Update README.md

README.md

@@ -79,9 +79,8 @@ thinking_budget = 2048
 
 # The implementation of thinking budget involves two-phase generation,
 # which is incompatible with the official transformers TextIteratorStreamer.
-# Hence we modified the streaming class. Could comment this part out if
-# not using thinking budget. See the commented lines below that involve
-# "streamer" for usage.
+# MUST use this new class for streaming whether thinking budget is used
+# or not. See the commented lines below that involve "streamer" for usage.
 from transformers import TextIteratorStreamer
 class MyTextIteratorStreamer(TextIteratorStreamer):
     def manual_end(self):