The Functor Model Architecture

I have been working on functor models for awhile now. The model learns by changes to functions rather than parameters. I am seeing possibilities for reducing AI energy usage. Functor models can be of most types and any size. I have put the functor LLM aside in favor of the LLM Offloading Pattern (LOP). Functor models learn one shot so a continual flow of information that is curated and learning routed to an SLM in a layer accessed by the LLM can establish a topical knowledge base. Agents discover the data and continue to do so this has a "self-evolving" aspect.

Documentation
GitHub repo for Functor ResNet demo