update

app.py

@@ -189,12 +189,7 @@ self_frame_idx = None
 
 
 @spaces.GPU()
-def run_interactive(first_frame, action, first_pose, curr_frame, device):
-    global self_frames
-    global self_poses
-    global self_memory_c2w
-    global self_frame_idx
-
+def run_interactive(first_frame, action, first_pose, curr_frame, device, self_frames, self_poses, self_memory_c2w, self_frame_idx):
     new_frame, self_frames, self_poses, self_memory_c2w, self_frame_idx = worldmem.interactive(first_frame,
                                     action,
                                     first_pose, 
@@ -204,7 +199,7 @@ def run_interactive(first_frame, action, first_pose, curr_frame, device):
                                     self_poses=self_poses,
                                     self_memory_c2w=self_memory_c2w,
                                     self_frame_idx=self_frame_idx)
-    return new_frame
+    return new_frame, self_frames, self_poses, self_memory_c2w, self_frame_idx
 
 def set_denoising_steps(denoising_steps, sampling_timesteps_state):
     worldmem.sampling_timesteps = denoising_steps
@@ -218,14 +213,23 @@ def generate(keys):
     actions = parse_input_to_tensor(keys)
     global input_history
     global memory_curr_frame
+    global self_frames
+    global self_poses
+    global self_memory_c2w
+    global self_frame_idx
+
     for i in range(len(actions)):
         memory_curr_frame += 1
 
-        new_frame = run_interactive(memory_frames[0],
+        new_frame, self_frames, self_poses, self_memory_c2w, self_frame_idx = run_interactive(memory_frames[0],
                                       actions[i],
                                       None, 
                                       memory_curr_frame, 
-                                      device=device)
+                                      device=device,
+                                      self_frames=self_frames,
+                                      self_poses=self_poses,
+                                      self_memory_c2w=self_memory_c2w,
+                                      self_frame_idx=self_frame_idx)
         
         # print("algo frame:", len(runner.algo.frames))
 
@@ -248,6 +252,10 @@ def reset():
     global memory_curr_frame
     global input_history
     global memory_frames
+    global self_frames
+    global self_poses
+    global self_memory_c2w
+    global self_frame_idx
 
     # worldmem.reset()
     self_frames = None
@@ -259,11 +267,15 @@ def reset():
     memory_curr_frame = 0
     input_history = ""
 
-    _ = run_interactive(memory_frames[0],
+    _, self_frames, self_poses, self_memory_c2w, self_frame_idx = run_interactive(memory_frames[0],
                                 actions[0],
                                 poses[0],
                                 memory_curr_frame,
-                                device=device)
+                                device=device,
+                                self_frames=self_frames,
+                                self_poses=self_poses,
+                                self_memory_c2w=self_memory_c2w,
+                                self_frame_idx=self_frame_idx)