llama.cpp

7a221b67 - llama : use F32 precision in Qwen2 attention and no FA (#8412)