adefossez/julius
A PyTorch library providing differentiable DSP algorithms for audio and 1D signals with CUDA support.
Not currently ranked — collecting fresh signals.
star history
Julius implements common Digital Signal Processing operations including resampling, FFT convolutions, and filter banks, all in PyTorch so they remain differentiable and GPU-compatible. This enables signal processing operations to be incorporated directly into neural network training pipelines. The library supports TorchScript export and focuses on efficient computation for audio processing workflows.
Frequently asked
- What is adefossez/julius?
- A PyTorch library providing differentiable DSP algorithms for audio and 1D signals with CUDA support.
- Is julius open source?
- Yes — adefossez/julius is open source, released under the MIT license.
- What language is julius written in?
- adefossez/julius is primarily written in Python.
- How popular is julius?
- adefossez/julius has 458 stars on GitHub.
- Where can I find julius?
- adefossez/julius is on GitHub at https://github.com/adefossez/julius.