fixed pre-commit errors

Author: pakerimus

llm_guard/input_scanners/anonymize_helpers/transformers_recognizer.py

@@ -139,8 +139,8 @@ def _load_pipeline(
         self.model.pipeline_kwargs["ignore_labels"] = self.ignore_labels
 
         transformers = cast("transformers", lazy_load_dep("transformers"))
-        self.pipeline = transformers.pipelines.pipeline(
-            "ner",
+        self.pipeline = transformers.pipelines.pipeline(  # type: ignore[arg-type]
+            "ner",  # type: ignore[arg-type]
             model=tf_model,
             tokenizer=tf_tokenizer,
             **self.model.pipeline_kwargs,

llm_guard/output_scanners/relevance.py

@@ -91,16 +91,20 @@ def __init__(
                 ),
             )
             assert model.onnx_path is not None
-            self._model = optimum_onnxruntime.ORTModelForFeatureExtraction.from_pretrained(
-                model.onnx_path,
-                export=False,
-                subfolder=model.onnx_subfolder,
-                file_name=model.onnx_filename,
-                revision=model.onnx_revision,
-                provider=(
+            onnx_kwargs = {
+                "export": False,
+                "subfolder": model.onnx_subfolder,
+                "file_name": model.onnx_filename,
+                "provider": (
                     "CUDAExecutionProvider" if device().type == "cuda" else "CPUExecutionProvider"
                 ),
                 **model.kwargs,
+            }
+            if model.onnx_revision is not None:
+                onnx_kwargs["revision"] = model.onnx_revision
+            self._model = optimum_onnxruntime.ORTModelForFeatureExtraction.from_pretrained(
+                model.onnx_path,
+                **onnx_kwargs,
             )
             LOGGER.debug("Initialized ONNX model", model=model, device=device())
         else:

llm_guard/util.py

@@ -198,7 +198,7 @@ def truncate_tokens_head_tail(tokens, max_length=512, head_length=128, tail_leng
 
 
 url_pattern = re.compile(
-    r"http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\\(\\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+"
+    r"http[s]?://(?:[a-zA-Z]|[0-9]|[$-_@.&+]|[!*\(\),]|(?:%[0-9a-fA-F][0-9a-fA-F]))+"
 )
 
 
@@ -225,7 +225,7 @@ def remove_markdown(text):
     clean_text = text
     for pattern in patterns:
         # Use substitution to preserve the text inside ** and *
-        if "([^\*]+)" in pattern:
+        if r"([^\*]+)" in pattern:
             clean_text = re.sub(pattern, r"\1", clean_text)
         else:
             clean_text = re.sub(pattern, "", clean_text)