utils/data_loader.py

import pandas as pd
import torch
from torch.utils.data import Dataset, DataLoader
from torch.nn.utils.rnn import pad_sequence
from esm_utils import get_latents, load_esm2_model
import config

class ProteinDataset(Dataset):
    def __init__(self, csv_file, tokenizer, model):
        self.data = pd.read_csv(csv_file).head(4)
        self.tokenizer = tokenizer
        self.model = model

    def __len__(self):
        return len(self.data)

    def __getitem__(self, idx):
        sequence = self.data.iloc[idx]['Sequence']
        latents = get_latents(self.model, self.tokenizer, sequence)

        attention_mask = torch.ones_like(latents)
        attention_mask = torch.mean(attention_mask, dim=-1)

        return latents, attention_mask

def collate_fn(batch):
    latents, attention_mask = zip(*batch)
    latents_padded = pad_sequence([torch.tensor(latent) for latent in latents], batch_first=True, padding_value=0)
    attention_mask_padded = pad_sequence([torch.tensor(mask) for mask in attention_mask], batch_first=True, padding_value=0)
    return latents_padded, attention_mask_padded

def get_dataloaders(config):
    tokenizer, masked_model, embedding_model = load_esm2_model(config.MODEL_NAME)
    
    train_dataset = ProteinDataset(config.Loader.DATA_PATH + "/train.csv", tokenizer, embedding_model)
    val_dataset = ProteinDataset(config.Loader.DATA_PATH + "/val.csv", tokenizer, embedding_model)
    test_dataset = ProteinDataset(config.Loader.DATA_PATH + "/test.csv", tokenizer, embedding_model)
    
    train_loader = DataLoader(train_dataset, batch_size=config.Loader.BATCH_SIZE, num_workers=0, shuffle=True, collate_fn=collate_fn)
    val_loader = DataLoader(val_dataset, batch_size=config.Loader.BATCH_SIZE, num_workers=0, shuffle=False, collate_fn=collate_fn)
    test_loader = DataLoader(test_dataset, batch_size=config.Loader.BATCH_SIZE, num_workers=0, shuffle=False, collate_fn=collate_fn)
    
    return train_loader, val_loader, test_loader