PFMBench/zeroshot/data_interface.py

from torch.utils.data import DataLoader
import torch
from torch.utils.data import DataLoader, DistributedSampler
from src.interface.data_interface import DInterface_base
from src.model.pretrain_model_interface import PretrainModelInterface
from src.data.proteingym_dataset import ProteinGYMDataset
from src.data.msa_dataset import MSADataset

class DInterface(DInterface_base):
    def __init__(self,  **kwargs):
        super().__init__(**kwargs)
        self.save_hyperparameters()

    def setup(self, stage=None):
        pass
    
    def data_setup(self, type="proteingym"):
        if type == "proteingym":
            self.mut_dataset = ProteinGYMDataset(
                dms_csv_dir = self.hparams.dms_csv_dir,
                dms_pdb_dir = self.hparams.dms_pdb_dir,
                dms_reference_csv_path = self.hparams.dms_reference_csv_path,
            )
        elif type == "msa":
            self.msa_dataset = MSADataset(
                msa_csv_path = msa_csv_path
            )

    def train_dataloader(self):
        return DataLoader(self.mut_dataset, batch_size=1, shuffle=True, num_workers=self.hparams.num_workers, pin_memory=True)

    def val_dataloader(self):
        return DataLoader(self.mut_dataset, batch_size=1, shuffle=False, num_workers=self.hparams.num_workers, pin_memory=True)

    def test_dataloader(self):
        return DataLoader(self.mut_dataset, batch_size=1, shuffle=False, num_workers=self.hparams.num_workers, pin_memory=True)