vit-vqgan icon indicating copy to clipboard operation
vit-vqgan copied to clipboard

JAX implementation ViT-VQGAN

vit-vqgan

JAX implementation of ViT-VQGAN.

Acknowledgements

  • Jiahui Yu for his advice on the implementation based on the work on ViT-VQGAN and Parti.
  • Phil Wang for suggesting using convolutions after self-attention layers.
  • Katherine Crowson for suggesting using convolutions in feed-forward layers.