.

main.py

@@ -512,7 +512,8 @@ async def websocket_endpoint(websocket: WebSocket):
                 #    print ('predicting', f"record_10003/image_{117+len(previous_frames)}.png")
                 print ('previous_actions', previous_actions)
                 next_frame, next_frame_append = predict_next_frame(previous_frames, previous_actions)
-                previous_frames.append(next_frame_append)
+                #previous_frames.append(next_frame_append)
+                previous_actions = []
                 # Load and append the corresponding ground truth image instead of model output
                 #print ('here4', len(previous_frames))
                 #if DEBUG_TEACHER_FORCING:

utils.py

@@ -71,6 +71,7 @@ def sample_frame(model: LatentDiffusion, prompt: str, image_sequence: torch.Tens
         print ('finished sleeping')
         DDPM = False
         DDPM = True
+        DDPM = False
 
         if DEBUG:
             #c['c_concat'] = c['c_concat']*0