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 open issues indexed

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.