Graph-connected issue/PR units with referenced context and review or CI gates attached.
| Thread | Score | Issues | PRs | Next Action | Edges |
|---|---|---|---|---|---|
| PR #188801 - [MPS] Leak MetalShaderLibrary bundled singleton to avoid exit-time destructor crashes waiting on contributor | 151 | #188812 | PR #188801 | Wait for contributor update on PR #188801; keep related issue/PR context attached. | 1 |
| PR #189122 - Fix sparse-dense mul dropping data when broadcasting a size-1 sparse dim ready for maintainer decision | 139 | #158861#188900 | PR #189122 | Review active PRs and linked issues as one maintainer work thread. | 4 |
| Issue #186535 - Windows, gloo: Access violation (0xC0000005) in ProcessGroupGloo::enqueue when calling allreduce on CUDA tensors — GlooAllreduceRegistry has no kCUDA creator needs design decision | 139 | #186535 | none | needs triage | 0 |
| Issue #188323 - [Inductor][CPU] dynamic=True convolution lowering crashes with ValueError: Exponent must be non-negative needs design decision | 139 | #188323 | none | needs triage | 0 |
| Issue #157668 - NCCL error caused due to use of NVLS in torch 2.7.1-cu128 on aarch64 gb200 cluster needs design decision | 127 | #157668 | none | needs triage | 0 |
| Issue #187912 - [CPU] Concurrent `cpublas::brgemm` calls can crash in the AMX path when the underlying oneDNN ukernel is shared needs design decision | 127 | #187912 | none | needs triage | 0 |
| PR #185730 - Fix dynamic shapes for variadic kwargs waiting on CI/check fix | 121 | #150022#150371 | PR #185730 | Review active PRs and linked issues as one maintainer work thread. | 5 |
| Issue #116254 - C++ API `at::quantized_max_pool2d`: Heap-buffer-overflow needs design decision | 121 | #116254#162476 | none | needs triage | 1 |
| Issue #162422 - Runtime failure when running torch.compile() & using GCC 11.5.0 on Neoverse V1 needs design decision | 121 | #162422 | none | needs triage | 0 |
| Issue #154297 - Hangs and timeouts on dist.reduce_scatter on B200 GPU needs maintainer decision | 115 | #154297#162178#162745#162748#162820#162871#162897#162917#162940#163429#165170#165685#187158#189065 | none | needs maintainer decision | 21 |
| Issue #55655 - Performance debugging / warning mode stale/low urgency | 115 | #55655#57118#68768#72948#75725 | none | stale/low urgency | 4 |
| Issue #66504 - BatchNorm runtimeError: one of the variables needed for gradient computation has been modified by an inplace operation needs design decision | 115 | #66504#68407#73332 | none | needs triage | 3 |
| Issue #144965 - RuntimeError "global alloc not supported yet" when using TorchScript optimization. needs design decision | 115 | #69078#144965 | none | needs triage | 1 |
| Issue #106164 - distributed.batch_isend_irecv() crash when send/recv refers to itself needs design decision | 115 | #106164 | none | needs triage | 0 |
| Issue #116423 - PyTorch Distributed Elastic Launch Segmentation Fault with Python 3.12 needs design decision | 115 | #116423 | none | needs triage | 0 |
| Issue #119845 - Segmentation fault in dataloader worker sub-process needs design decision | 115 | #119845 | none | needs triage | 0 |
| Issue #145610 - mmap fails on 64k page aarch64 systems for AOTI model loading needs design decision | 115 | #145610 | none | needs triage | 0 |
| Issue #153517 - [CI][CUDA][Distributed] test_non_blocking_with_eager_init timeout needs design decision | 115 | #153517 | none | needs triage | 0 |
| Issue #162731 - DTensor cached op propagation results can be mutated when propagating other ops needs design decision | 115 | #162731 | none | needs triage | 0 |
