👽️ update api

Signed-off-by: Peter <[email protected]>

converse.py

@@ -162,7 +162,6 @@ def gen_response(
         repetition_penalty=3.5,
         length_penalty=length_penalty,
         clean_up_tokenization_spaces=True,
-        skip_special_tokens=True,
         remove_invalid_values=True,
         **kwargs,
     )  # the likely better beam-less method

grammar_improve.py

@@ -501,7 +501,7 @@ def correct_grammar(
         temperature=temp,
     )
 
-    tgt_text = tokenizer.batch_decode(translated, skip_special_tokens=True)
+    tgt_text = tokenizer.batch_decode(translated)
     rt_min = (time.perf_counter() - st) / 60
     print(f"\n\ncorrected in {rt_min} minutes")