Sinkhorn: Make LLMs even smaller through quantisation while maintaining accuracy(github.com)4 points by ilitirit 364 days ago | 1 comment