Updating the TensorFlow models to work as expected with tokenizers v3.0.0 (#3684)

* Updating modeling tf files; adding tests

* Merge `encode_plus` and `batch_encode_plus`

tests/test_tokenization_distilbert.py

@@ -14,7 +14,7 @@
 # limitations under the License.
 
 
-from transformers.tokenization_distilbert import DistilBertTokenizer
+from transformers.tokenization_distilbert import DistilBertTokenizer, DistilBertTokenizerFast
 
 from .test_tokenization_bert import BertTokenizationTest
 from .utils import slow
@@ -27,6 +27,9 @@ class DistilBertTokenizationTest(BertTokenizationTest):
     def get_tokenizer(self, **kwargs):
         return DistilBertTokenizer.from_pretrained(self.tmpdirname, **kwargs)
 
+    def get_rust_tokenizer(self, **kwargs):
+        return DistilBertTokenizerFast.from_pretrained(self.tmpdirname, **kwargs)
+
     @slow
     def test_sequence_builders(self):
         tokenizer = DistilBertTokenizer.from_pretrained("distilbert-base-uncased")