from transformers import ViltProcessor, ViltForQuestionAnswering
import requests
from PIL import Image
import gradio as gr
import torch


processor = ViltProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")


def vqa(image, question):
    inp = Image.fromarray(image.astype('uint8'), 'RGB')
    inputs = processor(inp, question, return_tensors="pt")
    
    outputs = model(**inputs)
    logits = outputs.logits
    idx = logits.argmax(-1).item()

    return model.config.id2label[idx]


demo = gr.Interface(fn=vqa,
                    inputs=['image', 'text'],
                    outputs='text',
                    title = "vqa",
                    examples = [['soccer.jpg', 'how many people in the picture?']])


if __name__ == "__main__":
    demo.queue(concurrency_count=3).launch(server_name = "0.0.0.0", server_port = 7023)