llama.cpp

f64e4f04 - ggml : testing GPU FP precision via quantized CPY