lucidrains / DALLE2-pytorch

Implementation of DALL-E 2, OpenAI's updated text-to-image synthesis neural network, in Pytorch
MIT License
11.03k stars 1.07k forks source link

Decoder Training - Warn when image is smaller than clip image size #190

Closed Veldrovive closed 2 years ago

Veldrovive commented 2 years ago

When using on the fly image embedding generation, it hurts the quality of the embedding if the preprocessing has cropped the image to smaller than the clip input size. As an improvement to the decoder training script, it should warn the user if the image has been cropped too much.

Extra requirements:

I can handle this when I have time, but it could also be worked on by anyone who has a working knowledge of python.

lucidrains commented 2 years ago

@Veldrovive went with the most straightforward way to address this https://github.com/lucidrains/DALLE2-pytorch/commit/30706102312a22d2d327a98c2cbd06cd62d99473

Veldrovive commented 2 years ago

Yep, that'll do it. Thanks!