fixup

src/transformers/models/gemma/configuration_gemma.py

@@ -15,7 +15,6 @@
 # limitations under the License.
 
 
-
 from ...utils import (
     is_flash_attn_2_available,
 )

src/transformers/models/gemma/modeling_gemma.py

@@ -812,7 +812,6 @@ _CONFIG_FOR_DOC = "GemmaConfig"
     GEMMA_START_DOCSTRING,
 )
 class GemmaModel(GemmaPreTrainedModel):
-
     def __init__(self, config: GemmaConfig):
         super().__init__(config)
         self.padding_idx = config.pad_token_id
@@ -1032,7 +1031,6 @@ class GemmaModel(GemmaPreTrainedModel):
 
 
 class GemmaForCausalLM(GemmaPreTrainedModel):
-
     def __init__(self, config):
         super().__init__(config)
         self.model = GemmaModel(config)

utils/diff_model_converter.py

@@ -11,8 +11,10 @@ from libcst.metadata import MetadataWrapper, ParentNodeProvider, PositionProvide
 
 from transformers import logging
 
+
 logger = logging.get_logger(__name__)
 
+
 def get_module_source_from_name(module_name: str) -> str:
     spec = importlib.util.find_spec(module_name)
     if spec is None or spec.origin is None:
@@ -296,7 +298,7 @@ class DiffConverterTransformer(CSTTransformer):
         return node
 
     def visit_ImportFrom(self, node: cst.ImportFrom) -> None:
-        """ When visiting imports from `transformers.models.xxx` we need to:
+        """When visiting imports from `transformers.models.xxx` we need to:
         1. Get the original source code
         2. Parse it into an AST Tree
         3. Add this import to `self.transformers_imports` as visited to not parse it twice
@@ -358,7 +360,9 @@ class DiffConverterTransformer(CSTTransformer):
             if super_file_name not in self.visited_module:  # only extract classes once
                 class_finder = find_classes_in_file(self.transformers_imports[super_file_name], old_name, new_name)
                 self.visited_module[super_file_name] = class_finder
-                self.node_mapping[class_name] = class_finder.classes[class_name] # here we get the new node form the parent class
+                self.node_mapping[class_name] = class_finder.classes[
+                    class_name
+                ]  # here we get the new node form the parent class
             else:  # we are re-using the previously parsed data
                 class_finder = self.visited_module[super_file_name]
 
@@ -435,6 +439,7 @@ def convert_file(diff_file, cst_transformers=None):
     # TODO optimize by re-using the class_finder
     return cst_transformers
 
+
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument(