.. image:: https://autrainer.github.io/autrainer/_static/logo_banner.png :alt: autrainer — A Modular and Extensible Deep Learning Toolkit for Computer Audition Tasks :align: center

autrainer

|pypi| |python_versions| |hugging_face| |license| |preprint|

A Modular and Extensible Deep Learning Toolkit for Computer Audition Tasks.

autrainer is built on top of PyTorch <https://pytorch.org/>_ and Hydra <https://hydra.cc/>_, offering a modular and extensible way to perform reproducible deep learning experiments for computer audition tasks using YAML configuration files and the command line.

.. _installation:

Installation

It is recommended to install autrainer within a virtual environment. To create a new virtual environment, refer to the Python venv documentation <https://docs.python.org/3/library/venv.html>_.

Next, install autrainer using pip.

.. code-block:: pip

pip install autrainer

.. note::

If you are using Windows, make sure that PyTorch (along with torchvision and torchaudio) version 2.0 or higher is installed beforehand. For installation instructions, refer to the PyTorch documentation <https://pytorch.org/get-started/locally/>_.

The following optional dependencies can be installed to enable additional features:

• latex for LaTeX plotting (requires a LaTeX installation).
• mlflow for MLflow <https://mlflow.org/>_ logging.
• tensorboard for TensorBoard <https://www.tensorflow.org/tensorboard>_ logging.
• opensmile for audio feature extraction with openSMILE <https://audeering.com/opensmile/>_.
• albumentations for image augmentations with Albumentations <https://albumentations.ai/>_.
• audiomentations for audio augmentations with audiomentations <https://github.com/iver56/audiomentations>_.
• torch-audiomentations for audio augmentations with torch-audiomentations <https://github.com/asteroid-team/torch-audiomentations>_.

.. code-block:: pip

pip install autrainer[latex] pip install autrainer[mlflow] pip install autrainer[tensorboard] pip install autrainer[opensmile] pip install autrainer[albumentations] pip install autrainer[audiomentations] pip install autrainer[torch-audiomentations]

To install autrainer with all optional dependencies, use the following command:

.. code-block:: pip

pip install autrainer[all]

To install autrainer from source, refer to the contribution guide <https://autrainer.github.io/autrainer/development/contributing>_.

Next Steps

To get started using autrainer, the quickstart guide <https://autrainer.github.io/autrainer/usage/quickstart>_ outlines the creation of a simple training configuration and tutorials <https://autrainer.github.io/autrainer/usage/tutorials>_ provide examples for implementing custom modules including their configurations.

For a complete list of available CLI commands, refer to the CLI reference <https://autrainer.github.io/autrainer/usage/cli_reference>_ or the CLI wrapper <https://autrainer.github.io/autrainer/usage/cli_wrapper>_.

.. |pypi| image:: https://img.shields.io/pypi/v/autrainer?logo=pypi&logoColor=b4befe&color=b4befe :target: https://pypi.org/project/autrainer/ :alt: autrainer PyPI Version

.. |python_versions| image:: https://img.shields.io/pypi/pyversions/autrainer?logo=python&logoColor=b4befe&color=b4befe :target: https://pypi.org/project/autrainer/ :alt: autrainer Python Versions

.. |hugging_face| image:: https://img.shields.io/badge/Hugging_Face-autrainer-b4befe?logo=huggingface&logoColor=b4befe :target: https://huggingface.co/autrainer :alt: autrainer Hugging Face

.. |license| image:: https://img.shields.io/badge/license-MIT-b4befe?logo=c :target: https://github.com/autrainer/autrainer/blob/main/LICENSE :alt: autrainer GitHub License

.. |preprint| image:: https://img.shields.io/badge/arXiv-2412.11943-AD1C18?logoColor=b4befe&color=b4befe :target: https://arxiv.org/abs/2412.11943 :alt: autrainer arXiv preprint