Building GNN prediction model for chemical descriptors to facailitate IFLP design

The trained models can be found in "model" folder.

Installation

conda env create -f ai4chem.yml

Usages

To view all configurations, run the following command:

python main_iflp.py --help

Train the GNN model from scratch

python main_iflp.py mode=predictive

Fine-tune from a pre-trained GNN model

python main_iflp.py mode=ft_predictive paths.model_path=[pretrained_model_path]

Predict the chemical descriptors

python eval_iflp.py --p [pretrained_model_path] --i [input_file_path] --t FEPA FEHA

Analyze the trained model

python utilis/model_analysis.py --d [pretrained_model_path] --t FEPA FEHA