Update app.py

app.py

@@ -6,6 +6,8 @@ from transformers import AutoModelForCausalLM, AutoTokenizer
 #Setting device to cuda
 torch.set_default_device("cuda")
 
+torch.hub.download_url_to_file('https://github.com/manishkumart/SparrowVQE/blob/main/data/Images/week_01/week_01_page_024.png', 'week_01_page_024.png')
+
 # # Ensure GPU usage if available
 # device = "cuda" if torch.cuda.is_available() else "cpu"
 # torch.set_default_tensor_type('torch.cuda.FloatTensor' if device=='cuda' else 'torch.FloatTensor')
@@ -42,6 +44,7 @@ def gradio_predict(image, question, max_tokens):
     answer = predict_answer(image, question, max_tokens)
     return answer
 
+examples = [["week_01_page_024.png", "Can you explain the slide?"]]
 # Define the Gradio interface
 iface = gr.Interface(
     fn=gradio_predict,
@@ -49,6 +52,7 @@ iface = gr.Interface(
             gr.Textbox(label="Question", placeholder="e.g. Can you explain the slide?", scale=4),
             gr.Slider(2, 500, value=25, label="Token Count", info="Choose between 2 and 500")],
     outputs=gr.TextArea(label="Answer"),
+    examples=examples
     title="Sparrow - Tiny 3B | Visual Question Answering",
     description="An interactive chat model that can answer questions about images in an Academic context. \n We can input images, and the system will analyze them to provide information about their contents. I've utilized this capability by feeding slides from PowerPoint presentations used in classes and the lecture content passed as text. Consequently, the model now mimics the behavior and responses of my professors. So, if I present any PowerPoint slide, it explains it just like my professor would, further it can be personalized.",
 )