Issues

• Portable div(rounding_mode="trunc") and fmod round int64 values through double

  pytorch/executorch#24169 · Filter by author usmanmateen opened on Oct 27, 2026

• Float scalars in float64 programs are rounded to float32 by ScalarToTensorPass

  pytorch/executorch#24167 · Filter by author usmanmateen opened on Oct 27, 2026

• [QNN] 16a8w: gathered tensor read wrong by a consumer that only feeds a graph output

  pytorch/executorch#24125 · Filter by author msluszniak opened on Oct 23, 2026

• Elastic on-device inference - swap LLM mid-generation and carry KV cache (molt / kvine)

  pytorch/executorch#24106 · Filter by author NagaYu opened on Oct 23, 2026

• Arm/TOSA: allow int64 through when the spec declares the int64 extension

  bugmodule: armpartner: arm
  Bug

  pytorch/executorch#24095 · Filter by author abm-77 opened on Oct 22, 2026
  4

  

• Windows: _get_flatc_path() looks for data/bin/flatc but the wheel ships flatc.exe (WinError 2)

  bugmodule: exirtriaged
  Bug

  pytorch/executorch#23989 · Filter by author ErenAta16 opened on Oct 19, 2026
  

• executor_runner segfaults in portable kernels when a .pte tensor has neither a data buffer nor a memory plan (null data pointer)

  bugmodule: runtime
  Bug

  pytorch/executorch#23955 · Filter by author professor-moody opened on Oct 18, 2026
  

• [QNN] 16a4w_block stride-2 conv + ReLU fails HTP graph prepare on SM8850 (Relu16bParamsPlaceholder)

  bugmodule: qnnpartner: qualcomm
  Bug

  pytorch/executorch#23922 · Filter by author msluszniak opened on Oct 16, 2026
  

• Cortex-M backend replaces channels-first dw conv with channels-last dw conv without handling the changed layout, fails at runtime

  bugmodule: microcontrollerspartner: arm
  pytorch/executorch#23812 · Filter by author Erik-Lundell opened on Oct 14, 2026
  2

• [muse-glimmer] interpolate_pos_emb returns corrupted bf16 values on CPU and can segfault

  bugmodule: examplestriaged
  Bug

  pytorch/executorch#23797 · Filter by author AxelNoun opened on Oct 14, 2026
  2

  

• macOS Python wheel _portable_lib has non-relocatable libtorch rpath

  pytorch/executorch#23742 · Filter by author robell opened on Oct 13, 2026
  1

  

• torch.var_mean and torch.std_mean fail in to_edge: aten.var_mean.correction is not core ATen and has no core decomposition

  bugmodule: exirtriaged
  Bug

  pytorch/executorch#23680 · Filter by author usmanmateen opened on Oct 10, 2026
  

• Prompts of max_seq_len tokens fail to prefill: export_llm bounds the KV-cache token input at max_seq_len - 1 but publishes get_max_seq_len = max_seq_len

  bugmodule: extensionmodule: llmtriaged
  Bug

  pytorch/executorch#23678 · Filter by author alpharomercoma opened on Oct 10, 2026

• [v1.6.0] Release Schedule and Tracker

  pytorch/executorch#23608 · Filter by author JacobSzwejbka opened on Oct 8, 2026
  

• QNN sharded LLM export fails: ResolveDebugHandle is not the last edge pass (dep_table entry overwritten by SplitGraph registration)

  bugmodule: llmmodule: qnntriaged
  Bug

  pytorch/executorch#23580 · Filter by author SabYic opened on Oct 8, 2026
  

• yolo26 example breaks on portable with ultralytics newer than 8.4.14

  bugmodule: examples
  Task

  pytorch/executorch#23453 · Filter by author shoumikhin opened on Oct 6, 2026
  3

  

• Add Arm export and tests for MLPerf Tiny streaming wakeword

  Feature

  pytorch/executorch#23440 · Filter by author rascani opened on Oct 5, 2026
  2

• Add Cortex-M export and tests for MLPerf Tiny streaming wakeword

  Feature

  pytorch/executorch#23434 · Filter by author rascani opened on Oct 5, 2026
  

• Arm backend: int16 (A16W8) rsqrt lowered to a 513-point TOSA TABLE is several times off for small inputs, which breaks A16W8 RMSNorm/LayerNorm on Ethos-U85

  module: armtriaged
  Task

  pytorch/executorch#23356 · Filter by author tiago-vello opened on Oct 2, 2026
  3

  

• XNNPACK never delegates slice_copy with stride != 1, but op-support.csv only excludes zero-dim/dynamic shapes

  module: xnnpack
  pytorch/executorch#23258 · Filter by author JakeStevens opened on Sep 29, 2026
  

• [RFC] ExecuTorch Persisting Device Specialized Delegate Artifacts

  pytorch/executorch#23192 · Filter by author JacobSzwejbka opened on Sep 28, 2026
  6

  

• [QNN] Enable ConvTranspose + BatchNorm fusion after #23170

  module: qnnpartner: qualcomm
  pytorch/executorch#23185 · Filter by author mergennachin opened on Sep 28, 2026
  1

  

• [Tracking] Complete audio task coverage across CPU, GPU, and NPU backends

  enhancementmodule: examples
  pytorch/executorch#23164 · Filter by author mergennachin opened on Sep 25, 2026
  8

• Qualcomm: 8-bit per-channel weight scales are floored at the 16-bit eps, and the HTP miscomputes near-zero channels

  module: qnnmodule: quantizationpartner: qualcomm
  Task

  pytorch/executorch#23160 · Filter by author msluszniak opened on Sep 25, 2026
  1

  

• [cpu kernels] native_layer_norm: layer_norm_scalar returns NaN on large-mean rows; Half/BF16 at N>=256 slow after #23153

  module: kernels
  pytorch/executorch#23159 · Filter by author mergennachin opened on Sep 25, 2026
  2