More for Keys, Less for Values: Adaptive KV Cache Quantization

Open in new window