app.py

import torch
import gradio as gr
from PIL import Image
from transformers import Qwen2_5_VLForConditionalGeneration, Qwen2_5_VLProcessor

# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------
MODEL_ID = "Qwen/Qwen2.5-VL-3B-Instruct"
DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
DTYPE = torch.bfloat16 if torch.cuda.is_available() else torch.float32

DEFAULT_PROMPT = (
    "Do you see any abnormality in the chest? Write briefly. "
    "If yes, also tell where the abnormality is in which part of the chest. "
    "The chest parts include lungs, heart and vessels, spine, diaphragm, "
    "soft tissues, Mediastinum and bones of chest shown in image. "
    "Respond only in English. Do NOT use any other language. "
    "**Do not use Chinese language.**"
)

# ---------------------------------------------------------------------------
# Load model & processor
# ---------------------------------------------------------------------------
print(f"Loading model: {MODEL_ID}")
model = Qwen2_5_VLForConditionalGeneration.from_pretrained(
    MODEL_ID,
    torch_dtype=DTYPE,
    device_map="auto" if torch.cuda.is_available() else None,
)
if not torch.cuda.is_available():
    model = model.to(DEVICE)

# Skip video_processor attribute to avoid torchvision dependency
_orig_attrs = Qwen2_5_VLProcessor.attributes[:]
Qwen2_5_VLProcessor.attributes = [a for a in _orig_attrs if a != "video_processor"]
processor = Qwen2_5_VLProcessor.from_pretrained(MODEL_ID)
Qwen2_5_VLProcessor.attributes = _orig_attrs
print("Model loaded successfully.")


# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def pad_to_square(image: Image.Image) -> Image.Image:
    width, height = image.size
    if width == height:
        return image
    max_dim = max(width, height)
    new_image = Image.new("RGB", (max_dim, max_dim), (0, 0, 0))
    new_image.paste(image, ((max_dim - width) // 2, (max_dim - height) // 2))
    return new_image


# ---------------------------------------------------------------------------
# Inference
# ---------------------------------------------------------------------------
def predict(image: Image.Image, prompt: str, max_new_tokens: int, temperature: float):
    if image is None:
        return "Please upload a chest X-ray image."

    if image.mode != "RGB":
        image = image.convert("RGB")

    image = pad_to_square(image)

    if not prompt.strip():
        prompt = DEFAULT_PROMPT

    messages = [
        {
            "role": "user",
            "content": [
                {"type": "image", "image": image},
                {"type": "text", "text": prompt},
            ],
        }
    ]

    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    inputs = processor(
        text=[text], images=[image], return_tensors="pt", padding=True
    ).to(model.device)

    with torch.no_grad():
        generated_ids = model.generate(
            **inputs,
            max_new_tokens=int(max_new_tokens),
            do_sample=temperature > 0,
            temperature=temperature if temperature > 0 else 1.0,
        )

    generated_ids_trimmed = [
        out_ids[len(in_ids):]
        for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
    ]

    return processor.batch_decode(
        generated_ids_trimmed,
        skip_special_tokens=True,
        clean_up_tokenization_spaces=False,
    )[0]


# ---------------------------------------------------------------------------
# Gradio UI
# ---------------------------------------------------------------------------
with gr.Blocks(
    title="Chest X-Ray Analysis — Qwen2.5-VL-3B",
    theme=gr.themes.Soft(),
) as demo:
    gr.Markdown(
        "# Chest X-Ray Analysis\n"
        "Upload a chest X-ray and get an automated report "
        "powered by **Qwen2.5-VL-3B-Instruct**."
    )

    with gr.Row():
        with gr.Column(scale=1):
            image_input = gr.Image(type="pil", label="Upload Chest X-Ray")
            prompt_input = gr.Textbox(
                label="Prompt",
                value=DEFAULT_PROMPT,
                lines=4,
            )
            with gr.Row():
                max_tokens_slider = gr.Slider(
                    minimum=64, maximum=1024, value=512, step=64,
                    label="Max New Tokens",
                )
                temperature_slider = gr.Slider(
                    minimum=0.0, maximum=1.5, value=0.3, step=0.05,
                    label="Temperature (0 = greedy)",
                )
            submit_btn = gr.Button("Analyze", variant="primary")

        with gr.Column(scale=1):
            output_text = gr.Textbox(label="Model Report", lines=20)

    submit_btn.click(
        predict,
        inputs=[image_input, prompt_input, max_tokens_slider, temperature_slider],
        outputs=output_text,
    )

    gr.Markdown(
        "---\n"
        "*Research purposes only — not a substitute for professional medical diagnosis.*"
    )

if __name__ == "__main__":
    demo.launch()