docs: pre-commit style

data_juicer/utils/constant.py

@@ -1,5 +1,6 @@
 DEFAULT_PREFIX = '__dj__'
 
+
 class Fields(object):
     stats = DEFAULT_PREFIX + 'stats__'
     meta = DEFAULT_PREFIX + 'meta__'

data_juicer/utils/mm_utils.py

@@ -1,8 +1,8 @@
-
 from datasets import Image
 
 from data_juicer.utils.constant import DEFAULT_PREFIX
 
+
 # A class to keep special tokens for multimodal information in the texts
 # The tokens in this class can be updated by corresponding arguments in config
 class SpecialTokens(object):
@@ -12,9 +12,11 @@ class SpecialTokens(object):
     # others
     eoc = f'<|{DEFAULT_PREFIX}eoc|>'
 
+
 def load_images(paths):
     return [load_image(path) for path in paths]
 
+
 def load_image(path):
     img_feature = Image()
     img = img_feature.decode_example(img_feature.encode_example(path))

data_juicer/utils/model_utils.py

@@ -169,12 +169,13 @@ def prepare_huggingface_tokenizer(tokenizer_name):
                                               trust_remote_code=True)
     return tokenizer
 
+
 def prepare_huggingface_clip(clip_name):
     """
     Prepare and load a clip and processor from HuggingFace.
 
-    :param tokenizer_name: input tokenizer name
-    :return: a tokenizer instance.
+    :param clip_name: input clip name
+    :return: a pair of clip instance and processor instance.
     """
     from transformers import CLIPProcessor, CLIPModel