diffusers-rs icon indicating copy to clipboard operation
diffusers-rs copied to clipboard

An implementation of the diffusers api in Rust

Results 20 diffusers-rs issues
Sort by recently updated
recently updated
newest added

When stable diffusion was released, the most requested features were always about reducing GPU requirements as this makes it available to more users with cheaper GPUs. I think this is...

Hi @LaurentMazare, now that this repository has expanded more and more, I was thinking we could move most of the code of the examples inside a `StableDiffusion` pipeline, as HF...

Hi, since my NVidia GPU doesn't have enough memory, I resorted to Google Colab and wrote a notebook to run diffusion experiments on the 16GB GPU available there. [Here](https://colab.research.google.com/drive/1R875ar1ElP4NAWcwnyw2K8qY0ssmlLLF?usp=sharing) you...

Right now we are adding the schedulers, but it is difficult to work with since swapping the scheduler doesn't work well. This also slows down testing and evaluation of the...

Mostly a copy of #22 though walked through the HF implementation. Requires the same changes to the stable diffusion example to run. - Also skipped thresholding. - Moved scheduler in...

Would it be possible to use models based on the CompVis style used by stabilityai and supported in HF diffusers? My personal goals are: - Support for [1.5](https://huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned.ckpt) and [2.1](https://huggingface.co/stabilityai/stable-diffusion-2-1/blob/main/v2-1_768-ema-pruned.ckpt)...

Currently, it prints out something like this when generating an image: ``` Building the Clip transformer. Building the autoencoder. Building the unet. Timestep 0/5 Timestep 1/5 Timestep 2/5 Timestep 3/5...

It would be very useful to have a simple flag like `--H 512` and `--W 512` for setting the resolution to generate an image at.

When running this on the CPU, it would be nice to have a flag for setting how many threads it should be allowed to use. Sometimes it might make sense...

As we can see here https://github.com/huggingface/diffusers/issues/556, adding support for padding_mode allows to generate images that can be seamlessly tiled. Ideally, we would like to be able to reconfigure the padding_mode...