LycheeDecode: Accelerating Long-Context LLM Inference via Hybrid-Head Sparse Decoding

Re-assign community