SentimentProcessor ought to be able to save the trainer now... will m…

…ake it easier to access the whole model state elsewhere

stanza/pipeline/sentiment_processor.py

@@ -55,6 +55,7 @@ def _set_up_model(self, config, pipeline, device):
         trainer = Trainer.load(filename=filename,
                                args=args,
                                foundation_cache=pipeline.foundation_cache)
+        self._trainer = trainer
         self._model = trainer.model
         self._model_type = self._model.config.model_type
         # batch size counted as words