PyTorch Inc Backend

Welcome to the PyTorch Webgpu Backend repository. This project provides custom webgpu computing backend support for PyTorch.

Table of Contents

• Installation
• Usage

Installation

To install the necessary dependencies, install vcpkg first.

git clone https://github.com/microsoft/vcpkg.git
cd vcpkg

./bootstrap-vcpkg.sh  # Linux/macOS
.\bootstrap-vcpkg.bat  # Windows

Add the install directory to VCPKG_ROOT

Then create a virtual environment (At the time of writing, pytorch doesn't yet support python >3.13. Therefore it is better to use something less that 3.13):

python3.12 -m venv venv
source ./venv/bin/activate

Install the dependencies for both c++ and python:

vcpkg install
pip install -r requirements.txt

Usage

The setup.py by default points to the default vcpkg install folder: vcpkg_installed for include and lib folders. The triplet is detected from within setup.py.

To build and install the extension:

python setup.py develop

Run the test.py file to see a simple demo.

python test.py

Potential Errors and fixes:

• ImportError: dlopen: symbol not found in flat namespace - Linker error: Check if any new files that are added in c++ are included in the build.