Fix mask token handling (#14364)

* Fix mask token handling

* Revert "Fix mask token handling"

This reverts commit daaa3f5291b1f71e5bc3604ca281c000000c4648.

* Fix FNet mask token tokenization

src/transformers/models/fnet/tokenization_fnet.py

@@ -113,8 +113,13 @@ class FNetTokenizer(PreTrainedTokenizer):
         sp_model_kwargs: Optional[Dict[str, Any]] = None,
         **kwargs
     ) -> None:
-        # Mask token behave like a normal word, i.e. include the space before it
-        mask_token = AddedToken(mask_token, lstrip=True, rstrip=False) if isinstance(mask_token, str) else mask_token
+        # Mask token behave like a normal word, i.e. include the space before it and
+        # is included in the raw text, there should be a match in a non-normalized sentence.
+        mask_token = (
+            AddedToken(mask_token, lstrip=True, rstrip=False, normalized=False)
+            if isinstance(mask_token, str)
+            else mask_token
+        )
 
         self.sp_model_kwargs = {} if sp_model_kwargs is None else sp_model_kwargs
 

src/transformers/models/fnet/tokenization_fnet_fast.py

@@ -107,8 +107,13 @@ class FNetTokenizerFast(PreTrainedTokenizerFast):
         mask_token="[MASK]",
         **kwargs
     ):
-        # Mask token behave like a normal word, i.e. include the space before it
-        mask_token = AddedToken(mask_token, lstrip=True, rstrip=False) if isinstance(mask_token, str) else mask_token
+        # Mask token behave like a normal word, i.e. include the space before it and
+        # is included in the raw text, there should be a match in a non-normalized sentence.
+        mask_token = (
+            AddedToken(mask_token, lstrip=True, rstrip=False, normalized=False)
+            if isinstance(mask_token, str)
+            else mask_token
+        )
 
         super().__init__(
             vocab_file,