update

app.py

@@ -62,7 +62,7 @@ def handle_image_selection(image_name, state):
     return initialize_simulator(image_name, state)
 
 def init_model():
-    return  GenieSimulator(
+    genie =  GenieSimulator(
                 image_encoder_type='temporalvae',
                 image_encoder_ckpt='stabilityai/stable-video-diffusion-img2vid',
                 quantize=False,
@@ -73,6 +73,14 @@ def init_model():
                 domain='language_table',
             )
 
+    image = Image.open("sim/assets/langtable_prompt/frame_06.png")
+    prompt_image = np.tile(
+        np.array(image), (genie.prompt_horizon, 1, 1, 1)
+    ).astype(np.uint8)
+    prompt_action = np.zeros(
+        (genie.prompt_horizon, genie.action_stride, 2)
+    ).astype(np.float32)
+    return genie
 
 if __name__ == '__main__':
 
@@ -89,6 +97,10 @@ if __name__ == '__main__':
         with gr.Row():
             image_display = gr.Image(type="pil", label="Generated Image")
 
+        select_button.click(
+            fn=handle_image_selection, inputs=[image_selector, genie_instance], outputs=image_display, show_progress='hidden'
+        )
+
         with gr.Row():
             up = gr.Button("↑ Up")
         with gr.Row():
@@ -96,9 +108,7 @@ if __name__ == '__main__':
             down = gr.Button("↓ Down")
             right = gr.Button("→ Right")
 
-        select_button.click(
-            fn=handle_image_selection, inputs=[image_selector, genie_instance], outputs=image_display, show_progress='hidden'
-        )
+
         up.click(fn=lambda state: handle_input("up", state), inputs=[genie_instance], outputs=image_display, show_progress='hidden')
         down.click(fn=lambda state: handle_input("down", state), inputs=[genie_instance], outputs=image_display, show_progress='hidden')
         left.click(fn=lambda state: handle_input("left", state), inputs=[genie_instance], outputs=image_display, show_progress='hidden')