pytorch/executorch

On-device AI across mobile, embedded and edge for PyTorch

Stars: 5k
Forks: 1.2k
Open beginner issues: 0
Indexed issues: 851
Avg merge: 2d 10h
Merged PRs (30d): 581

Dominant language: Python
License: No license data
Last GitHub push: Sep 19, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: good first issue help wanted

Beginner friendly Everything

Open All

851 issues indexed so far

Some ET workflows running on PRs are not cancelled when the PRs are updated Open

pytorch/executorch#22845 · 1 assignee · Opened Sep 15, 2026
[ET-VK] Static activation quantization passes the int8 tensor to linear_q8ta_q8csw, which expects the float Open

bug module: vulkan triaged

pytorch/executorch#22775 · 3 comments · 1 assignee · Opened Sep 13, 2026
[ET-VK] Conv2d with padding="valid" asserts in make_ivec2 (breaks every SigLIP/CLIP patch embed) Open

bug module: vulkan triaged

pytorch/executorch#22774 · 1 assignee · Opened Sep 13, 2026
[ET-VK] RMS-norm fusion folds a non-leaf multiplier and then prepacks it, breaking adaptive norms Open

bug module: vulkan triaged

pytorch/executorch#22773 · 1 assignee · Opened Sep 13, 2026
[iOS] Split the image processor into its own SwiftPM product instead of bundling it in executorch Open

module: build/install module: ios

pytorch/executorch#22722 · 1 assignee · Opened Sep 11, 2026
[Docs] Add a structured issue form for ExecuTorch deployment stories Open

enhancement module: doc module: user experience

pytorch/executorch#22687 · 1 comment · 1 reaction · 1 assignee · Opened Sep 10, 2026
[LLM runner] Correct the deprecated seq_len CLI help text Open

bug module: llm

pytorch/executorch#22681 · 2 assignees · Opened Sep 10, 2026
[Android] Honor standard SDK environment variables in build_android_library.sh Open

bug module: android module: build/install

pytorch/executorch#22682 · 2 assignees · Opened Sep 10, 2026
[MLX] Probe the usable Metal compiler in Apple CMake presets Open

bug module: build/install module: mlx

pytorch/executorch#22683 · 3 assignees · Opened Sep 10, 2026
[Core ML] Avoid unnecessary portable-kernel linkage from coremldelegate Open

bug module: build/install module: coreml

pytorch/executorch#22684 · 1 assignee · Opened Sep 10, 2026
[LLM runner] Align make_image_input HWC contract with runtime behavior Open

bug module: llm

pytorch/executorch#22685 · 3 comments · 3 assignees · Opened Sep 10, 2026
[ET-VK] is_coopmat_eligible() disables cooperative matrix on all integrated GPUs Open

enhancement module: vulkan needs reproduction

pytorch/executorch#22665 · 6 comments · 1 reaction · 1 assignee · Opened Sep 10, 2026
Libtorch backend Open

module: backend rfc triaged

pytorch/executorch#22598 · 6 comments · 2 assignees · Opened Sep 8, 2026
[Arm/Ethos-U] vela_bin_stream parser does not bounds-check block sizes or IO descriptors Open

bug module: arm partner: arm security-fix triaged

pytorch/executorch#22579 · 1 comment · 2 assignees · Opened Sep 5, 2026
Depthwise separable convolution computation by portable kernels expects wrong dim_order Open

bug module: kernels module: nxp module: runtime

pytorch/executorch#22520 · 2 assignees · Opened Sep 3, 2026
MLX encode allocates ~0.9 GB on iOS for a program that peaks at 208 MB on macOS Open

bug module: ios module: mlx triaged

pytorch/executorch#22513 · 3 comments · 2 assignees · Opened Sep 3, 2026
[ET-VK] TagMemoryMetaPass skips a tensor that sits in a list alongside None Open

