Support optionally using karras sigmas in the k_diffusion wrapper pipeline

[apolinario]

Is your feature request related to a problem? Please describe.
K-Diffusion enables users to generate sigmas with k_diffusion.sampling.get_sigmas_karras. diffusers currently do not support using those to generate the sigmas with its k_diffusion wrapper

Describe the solution you'd like
Enable users to optionally generate karras sigmas similarly to how AUTO1111 does it https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/44c46f0ed395967cd3830dd481a2db759fda5b3b/modules/sd_samplers.py#L453

[patrickvonplaten]

I will try to allocate some time in the next weeks to look into it (can't promise to get it done soon though). Just wondering: do we have some results/benchmarks to know whether karras sigmas work better ?

[apolinario]

I don't think they necessarily work better.

It just works differently and the community enjoys having the possibility of having this variations for artistic purposes when using tools that support it such as AUTO1111 and any other based on k_diffusion so I think it would be great for diffusers to also enable such variations. But I can't say it works better (some people believe so subjectively)

[patrickvonplaten]

In case one is interested in taking a look? @williamberman @yiyixuxu @pcuenca @patil-suraj

[patrickvonplaten]

Here a solution that should work well: #1887 (comment) . Hope to soon find time to implement it :-)