Add example

app.py

@@ -2,8 +2,11 @@ import gradio as gr
 from transformers import ViltProcessor, ViltForNaturalLanguageVisualReasoning
 import torch
 
+# NLRV2 example images
 torch.hub.download_url_to_file('https://lil.nlp.cornell.edu/nlvr/exs/ex0_0.jpg', 'image1.jpg')
 torch.hub.download_url_to_file('https://lil.nlp.cornell.edu/nlvr/exs/ex0_1.jpg', 'image2.jpg')
+torch.hub.download_url_to_file('https://lil.nlp.cornell.edu/nlvr/exs/acorns_1.jpg', 'image3.jpg')
+torch.hub.download_url_to_file('https://lil.nlp.cornell.edu/nlvr/exs/acorns_6.jpg', 'image4.jpg')
 
 processor = ViltProcessor.from_pretrained("nielsr/vilt-b32-finetuned-nlvr2")
 model = ViltForNaturalLanguageVisualReasoning.from_pretrained("nielsr/vilt-b32-finetuned-nlvr2")
@@ -26,8 +29,9 @@ images = [gr.inputs.Image(type="pil"), gr.inputs.Image(type="pil")]
 text = gr.inputs.Textbox(lines=2, label="Sentence")
 answer = gr.outputs.Textbox(label="Predicted answer")
 
-example_sentence = "The left image contains twice the number of dogs as the right image, and at least two dogs in total are standing."
-examples = [["image1.jpg", "image2.jpg", example_sentence]]
+example_sentence_1 = "One image contains twice the number of dogs as the other image, and at least two dogs in total are standing."
+example_sentence_2 = "One image shows exactly two brown acorns in back-to-back caps on green foliage."
+examples = [["image1.jpg", "image2.jpg", example_sentence_1], ["image1.jpg", "image2.jpg", example_sentence_2]]
 
 title = "Interactive demo: natural language visual reasoning with ViLT"
 description = "Gradio Demo for ViLT (Vision and Language Transformer), fine-tuned on NLVR2. To use it, simply upload a pair of images and type a sentence and click 'submit', or click one of the examples to load them. The model will predict whether the sentence is true or false, based on the 2 images. Read more at the links below."