Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

Open in new window