First make sure you have torch
and rdkit
installed.
Then run
pip install -e .
for development mode installation.
First we download preprocessed data:
cd Data/NMR/
wget https://homes.esat.kuleuven.be/~jsimm/nmr_data/folding0.npy
wget https://homes.esat.kuleuven.be/~jsimm/nmr_data/folding1.npy
wget https://homes.esat.kuleuven.be/~jsimm/nmr_data/folding2.npy
wget https://homes.esat.kuleuven.be/~jsimm/nmr_data/nmr_webo.npy
Then train a GraphInformer model on the NMR shift data for 100 epochs using 3-distance attention heads:
python train.py --data_file nmr_webo.npy --folding_file folding0.npy \
--head_radius 3 3 3 3 3 3 \
--layers 2 \
--hidden_size 256 \
--hidden_dropout 0.1 \
--num_epochs 100 \
--lr_steps 50 80