| Issue #167693 - DISABLED test_side_stream_backward_overlap_cuda (__main__.TestAutogradStreamSynchronizationCUDA) needs design decision | 115 | #167693 | none | needs triage | 0 |
| Issue #170003 - batch_isend_irecv with nccl causes illegal memory access depending on P2P ordering needs design decision | 115 | #170003 | none | needs triage | 0 |
| Issue #173921 - [Inductor] Significant numerical divergence (4.7% relative error) in Conv2d on CPU with torch.compile needs design decision | 115 | #173921 | none | needs triage | 0 |
| Issue #174288 - [distributed] Batched isend/irecv with NCCL backend hangs on high load needs design decision | 115 | #174288 | none | needs triage | 0 |
| Issue #179502 - [DTensor] sharded view incorrectly passes when redistribution is needed needs design decision | 115 | #179502 | none | needs triage | 0 |
| Issue #180088 - [DTensor] RNG tracker does not advance state for CPU tensors on CUDA mesh, causing trunc_normal_ infinite loop needs design decision | 115 | #180088 | none | needs triage | 0 |
| Issue #183459 - Regression in torch.distributed._functional_collectives.all_to_all_single in 2.9.1 -> 2.11 needs design decision | 115 | #183459 | none | needs triage | 0 |
| Issue #189126 - Report a bug, as requested by the error message needs design decision | 115 | #189126 | none | needs triage | 0 |
| Issue #189144 - [XPU][B580] `flex-attn-causal` performance drop with `last_level_cache_size` cache clear between runs needs design decision | 115 | #189144 | none | needs triage | 0 |
| Issue #189150 - SIGSEGV / cudaErrorIllegalAddress when replaying a CUDA graph that captured multiple training iterations of two trainers over deeply unrolled recurrent modules needs design decision | 115 | #189150 | none | needs triage | 0 |
| Issue #189194 - [ROCm] torch 2.13 wheel: "Can't detect vectorized ISA for CPU" in torch.compile smoke test on non-ROCm image (regression vs 2.12.1) needs design decision | 115 | #189194 | none | needs triage | 0 |
| Issue #189239 - [Inductor]Unable to apply layout optimization on convolution backward causes training performance regression. needs design decision | 115 | #189239 | none | needs triage | 0 |
| Issue #189281 - [CPU] operator_benchmark: embedding / embeddingbag ~70-130x slower since ~Nov 2025 (x86 + aarch64; possible measurement issue) needs design decision | 115 | #189281 | none | needs triage | 0 |
| Issue #189282 - [CPU] operator_benchmark: backward-pass (add, + batchnorm on aarch64) ~200-500x slower since ~Nov 2025 (x86 + aarch64; possible measurement artifact) needs design decision | 115 | #189282 | none | needs triage | 0 |
| Issue #71303 - [RFC] Cross-Process Performance Analysis: Straggler Detection needs design decision | 115 | #71303 | none | needs triage | 0 |
| Issue #76449 - Enhance _verify_param_shape_across_processes needs design decision | 115 | #76449 | none | needs triage | 0 |
| Issue #81684 - Message exchange failure when perform alltoallv (cpus) needs design decision | 115 | #81684 | none | needs triage | 0 |
| Issue #85088 - reentrant torch.utils.checkpoint does not work with NamedTuple outputs needs design decision | 115 | #85088 | none | needs triage | 0 |
| PR #186461 - Fix SymBool equality handling in symbolic shapes ready for maintainer decision | 112 | #124110 | PR #186461 | Review active PRs and linked issues as one maintainer work thread. | 1 |
| PR #186429 - Fix ONNX export of pad_sequence with symbolic split lengths ready for maintainer decision | 112 | #127153 | PR #186429 | Review active PRs and linked issues as one maintainer work thread. | 1 |
