app.py

# #Inference using gradio
from peft import PeftModel
from transformers import Qwen2VLForConditionalGeneration
from transformers import AutoProcessor
import gradio as gr
from transformers import Qwen2VLProcessor
from qwen_vl_utils import process_vision_info

#load the base model and finetuned adapter
base_model = Qwen2VLForConditionalGeneration.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")
model = PeftModel.from_pretrained(base_model, "vignesha7/qwen2-2b-instruct-Brain-MRI-Description")

processor = AutoProcessor.from_pretrained("Qwen/Qwen2-VL-2B-Instruct")

#inference function
def generate_description(sample):
    system_message = "You are an expert MRI radiographer. you can describe what you see in the mri image"

    prompt = "Describe accurately what you see in this radiology image."
    messages = [
        {  "role": "system",
           "content": [{"type": "text", "text": system_message}]
         },
        { "role": "user",
          "content" : [
            {"type" : "text",  "text"  : prompt},
            {"type" : "image", "image" : sample}]
        },
    ]
    # Preparation for inference
    text = processor.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=True
    )
    image_inputs, video_inputs = process_vision_info(messages)
    inputs = processor(
        text=[text],
        images=image_inputs,
        videos=video_inputs,
        padding=True,
        return_tensors="pt",
    )
    inputs = inputs.to(model.device)
    # Inference: Generation of the output
    generated_ids = model.generate(**inputs, max_new_tokens=256, top_p=1.0, do_sample=True, temperature=0.8)
    generated_ids_trimmed = [out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)]
    output_text = processor.batch_decode(
        generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
    )
    return output_text[0]

### Gradio app ###
title = "BrainMRI Radiology Expert"
description = "An Qwen2-VL-2B-Instruct model fine tuned on brain mri images.Describes the brain image"

demo = gr.Interface(
    fn=generate_description,
    inputs=gr.Image(type='pil'),
    outputs='text',
    title=title,
    description=description,    
)

demo.launch()