ExplainingAI

Thank you @soulslicer (have now updated the readme to indicate the same) @Recoil77 So sorry for completely missing to respond on this. I know its been long but were you...

Hi, In this repo, only gaussian noise is added progressively. It seems that based on your requirement, this [paper](https://arxiv.org/abs/2208.09392), might be exactly what you are looking for. They have an...

Thank you for the appreciation. I have never trained ldm for the inpainting task but from what I could figure out from the code/checkpoints provided by authors, you would pass...

When I trained this on Celeb captions, I also found that for the captions that are very common(like hair color), the trained text conditioned diffusion model was performing very well...

Hello @Vadori , Can you share your config file for vqvae training once ? Regarding the green spots , I feel that it might be because the pixels are not...

I see that the disc_start parameter has been set to 10 iterations. Have you tried training vqvae with a higher value of disc start. Like maybe 5000? If not, then...

@Vadori You can try with more vectors, I didnt find much help on the dataset that I worked with(CelebHQ) but give it a try, might help for your case. I...

@Vadori Went through everything and your understanding for all of them is correct. Just one thing, I dont have any `num_headblocks` in config. I am assuming thats a parameter that...

Hello @Vadori , Yes in the current implementation the only image conditioning that is done is spatial conditioning(for mask to image task). For spatial conditioning, the authors use the mechanism...

By better, I am mainly referring to how easy it is for the model to learn 'how to use spatial conditioning' . In concatenation, because of the convolution layer, each...