API_PORT=8000 llamafactory-cli api examples/inference/llama3_vllm.yaml