Fix Gemma3n not executed as CUDA_GRAPH on NVGPUs#14741

slaren merged 3 commits intoggml-org:masterggml-org/llama.cpp:masterfrom

ORippler:hotfix_perf_issues_of_gemma3nORippler/llama.cpp:hotfix_perf_issues_of_gemma3nCopy head branch name to clipboard