ussoewwin/torchaudio-built-on-cu132

🎵 torchaudio for Windows/Linux (Built on over PyTorch 2.12.0 + cu132)

This repository archives and distributes a custom build (wheel file) of torchaudio specifically compiled for over PyTorch 2.12.0 + CUDA 13.2.

📌 Background: Why this build?

As the machine learning ecosystem evolves rapidly, the provision of official pre-compiled binaries and support for specific version combinations—especially for Windows—is gradually being reduced or discontinued.

Once official build support ends for a specific environment, users are forced to manually configure C++ compilers and the CUDA Toolkit to build torchaudio from source. This process is complex, time-consuming, and presents a significant hurdle for many developers.

This project was created to solve a specific problem: "I want to use PyTorch 2.12.0 with the CUDA 13.2 environment on Windows, but the official binaries are no longer available."

By providing a pre-compiled, independently tested binary (.whl), this repository allows you to skip the tedious build process and jump straight into your audio processing projects.

💻 Requirements

This build is compiled to work optimally in the following environment:

• OS: Windows 10 / Windows 11 / Linux
• PyTorch: over 2.12.0+cu132
• CUDA Toolkit: 13.2
• Python: 3.12-3.14

🚀 Installation

1. Download the .whl file that matches your Python version from the Files and versions tab in this repository.
2. Open Command Prompt or PowerShell, navigate to the directory where you downloaded the file, and run the following pip command:

Note: Please replace the filename below with the actual name of the file you downloaded.