Unlearn API for Diffusers #12373
Description
Is your feature request related to a problem? Please describe.
Currently, it appears there in no unlearn API that is part of diffusers library specifically in the context of T2I models. Researchers always need to fork existing repos released by various works , plumb them for execution. This is not straight forward. (Ex:- Erasing Concepts from Diffusion Models, Ablating Concepts in Text-to-Image Diffusion Models, Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models and others)
Describe the solution you'd like.
An Unlearn API, that can take inputs as diffusion model, method, concept to unlearn and produce and unlearned model will be very helpful.
Describe alternatives you've considered.
Researchers always need to fork existing repos released by various works , plumb them for execution. This is not straight forward. (Ex:- Erasing Concepts from Diffusion Models, Ablating Concepts in Text-to-Image Diffusion Models, Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models and others)
Additional context.
Similar API may also be applicable for transformers library. Let me know if another issue needs to be submitted there.