Is padding needed for prediction in LSTM keras

Question

I am trying to do sentence generator using 50D word embedding. If my training sentence is "hello my name is abc" here max words is 5. So my first training x is