A Python library to train neural network emulators of atmospheric transport models.
conda create -n neuraltransport -c conda-forge python=3.12
conda activate neuraltransport
conda install -c conda-forge ffmpeg pkg-config libjpeg-turbo opencv cupy
conda install -c conda-forge numpy pandas xesmf cdo python-cdo xarray dask zarr netCDF4 bottleneck matplotlib seaborn cartopy shapely xskillscore xrft pyarrow
pip3 install torch torchvision torchaudio
pip install torch_geometric
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.7.0+cu126.html
pip3 install lightning cdsapi pypdf2 trimesh rtree ipykernel ipywidgets tensorboard einops timm ecmwf-api-client eccodes dm-tree cfgrib pynvml wandb ruamel.yaml moviepy torch_harmonics tensorly tensorly-torch
pip3 install git+https://github.com/jbusecke/xmovie.git
# Go inside your neural_transport folder (cd neural_transport)
pip install -e .
In case you use NeuralTransport in your research or work, it would be highly appreciated if you include a reference to our paper in any kind of publication.
@article{benson2024neuraltransport,
title = {Atmospheric Transport Modeling of CO2 with Neural Networks},
author = {Vitus Benson, Ana Bastos, Christian Reimers, Alexander J. Winkler,
Fanny Yang and Markus Reichstein},
eprint={2408.11032},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2408.11032},
}For questions or comments regarding the usage of this repository, please use the discussion section on Github. For bug reports and feature requests, please open an issue on GitHub. In special cases, you can reach out to Vitus (find his email on his website).