This is a custom dataset for loading celebA dataset to be trained on google colab. Due to google drive limitations this could serve as a manual alternative.

customCelebA.py

# Download the data on your google drive by making a copy of the annotations 
# change the shareable link file/d/ into uc?id=/ 
# store those links in their respective url
# use the gdown command on the url to download straight to colab

import pandas as pd
import os
from skimage import io 
from torch.utils.data import DataLoader

identity=pd.read_csv('/content/identity_CelebA.txt',delim_whitespace=True, header=None)
bbox=pd.read_csv('/content/list_bbox_celeba.txt', delim_whitespace=True, header=1)
landmarks_align =pd.read_csv('/content/list_landmarks_align_celeba.txt',delim_whitespace=True, header=1)
attr=pd.read_csv('/content/list_attr_celeba.txt', delim_whitespace=True,header=1)

# Define the custom class

class celebADataset(Dataset):
  
    def __init__(self,root_dir,labels=None,target_type="identity", transforms=None):
        '''
         initilaize the important passed variables
        '''
        self.annotations = labels
        self.root_dir=root_dir
        self.transforms=transforms
        self.target_type=target_type
        
    def __len__(self):
        '''
        returns the no of dataset
        '''
        return len(self.annotations)
    
    def __getitem__(self, index):
        '''
        returns image in numpy with labels
        '''
        target=[]
        # due to different files being read differently some will raise error due to how it's stored
        if self.target_type == "identity":
            img_path = os.path.join(self.root_dir, str(self.annotations.iloc[index, 0]))
            image = io.imread(img_path)
            target.append(self.annotations.iloc[index, 1])
        elif self.target_type == "attr":
            img_path = os.path.join(self.root_dir, str(self.annotations.index[index]))
            image = io.imread(img_path)
            target.append(self.annotations.iloc[index,:].to_numpy())
        elif self.target_type== "bbox":
            img_path = os.path.join(self.root_dir, self.annotations.loc[:,'image_id'][index])
            image = io.imread(img_path)
            target.append(self.annotations[index,:].to_numpy())
        elif self.target_type == "landmarks":
            img_path = os.path.join(self.root_dir, self.annotations.index[index])
            image = io.imread(img_path)
            target.append(self.annotations.iloc[index,:].to_numpy())
        else:
            raise ValueError("Target type \"{}\" is not recognized.".format(target_type))

        if self.transforms:
          image = self.transforms(image)

        if target:
            target = tuple(target) if len(target) > 1 else target[0]

        else:
            target = None

        return (image,target)
      
# example of loading the dataset      
image_size = 64
root='/content/img_align_celeba'
transform = transforms.Compose(
    [transforms.ToPILImage(),
     transforms.Resize(image_size),
     transforms.CenterCrop(image_size),
     transforms.ToTensor(),
     transforms.Normalize((0.5,0.5,0.5 ), (0.5,0.5,0.5 ))
])

dataset = celebADataset(
    labels=identity,
    root_dir=root,
    transforms=transform
)