bug module: vulkan triaged

pytorch/executorch#22510 · 3 comments · 1 assignee · Opened Sep 3, 2026
Memory planning asserts `graph_output_allocated not set` when a graph output carries no tensors Open

bug module: exir

pytorch/executorch#22477 · 2 assignees · Opened Sep 2, 2026
[ET-VK] VulkanQuantizer dynamic activation quantization is per tensor, and very lossy on transformer encoders Open

bug module: quantization module: vulkan triaged

pytorch/executorch#22431 · 1 assignee · Opened Sep 2, 2026
Core ML llama export ignores enable_state and preserve_sdpa Open

bug module: coreml module: llm

pytorch/executorch#22352 · 2 comments · 1 assignee · Opened Aug 31, 2026
Building a backend tailored to Espressif systems Open

module: backend module: microcontrollers need-user-input rfc triaged

pytorch/executorch#22334 · 6 comments · 2 assignees · Opened Aug 30, 2026
[ET-VK] Embedding with vocab > 16384 silently returns wrong values on the texture path; force_fp16 breaks all-MiniLM-L6-v2 Open

bug high priority module: vulkan triaged

pytorch/executorch#22333 · 1 comment · 1 assignee · Opened Aug 30, 2026
[ET-VK][Adreno] fp32 nn.Linear returns wrong results on ~23% of executions at large M Open

bug high priority module: vulkan triaged

pytorch/executorch#22327 · 1 assignee · Opened Aug 30, 2026
Arm backend: requirements-arm-tosa.txt is uninstallable on Python 3.13/3.14 although pyproject.toml declares support up to <3.15 Open

bug dependencies module: arm module: build/install partner: arm

pytorch/executorch#22314 · 2 assignees · Opened Aug 29, 2026
ExportPass opens a second FakeTensorMode when every input is a lifted constant, aborting lowering Open

bug module: exir triaged

pytorch/executorch#22309 · 2 assignees · Opened Aug 29, 2026
examples/ is outside the mypy lint scope, which let a missing-attribute bug reach main Open

module: ci module: dev infra

pytorch/executorch#22302 · 1 comment · 2 assignees · Opened Aug 29, 2026
[v1.5.0] Release Schedule and Tracker Open

release tracker

pytorch/executorch#22188 · 5 comments · 1 reaction · 1 assignee · Opened Aug 26, 2026
[Windows] CMake preset llm-debug-vulkan selects MSVC instead of Clang, causing XNNPACK compilation failures Open

module: vulkan module: windows module: xnnpack triaged

pytorch/executorch#22084 · 1 comment · 1 reaction · 1 assignee · Opened Aug 24, 2026
Remove the need for symlinks when building ExecuTorch on Windows Open

module: windows triaged

pytorch/executorch#22081 · 1 comment · 1 assignee · Opened Aug 24, 2026
to_edge_transform_and_lower does not constant-fold parameter-only subgraphs, so weight_norm'd convolutions never reach the delegate (941x on one module) Open

module: exir module: xnnpack triaged

pytorch/executorch#22078 · 2 comments · 1 assignee · Opened Aug 24, 2026
macOS wheel: libquantized_ops_aot_lib.dylib never loads, so PT2E models fail later with "Missing out variants" Open

module: quantization triaged

pytorch/executorch#22071 · 2 comments · 2 assignees · Opened Aug 24, 2026
QNN-delegated static-cache decoder produces correct output via native C++ Module, but degenerate/repeating output via Android JNI Module (same .pte, same device) Open

module: android module: qnn triaged

pytorch/executorch#22061 · 5 comments · 3 reactions · 5 assignees · Opened Aug 23, 2026
8da4w silently skips layers that do not divide the group size, and 8da8w is implemented but rejected by config validation Open

module: llm module: quantization triaged

pytorch/executorch#22044 · 1 assignee · Opened Aug 22, 2026
no_rope_layer_interval is declared but never applied, and SmolLM3's own config sets it Open

