How do you initialize a recurrent layer if you don't know the exact size of the input yet?

Web: https://www.reddit.com/r/reinforcementlearning/comments/uj6qcl/how_do_you_initialize_a_recurrent_layer_if_you/

Let's say I have a function that processes my observation in a way that makes it difficult to predict the size of the input for the subsequent recurrent layer. In the init method, how do I initialize such recurrent layer?

reinforcementlearning