app.py

import gradio as gr
from transformers import pipeline
from PIL import Image

# โหลดโมเดลและ pipeline
model, tokenizer = FastVisionModel.from_pretrained(
    "unsloth/Llama-3.2-11B-Vision-Instruct",
    load_in_4bit=True,
    use_gradient_checkpointing="unsloth",
    device_map="auto"
)

vision_pipeline = pipeline(
    "image-to-text",
    model=model,
    tokenizer=tokenizer,
    device=0
)

# ฟังก์ชันสำหรับ Gradio
def analyze_image(image, instruction):
    result = vision_pipeline(images=image, instruction=instruction, max_new_tokens=128)
    return result[0]["generated_text"]

# สร้าง UI ด้วย Gradio
interface = gr.Interface(
    fn=analyze_image,
    inputs=[
        gr.Image(type="pil"),
        gr.Textbox(label="Instruction", value="You are an expert radiographer. Describe accurately what you see in this image.")
    ],
    outputs=gr.Textbox(label="Generated Text"),
    title="Medical Image Analysis",
    description="Analyze X-ray images and generate a diagnostic report."
)

# รันแอป
if __name__ == "__main__":
    interface.launch()