This code is a subset of the original TNDM repo, which can be found here: https://github.com/HennigLab/tndm. The code is MIT licensed.

Training and sample generation is performed in tndm_reach.ipynb. To sample unconditional samples, prior variance is set to 1, and then samples are drawn from the Gaussian prior before passed through the RNNs and decoder.

To run this, we require tensorflow 2.x. Do this by creating a conda environment with tensorflow 2.x.

conda create -n tndm python=3.9
conda activate tndm
pip install -e .