Update

app.py

@@ -144,12 +144,25 @@ with gr.Blocks(css="style.css") as demo:
         visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1",
     )
 
-    image = gr.Image(type="pil")
+    with gr.Box():
+        image = gr.Image(type="pil")
+        with gr.Tabs():
+            with gr.Tab(label="Image Captioning"):
+                caption_button = gr.Button("Caption it!")
+                caption_output = gr.Textbox(label="Caption Output", show_label=False, container=False)
+            with gr.Tab(label="Visual Question Answering"):
+                chatbot = gr.Chatbot(label="VQA Chat")
+                history_orig = gr.State(value=[])
+                history_qa = gr.State(value=[])
+                vqa_input = gr.Text(label="Chat Input", show_label=False, max_lines=1, container=False)
+                with gr.Row():
+                    clear_chat_button = gr.Button("Clear")
+                    chat_button = gr.Button("Submit", variant="primary")
     with gr.Accordion(label="Advanced settings", open=False):
         sampling_method = gr.Radio(
             label="Text Decoding Method",
             choices=["Beam search", "Nucleus sampling"],
-            value="Beam search",
+            value="Nucleus sampling",
         )
         temperature = gr.Slider(
             label="Temperature",
@@ -175,20 +188,6 @@ with gr.Blocks(css="style.css") as demo:
             value=1.5,
             step=0.5,
         )
-    with gr.Row():
-        with gr.Column():
-            with gr.Box():
-                caption_button = gr.Button(value="Caption it!")
-                caption_output = gr.Textbox(label="Caption Output", show_label=False)
-        with gr.Column():
-            with gr.Box():
-                chatbot = gr.Chatbot(label="VQA Chat")
-                history_orig = gr.State(value=[])
-                history_qa = gr.State(value=[])
-                vqa_input = gr.Text(label="Chat Input", show_label=False, max_lines=1)
-                with gr.Row():
-                    clear_chat_button = gr.Button(value="Clear")
-                    chat_button = gr.Button(value="Submit")
 
     gr.Examples(
         examples=examples,
@@ -227,12 +226,22 @@ with gr.Blocks(css="style.css") as demo:
         fn=chat,
         inputs=chat_inputs,
         outputs=chat_outputs,
+    ).success(
+        fn=lambda: "",
+        outputs=vqa_input,
+        queue=False,
+        api_name=False,
     )
     chat_button.click(
         fn=chat,
         inputs=chat_inputs,
         outputs=chat_outputs,
         api_name="chat",
+    ).success(
+        fn=lambda: "",
+        outputs=vqa_input,
+        queue=False,
+        api_name=False,
     )
     clear_chat_button.click(
         fn=lambda: ("", [], [], []),

style.css

@@ -8,3 +8,9 @@ h1 {
   background: #1565c0;
   border-radius: 100vh;
 }
+
+#component-0 {
+  max-width: 730px;
+  margin: auto;
+  padding-top: 1.5rem;
+}