- uv: running scripts; managing virtual environments
- ffmpeg: optional video generation during evaluation
If you'd just like to use the code in autocast:
# Clone the repo
git clone https://github.com/alan-turing-institute/autocast.git
cd autocast
# Install dependencies
uv sync
This will allow you to run uv run autocast from within the repository.
The main documentation contains a tutorial on how to use autocast to train and evaluate models on a dataset of your choice.
If you want to contribute to the autocast codebase, please see the contributing section of the docs.
See ETHICAL_GUIDANCE.md for guidance on the intended scope and use of AutoCast.
@misc{greenbury-etal-2026-reliability,
title={Reliability of Probabilistic Emulation of Physical Systems},
author={Sam F. Greenbury and Radka Jersakova and Paolo Conti and Marjan Famili and Christopher Iliffe Sprague and Edwin Brown and Jason D. McEwen},
year={2026},
eprint={2606.12997},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2606.12997},
}Thanks goes to these wonderful people (emoji key):
This project follows the all-contributors specification. Contributions of any kind welcome!
