Beyond KV Caching: Shared Attention for Efficient LLMs

Re-assign community