39cc500e4
issue-collapse:4d5dd70bc0413a87,module: deadlock needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,25904,,,,module: deadlock,"module: deadlock, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#25904 Python hang after using torch.exp(),adb87219cbf499e666f2866924d60e62e73b442674aca270075255fef77733ed
issue-collapse:38a235a350d47882,module: data needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,67740,,,,module: data,"module: data, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#67740 `torch.utils.data.random_split` example broken in 1.10,11044383023a52f8a456e670cf904524eab8d9ef2edee7e9099c8fbd390cbd22
issue-collapse:66b7a72ea9a9de47,module: complex needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,127978,,,,"module: complex, module: nans and infs","module: NaNs and Infs, module: complex, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#127978 torch.addr outputs NaN when computing on inf vectors,4757f37b5b35e28b8a5fd6b68ec26dc98c7a46e862e29934d101e76f2f604c19
issue-collapse:26583f0581d9caff,module: advanced indexing needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,110946,,,,module: advanced indexing,"module: advanced indexing, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#110946 RuntimeError: out_ptr == out_accessor[thread_count_nonzero[tid + 1]].data() INTERNAL ASSERT FAILED,c707c24352e87be76fb26f2fb2eeebde2889a6cb98b8313bb06839a172524e5b
issue-collapse:080280e3132f1c84,module: activation checkpointing needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,163923,,,,module: activation checkpointing,"module: activation checkpointing, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#163923 Enabling activation checkpoint debug makes throughput gradually slow down to crawl.,24aa366a810a27ebe66591ab57de4a20510526f96f57fdf87db5b8d25300bf0f
issue-collapse:6151d10a155204ba,module: abi needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,124197,,,,module: abi,"module: abi, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#124197 Libtorch crashes docker when included in header file,522cba2ee489d083ced2d056d3522e82bc6dd301b4bc839e2d91e8e556246210
issue-collapse:a3f2448f1f9319c6,module: __torch_function__ needs reproduction,single issue review,needs reproduction,14,low,bucket_component_queue,1,0,0,0,103752,,,,"module: __torch_function__, module: docker","module: __torch_function__, module: docker, needs reproduction, triaged",,Kept as a singleton so the audit trail still proves this issue was represented.,Needs human review.,1,Represents 1 open issue; raw issue rows remain available for audit.,#103752 Pytorch not calling to C code from a docker container,aa1cfa460ed0018846f1997f9ecfc06b671bade9cf6c279a372101651c9f024d
issue-collapse:c108e50d3708df40,oncall: fx stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,97659,,,,oncall: fx,"oncall: fx, triaged",,Handle this oncall: fx stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#97659 How do I get the original object wrapped by the torch.fx.proxy class？,b5392ec8a4cbff38ca387307041f81ce9032a2db905f18b77e70feff4306d316
issue-collapse:f25b576ee4e05e30,oncall: export stale/low urgency  batch 3,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,106137,,,,oncall: export,"oncall: export, pre_dispatch tracing, triaged",,Handle this oncall: export stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#106137 `export(..., pre_dispatch=True)` for model in eval mode still inserts autograd ops",396b4995fb423405b951411e211bd7fbb81de30fa121b5f770eb2624a24dcd6c
issue-collapse:362fcb0a7ea41af0,module: xnnpack stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,36517,,,,module: xnnpack,"better-engineering, module: xnnpack, triaged",,Handle this module: xnnpack stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#36517 XNNPACK operators are not actually registered under xnnpack namespace,d9e78cd1847f2509437ded9340547c7feffed6091f1cc8027c72c387b0ae093b
issue-collapse:31db396263432e79,module: unknown stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,70180,,,,module: unknown,"enhancement, module: unknown, triaged",,Handle this module: unknown stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#70180 Can you provide the torch.trt module to directly convert the pytorch weights to tensorrt?,26d2d1511483ac25c8434bd3ef2ecac2d27e32c8acbea0e443e247d97a6942ee
issue-collapse:17aaeeb76de0a8ca,module: tensorpipe stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,39082,,,,module: tensorpipe,"module: tensorpipe, triaged",,Handle this module: tensorpipe stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#39082 [TensorPipe] Avoid wrapping the future message in order to do atomic test-and-set,21042314c211cd9d9343371e6c086414bac0fbf2c8e7d34bd7eae50e1d4828fa
