Add _ignore_kwargs function to filter kwargs based on function signature

This new utility function filters keyword arguments to only those accepted by the specified function, logging any ignored keys. It is integrated into the model building process to ensure compatibility with model initialization.

Author: developer0hye

timm/models/_builder.py

@@ -1,6 +1,7 @@
 import dataclasses
 import logging
 import os
+import inspect
 from copy import deepcopy
 from pathlib import Path
 from typing import Any, Callable, Dict, List, Optional, Tuple, Type, TypeVar, Union
@@ -303,6 +304,19 @@ def _filter_kwargs(kwargs: Dict[str, Any], names: List[str]) -> None:
     for n in names:
         kwargs.pop(n, None)
 
+def _ignore_kwargs(func, kwargs):
+    """ Filter kwargs to those that func accepts.
+    """
+    sig = inspect.signature(func)
+    if any(p.kind == inspect.Parameter.VAR_KEYWORD for p in sig.parameters.values()):
+        return kwargs
+    filter_keys = [p.name for p in sig.parameters.values() if p.kind in (p.POSITIONAL_OR_KEYWORD, p.KEYWORD_ONLY)]
+    filtered_kwargs = {k: v for k, v in kwargs.items() if k in filter_keys}
+    ignored_keys = set(kwargs.keys()) - set(filtered_kwargs.keys())
+    if ignored_keys:
+        _logger.warning(
+            f'Ignored attempt to pass arguments ({", ".join(ignored_keys)}) to function {func}.')
+    return filtered_kwargs
 
 def _update_default_model_kwargs(pretrained_cfg, kwargs, kwargs_filter) -> None:
     """ Update the default_cfg and kwargs before passing to model
@@ -441,6 +455,7 @@ def build_model_with_cfg(
             feature_cfg['feature_cls'] = kwargs.pop('feature_cls')
 
     # Instantiate the model
+    kwargs = _ignore_kwargs(model_cls.__init__, kwargs)
     if model_cfg is None:
         model = model_cls(**kwargs)
     else: