from lmdeploy import pipeline, PytorchEngineConfig

messages = [
        dict(role='user', content=[
            dict(type='text', text='Describe the following images in detail'),
            dict(type='image_url', image_url=dict(url='https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/tests/data/tiger.jpeg')),
            dict(type='image_url', image_url=dict(url='https://raw.githubusercontent.com/open-mmlab/mmdeploy/main/tests/data/tiger.jpeg')),
            dict(type='text', text='How many tigers are there in total?')
        ]),
    ]

with pipeline('internlm/Intern-S1-mini', backend_config=PytorchEngineConfig()) as pipe:
    res = pipe(messages)
    print(res)