GPyTorch is a Gaussian process library implemented using PyTorch. GPyTorch is designed for creating scalable, flexible, and modular Gaussian process models with ease.
GPyTorch primarily works by creating an accurate kernel approximation (typically via SKI) which admits fast matrix vector multiplies (MVMs) for iterative pre-conditioned MVM-based scalable inference and learning.
GPyTorch provides (1) significant GPU acceleration (through MVM based inference); (2) state-of-the-art implementations of the latest algorithmic advances for scalability and flexibility (SKI/KISS-GP, stochastic Lanczos expansions, LOVE, SKIP, stochastic variational deep kernel learning, ...); (3) easy integration with deep learning frameworks.
This package is currently under development, and is likely to change. Some things you can do right now:
- Simple GP regression (example here)
- Simple GP classification (example here)
- Multitask GP regression (example here)
- Scalable GP regression using kernel interpolation (example here)
- Scalable GP classification using kernel interpolation (example here)
- Deep kernel learning (example here)
- And (more!)
If you use GPyTorch, please cite the following papers:
@inproceedings{gardner2018product,
title={Product Kernel Interpolation for Scalable Gaussian Processes},
author={Gardner, Jacob R and Pleiss, Geoff and Wu, Ruihan and Weinberger, Kilian Q and Wilson, Andrew Gordon},
booktitle={AISTATS},
year={2018}
}
@article{pleiss2018constant,
title={Constant-Time Predictive Distributions for Gaussian Processes},
author={Pleiss, Geoff and Gardner, Jacob R and Weinberger, Kilian Q and Wilson, Andrew Gordon},
journal={arXiv preprint arXiv:1803.06058},
year={2018}
}
The easiest way to install GPyTorch is by installing PyTorch >= 0.4.0
using the appropriate command from here, and then installing
GPyTorch using pip:
pip install git+https://github.com/cornellius-gp/gpytorch.git
To use packages globally but install GPyTorch as a user-only package, use pip install --user
above.
We also provide two conda environment files, environment.yml
and environment_cuda90.yml
. As an example, to install GPyTorch in a conda environment with cuda support, run:
git clone git+https://github.com/cornellius-gp/gpytorch.git
conda create -f gpytorch/environment_cuda.yml
source activate gpytorch
pip install gpytorch/
Still a work in progress. For now, please refer to the following example Jupyter notebooks.
To run the unit tests:
python -m unittest
By default, the random seeds are locked down for some of the tests. If you want to run the tests without locking down the seed, run
UNLOCK_SEED=true python -m unittest
Please lint the code with flake8
.
pip install flake8 # if not already installed
flake8
GPyTorch is developed at Cornell University by
- Jake Gardner (lead developer)
- Geoff Pleiss (lead developer)
- Kilian Weinberger
- Andrew Gordon Wilson
We would like to thank our other contributors including (but not limited to) Max Balandat, Ruihan Wu, Bram Wallace, Jared Frank.
Development of GPyTorch is supported by funding from the Bill and Melinda Gates Foundation.