Update tokenization_kimi.py

Pass added_tokens_decoder to the base class so that the non-special added tokens will be saved during `save_pretrained`.

tokenization_kimi.py

@@ -151,6 +151,7 @@ class TikTokenTokenizer(PreTrainedTokenizer):
             unk_token=unk_token,
             pad_token=pad_token,
             additional_special_tokens=additional_special_tokens,
+            added_tokens_decoder=added_tokens_decoder,
             **kwargs,
         )
         self.all_special_ids_set = set(self.all_special_ids)