@saf3synt: AutoRound shrinks LLMs down to 2 to 4 bits AutoRound is an open source toolkit by Intel that quantizes large language and vision language models to as low as two bits using sign-gradient descent. Useful for researchers and builders who need to run large models with less memory. https://github.com/intel/auto-round #AI #Opensource #LLMQuantization #AutoRound Follow for daily open source drops 👇