What code should be added so that the word embedding and confusion matrix at each step could be visualized in tensorboard?