torchfx 0.3.0

A GPU accelerated library for audio DSP based on PyTorch

TorchFX

GPU-Accelerated Audio DSP with PyTorch

Documentation | Getting Started | API Reference | Blog

---

TorchFX is a modern Python library for high-performance digital signal processing in audio, leveraging PyTorch and GPU acceleration. Built for researchers, engineers, and developers who need fast, flexible, and differentiable audio processing.

✨ Key Features

• ⚡ GPU Acceleration - Built on PyTorch for high-performance audio processing on CUDA-enabled devices
• 🔗 Composable Pipelines - Chain filters with the pipe operator (|) for sequential processing
• ➕ Parallel Processing - Combine filters with the add operator (+) for parallel filter combination
• 🧠 PyTorch Native - All filters are torch.nn.Module subclasses, enabling gradient-based optimization
• 🎯 Simple & Intuitive - Clean, object-oriented API designed for ease of use
• ⚙️ Highly Extensible - Create custom filters and effects by extending base classes
• 📊 Performance Optimized - Substantial performance gains over SciPy for long and multichannel signals

🚀 Quick Start

Installation

pip install torchfx

Or install from source:

git clone https://github.com/matteospanio/torchfx
cd torchfx
pip install -e .

Basic Example

import torch
from torchfx import Wave
from torchfx.filter import LoButterworth, ParametricEQ

# Load audio
wave = Wave.from_file("audio.wav")

# Create filters
lowpass = LoButterworth(cutoff=5000, order=4, fs=wave.fs)
eq = ParametricEQ(frequency=1000, q=2.0, gain=3.0, fs=wave.fs)

# Sequential processing with pipe operator (|)
processed = wave | lowpass | eq

# Parallel processing with add operator (+)
stereo_enhancer = lowpass + eq
enhanced = wave | stereo_enhancer

# Save result
processed.save("output.wav")

📚 Documentation

• Full Documentation - Complete guides and API reference
• Getting Started - Installation and first steps
• Tutorials - Practical examples and use cases
• API Reference - Detailed API documentation
• Blog - Updates, releases, and insights

🛠️ Development

• Contributing Guide - How to contribute to TorchFX
• Style Guide - Coding standards and conventions
• Roadmap - Future plans and priorities

We welcome contributions from everyone! Please read our Contributing Guide to get started.

📖 Citation

If you use TorchFX in your research, please cite our paper:

@misc{spanio2025torchfxmodernapproachaudio,
  title={TorchFX: A modern approach to Audio DSP with PyTorch and GPU acceleration},
  author={Matteo Spanio and Antonio Rodà},
  year={2025},
  eprint={2504.08624},
  archivePrefix={arXiv},
  primaryClass={eess.AS},
  url={https://arxiv.org/abs/2504.08624},
}

License

This project is licensed under the terms of the GNU General Public License v3.0. See the LICENSE file for details.

Third-Party Acknowledgments

This project uses the following third-party libraries:

• PyTorch – BSD-style license
• Torchaudio – BSD 2-Clause License
• NumPy – BSD 3-Clause License
• SciPy – BSD 3-Clause License
• SoundFile – BSD 3-Clause License

Their respective license texts are included in the licenses/ directory.