Torch Utilities is a collection of tools and utilities for working with PyTorch in the audio domain.
Torch Utilities is developed as a personal set of tools and is provided as-it-is, without any warranties or guarantees. The API and utilities may change in the future as the module continues to evolve. While effort is made to maintain compatibility with previous versions, users are advised to carefully consider the stability of the API before using Torch Utilities in production environments.
You can install Torch Utilities using pip.
pip install torch_utilitiesTo run the tests you need to clone the repository locally and run use pytest.
git clone git@github.com:FedericoDiMarzo/torch_utilities.git
pip install -e torch_utilities[dev]
pytest torch_utilities/testsIf any tests fail, it may indicate that there is a bug in the code or that some aspect of the API has changed. In such cases, we encourage you to open an issue on the repository so that we can help resolve the problem.
To read the API documentation of the module after cloning the repository, you can use pdoc to generate the documentation and serve it locally.
pdoc --docformat numpy torch_utilities/torch_utilitiesThe documentation will then be accessible at the address http://localhost:8080 .
The most relevant function and classes are available in the main namespace of the module. You can import them directly and use them in your code.
The source code is divided into
audio: Utilities working mostly on waveformsaugmentation: Data augmentation for audio signalsio: Input/output utilitiesmetrics: Various metrics for audio signalsmodules: PyTorch modules for audio processingutilities: General utilities.