Implementation for paper Co-modeling the Sequential and Graphical Routes for Peptide Representation Learning
To train a RepCon model on the example dataset AP, please use the following command:
python methods/co-modeling\ contrastive/main.py --dataset AP
Make sure the hyperparameter args.mode is set as 'train' before a trained model has been stored.
The predictive results can be found in the 'results' folder in the root directory.
args.seq_lr the learning rate of the sequential encoder & predictor.
args.graph_lr the learning rate of the graphical encoder & predictor.
args.nce_weight the weight which balance the supervised loss and the contrastive loss.
The other hyperparameters are relatively insensitive to downstream tasks, and the user can keep the default settings.