llama.cpp

fabdc3bd - ggml/ex: calculate accuracy in graph, adapt MNIST (ggml/980)