llama.cpp

ggml : use kernel type for conv_1d im2col to support f32

#25323

Open