Update pipeline.py

Hi, I'm [Hyoungwon Cho](https://huggingface.co/hyoungwoncho), author of Perturbed-Attention Guidance. There was an error in implementation, so I corrected that part. I'd appreciate it if you could reflect it. Thank you!

pipeline.py

@@ -157,8 +157,8 @@ class PAGIdentitySelfAttnProcessor:
 
         value = attn.to_v(hidden_states_ptb)
         
-        hidden_states_ptb = torch.zeros(value.shape).to(value.get_device())
-        #hidden_states_ptb = value
+        # hidden_states_ptb = torch.zeros(value.shape).to(value.get_device())
+        hidden_states_ptb = value
         
         hidden_states_ptb = hidden_states_ptb.to(query.dtype)