Pad coreference model input to 5 (#2933)

* Pad coreference model input to 5.

It has a 5-wide Conv1D layer.

* mypy

allennlp/pretrained.py

@@ -61,7 +61,10 @@ def neural_coreference_resolution_lee_2017() -> predictors.CorefPredictor:
         warnings.simplefilter(action="ignore", category=DeprecationWarning)
         model = PretrainedModel('https://s3-us-west-2.amazonaws.com/allennlp/models/coref-model-2018.02.05.tar.gz',
                                 'coreference-resolution')
-        return model.predictor() # type: ignore
+        predictor = model.predictor()
+        # pylint: disable=protected-access
+        predictor._dataset_reader._token_indexers['token_characters']._min_padding_length = 5  # type: ignore
+        return predictor  # type: ignore
 
 def named_entity_recognition_with_elmo_peters_2018() -> predictors.SentenceTaggerPredictor:
     with warnings.catch_warnings():