-
Notifications
You must be signed in to change notification settings - Fork 28.8k
All issues
Issue creation is restricted in this repository
Issues
is:issue state:open
is:issue state:open
Search results
torch.compilesilently computes wrong gradients for chained modulated LayerNorms with a shared scale/shift parameter (regression 2.9.1 → 2.10.0)bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botStatus: Open.#193061 In pytorch/pytorch;MPS ignores storage offset of zero-dimensional CPU tensor views in mixed-device arithmetic
bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: correctness (silent)issue that returns an incorrect result silentlyissue that returns an incorrect result silentlymodule: mpsRelated to Apple Metal Performance Shaders frameworkRelated to Apple Metal Performance Shaders frameworkStatus: Open.#193057 In pytorch/pytorch;Upstream dispatchless custom ops to PyTorch
module: custom-operatorscustom operators, custom ops, custom-operators, custom-opscustom operators, custom ops, custom-operators, custom-opsmodule: pt2-dispatcherPT2 dispatcher-related issues (e.g., aotdispatch, functionalization, faketensor, custom-op,PT2 dispatcher-related issues (e.g., aotdispatch, functionalization, faketensor, custom-op,triagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.#193056 In pytorch/pytorch;[triage bot] validate_labels.py hard-blocks and silently skips triage when all proposed labels are filtered
bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: ciRelated to continuous integrationRelated to continuous integrationmodule: regressionIt used to work, and now it doesn'tIt used to work, and now it doesn'tStatus: Open.#193045 In pytorch/pytorch;torch.atan2backward returns±inf(then silently0) in float16 where the gradient is an ordinary numberbot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: derivativesRelated to derivatives of operatorsRelated to derivatives of operatorsmodule: halfRelated to float16 half-precision floatsRelated to float16 half-precision floatsmodule: numerical-stabilityProblems related to numerical stability of operationsProblems related to numerical stability of operationstriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.#193040 In pytorch/pytorch;[v.2.14.0] Release Tracker
release trackerAdd this label to release tracker issuesAdd this label to release tracker issuestriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.Async TP still requires deprecated enable_symm_mem_for_group
bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: symm_memIssues and PRs of Symmetric MemoryIssues and PRs of Symmetric Memoryoncall: distributedAdd this issue/PR to distributed oncall triage queueAdd this issue/PR to distributed oncall triage queuetriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.#193027 In pytorch/pytorch;torch.autograd.functional.hessianreturnsinf/nanin float16 wheretorch.func.hessianreturns the correct valuebot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: autogradRelated to torch.autograd, and the autograd engine in generalRelated to torch.autograd, and the autograd engine in generalmodule: double backwardsProblem is related to double backwards definition on an operatorProblem is related to double backwards definition on an operatormodule: NaNs and InfsProblems related to NaN and Inf handling in floating pointProblems related to NaN and Inf handling in floating pointtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.#193012 In pytorch/pytorch;rocm-navi31has been 100% red since 2026-07-30: Triton AMD backend fails to link hsaco (ld.lld: target emulation unknown)module: ciRelated to continuous integrationRelated to continuous integrationmodule: rocmAMD GPU support for PytorchAMD GPU support for PytorchStatus: Open.#193009 In pytorch/pytorch;UNSTABLE rocm-navi31 / linux-jammy-rocm-py3.10-navi31 / test (default)
bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: ciRelated to continuous integrationRelated to continuous integrationmodule: rocmAMD GPU support for PytorchAMD GPU support for PytorchtriagedThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleThis issue has been looked at a team member, and triaged and prioritized into an appropriate moduleStatus: Open.#193008 In pytorch/pytorch;torch.linalg.normreturnsinffor norms well inside the dtype range, and its gradient is silently0(float32/float64)bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: linear algebraIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulIssues related to specialized linear algebra operations in PyTorch; includes matrix multiply matmulmodule: numerical-stabilityProblems related to numerical stability of operationsProblems related to numerical stability of operationsStatus: Open.#193006 In pytorch/pytorch;DISABLED test_flattened_submesh_no_getattr_compile_on_one_rank (__main__.TestDTensorCompileWithCompiledAutograd)
bot-triagedThis is a label only to be used by the auto triage botThis is a label only to be used by the auto triage botmodule: xpuIntel XPU related issuesIntel XPU related issuesoncall: distributedAdd this issue/PR to distributed oncall triage queueAdd this issue/PR to distributed oncall triage queueskippedDenotes a (flaky) test currently skipped in CI.Denotes a (flaky) test currently skipped in CI.Status: Open.#193002 In pytorch/pytorch;