issue-collapse:c1698ab112b48232,module: tbb stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,90555,,,,module: tbb,"enhancement, module: tbb, triaged",,Handle this module: tbb stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#90555 [RFC] Add torch.backends.tbb.is_available(),0a3227f255b8c71afdf814fe3308c7baabd84889611f4dd5381bedfd0d0d938a
issue-collapse:9dd75edce8b6282a,module: symm_mem stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,167765,,,,module: symm_mem,"module: symm_mem, triaged",,Handle this module: symm_mem stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#167765 [RFC] Extend MemPool support to SymmMem CUDA backend,4b29c67883e2ca7390d53f4927f4c66f25d92668d3a2533c504a522f25427177
issue-collapse:edf3f2d73a437360,module: shape checking stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,56542,,,,module: shape checking,"module: shape checking, triaged",,Handle this module: shape checking stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#56542 CrossEntropyLoss target shape expectation is inconsistent with BCEWithLogits,f077e209acd0124f6fcf54afcf6c1e1a556a19faa67a78194f01c19a5b69844d
issue-collapse:b4efd143ce727d07,module: scipy compatibility stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,119462,,,,"module: scipy compatibility, module: tests","module: scipy compatibility, module: tests, triaged",,Handle this module: scipy compatibility stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#119462 Python 3.12 CPU TestBinaryUfuncsCPU.test_cumulative_trapezoid_cpu is failing,b1b56dccfcc31228ab883cefe5e02a4b22e241fff7045442aa05009435185599
issue-collapse:f1ca357af52034e3,module: optimizer stale/low urgency  batch 7,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,6083,,,,"module: optimizer, oncall: distributed","module: optimizer, oncall: distributed, triaged",,Handle this module: optimizer stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#6083 [feature request] DC-ASGD (Delay Compensated Asynchronous Stochastic Gradient Descent),dd575ecd04f5a71e7e3e81b8b90f668cb21e724bcaca377b89abebf06802656b
issue-collapse:d79d3024215a537f,module: openmp stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,101188,,,,module: openmp,"module: openmp, triaged",,Handle this module: openmp stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#101188 Weird dataloader performance degradation caused by torch and numpy import order,846e5a6de6f080ef4f36164a2d7c872a45f14a618e46f7288829a9518d027ad0
issue-collapse:930dade55609cb1c,module: numba stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,23067,,,,module: numba,"feature, low priority, module: numba, triaged",,Handle this module: numba stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#23067 creation of a tensor from a numba.cuda array,3fabf9655b6892d0d8ccf49e03c6a186cc734dcf73279996ff2d9ea4e2f63cf1
issue-collapse:443bc1d51273e0b2,module: mps stale/low urgency  batch 4,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,84813,,,,module: mps,"enhancement, module: mps, triaged",,Handle this module: mps stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#84813 MPS: allow selecting specific MTLDevice by registryID via environment variable,d17063d1e7a7d26b41e73e939f0b085dcff8a1c36e8960961b7e00e15c4e5c0d
issue-collapse:d157b1cb4d3c6508,module: minifier stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,114356,,,,module: minifier,"module: minifier, triaged",,Handle this module: minifier stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#114356 [WARNING] could not load None, generating random data instead is too spammy",e6eb7eb7f021d08e2cc4389d6607c7c34ed1afee06f507a9544edfaed7f8c3a9
issue-collapse:d7f6ecf4067d42c1,module: m1 stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,88614,,,,"module: m1, module: mps","actionable, module: m1, module: mps, triaged",,Handle this module: m1 stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#88614 The libtorch test TestScalarTensor.TestScalarTensorMPS fails on Apple Silicon,893af6de7cfe0582d4784b726366951b1963ecf489d28fd01982dc835de867d8
issue-collapse:8621def8707b3711,module: lint stale/low urgency  batch 3,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,20466,,,,module: lint,"module: lint, triaged",,Handle this module: lint stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#20466 Lint rule to prevent direct use of #pragma omp,c47ec03a19271c9e96b3c3d28cf45741577d9be20744bc52c5d165e025cd15c9
