Update app.py

app.py

@@ -65,10 +65,10 @@ class StarlingBot:
             response_text = tokenizer.decode(response[0], skip_special_tokens=True)
 #           response_text = response.split("<|assistant|>\n")[-1]
             return response_text       
-            finally:
-                del input_ids, attention_mask, output_ids
-                gc.collect()
-                torch.cuda.empty_cache()
+        finally:
+            del input_ids, attention_mask, output_ids
+            gc.collect()
+            torch.cuda.empty_cache()
 
 starling_bot = StarlingBot()