ProductGen !

This is the official repo for Warhol.

Setting up the environment !

Follows the instructions in install_warhol.sh

How to run the notebook ?

Run it with the Warhol venv that has just been created.

How to launch a training

python train_warhol.py --image_text_folder path_to_folder --vqgan_model_path path_vqgan_model.ckpt --vqgan_config_path path_to_vqgan_config.yaml --taming --keep_n_checkpoints 4 --truncate_captions --use_of_clip_embed both

What about the dataset ?

You need a dataset with files

• 0000.jpg
• 0000.txt
• 0000.npy (with both img & txt embeddings)

What about clip embeddings ?

If you don't have them, use --inferring_clip_embeddings or run the file infer_clip_embeddings: python infer_clip_embeddings.py --inferring_clip_embeddings --image_text_folder path_to_folder --truncate_captions