diff --git a/modelscope/utils/hf_util/patcher.py b/modelscope/utils/hf_util/patcher.py
index 0174cf93..418f65f3 100644
--- a/modelscope/utils/hf_util/patcher.py
+++ b/modelscope/utils/hf_util/patcher.py
@@ -243,13 +243,12 @@ def _patch_pretrained_class(all_imported_modules, wrap=False):
                     from transformers.tokenization_utils import SPECIAL_TOKENS_MAP_FILE
                     from transformers.tokenization_utils import TOKENIZER_CONFIG_FILE
                     from transformers.tokenization_utils_base import FULL_TOKENIZER_FILE
-                    from transformers.tokenization_utils_base import CHAT_TEMPLATE_FILE
                     extra_allow_file_pattern = list(
                         (cls.vocab_files_names.values()) if cls is not None
                         and hasattr(cls, 'vocab_files_names') else []) + [
                             ADDED_TOKENS_FILE, SPECIAL_TOKENS_MAP_FILE,
                             TOKENIZER_CONFIG_FILE, FULL_TOKENIZER_FILE,
-                            CHAT_TEMPLATE_FILE, r'*.py'
+                            'chat_template.jinja', r'*.py'
                         ]  # noqa
                 elif 'Processor' in module_class.__name__:
                     from transformers.utils import FEATURE_EXTRACTOR_NAME