双 GEMM
作者: WalrusWTQ创建于 2026年3月6日更新于 2026年7月13日
Hi, I'm studying the source code of flashmla. I noticed that in the implementation of **sparse decode head64**, it uses "dual gemm" to compute P=QK^T. I have a few questions about this design.
内容来源: deepseek-ai/FlashMLA