module: llm triaged

pytorch/executorch#22043 · 2 comments · 2 assignees · Opened Aug 22, 2026
SDPA decomposition does not preserve intermediate dtype for fp16/bf16 inputs Open

pytorch/executorch#22012 · 4 comments · 1 assignee · Opened Aug 21, 2026
aten.bmm, aten.mul, aten._softmax doesn't get quantized/replaced with Cortex-M kernels in Transformer model Open

bug good first issue module: arm module: quantization triaged

pytorch/executorch#21943 · 9 comments · 1 reaction · 3 assignees · Opened Aug 19, 2026
[ET-VK][Adreno] Low-rate non-deterministic CNN output, dependent on GPU power state; clean on Mali Open

module: vulkan

pytorch/executorch#21939 · 1 comment · 1 assignee · Opened Aug 19, 2026
[ET-VK][Adreno] Non-deterministic conv2d output for c_out in {168,176,184,192}; clean on Mali Open

module: vulkan

pytorch/executorch#21938 · 2 comments · 1 assignee · Opened Aug 19, 2026
AtenToCortexMPass crashes with "unsupported param type, call_function" when aten.linear.default is in preserve_ops Open

bug module: arm module: microcontrollers partner: arm triaged

pytorch/executorch#21904 · 5 comments · 1 assignee · Opened Aug 18, 2026
TextLLMRunner degenerates on an LFM2.5 export that steps correctly token by token Open

module: llm

pytorch/executorch#21858 · 1 assignee · Opened Aug 14, 2026
`export_llm` Core ML builds fail at execute: "must be an MLState" even though `take_over_mutable_buffer=False` Open

bug module: coreml module: llm partner: apple

pytorch/executorch#21855 · 6 comments · 1 assignee · Opened Aug 14, 2026
Runtime accepts a channels-last input for a contiguous method and reads it as contiguous (silent wrong output) Open

bug high priority module: runtime triage review

pytorch/executorch#21837 · 3 comments · 1 assignee · Opened Aug 14, 2026
[Muse Glimmer][CUDA] SM80+PTX vision kernel exceeds RTX 5090 shared-memory limit; native SM120 export OOMs Open

bug module: cuda module: llm

pytorch/executorch#21827 · 3 comments · 1 assignee · Opened Aug 13, 2026
iOS swiftpm-1.4.0 prebuilts: HF tokenizers with lookahead regex abort the app (signal 6) — regex_lookahead still missing Open

bug module: ios partner: apple

pytorch/executorch#21805 · 1 comment · 1 assignee · Opened Aug 13, 2026
The Python bindings cannot run a program that uses non-CPU planned memory Open

bug module: cuda module: extension

Difficulty 4/5 3-5 days Newbie friendliness 72/100

pytorch/executorch#21682 · 1 comment · Opened Aug 7, 2026
Automate dependency updates with a scheduled Action Open

enhancement triaged

Difficulty 2/5 Half a day Newbie friendliness 68/100

pytorch/executorch#21640 · Opened Aug 7, 2026
coreml_static_int8 operators suite never finishes on macOS (hits the 180 min cap) Open

bug module: ci module: coreml

Difficulty 4/5 3-5 days Newbie friendliness 48/100

pytorch/executorch#21623 · Opened Aug 6, 2026
yolo26 does not export on the portable backend Open

bug module: examples module: kernels

Difficulty 4/5 3-5 days Newbie friendliness 58/100

pytorch/executorch#21621 · Opened Aug 6, 2026
CUDA delegate: a device-resident input is copied with a host memcpy, crashing the no-copy export mode Open

module: cuda

Difficulty 4/5 3-5 days Newbie friendliness 55/100

pytorch/executorch#21580 · 2 comments · Opened Aug 4, 2026
to_executorch segfaults when constant tensors are in device memory Open

bug module: cuda module: exir

pytorch/executorch#21545 · 3 comments · 1 assignee · Opened Aug 2, 2026

A short digest of beginner-friendly GitHub issues.