Time series forecasting with PyTorch
Project description
TorchTS is a PyTorch-based library for time series data.
Currently under active development!
Why Time Series?
Time series data modeling has broad significance in public health, finance and engineering. Traditional time series methods from statistics often rely on strong modeling assumptions, or are computationally expensive. Given the rise of large-scale sensing data and significant advances in deep learning, the goal of the project is to develop an efficient and user-friendly deep learning library that would benefit the entire research community and beyond.
Why TorchTS?
Existing time series analysis libraries include statsmodels and sktime. However, these libraries only include traditional statistics tools such as ARMA or ARIMA, which do not have the state-of-the-art forecasting tools based on deep learning. GluonTS is an open-source time series library developed by Amazon AWS, but is based on MXNet. Pyro is a probabilistic programming framework based on PyTorch, but is not focused on time series forecasting.
Installation
Installation Requirements
TorchTS supports Python 3.7+ and has the following dependencies:
Installing the latest release
The latest release of TorchTS is easily installed either via pip
:
pip install torchts
or via conda from the conda-forge channel (coming soon):
conda install torchts -c conda-forge
You can customize your PyTorch installation (i.e. CUDA version, CPU only option) by following the PyTorch installation instructions.
Important note for MacOS users:
- Make sure your PyTorch build is linked against MKL (the non-optimized version of TorchTS can be up to an order of magnitude slower in some settings). Setting this up manually on MacOS can be tricky - to ensure this works properly, please follow the PyTorch installation instructions.
- If you need CUDA on MacOS, you will need to build PyTorch from source. Please consult the PyTorch installation instructions above.
Getting Started
Check out our documentation and tutorials (coming soon).
Citing TorchTS
If you use TorchTS, please cite the following paper (coming soon):
@inproceedings{TBD,
title={{TorchTS: A Framework for Efficient Time Series Modeling}},
author={TBD},
booktitle = {TBD},
year={TBD},
url = {TBD}
}
See here (coming soon) for a selection of peer-reviewed papers that either build off of TorchTS or were integrated into TorchTS.
Contributing
Interested in contributing to TorchTS? Please see the contributing guide to learn how to help out.
License
TorchTS is MIT licensed.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.