Hrishi15/

Mulit-model_Contrastive_learning

The models trained during the project are provided

Datasets - contains all the data used during the course of the project

For each model excpet for GNN pairwise and GNN Text graph concat, Contrastive model is saved as "contrastive_model" in "full_model_checkpoint.pt" Space Mapper model is saved as "model_state_dict" in "space_mapper_zinc250k_final.pt"

For GNN pairwise and GNN text graph concat, Contrastive model and space mapper can be obtained from respective .pt files thorugh "contrastive_model" and "space_mapper" respectively

bartsmiles folder contains the finetuned smiles tokenizer, and standard autoencoder bartsmiles-checkpoint-5700 contains the finetuned smiles model, noisy decoder

Code for the project can be obtained at https://github.com/HrishikeshA15/Multi-Modal-Contrastive-Learning-for-Text-Guided-Molecule-Optimization