One more question about vector search with self-hosted implementation, is there any documentation around utilizing GPU resources on VM for generating embeddings? I'm running my typesense instance on nvidia runtime, I don't see the allocated when running

nvidia-smi

however