[Tokenizer] Fix decode output with space in decode_token (#9010)

* fix

* fix

paddlenlp/transformers/tokenizer_utils.py

@@ -1881,10 +1881,14 @@ def decode_token(
         """tokenizer decoding for the streaming generation use case. This method can be overrided for tokenizer that doesn't follow this API"""
         # The prefix text is necessary only to defeat cleanup algorithms in the decode
         # which decide to add a space or not depending on the surrounding ids.
-        prefix_text = self.decode(all_input_ids[prefix_offset:read_offset], skip_special_tokens=False)
-        new_text = self.decode(all_input_ids[prefix_offset:], skip_special_tokens=False)
+        prefix_text = self.decode(
+            all_input_ids[prefix_offset:read_offset], skip_special_tokens=False, clean_up_tokenization_spaces=False
+        )
+        new_text = self.decode(
+            all_input_ids[prefix_offset:], skip_special_tokens=False, clean_up_tokenization_spaces=False
+        )
 
-        if len(new_text) > len(prefix_text) and not new_text.endswith("�"):
+        if len(new_text) > len(prefix_text) and not prefix_text.endswith("�") and not new_text.endswith("�"):
             # utf-8 char at the end means it's a potential unfinished byte sequence
             # from byte fallback tokenization.
             # If it's in the middle, it's probably a real invalid id generated