app.py

from vllm import LLM, SamplingParams
import gradio as gr
repo_id = "mistral-community/pixtral-12b-240910" #Replace to the model you would like to use
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)

# @spaces.GPU #[uncomment to use ZeroGPU]
def infer(image_url, prompt, progress=gr.Progress(track_tqdm=True)):
    # tokenize image urls and text
    llm = LLM(model="mistralai/Pixtral-12B-2409")  # Name or path of your model
    messages = [
        {
            "role": "user",
            "content": [{"type": "text", "text": prompt}, {"type": "image_url", "image_url": {"url": image_url}}]
        },
    ]

    outputs = llm.chat(messages, sampling_params=sampling_params)

    print(outputs[0].outputs[0].text)
    return outputs


example_images = ["https://picsum.photos/id/237/200/300"]
example_prompts = ["What do you see in this image?"]

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(f"""
        # Text-to-Image Gradio Template
        """)

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=2,
                placeholder="Enter your prompt",
                container=False,
            )

            image_url = gr.Text(
                label="Image URL",
                show_label=False,
                max_lines=1,
                placeholder="Enter your image URL",
                container=False,
            )

            run_button = gr.Button("Run", scale=0)

        result = gr.Textbox(
            show_label=False
        )

        gr.Examples(
            examples=example_images,
            inputs=[image_url]
        )

        gr.Examples(
            examples=example_prompts,
            inputs=[prompt]
        )
    gr.on(
        triggers=[run_button.click, image_url.submit, prompt.submit],
        fn=infer,
        inputs=[image_url, prompt],
        outputs=[result]
    )

demo.queue().launch()