Experience Replay Buffer save environment transition in a continuous RAM for high performance training we save trajectory in order and save state and other (action, reward, mask, ...) separately. `int max_len` the maximum capacity of ReplayBuffer. First In First Out
(self, max_len, state_dim, action_dim, if_discrete, if_on_policy, if_per_or_gae)
source not stored for this graph (policy: none)
nothing calls this directly
no test coverage detected