Repository for reproducibility Challenge 2020
Jiang, C., Kashinath, K., Prabhat, & Marcus, P. (2020). Enforcing physical constraints in neural networks through diffenrentiable PDE layer. ICLR 2020 Workshop DeepDiffEq. OpenReview
The code base is mostly adapted from https://github.com/NVlabs/SPADE.
The model requires the Synchronized-BatchNorm-PyTorch, it is included as git submodule in folder model/networks/sync_batchnorm/
use
git clone --recurse-submodules https://github.com/zitkat/PDEL.git
to painlessly get it cloned along with the codebase itself. This code requires PyTorch>=1.7 and Python 3+, additionally data downloader for turbulence dataset requires pyJHTDB. You can install dependencies by
pip install -r requirements.txt