Fix idx2sym not loaded from pretrained vocab file in Transformer XL (#27589)

* Load idx2sym from pretrained vocab file in Transformer XL

When loading vocab file from a pretrained tokenizer for Transformer XL,
although the pickled vocabulary file contains a idx2sym key, it isn't
loaded, because it is discarded as the empty list already exists as
an attribute.

Solution is to explicitly take it into account, just like for sym2idx.

* ran make style

tests/models/transfo_xl/test_tokenization_transfo_xl.py

@@ -15,7 +15,9 @@
 
 
 import os
+import pickle
 import unittest
+from collections import Counter, OrderedDict
 
 from transformers.models.transfo_xl.tokenization_transfo_xl import VOCAB_FILES_NAMES, TransfoXLTokenizer
 
@@ -47,6 +49,25 @@ class TransfoXLTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
         with open(self.vocab_file, "w", encoding="utf-8") as vocab_writer:
             vocab_writer.write("".join([x + "\n" for x in vocab_tokens]))
 
+        saved_dict = {
+            "eos_idx": 0,
+            "min_freq": 0,
+            "vocab_file": None,
+            "counter": Counter(["welcome home"]),
+            "sym2idx": OrderedDict([("<eos>", 0), ("welcome", 1), ("home", 2)]),
+            "delimiter": None,
+            "idx2sym": ["<eos>", "welcome", "home"],
+            "max_size": None,
+            "lower_case": False,
+            "special": ["<eos>"],
+        }
+        self.pretrained_vocab_file = os.path.join(
+            self.tmpdirname, "mock_folder", VOCAB_FILES_NAMES["pretrained_vocab_file"]
+        )
+        os.makedirs(os.path.dirname(self.pretrained_vocab_file), exist_ok=True)
+        with open(self.pretrained_vocab_file, "wb") as f:
+            pickle.dump(saved_dict, f)
+
     def get_tokenizer(self, **kwargs):
         kwargs["lower_case"] = True
         return TransfoXLTokenizer.from_pretrained(self.tmpdirname, **kwargs)
@@ -128,3 +149,8 @@ class TransfoXLTokenizationTest(TokenizerTesterMixin, unittest.TestCase):
         # Check that token is moved to specified id
         self.assertEqual(tokenizer.encode("new1"), [1])
         self.assertEqual(tokenizer.decode([1]), "new1")
+
+    def test_from_pretrained_vocab_file(self):
+        tokenizer = TransfoXLTokenizer.from_pretrained(os.path.join(self.tmpdirname, "mock_folder"))
+        sentence = "welcome home"
+        self.assertEqual(tokenizer.decode(tokenizer.encode(sentence)), sentence)