How to train a keras lstm model more than one training data set but without letting it forget its previous training result?

Question

Normally with I got only ONE training data, I just do model.fit(x_train1, y_train1).

What should I do if I want to train the model with Two sets of train