llama.cpp

[SYCL] Fix batched impl for NVidia GPU

#6164

Merged