fix(ggml-cuda): skip sm_120→sm_120a for consumer Blackwell (no FP4 MMA)#3

davide221 merged 1 commit into

Luce-Org:luce-dflashLuce-Org/llama.cpp-dflash-ggml:luce-dflashfrom

easel:fix/consumer-blackwell-sm120easel/llama.cpp-dflash-ggml:fix/consumer-blackwell-sm120Copy head branch name to clipboard