llama.cpp
ggml-zdnn: accept non-zDNN tensors, invalidate compute tensors after every operation
#31824
Merged

Loading