diff --git a/src/transformers/tokenization_utils_base.py b/src/transformers/tokenization_utils_base.py
index bfb0015b894574..543e089997406a 100644
--- a/src/transformers/tokenization_utils_base.py
+++ b/src/transformers/tokenization_utils_base.py
@@ -2172,7 +2172,6 @@ def _from_pretrained(
 
         additional_special_tokens = init_kwargs.pop("additional_special_tokens", None) or []
         added_tokens_decoder = {}
-        # added_tokens_map = {}
         legacy_saved = "added_tokens_decoder" not in init_kwargs
         if not legacy_saved:
             for idx, token in init_kwargs["added_tokens_decoder"].items():
@@ -2180,14 +2179,10 @@ def _from_pretrained(
                     token = AddedToken(**token)
                 if isinstance(token, AddedToken):
                     added_tokens_decoder[int(idx)] = token
-                    # added_tokens_map[str(token)] = token
                 else:
                     raise ValueError(
                         f"Found a {token.__class__} in the saved `added_tokens_decoder`, should be a dictionary."
                     )
-            # for key in cls.SPECIAL_TOKENS_ATTRIBUTES:
-            #     if key in init_kwargs:
-            #         init_kwargs[key] = added_tokens_map[init_kwargs[key]]
         else:
             # begin legacy: read the added_tokens_file and update kwargs with special_tokens_map if modified
             if special_tokens_map_file is not None:
@@ -2200,22 +2195,17 @@ def _from_pretrained(
                             continue
                         if isinstance(value, dict):
                             value = AddedToken(**value)
-                            value.special = True
                             init_kwargs[key] = value
                         elif key == "additional_special_tokens" and isinstance(value, list):
                             for token in value:
-                                token = AddedToken(**token, special=True) if isinstance(token, dict) else token
+                                token = AddedToken(**token) if isinstance(token, dict) else token
                                 if token not in additional_special_tokens:
                                     additional_special_tokens.append(token)
                         else:
                             init_kwargs[key] = value
             all_special_strings = [str(token) for token in additional_special_tokens]
             # also add the other special tokens
-            all_special_strings += [
-                str(init_kwargs.get(key, ""))
-                for key in cls.SPECIAL_TOKENS_ATTRIBUTES
-                if init_kwargs.get(key, None) not in all_special_strings
-            ]
+            all_special_strings += [str(init_kwargs[key]) for key in cls.SPECIAL_TOKENS_ATTRIBUTES & init_kwargs.keys()]
             # slow -> slow|fast, legacy: convert the `"added_tokens.json"` file to `added_tokens_decoder`.
             if added_tokens_file is not None:
                 with open(added_tokens_file, encoding="utf-8") as added_tokens_handle:
diff --git a/tests/test_tokenization_common.py b/tests/test_tokenization_common.py
index 308963ae1bc73f..3eb884ae045dc3 100644
--- a/tests/test_tokenization_common.py
+++ b/tests/test_tokenization_common.py
@@ -4071,7 +4071,7 @@ def test_added_tokens_serialization(self):
                     if self.rust_tokenizer_class is not None:
                         # fast from pretrained ignore the lstrip rstrip
                         tokenizer_fast = self.rust_tokenizer_class.from_pretrained(tmp_dir_2)
-                        self.assertEquals(tokenizer_fast._eos_token, str(new_eos))
+                        self.assertEquals(tokenizer.added_tokens_decoder[tokenizer.eos_token_id], new_eos)
                         self.assertIn(new_eos, tokenizer_fast.added_tokens_decoder.values())
                         self.assertEqual(EXPECTED_ADDED_TOKENS_DECODER, tokenizer_fast.added_tokens_decoder)
                         with tempfile.TemporaryDirectory() as tmp_dir_3:
@@ -4079,12 +4079,12 @@ def test_added_tokens_serialization(self):
                             tokenizer_fast = self.rust_tokenizer_class.from_pretrained(
                                 pretrained_name, eos_token=new_eos, use_fast=True
                             )
-                            self.assertEquals(tokenizer._eos_token, new_eos)
+                            self.assertEquals(tokenizer.added_tokens_decoder[tokenizer.eos_token_id], new_eos)
                             self.assertIn(new_eos, tokenizer.added_tokens_decoder.values())
                             self.assertEqual(EXPECTED_ADDED_TOKENS_DECODER, tokenizer.added_tokens_decoder)
 
                             tokenizer = self.tokenizer_class.from_pretrained(tmp_dir_2)
-                            self.assertEquals(tokenizer._eos_token, new_eos)
+                            self.assertEquals(tokenizer.added_tokens_decoder[tokenizer.eos_token_id], new_eos)
                             self.assertIn(new_eos, tokenizer.added_tokens_decoder.values())
                             self.assertEqual(EXPECTED_ADDED_TOKENS_DECODER, tokenizer.added_tokens_decoder)