CUDA: skip fusion for repeating adds in bias#17080

am17an merged 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

am17an:cuda-fix-bias-fusionam17an/llama.cpp:cuda-fix-bias-fusionCopy head branch name to clipboard