Handy PyTorch models and utilities which should save you time—and heartache.
Please read the documentation.
This Python library contains a collection of neural networks, implemented in PyTorch, and a generic DataSet class with useful features to help you get started with your deep-learning problem more quickly. Hopefully, the user will spend less time engineering and testing—because researchers always test —their code. There is also the hope that you will write and duplicate less code by being able to easily set and modify these models' architectures at instantiation.
This repository is for anyone intested in using the tools—deep learning or otherwise—it contains. Please cite the repository if you use it in any work you publish. Please also contact me, by creating an issue on GitHub, with details of the publication so I can maintain a list of publications in this README.
This project was created by J. Denholm.
J.D. kindly thanks:
for their help and time in reviewing demos for this repo.
If you would like to contribute to this repository, or request something—such as a particular model—be added to it, simply create an issue on GitHub and I'll be glad to discuss collaborating or adding the model.
Distributed under the MIT License. See LICENSE.md for more information.
To install the most up-to-date version, use
pip install git+https://github.com/jdenholm/TorchTools.gitTo install a specific version, say v0.1.0, use
pip install git+https://github.com/jdenholm/[email protected]To run the tests (with the dev env):
pytest tests/There are demos for each of the models in the demos/ directory. Note: the demos require a different python environment, which is specified in requirements-dev.conda.yaml. To build it, use:
cd /path/to/TorchTools/
conda env create -f requirements-dev.conda.yaml