diff --git a/modelscope/models/nlp/T5/text2text_generation.py b/modelscope/models/nlp/T5/text2text_generation.py
index 0b695589..bead9e25 100644
--- a/modelscope/models/nlp/T5/text2text_generation.py
+++ b/modelscope/models/nlp/T5/text2text_generation.py
@@ -57,7 +57,7 @@ class T5ForConditionalGeneration(T5PreTrainedModel):
         r'decoder\.block\.0\.layer\.1\.EncDecAttention\.relative_attention_bias\.weight',
     ]
 
-    def __init__(self, config: T5Config):
+    def __init__(self, config: T5Config, **kwargs):
         super().__init__(config)
         self.model_dim = config.d_model
 
diff --git a/modelscope/models/nlp/bert/document_segmentation.py b/modelscope/models/nlp/bert/document_segmentation.py
index 36c39f43..0f2f2880 100644
--- a/modelscope/models/nlp/bert/document_segmentation.py
+++ b/modelscope/models/nlp/bert/document_segmentation.py
@@ -24,7 +24,7 @@ class BertForDocumentSegmentation(BertPreTrainedModel):
 
     _keys_to_ignore_on_load_unexpected = [r'pooler']
 
-    def __init__(self, config):
+    def __init__(self, config, **kwargs):
         super().__init__(config)
         self.num_labels = config.num_labels
         self.sentence_pooler_type = None
diff --git a/modelscope/models/nlp/bert/sentence_embedding.py b/modelscope/models/nlp/bert/sentence_embedding.py
index f4c2620e..18cecd3c 100644
--- a/modelscope/models/nlp/bert/sentence_embedding.py
+++ b/modelscope/models/nlp/bert/sentence_embedding.py
@@ -11,7 +11,7 @@ from .backbone import BertModel, BertPreTrainedModel
 @MODELS.register_module(Tasks.sentence_embedding, module_name=Models.bert)
 class BertForSentenceEmbedding(BertPreTrainedModel):
 
-    def __init__(self, config):
+    def __init__(self, config, **kwargs):
         super().__init__(config)
         self.config = config
         setattr(self, self.base_model_prefix,
diff --git a/modelscope/models/nlp/bert/text_classification.py b/modelscope/models/nlp/bert/text_classification.py
index 32aab7b2..df227064 100644
--- a/modelscope/models/nlp/bert/text_classification.py
+++ b/modelscope/models/nlp/bert/text_classification.py
@@ -66,7 +66,7 @@ class BertForSequenceClassification(BertPreTrainedModel):
             weights.
     """
 
-    def __init__(self, config):
+    def __init__(self, config, **kwargs):
         super().__init__(config)
         self.num_labels = config.num_labels
         self.config = config
diff --git a/modelscope/models/nlp/ponet/document_segmentation.py b/modelscope/models/nlp/ponet/document_segmentation.py
index 5e933491..e2cb0812 100644
--- a/modelscope/models/nlp/ponet/document_segmentation.py
+++ b/modelscope/models/nlp/ponet/document_segmentation.py
@@ -25,7 +25,7 @@ __all__ = ['PoNetForDocumentSegmentation']
 class PoNetForDocumentSegmentation(PoNetPreTrainedModel):
     _keys_to_ignore_on_load_unexpected = [r'pooler']
 
-    def __init__(self, config):
+    def __init__(self, config, **kwargs):
         super().__init__(config)
         self.num_labels = config.num_labels