Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
Why need unnormalize img when apply openai's clip `embed_image`
See original GitHub issueHi Phil,
I doubt whether need translate image from (-1,1) -> (0,1) here. Since the image parsed in is from the original dataset(PIL), and its range is (0,1), I think here should avoid using unnormalize()
https://github.com/lucidrains/DALLE2-pytorch/blob/2db0c9794c33e98df25b84f557a683a8900dfc61/dalle2_pytorch/dalle2_pytorch.py#L281
When I do an experiment on decoder training, the sampled image from Decoder turns out too bright (the whole pixel value shifts to a higher range)
Best,
Issue Analytics
- State:
- Created a year ago
- Comments:6 (5 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
wow what a speed! I just took a snap haha
Yeah, do normalize / unnormalize within the decoder class is enough (operate in
p_sample_loopandp_losses)ok all done in the latest! now nobody has to worry about this normalization / inverse normalization business 😃