[Open PR] llama: add --n-cpu-ffn option by John-194 · Pull Request #26622 · ggml-org/llama.cpp

Collecting trend data

The percentage is based on available heat signal, not comment count or independent people.