Processed dataset for the notebook demo

[masud99r]

Thank you for the great work!
I am trying to run the notebooks demo notebooks/Demo_UNet_CelebA80x80.ipynb. However, this demo seems to require a processed image dataset "train80x80_no_repeats.pt". Is there any script available to generate them or do you have the processed image that can be downloaded?

[Zahra-Kadkhodaie]

I just added a script, code/celebA_to_torch.py , to pre-process celebA images to 80x80 images we used in our work. To use this you will need to download the celebA dataset, which contains around 202K images. The script then crops and downsamples the images and removes the repeated ones.