Values of 17 from policy network although it should be between -1-1

Hello,

my policy network give values between 0-17 although it should be between -1 - 1.
The network consists of 4 linear layers and relu activations. I am taking the raw data of the network. How can i let the network predict values between -1 and 1 and not between -inf to +inf ?

Do i have to add layernom between the layers?