Performance issue in the definition of build_caption_greedy_inference, model.py(P1)

[DLPerf]

Hello, I found a performance issue in the definition of build_caption_greedy_inference, model.py, tf.reshape(event_hidden_feats_tile, [-1, 2*self.options['rnn_size']]) will be created repeatedly during program execution, resulting in reduced efficiency. I think event_hidden_feats_tile and event_hidden_feats_reshape should be created before the loop.

Looking forward to your reply. Btw, I am very glad to create a PR to fix it if you are too busy.