Feature Request: Implement Diffusers Pipeline

[johko]

trafficstars

Contact Details

[email protected]

Feature Description

In the spirit of the implementation of the transformers pipeline functionality, I think it would be a great addition to also have the diffusers (https://github.com/huggingface/diffusers) DiffusionPipeline and with this, adding a lot of Diffusion Model capabilities to SuperDuperDB.

Pipelines Overview: https://huggingface.co/docs/diffusers/api/pipelines/overview

What do you think @blythed ? It would be something I'm up to implement, even though the scope might be a bit bigger.

Use Case Description

There are a lot of different use-cases for the diffusers pipeline

• text-to-image with various models
• text-to-3D (Shap-E)
• image editing (DiffEdit)
• text-to-audio

Organization

It coudl be used by inidividuals as well as companies to quickly experiment on a set of prompts. E.g. you could have database with slightly varying prompts. To see which gives you the best result you simply run the diffusers pipeline and get the created images in your database.

Who are the stake-holders?

No response