k-diffusion
k-diffusion copied to clipboard
crowsonkb
Karras et al. (2022) diffusion models for PyTorch
Fixed the UnboundLocalError in the sample_dpmpp_2m_sde and sample_dpmpp_3m_sde functions. ``` File "/workspace/stable-diffusion-webui/modules/processing.py", line 867, in process_images_inner samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts) File "/workspace/stable-diffusion-webui/modules/processing.py", line 1528, in sample...
In Readme it is claimed that k-diffusion supports log likelihood calculation, is there any demo code in this repo?
Hello, I am trying to use the sampler with a custom openAI model, so I reconstruct the Denoiser and sample function as below, while it seems to lead to a...
https://arxiv.org/abs/2305.08891 I think these might be helpful
Hi, I've been playing around with the sample_dpmpp_2m sampling and found that swapping one variable changes/fixes blur. I don't know the math formula for this, so I might be wrong....
When Using the [AUTOMATIC1111 ](https://github.com/AUTOMATIC1111/stable-diffusion-webui) or [vladmandic/automatic](https://github.com/vladmandic/automatic) stable diffusion webui, the Sampler `DPM++ SDE` spits out an error when a seed `3106912320` is defined in the `Prompt Matrix` script: `SeedSequence...
torch.Generator is not supported on Intel so we have to run BrownianTreeNoiseSampler on the CPU with transform and send results back to the GPU with transform_last. It will throw this...
Was running into an issue with this at times when using Tiled VAE to upscale, however it does make sense to check here does it not?
 In k-diffusion sampling https://github.com/crowsonkb/k-diffusion/blob/master/k_diffusion/sampling.py#LL56C1-L56C1, it takes 170ms
