Quickstart#
Installation#
tsl is compatible with Python>=3.8. We recommend installation
on a Anaconda or Miniconda
environment or a virtual env.
Before installation
tsl is built upon PyTorch>=1.9 and
PyG>=2.0.3. Make sure you have
both installed in your environment before installing
tsl. In the following,
we provide instructions on how to install them for the chosen installation
procedure.
Installing using conda#
Tip
Using conda allows to automatically solve PyTorch and PyG dependencies, choosing the latest CUDA version available supported by the system.
To install tsl using conda, clone the repository, navigate to the library root
directory and create a new conda environment using the provided conda configuration:
git clone https://github.com/TorchSpatiotemporal/tsl.git
cd tsl
conda env create -f conda_env.yml
Then, activate the environment and install tsl using
pip
.
conda activate tsl
python setup.py install # Or 'pip install .'
Note
Installation of tsl directly from conda is on the roadmap!
Installing using pip#
Alternatively, you can install the library directly from pip
. Please
refer to PyTorch and PyG installation guidelines
for installation without conda. After having installed the libraries, install
torch-spatiotemporal
using pip. For the latest version:
pip install git+https://github.com/TorchSpatiotemporal/tsl.git
For the stable version:
pip install torch-spatiotemporal
Example scripts#
The github repository hosts example scripts and notebooks on how to use the library for different use cases, such as spatiotemporal predictions and imputations. You can refer to the notebook A Gentle Introduction to tsl for an introductory overview of the library main functionalities.
Citing#
If you use Torch Spatiotemporal for your research, please consider citing the library
@software{Cini_Torch_Spatiotemporal_2022,
author = {Cini, Andrea and Marisca, Ivan},
license = {MIT},
month = {3},
title = {{Torch Spatiotemporal}},
url = {https://github.com/TorchSpatiotemporal/tsl},
year = {2022}
}