src/dataset.py

import torch
from torch.utils.data import DataLoader
from torchvision import datasets,transforms
from src.path import DATA_DIR

def get_dataloader(config):
    batch_size = config.get('batch_size',128)
    data_path = config.get('data_path',DATA_DIR)
    num_workers = config.get('num_workers',8)

    train_transform = transforms.Compose([
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
    ])
    val_transform = transforms.Compose([
        transforms.ToTensor(),
    ])

    print("📂Loading CIFAR100 dataset...")

    train_data = datasets.CIFAR100(root=data_path,train=True,download=True,transform=train_transform)
    test_data = datasets.CIFAR100(root=data_path,train=False,download=True,transform=val_transform)

    train_loader = DataLoader(
        train_data,
        batch_size=batch_size,
        num_workers=num_workers,
        shuffle=True,
        pin_memory=True,
        persistent_workers=True,
    )

    test_loader = DataLoader(
        test_data,
        batch_size=batch_size,
        num_workers=num_workers,
        shuffle=False,
        pin_memory=True,
        persistent_workers=True,
    )

    return train_loader,test_loader