llama.cpp

Enable per-conversation loading states to allow having parallel conversations

#16327

Open