from llama_cpp import Llama llm = Llama.from_pretrained( repo_id="BeaverAI/Gemmasutra-Mini-2B-v2aa-GGUF", filename="Gemmasutra-Mini-2B-v2aa-BF16.gguf", ) llm.create_chat_completion( messages = [ { "role": "user", "content": "What is the capital of France?" } ] )