pytorch / pytorch/executorch

Arm backend: requirements-arm-tosa.txt is uninstallable on Python 3.13/3.14 although pyproject.toml declares support up to <3.15

Open
#22,314 0 comments 0 reactions 2 assignees View on GitHub

@usamahz is already working on this.

Since Sep 9, 2026.

bug dependencies module: arm module: build/install partner: arm
Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

🐛 Describe the bug

Environment

  • macOS 26.5 (Apple Silicon), Python 3.14.6
  • executorch main @ d614f9dba4

Repro

Install executorch on Python 3.14 (allowed by pyproject.toml: requires-python = ">=3.10,<3.15" — the install works), then run the Arm setup:

./examples/arm/setup.sh --i-agree-to-the-contained-eula

It fails installing backends/arm/requirements-arm-tosa.txt:

ERROR: Ignored the following versions that require a different python version:
  0.0.1 Requires-Python >=3.10,<3.13; 0.1.0 Requires-Python >=3.10,<3.13
ERROR: Could not find a version that satisfies the requirement tosa-adapter-model-explorer==0.1.0 (from versions: none)

Root cause

tosa-adapter-model-explorer publishes only for Python <3.13, but it is pinned unconditionally in backends/arm/requirements-arm-tosa.txt — whose own header comment says "These dependencies need to match pyproject.toml". The two version envelopes contradict each other, so the effective supported range for the Arm flow is <3.13, which users only discover mid-setup.

Suggested fix

One of:

  1. Publish tosa-adapter-model-explorer wheels for 3.13/3.14 (it's an Arm-maintained package), or
  2. Make it an environment-conditional dependency (; python_version < "3.13") — it's a Model Explorer visualization adapter, not needed for the core AOT flow, or
  3. Lower requires-python (or document the Arm backend's narrower envelope) so the failure happens at install time with a clear message instead of mid-setup.

Everything worked after rebuilding the venv on Python 3.12.

Versions

PyTorch version: 2.13.0
Is debug build: False
CUDA used to build PyTorch: None
ROCM used to build PyTorch: N/A

OS: macOS 26.5.2 (arm64)
GCC version: Could not collect
Clang version: 21.0.0 (clang-2100.1.1.101)
CMake version: version 3.31.10
Libc version: N/A

Python version: 3.12.14 (main, Aug 12 2026, 13:57:54) [Clang 21.0.0 (clang-2100.1.1.101)] (64-bit runtime)
Python platform: macOS-26.5.2-arm64-arm-64bit
Is CUDA available: False
CUDA runtime version: No CUDA
CUDA_MODULE_LOADING set to: N/A
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA
Is XPU available: False
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: False
Caching allocator config: N/A

CPU:
Apple M4 Pro

Versions of relevant libraries:
[pip3] executorch==1.5.0+d614f9d
[pip3] numpy==2.5.2
[pip3] pytorch_tokenizers==1.3.0
[pip3] torch==2.13.0
[pip3] torchao==0.18.0.dev20260729+cpu
[pip3] torchaudio==2.11.0
[pip3] torchdata==0.11.0+cpu
[pip3] torchsr==1.0.4
[pip3] torchtune==0.0.0
[pip3] torchvision==0.28.0
[conda] Could not collect

cc @larryliu0820 @GregoryComer @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.