Update app.py

app.py

@@ -215,7 +215,7 @@ def localized_query(img, x, y, question):
         outline="blue",
     )
     
-    yield answer, {"text": "", "entities": []}, gr.update(visible=True, value=img_clone)
+    yield answer, gr.update(visible=True, value=img_clone)
 
 
 js = ""
@@ -243,7 +243,7 @@ with gr.Blocks(title="moondream vl (new)", css=css, js=js) as demo:
 
     gr.Markdown(
         """
-        # 🌔 grounded visual question asnwering
+        # 🌔 grounded visual question answering
         """
     )
 
@@ -251,72 +251,30 @@ with gr.Blocks(title="moondream vl (new)", css=css, js=js) as demo:
 
     with gr.Row():
         with gr.Column():
-
-            @gr.render(inputs=[mode_radio])
-            def show_inputs(mode):
-                with gr.Group():
-                    with gr.Row():
-                        prompt = gr.Textbox(
-                            label="Input",
-                            value="What is this?",
-                            scale=4,
-                        )
-                        submit = gr.Button("Submit")
-                    img = gr.Image(type="pil", label="Upload an Image")
-                    x_slider = gr.Slider(label="x", minimum=0, maximum=1)
-                    y_slider = gr.Slider(label="y", minimum=0, maximum=1)
-                submit.click(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
-                prompt.submit(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
-                x_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
-                y_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
-                img.change(localized_query, [img, x_slider, y_slider, prompt], [output, thought, ann])
-                def select_handler(image, evt: gr.SelectData):
-                    w, h = img.size
-                    return [evt.index[0] / w, evt.index[1] / h]
-                img.select(select_handler, img, [x_slider, y_slider])
+            with gr.Group():
+                with gr.Row():
+                    prompt = gr.Textbox(
+                        label="Input",
+                        value="What is this?",
+                        scale=4,
+                    )
+                    submit = gr.Button("Submit")
+                img = gr.Image(type="pil", label="Upload an Image")
+                x_slider = gr.Slider(label="x", minimum=0, maximum=1)
+                y_slider = gr.Slider(label="y", minimum=0, maximum=1)
+            submit.click(localized_query, [img, x_slider, y_slider, prompt], [output, ann])
+            prompt.submit(localized_query, [img, x_slider, y_slider, prompt], [output, ann])
+            x_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, ann])
+            y_slider.change(localized_query, [img, x_slider, y_slider, prompt], [output, ann])
+            img.change(localized_query, [img, x_slider, y_slider, prompt], [output, ann])
+            def select_handler(image, evt: gr.SelectData):
+                w, h = img.size
+                return [evt.index[0] / w, evt.index[1] / h]
+            img.select(select_handler, img, [x_slider, y_slider])
 
         with gr.Column():
-            thought = gr.HighlightedText(
-                elem_classes=["chain-of-thought"],
-                label="Thinking tokens",
-                interactive=False,
-            )
             output = gr.Markdown(label="Response", elem_classes=["output-text"], line_breaks=True)
             ann = gr.Image(visible=False)
 
-        def on_select(img, evt: gr.SelectData):
-            if img is None or evt.value[1] is None:
-                return gr.update(visible=False, value=None)
-
-            w, h = img.size
-            if w > 768 or h > 768:
-                img = Resize(768)(img)
-                w, h = img.size
-
-            points = json.loads(evt.value[1])
-
-            img_clone = img.copy()
-            draw = ImageDraw.Draw(img_clone)
-            
-            for point in points:
-                x = int(point[0] * w)
-                y = int(point[1] * h)
-                draw.ellipse(
-                    (x - 3, y - 3, x + 3, y + 3),
-                    fill="red",
-                    outline="red",
-                )
-
-
-            return gr.update(visible=True, value=img_clone)
-
-        thought.select(on_select, [input_image], [ann])
-        input_image.change(lambda: gr.update(visible=False), [], [ann])
-
-    mode_radio.change(
-        lambda: ("", "", gr.update(visible=False, value=None)),
-        [],
-        [output, thought, ann],
-    )
 
 demo.queue().launch()