https://github.com/morton-li/pytorch-macos-builder
The project is dedicated to extending the PyTorch support cycle for Intel-based macOS users by providing x86-compatible builds of the official PyTorch releases. 项目致力于为 Intel 架构 macOS 用户延续 PyTorch 支持周期，提供官方原版 PyTorch 的 x86 兼容构建。
https://github.com/morton-li/pytorch-macos-builder
macos pytorch x86
Last synced: 4 months ago
JSON representation
The project is dedicated to extending the PyTorch support cycle for Intel-based macOS users by providing x86-compatible builds of the official PyTorch releases. 项目致力于为 Intel 架构 macOS 用户延续 PyTorch 支持周期，提供官方原版 PyTorch 的 x86 兼容构建。
- Host: GitHub
- URL: https://github.com/morton-li/pytorch-macos-builder
- Owner: Morton-Li
- Created: 2025-05-15T03:35:07.000Z (about 1 year ago)
- Default Branch: main
- Last Pushed: 2026-01-22T06:07:57.000Z (5 months ago)
- Last Synced: 2026-01-22T15:35:58.803Z (5 months ago)
- Topics: macos, pytorch, x86
- Homepage:
- Size: 29.3 KB
- Stars: 7
- Watchers: 1
- Forks: 3
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# PyTorch macOS Builder
> 🔧 自动构建 PyTorch 官方发行版本的 macOS 平台兼容 `.whl` 安装包。
> 🔧 Automatically build official PyTorch release `.whl` packages for macOS
---
## 📦 项目简介 Project Introduction
[本项目](https://github.com/Morton-Li/PyTorch-MacOS-Builder)通过 GitHub Actions 定期获取 [PyTorch 官方仓库](https://github.com/pytorch/pytorch) 的最新稳定版本，并自动构建适用于 **macOS** 的 Python wheel 安装包。
构建产物为**多 Python 版本**的 `.whl` 文件，便于在老款 Mac 上继续使用高版本 PyTorch。
[This project](https://github.com/Morton-Li/PyTorch-MacOS-Builder) uses GitHub Actions to periodically fetch the latest stable release from the [official PyTorch repository](https://github.com/pytorch/pytorch), and automatically builds Python wheel packages for **macOS**.
The output includes `.whl` files for **multiple Python versions**, allowing users to continue using newer versions of PyTorch on older Mac machines.
---
## 🛠 使用方式 How to Use
1. 从 [Releases 页面](../../releases) 下载你所需版本的 `.whl`：
示例文件名：`torch-2.10.0-cp313-cp313-macosx_11_0_x86_64.whl`
2. 使用 `pip` 安装：
```bash
pip install torch-2.10.0-cp313-cp313-macosx_11_0_x86_64.whl
```
3. 验证是否安装成功：
```bash
python -c "import torch; print(torch.__version__)"
```
---
## 💡 为什么需要这个项目 Why This Project Matters
由于 PyTorch 从 2.3 版本开始不再支持 Intel 架构的 macOS，而部分旧款 Mac 用户仍需使用较新 PyTorch，本项目旨在填补官方构建缺失，**延续 PyTorch 对 Intel Mac 的支持周期**。
Starting from version 2.3, PyTorch has dropped support for Intel-based macOS. This project aims to fill the gap by **extending PyTorch support for Intel Macs**, helping users continue to benefit from newer versions.
---
## ⚙️ 构建特性 Build Features
本项目构建的 PyTorch 安装包具有以下特性：
* ✅ 使用 **Intel MKL** 作为 BLAS 后端，提供更高的数值计算性能
* ✅ 启用 **OpenMP** 并行支持，充分利用多核 CPU 加速张量运算
* ✅ 启用 **oneDNN** 支持，自动加速部分 CNN、线性层等算子，提升 CPU 上的推理性能
* ✅ 启用 **Metal Performance Shaders (MPS)** 后端，在 macOS 上提供 GPU 加速能力，通过 Metal 后端优化张量运算效率
The PyTorch wheels built by this project have the following features:
* ✅ **Intel MKL** as the BLAS backend for high-performance numerical operations
* ✅ **OpenMP** support enabled to accelerate tensor computations on multi-core CPUs
* ✅ **oneDNN** support enabled to accelerate CPU inference for convolution and linear layers
* ✅ **Metal Performance Shaders (MPS)** backend enabled to provide GPU acceleration on macOS, leveraging the Metal framework to optimize tensor computation performance
---
## 🤝 鸣谢 Acknowledgements
- [PyTorch](https://github.com/pytorch/pytorch)
- [GitHub Actions](https://github.com/features/actions)