Applied more comments

Author: AlexKoff88

optimum/intel/openvino/modeling_base.py

@@ -164,6 +164,7 @@ def _from_pretrained(
         from_onnx: bool = False,
         local_files_only: bool = False,
         load_in_8bit: bool = False,
+        load_in_4bit: bool = False,
         **kwargs,
     ):
         """
@@ -193,8 +194,11 @@ def _from_pretrained(
                 Whether or not to only look at local files (i.e., do not try to download the model).
             load_in_8bit (`bool`, *optional*, defaults to `False`):
                 Whether or not to apply 8-bit weight quantization.
+            load_in_4bit (`bool`, *optional*, defaults to `False`):
+                Whether or not to apply 4-bit weight quantization.
         """
-
+        if load_in_4bit:
+            raise ValueError("load_in_4bit is available for OVModelForCausalLM only.")
         model_path = Path(model_id)
         default_file_name = ONNX_WEIGHTS_NAME if from_onnx else OV_XML_FILE_NAME
         file_name = file_name or default_file_name
@@ -262,6 +266,7 @@ def _from_transformers(
         task: Optional[str] = None,
         trust_remote_code: bool = False,
         load_in_8bit: Optional[bool] = None,
+        load_in_4bit: Optional[bool] = None,
         **kwargs,
     ):
         """

optimum/intel/openvino/modeling_decoder.py

@@ -580,6 +580,8 @@ def _from_pretrained(
             local_files_only=local_files_only,
         )
 
+        if load_in_8bit and load_in_4bit:
+            raise ValueError("Either load_in_8bit or load_in_4bit should be set to True.")
         model = cls.load_model(model_cache_path, load_in_8bit=False if load_in_4bit else load_in_8bit)
 
         model_type = config.model_type.replace("_", "-")