Hi LCM team,
I've been exploring the large_concept_model repository and noticed the implementations of BaseLCM, Single Tower Diffusion, and Two Tower Diffusion models. I am curious if the team has experimented with or considered other Transformer variants, such as recursive transformers or their recent evolutions like Relaxed Recursive Transformers or Mixture-of-Recursions models?
Given the efficiency and parameter-sharing advantages of recursive transformers researched in recent papers, I wonder if they could be beneficial or are planned for this codebase.
Would appreciate any insights or pointers regarding this!
Thank you for your work on this great project.