app.py

import spaces
import io
import os
import torch
from PIL import Image
import gradio as gr
from transformers import AutoModelForCausalLM, AutoProcessor, GenerationConfig

title = """# Welcome to🌟Tonic's CheXRay⚕⚛ !
You can use this ZeroGPU Space to test out the current model [StanfordAIMI/CheXagent-8b](https://huggingface.co/StanfordAIMI/CheXagent-8b). CheXRay⚕⚛ is fine tuned to analyze chest x-rays with a different and generally better results than other multimodal models. 
You can also useCheXRay⚕⚛  by cloning this space. 🧬🔬🔍 Simply click here: <a style="display:inline-block" href="https://huggingface.co/spaces/Tonic/CheXRay?duplicate=true"><img src="https://img.shields.io/badge/-Duplicate%20Space-blue?labelColor=white&style=flat&logo=data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAABAAAAAQCAYAAAAf8/9hAAAAAXNSR0IArs4c6QAAAP5JREFUOE+lk7FqAkEURY+ltunEgFXS2sZGIbXfEPdLlnxJyDdYB62sbbUKpLbVNhyYFzbrrA74YJlh9r079973psed0cvUD4A+4HoCjsA85X0Dfn/RBLBgBDxnQPfAEJgBY+A9gALA4tcbamSzS4xq4FOQAJgCDwV2CPKV8tZAJcAjMMkUe1vX+U+SMhfAJEHasQIWmXNN3abzDwHUrgcRGmYcgKe0bxrblHEB4E/pndMazNpSZGcsZdBlYJcEL9Afo75molJyM2FxmPgmgPqlWNLGfwZGG6UiyEvLzHYDmoPkDDiNm9JR9uboiONcBXrpY1qmgs21x1QwyZcpvxt9NS09PlsPAAAAAElFTkSuQmCC&logoWidth=14" alt="Duplicate Space"></a></h3> 

### How To use

Upload a medical image and enter a prompt to receive an AI-generated analysis.
simply upload an image with the right prompt (coming soon!) and anaylze your Xray !

Join us : 🌟TeamTonic🌟 is always making cool demos! Join our active builder's🛠️community 👻  [![Join us on Discord](https://img.shields.io/discord/1109943800132010065?label=Discord&logo=discord&style=flat-square)](https://discord.gg/GWpVpekp) On 🤗Huggingface: [TeamTonic](https://huggingface.co/TeamTonic) & [MultiTransformer](https://huggingface.co/MultiTransformer) On 🌐Github: [Tonic-AI](https://github.com/tonic-ai) & contribute to 🌟 [DataTonic](https://github.com/Tonic-AI/DataTonic) 🤗Big thanks to Yuvi Sharma and all the folks at huggingface for the community grant 🤗
"""

device = "cuda"
dtype = torch.float16

processor = AutoProcessor.from_pretrained("StanfordAIMI/CheXagent-8b", trust_remote_code=True)
generation_config = GenerationConfig.from_pretrained("StanfordAIMI/CheXagent-8b")
# model = AutoModelForCausalLM.from_pretrained("StanfordAIMI/CheXagent-8b", torch_dtype=dtype, trust_remote_code=True)

@spaces.GPU
def generate(image, prompt):
    model = AutoModelForCausalLM.from_pretrained("StanfordAIMI/CheXagent-8b", torch_dtype=dtype, trust_remote_code=True).to(device)
    if hasattr(image, "read"):
        image = Image.open(io.BytesIO(image.read())).convert("RGB")
    else:
        image = image
    images = [image]
    inputs = processor(images=images, text=f" USER: <s>{prompt} ASSISTANT: <s>", return_tensors="pt").to(device=device, dtype=dtype)
    output = model.generate(**inputs, generation_config=generation_config)[0]
    response = processor.tokenizer.decode(output, skip_special_tokens=True)
    return response


with gr.Blocks() as demo:
    gr.Markdown(title)

    with gr.Accordion("Custom Prompt Analysis"):
        with gr.Row():
            image_input_custom = gr.Image(type="pil")
            prompt_input_custom = gr.Textbox(label="Enter your custom prompt")
        generate_button_custom = gr.Button("Generate")
        output_text_custom = gr.Textbox(label="Response")

        def custom_generate(image, prompt):
            if isinstance(image, str) and os.path.exists(image):
                with open(image, 'rb') as file:
                    return generate(file, prompt)
            else:
                return generate(image, prompt)

        generate_button_custom.click(fn=custom_generate, inputs=[image_input_custom, prompt_input_custom], outputs=output_text_custom)

        example_prompt = "65 y/m Chronic cough and weight loss x 6 months. Chest X-rays normal. Consulted multiple pulmonologists with not much benefit. One wise pulmonologist thinks of GERD and sends him to the Gastro department. Can you name the classical finding here?"
        example_image_path = os.path.join(os.path.dirname(__file__), "hegde.jpg")
        gr.Examples(
            examples=[[example_image_path, example_prompt]],
            inputs=[image_input_custom, prompt_input_custom],
            outputs=[output_text_custom],
            fn=custom_generate,
            cache_examples=True
        )
    
    with gr.Accordion("Anatomical Feature Analysis"):
        anatomies = [
            "Airway", "Breathing", "Cardiac", "Diaphragm",
            "Everything else (e.g., mediastinal contours, bones, soft tissues, tubes, valves, and pacemakers)"
        ]
        with gr.Row():
            image_input_feature = gr.Image(type="pil")
            prompt_select = gr.Dropdown(label="Select an anatomical feature", choices=anatomies)
        generate_button_feature = gr.Button("Analyze Feature")
        output_text_feature = gr.Textbox(label="Response")
        generate_button_feature.click(fn=lambda image, feature: generate(image, f'Describe "{feature}"'), inputs=[image_input_feature, prompt_select], outputs=output_text_feature)

    with gr.Accordion("Common Abnormalities Analysis"):
        common_abnormalities = ["Lung Nodule", "Pleural Effusion", "Pneumonia"]
        with gr.Row():
            image_input_abnormality = gr.Image(type="pil")
            abnormality_select = gr.Dropdown(label="Select a common abnormality", choices=common_abnormalities)
        generate_button_abnormality = gr.Button("Analyze Abnormality")
        output_text_abnormality = gr.Textbox(label="Response")
        generate_button_abnormality.click(fn=lambda image, abnormality: generate(image, f'Analyze for "{abnormality}"'), inputs=[image_input_abnormality, abnormality_select], outputs=output_text_abnormality)

demo.launch()