Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters

Open in new window