Mixture of specialists aritechture By @Vio 0 Devlogs 0 Total hours This project is a different way of local hosting models splitting them into distinct specialist and using ultra low latency direct vram memory access and other tricks to get a final response
