How to initialize weights/bias of RNN LSTM GRU?

[Rafael_Valle]

How about these solutions below. Are they viable?

init.xavier_normal(GRU)
# or
for p in layer.params():
    init.xavier_normal(p)