Doron Adler

Had the same issue, so it happened that "all_files" was already sorted at this point, so simply commented out #all_files.sort. But I guess it's not something to count on for...

> Did you ever figure out a different solution besides just commenting out? I didn't try

@woshildh This repository contains an example of a paired-image training dataset under https://github.com/Norod/U-2-Net-StyleTransfer/tree/main/train_data folder. It's basically two folders with jpg files each pair has the same name

Heya, Make sure you have [git lfs ](https://git-lfs.github.com/) installed and [configured](https://stackoverflow.com/a/56284009/1495606), so the actual model files will be downloaded rather than just a reference pointer to their location Edit: If...

> Will it only work on faces? I did not try training it on other kinds of datasets. If you have a paired image dataset of something which isn't faces,...

Google changed something and it's breaking up for everyone, see here: https://github.com/hardikvasa/google-images-download/issues/280#issuecomment-582705697

Had the same issue. This happens on Windows. Solved by setting num_workers to 0 rather than 1 in train_lora_dreambooth.py ``` train_dataloader = torch.utils.data.DataLoader( train_dataset, batch_size=args.train_batch_size, shuffle=True, collate_fn=collate_fn, num_workers=0, ) ```

Thank you, I had the same problem and your PR fixes it well. Just note that you forgot to edit `pyproject.toml` to reflect the updated gradio and gradio_client versions. Thank...

> OK, I actually made it work by using the original stylegan2-ada repo with the following modfied `blend_models.py` script added to the main directory: > Indeed. Here is a Colab...

Thank you for letting us know. @justinpinkney fixed it.