Nit-added-tokens (#26538)

* fix stripping

* nits

* fix another test

* styling

* fix?

* update

* revert bad merge

* found the bug

* YES SIR

* is that change really required?

* make fast even faster

* re order functions

src/transformers/tokenization_utils.py

@@ -367,6 +367,25 @@ class PreTrainedTokenizer(PreTrainedTokenizerBase):
 
         self._decode_use_source_tokenizer = False
 
+    @property
+    def is_fast(self) -> bool:
+        return False
+
+    @property
+    def vocab_size(self) -> int:
+        """
+        `int`: Size of the base vocabulary (without the added tokens).
+        """
+        raise NotImplementedError
+
+    @property
+    def added_tokens_encoder(self) -> Dict[str, int]:
+        """
+        Returns the sorted mapping from string to index. The added tokens encoder is cached for performance
+        optimisation in `self._added_tokens_encoder` for the slow tokenizers.
+        """
+        return {k.content: v for v, k in sorted(self._added_tokens_decoder.items(), key=lambda item: item[0])}
+
     @property
     def added_tokens_decoder(self) -> Dict[int, AddedToken]:
         """
@@ -389,17 +408,6 @@ class PreTrainedTokenizer(PreTrainedTokenizerBase):
             self._added_tokens_decoder[index] = AddedToken(token) if isinstance(token, str) else token
             self._added_tokens_encoder[str(token)] = index
 
-    @property
-    def is_fast(self) -> bool:
-        return False
-
-    @property
-    def vocab_size(self) -> int:
-        """
-        `int`: Size of the base vocabulary (without the added tokens).
-        """
-        raise NotImplementedError
-
     def get_added_vocab(self) -> Dict[str, int]:
         """
         Returns the added tokens in the vocabulary as a dictionary of token to index. Results might be different from

src/transformers/tokenization_utils_base.py

@@ -846,15 +846,26 @@ class SpecialTokensMixin:
         # We directly set the hidden value to allow initialization with special tokens
         # which are not yet in the vocabulary. Necessary for serialization/de-serialization
         # TODO clean this up at some point (probably by switching to fast tokenizers)
+
         for key, value in kwargs.items():
             if value is None:
                 continue
             if key in self.SPECIAL_TOKENS_ATTRIBUTES:
                 if key == "additional_special_tokens":
+                    # TODO THIS IS NASTY! Will always reset tokens to default rstrip and lstrip because self.set_attr on strings
+                    # will not check the addedtokens decoder. WILL FIX TOMORROW
                     assert isinstance(value, (list, tuple)), f"Value {value} is not a list or tuple"
                     assert all(
                         isinstance(t, (str, AddedToken)) for t in value
                     ), "One of the tokens is not a string or an AddedToken"
+                    if hasattr(self, "added_tokens_encoder"):
+                        extended_token = []
+                        for token in value:
+                            if isinstance(token, str) and str(token) in self.added_tokens_encoder:
+                                extended_token.append(self.added_tokens_decoder[self.added_tokens_encoder[str(token)]])
+                            else:
+                                extended_token.append(token)
+                        value = extended_token
                     setattr(self, key, value)
                 elif isinstance(value, (str)):
                     value = AddedToken(value, normalized=False, special=True)
@@ -1674,14 +1685,6 @@ class PreTrainedTokenizerBase(SpecialTokensMixin, PushToHubMixin):
         """Sets processor class as an attribute."""
         self._processor_class = processor_class
 
-    @property
-    def added_tokens_encoder(self) -> Dict[str, int]:
-        """
-        Returns the sorted mapping from string to index. The added tokens encoder is cached for performance
-        optimisation in `self._added_tokens_encoder` for the slow tokenizers.
-        """
-        return {k.content: v for v, k in sorted(self._added_tokens_decoder.items(), key=lambda item: item[0])}
-
     @property
     def added_tokens_decoder(self) -> Dict[int, AddedToken]:
         raise NotImplementedError()
@@ -2196,9 +2199,13 @@ class PreTrainedTokenizerBase(SpecialTokensMixin, PushToHubMixin):
             for idx, token in init_kwargs["added_tokens_decoder"].items():
                 if isinstance(token, dict):
                     token = AddedToken(**token)
-
                 if isinstance(token, AddedToken):
                     added_tokens_decoder[int(idx)] = token
+                    if str(token) in additional_special_tokens:
+                        # at this point the token is in `additional_special_tokens` as an str, let's add the AddedToken info
+                        additional_special_tokens.remove(str(token))
+                    if token.special and token not in additional_special_tokens:
+                        additional_special_tokens.append(token)
                 else:
                     raise ValueError(
                         f"Found a {token.__class__} in the saved `added_tokens_decoder`, should be a dictionary."
@@ -2381,9 +2388,8 @@ class PreTrainedTokenizerBase(SpecialTokensMixin, PushToHubMixin):
 
         tokenizer_config = copy.deepcopy(self.init_kwargs)
 
-        # TODO: Ensure the modified attributes (those are also in the __init__ kwargs) will give identical tokenizers
-        # target_keys = self.init_kwargs.keys()
-        target_keys = ["model_max_length", "clean_up_tokenization_spaces", "additional_special_tokens"]
+        target_keys = list(self.init_kwargs.keys())
+        target_keys += ["model_max_length", "clean_up_tokenization_spaces", "additional_special_tokens"]
         for k in target_keys:
             if hasattr(self, k):
                 tokenizer_config[k] = getattr(self, k)

src/transformers/tokenization_utils_fast.py

@@ -185,6 +185,14 @@ class PreTrainedTokenizerFast(PreTrainedTokenizerBase):
     def vocab(self) -> Dict[str, int]:
         return self.get_vocab()
 
+    @property
+    def added_tokens_encoder(self) -> Dict[str, int]:
+        """
+        Returns the sorted mapping from string to index. The added tokens encoder is cached for performance
+        optimisation in `self._added_tokens_encoder` for the slow tokenizers.
+        """
+        return {k.content: v for v, k in sorted(self.added_tokens_decoder.items(), key=lambda item: item[0])}
+
     @property
     def added_tokens_decoder(self) -> Dict[int, AddedToken]:
         """
@@ -202,10 +210,7 @@ class PreTrainedTokenizerFast(PreTrainedTokenizerBase):
         Returns:
             `Dict[str, int]`: The added tokens.
         """
-        base_vocab = self._tokenizer.get_vocab(with_added_tokens=False)
-        full_vocab = self._tokenizer.get_vocab(with_added_tokens=True)
-        added_vocab = {tok: index for tok, index in full_vocab.items() if tok not in base_vocab}
-        return added_vocab
+        return {k.content: v for v, k in sorted(self.added_tokens_decoder.items(), key=lambda item: item[0])}
 
     def __len__(self) -> int:
         """