app.py

import gradio as gr
from transformers import pipeline

# Load the VQA model
oracle = pipeline(model="dandelin/vilt-b32-finetuned-vqa")

# Define a function to use the model
def answer_question(image, question):
    response = oracle(question=question, image=image)
    return response[0]['answer']

# Import the specific input and output components directly
from gradio import Image, Textbox

# Create the Gradio interface
interface = gr.Interface(
    fn=answer_question,
    inputs=[Image(type="filepath"), Textbox()],
    outputs=Textbox()
)

# Launch the app
interface.launch()