issue-collapse:a0624f9fe8ea2df4,module: jiterator stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,94443,,,,module: jiterator,"module: jiterator, triaged",,Handle this module: jiterator stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#94443 A better error msg for `cuda.jiterator` when input is on `cpu`,19af9f172bd627bc9d39cfdc328bbeb1fa80bd329f47e540d2baab598b2886e7
issue-collapse:2919c74d28fa84c8,module: graph breaks stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,114805,,,,module: graph breaks,"module: graph breaks, triaged",,Handle this module: graph breaks stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#114805 Unsupported: call_method UserDefinedObjectVariable(_profiler_enabled) __call__ [] {},030da74b4c5b8efef93d04f1ff469d080bdc03ddd1ebee9d15ca23ac26000522
issue-collapse:4c083401aee84fcd,module: functional ux stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,104768,,,,"module: functional ux, module: nn","module: functional UX, module: nn, triaged",,Handle this module: functional ux stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#104768 Support for `eval` in functional_call,6d623c9dd96fe25728dd97bc0e6a272661ae9c9f1a854f00fabf9e15ae394d85
issue-collapse:0071b1c330bdff7a,module: flaky-tests stale/low urgency  batch 2,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,8044,,,,"module: flaky-tests, module: memory usage","module: flaky-tests, module: memory usage, triaged",,Handle this module: flaky-tests stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#8044 [PyTorch] Windows CI CUDA mem leak check on BN tests are flaky,033f01346fda66977dbf604b62b07b6c4eed8ef75e198be7898e00a4d7d716e5
issue-collapse:aa8ebad977b7991c,module: elastic stale/low urgency  batch 2,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,60911,,,,module: elastic,"module: elastic, triaged",,Handle this module: elastic stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#60911 [elastic launcher] redirects/tee support for global rank,ccfd4eab88b6747c6c287c181d62fae9d4101d4aa6407cea85f205373e30a727
issue-collapse:221b86e58939ece8,module: core aten stale/low urgency,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,133941,,,,module: core aten,"module: core aten, triaged",,Handle this module: core aten stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,"#133941 Why core aten IR doesn’t contain scalar_tensor overload version of some ops, like bitwise-like ops, remainder and so on?",69eb9ffae223b89f1be8b8853d5e4a4ed7795c856b60ab0e0e36c17c82fcf4e2
issue-collapse:c94a786e6cad010c,module: boolean tensor stale/low urgency  batch 2,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,25480,,,,module: boolean tensor,"enhancement, module: boolean tensor, triaged",,Handle this module: boolean tensor stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#25480 Vectorize bool operations,31a8d6532f1bce7d5c6e249dcf99266edb74e36705afcea395a3dfb2103cc08d
issue-collapse:177e91dd9ae1daac,module: autograd stale/low urgency  batch 19,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,20272,,,,"module: autograd, module: bootcamp","low priority, module: autograd, module: bootcamp, triaged",,Handle this module: autograd stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#20272 Redundantly saving sizes of SavedVariables in autograd Function,555f9ba7adfac14cfb46fa54df1fdb0e58b39372d5925a36d728602ecb21a839
issue-collapse:318d0a1713a6449e,module: arm stale/low urgency  batch 5,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,42439,,,,"module: arm, module: binaries","module: arm, module: binaries, triaged",,Handle this module: arm stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#42439 How to use libtorch on Jetson TX2,3e8ab63d667733e70792047007c146ace4e8ed432e950524aea27ba396a7b4c0
issue-collapse:e2f8aeb24dfee121,module: advanced indexing stale/low urgency  batch 3,single issue review,stale/low urgency,2,low,bucket_component_queue,1,0,0,0,50737,,,,"module: advanced indexing, module: docs, module: scatter & gather ops","module: advanced indexing, module: docs, module: scatter & gather ops, triaged",,Handle this module: advanced indexing stale queue with one sweep policy instead of rediscovering old context issue by issue.,Batch sweep for close/reproduce/docs-update decisions; keep any live evidence attached.,1,Represents 1 open issue; raw issue rows remain available for audit.,#50737 Adding visualizations for indexing and other functions,de88a3e653951280e49922a58ce76e7232c58fa6411d1029233c4487ebc446a3