1 Commits

Author	SHA1	Message	Date
ViperEkura	9e63cb9ed0	feat: MMA head-packing decode kernel with scalar fallback dispatch ... - Add gqa_decode_attn_mma.cuh for tensor-core decode path - Add dispatch_decode<> selecting MMA vs scalar based on G and mask - Add TORCH_CHECK for unsupported head_dim instead of silent scalar launch	2026-07-07 22:56:02 +08:00