kwonminki

Did you solve the problem...? I have the same issue... :(

 50 step with your prompt : a photograph of a hamburger, highly detailed, 4k  3 step with original timesteps ([667,...

I want to share more experiments.  DDIM, 10 step sampling with previous timesteps.  DDIM, 10 step sampling with my modified...

- Diffusion Models already have a Semantic Latent Space - https://arxiv.org/abs/2210.10960 - https://kwonminki.github.io/Asyrp/ - 제 논문 소개입니다...ㅎ - Diffusion models의 bottleneck을 semantic latent space로 쓸 수 있다는 논문입니다  ...

It seems that SDEdit no longer provides checkpoints. Please check it for CelebA-HQ model. https://github.com/curryjung/DiffStyle_official/issues/1

Thank you for asking about my code. I apologize for the delayed response. After reviewing my previous work on the AFHQ-dog dataset, I realized that there may have been some...

After the Open-Sora update, --enable_flashattn occurs NaN loss on my side. @oahzxl

Hi. I tried 1. disable modulate kernel 2. use bf16 3. grad clip; and all possible other things. Only when I disabled the flashattn, the loss was not NaN. I...