I started learning RL quite recently and still playing around with different implementations. Your implementation of A2C does not seem to have any parallelized workers for collecting training samples which reduces the implementation to an Actor Critic that uses GAE.
Please correct me if I'm wrong.
I started learning RL quite recently and still playing around with different implementations. Your implementation of A2C does not seem to have any parallelized workers for collecting training samples which reduces the implementation to an Actor Critic that uses GAE.
Please correct me if I'm wrong.