RS-Coreset

Requirement

python3+
pytorch
numpy
sklearn
matplotlib
rdkit

You can also use our requirements.txt to install

pip install -r requirements.txt

Usage

For your own dataset

1. Prepare an xlsx file of reaction data in SMILES format according to the given style and convert the reaction space into a binary npy file with provided script in utils.
2. Or prepare your own molecular descriptor npz file with data store as train_data and yield(0-100) store as train_label
3. Modify the arguments in class Arguments of main.py
4. run main.py and the prediction result will be represent as result.csv

For our real world dataset

Download and unzip reaction93.zip in current dir from https://drive.google.com/file/d/1O_Qcn5Z2gwr5e93Uh694d7d5HK3spkJo/view?usp=drive_link

python yield_predict_real.py

For public HTE dataset

We prepare BH.py, SM.py and BH_Plus.py for performance test on these three public HTE dataset
Download and unzip datasets.zip in current dir from https://drive.google.com/drive/folders/1Dioh_fcPyMbhNNtEmNyUE4TrzxMnXgkV?usp=sharing

python BH.py
python SM.py
python BH_Plus.py