Language models with the flexibility of Stable Diffusion?

Hi all!

I have been really appreciating the increasing feeling of both leverage and wonder this class and the wide range of notebooks floating in the Twitter+Discord ML world have afforded me in having an actual creative relationship with Stable Diffusion and image models.

I was fine tuning GPT-3 this week for a project (teach the model how to tell a certain kind of short story,) and while I got ok results, I found I really missed the ability to tweak parameters and generally look as under the hood as I’d like that I’ve come to feel from SD with the support of the ecosystem & this course.

Does anyone know of an analogous language model that has a thriving ecosystem of teaching and experimentation around it like SD does for diffusion models? I don’t know of one yet, and am afraid it may just not exist, but I figure this would be a good place to start looking!

2 Likes

There isn’t one yet (at least that I know of), but this paper talks about something very similar to what you’re looking for, I believe:

Unfortunately, they don’t have any code that you can play around with …

4 Likes

Oh, thanks @Fahim !
Also nice to virtually meet you. I try to do a sweep of the forums at least weekly and glean what I can from them … I appreciate your posts!

1 Like

Thank you for the kind words :slight_smile: Nice to meet you too!