add example

app.py

@@ -120,7 +120,7 @@ def inference_caption(
 
 
 title = """<h1 align="center">BLIP-2</h1>"""
-description = """Gradio demo for BLIP-2, a multimodal chatbot from Salesforce Research. To use it, simply upload your image, or click one of the examples to load them. Please visit our <a href='https://github.com/salesforce/LAVIS/tree/main/projects/blip2' target='_blank'>project webpage</a>.</p> 
+description = """Gradio demo for BLIP-2, image-to-text generation from Salesforce Research. To use it, simply upload your image, or click one of the examples to load them. Please visit our <a href='https://github.com/salesforce/LAVIS/tree/main/projects/blip2' target='_blank'>project webpage</a>.</p> 
 <p> <strong>Disclaimer</strong>: This is a research prototype and is not intended for production use. No data including but not restricted to text and images is collected. </p>"""
 article = "<p style='text-align: center'><a href='https://arxiv.org/abs/2201.12086' target='_blank'>BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models</a>"
 
@@ -128,6 +128,7 @@ endpoint = Endpoint()
 
 examples = [
     ["house.png", "How could someone get out of the house?"],
+    ["forbidden_city.webp", "In what dynasties was this place build?"],
     # [
     #     "sunset.png",
     #     "Write a romantic message that goes along this photo.",
@@ -156,17 +157,18 @@ with gr.Blocks() as iface:
                     minimum=0.5,
                     maximum=1.0,
                     value=0.8,
+                    step=0.1,
                     interactive=True,
-                    label="Temperature (set to 0 for greedy decoding with nucleus sampling)",
+                    label="Temperature (used with nucleus sampling)",
                 )
 
                 len_penalty = gr.Slider(
-                    minimum=-2.0,
+                    minimum=-1.0,
                     maximum=2.0,
                     value=1.0,
-                    step=0.5,
+                    step=0.2,
                     interactive=True,
-                    label="Length Penalty (larger value encourages longer sequence with beam search)",
+                    label="Length Penalty (set to larger for longer sequence, used with beam search)",
                 )
 
                 rep_penalty = gr.Slider(
@@ -179,7 +181,7 @@ with gr.Blocks() as iface:
                 )
 
             with gr.Row():
-                caption_output = gr.Textbox(lines=2, label="Caption Output")
+                caption_output = gr.Textbox(lines=2, label="Caption Output (from OPT)")
                 caption_button = gr.Button(
                     value="Caption it!", interactive=True, variant="primary"
                 )
@@ -199,7 +201,7 @@ with gr.Blocks() as iface:
             chat_input = gr.Textbox(lines=2, label="Chat Input")
 
             with gr.Row():
-                chatbot = gr.Chatbot()
+                chatbot = gr.Chatbot(label="Chat Output (from FlanT5)")
                 image_input.change(lambda: (None, "", "", []), [], [chatbot, chat_input, caption_output, state])
 
             with gr.Row():