.

main.py

@@ -224,8 +224,8 @@ def predict_next_frame(previous_frames: List[np.ndarray], previous_actions: List
     prev_y = 0
     #print ('here')
 
-    prompt = 'N + 0 2 4 0 : + 0 0 3 2 N + 0 1 6 0 : + 0 0 2 4 N + 0 1 9 2 : + 0 0 9 6 N + 0 1 1 2 : + 0 0 9 6 N + 0 1 5 2 : + 0 1 7 6 N + 0 0 0 0 : + 0 2 3 2 N + 0 3 2 0 : + 0 2 7 2 N + 0 4 6 4 : + 0 3 2 8'
-    previous_actions = [('move', (240, 32)), ('move', (160, 24)), ('move', (192, 96)), ('move', (112, 96)), ('move', (152, 176)), ('move', (0, 232)), ('move', (320, 272)), ('move', (464, 328))]
+    prompt = 'N + 0 0 0 0 : + 0 0 4 8 N + 0 2 4 8 : + 0 1 2 0 N + 0 2 4 8 : + 0 3 1 2 N + 0 0 9 6 : + 0 0 1 6 L + 0 4 8 0 : + 0 1 2 0 N + 0 2 9 6 : + 0 0 2 4 N + 0 2 0 0 : + 0 3 0 4 N + 0 2 7 2 : + 0 1 5 2'
+    previous_actions = [('move', (0, 48)), ('move', (248, 120)), ('move', (248, 312)), ('move', (96, 16)), ('left_click', (480, 120)), ('move', (296, 24)), ('move', (200, 304)), ('move', (272, 152))]
     
     
     for action_type, pos in previous_actions: #[-8:]:

utils.py

@@ -37,6 +37,7 @@ def sample_frame(model: LatentDiffusion, prompt: str, image_sequence: torch.Tens
         #uc = model.enc_concat_seq(uc, u_dict, 'c_concat')
         
         c_dict = {'c_crossattn': prompt, 'c_concat': image_sequence}
+        model.eval()
         c = model.get_learned_conditioning(c_dict)
         print (c['c_crossattn'].shape)
         print (c['c_crossattn'][0])