llama.cpp
fix embeddings when using CUDA
#3657
Merged

Loading