【[254星]KVSplit：让你的Mac运行更大语言模型、更长上下文！亮点：1. 8位键+4位值的差异化精度，内存减少59%；2. 性能提升，比FP16快5.7%；3. 专为M1/M2/M3 Mac优化，支持Metal】
'KVSplit: Run larger LLMs with longer contexts on Apple Silicon by using differentiated precision for KV cache quantization. KVSplit enables 8-bit keys & 4-bit values, reducing memory by 59% with
