normalized attention score

tokenwise quantization (1) large language model (1) attention score (1) salient token identification (1) adaptive compression (1) kv cache quantization (1) flash attention (1) attention mechanism (1) model compression (1)