wxx07 / miDruglikeness

miDruglikeness

The webserver is available at miDruglikeness

Requirements

• Python 3.6.13
• rdkit
• chemprop 0.0.2
• alipy 1.2.5
• torch 1.8.0

Usage

The whole frame work supports train, prediction,al_ensemble

Training

To train a model, run:

python train.py --data_path <train_data> --separate_test_path <test_data> --save_dir <model_path> --s <output> --mode <training mode> 

where <train_data> is the path to a CSV file containing training data, <test_data> is the path to a CSV file containing test_data, <model_path> is the directory where trained model will be saved, is the directory where results will be saved, is one of "normal", "active" or "passive" for normal training, active learning, or passive learning.

For example:

python train.py --data_path ../datasets/market_approvability_train.csv --separate_test_path ../datasets/market_approvability_test.csv --dataset_type classification --save_dir ../pipeline/market-approvability_test --epochs 50  --s ../results/test_s --mode normal --target_columns label

Prediction

python predict.py --separate_test_path <test_data> --save_dir <model_path> --outputfile <outputfile> 

where <test_data> is the path to a CSV file containing test data, <model_path> is the directory where trained model is saved, and is the path for prediction results.

For example:

python predict.py --data_path ../datasets/market_approvability_train.csv --separate_test_path ../datasets/market_approvability_test.csv --dataset_type classification --save_dir ../pipeline/market-approvability --s ../results/predict_s --mode normal --target_columns label --outputfile output

Al ensemble

python al_ensemble.py --data_path <train_data> --separate_test_path <test_data> --save_dir <model_path> --s <output> 

where <train_data> is the path to a CSV file containing training data, <test_data> is the path to a CSV file containing test_data, <model_path> is the directory where trained model will be ensembled, is the directory where results will be saved.

For example:

python al_ensemble.py --data_path ../datasets/market_approvability_train.csv --separate_test_path ../datasets/market_approvability_test.csv --dataset_type classification --save_dir ../pipeline/market-approvability --s ../results/al_ensemble_s --mode active --start_iter 11 --end_iter 16