| PR #185691 - Avoid UB in float to signed integer casts ready for maintainer decision | 112 | #151510 | PR #185691 | Review active PRs and linked issues as one maintainer work thread. | 1 |
| PR #184716 - Fix conv_transpose2d meta output padding validation ready for maintainer decision | 112 | #178127 | PR #184716 | Review active PRs and linked issues as one maintainer work thread. | 1 |
| PR #188632 - Fix Dynamo opaque object staticmethod guards ready for maintainer decision | 112 | #188544 | PR #188632 | Review active PRs and linked issues as one maintainer work thread. | 3 |
| PR #184939 - Support SymInt steps for linspace/logspace export ready for maintainer decision | 112 | none | PR #184939 | ready for maintainer decision | 0 |
| PR #184648 - [Test] Add has_sufficient_memory() hook to DeviceTypeTestBase waiting on CI/check fix | 108 | none | PR #184648 | PR blocked | 1 |
| PR #185882 - [PrivateUse1] Return `False` instead of `None` when PU1 backend is not available waiting on CI/check fix | 108 | none | PR #185882 | PR blocked | 0 |
| PR #186927 - [MPS] Gemv kernels waiting on CI/check fix | 108 | none | PR #186927 | PR blocked | 0 |
| PR #187501 - [Stable C Shim] Use error message retrieval shim if available at runtime waiting on CI/check fix | 108 | none | PR #187501 | PR blocked | 0 |
| PR #188237 - Optimize CSR SpMM CPU grain size and inner accumulation waiting on CI/check fix | 108 | none | PR #188237 | PR blocked | 1 |
| PR #188301 - [inductor] Fix inductor dropping ordering dep between effectful ops with different kernel types waiting on CI/check fix | 108 | none | PR #188301 | PR blocked | 2 |
| PR #189307 - [CI/CD] Copy newer CUPTI headers into manywheel binary-build images waiting on CI/check fix | 108 | none | PR #189307 | PR blocked | 0 |
| PR #189312 - [CUDA][cuBLAS] Change cuBLAS default workspace size for SM 11.0 to 32 MiB waiting on CI/check fix | 108 | none | PR #189312 | PR blocked | 0 |
| PR #188948 - Fix triangular_solve for sparse CPU tensors on non-MKL platforms waiting on contributor | 107 | #153410 | PR #188948 | Wait for contributor update on PR #188948; keep related issue/PR context attached. | 2 |
| PR #181720 - [MPS] Make pin_memory return CPU-aliased storage backed by a unified MTLBuffer has linked issue | 106 | #180397#181374#188970 | PR #181720PR #189256 | Review active PRs and linked issues as one maintainer work thread. | 9 |
| PR #189043 - Preload full bundled cuDNN set with RTLD_GLOBAL to prevent sublibrary version mismatch waiting on contributor | 106 | #188892 | PR #189043 | Wait for contributor update on PR #189043; keep related issue/PR context attached. | 2 |
| PR #135631 - [scan] Autograd waiting on CI/check fix | 100 | none | PR #135631 | PR blocked | 0 |
| PR #188115 - [Inductor][TEST] Align `test_main_loop_scaling` with H100 support surface waiting on CI/check fix | 98 | none | PR #188115 | PR blocked | 0 |
| Issue #103625 - DISABLED test_backward_ddp_outside (__main__.TensorPipeDdpUnderDistAutogradTest) needs maintainer decision | 97 | #15963#16295#51062#58856#64093#69178#69179#94003#103625#104011#132845 | none | needs maintainer decision | 10 |
| Issue #104506 - TImeout in NCCL doesn't work stale/low urgency | 97 | #88103#104506#124950 | none | stale/low urgency | 2 |
| Issue #42258 - Cannot load certain function from dumped Torchscript file needs design decision | 97 | #42258#46944#46946 | none | needs triage | 2 |
| Issue #58111 - Static graph training fails if forward is called multiple times before backward needs design decision | 97 | #58111#62998#70865 | none | needs triage | 2 |