Processed training dataset for finetuning pocket-based molecular generation task in Token-Mol 1.0

Processed CrossDocked2020 dataset for training pocket-based molecular generation task in Token-Mol 1.0. 

`protein_represent.pkl` is the representation embedding of protein pocket encoded with ResGen encoder.

`mol_input.pkl` is the corresponding ligands to the pocket in the training set presented as SMILES strings, which haved been tokenized.