Fix conflicting key in init kwargs in PreTrainedTokenizerBase

src/transformers/tokenization_utils_base.py

@@ -1569,6 +1569,13 @@ class PreTrainedTokenizerBase(SpecialTokensMixin, PushToHubMixin):
     def __init__(self, **kwargs):
         # inputs and kwargs for saving and re-loading (see ``from_pretrained`` and ``save_pretrained``)
         self.init_inputs = ()
+        if "add_special_tokens" in kwargs:
+            kwargs["_add_special_tokens"] = kwargs.pop("add_special_tokens")
+
+        for key in kwargs:
+            if hasattr(self, key) and callable(getattr(self, key)):
+                raise AttributeError(f"{key} conflicts with the method {key} in {self.__class__.__name__}")
+
         self.init_kwargs = copy.deepcopy(kwargs)
         self.name_or_path = kwargs.pop("name_or_path", "")
         self._processor_class = kwargs.pop("processor_class", None)

tests/test_tokenization_common.py

@@ -4319,3 +4319,17 @@ def test_special_token_addition(self):
                         replace_additional_special_tokens=False,
                     )
                     self.assertEqual(tokenizer_2.additional_special_tokens, ["<other>", "<another>", "<tok>"])
+
+    def test_tokenizer_initialization_with_conflicting_key(self):
+        if self.test_rust_tokenizer:
+            tokenizer = self.get_rust_tokenizer(add_special_tokens=True)
+        else:
+            tokenizer = self.get_tokenizer(add_special_tokens=True)
+
+        with tempfile.TemporaryDirectory() as tmp_dir_1:
+            tokenizer.save_pretrained(tmp_dir_1)
+            loaded_tokenizer = tokenizer.from_pretrained(tmp_dir_1)
+            assert loaded_tokenizer.init_kwargs.get("_add_special_tokens")
+
+        with self.assertRaises(AttributeError, msg="conflicts with the method"):
+            self.get_tokenizer(get_vocab=True)