Paligemma Example

Source vllm-project/vllm.

 1from vllm import LLM
 2from vllm.assets.image import ImageAsset
 3
 4
 5def run_paligemma():
 6    llm = LLM(model="google/paligemma-3b-mix-224")
 7
 8    prompt = "caption es"
 9
10    image = ImageAsset("stop_sign").pil_image
11
12    outputs = llm.generate({
13        "prompt": prompt,
14        "multi_modal_data": {
15            "image": image
16        },
17    })
18
19    for o in outputs:
20        generated_text = o.outputs[0].text
21        print(generated_text)
22
23
24if __name__ == "__main__":
25    run_paligemma()