MixLLM: LLM Quantization with Global Mixed-precision between Output-features and Highly-efficient System Design

Open in new window