Hey guys - primarily a research release with working models, Not so much a model as a psuedo-new quantization technique. I've been experimenting with a modification of ISTALab's RCO algorithm that can quantize models on a strict VRAM budget. It's at its core an approximation algorithm that attempts
