This PR address the following logical bug in get_state. During get_state(complete=False) we should return the last state_size frames from agent.history. Previously we were returning the first agent.state_size frames.
At agent.optimize we get the first agent.state_size frames as state and last agent.state_size frames for next_state
This PR address the following logical bug in get_state. During
get_state(complete=False)
we should return the laststate_size
frames from agent.history. Previously we were returning the firstagent.state_size
frames.At
agent.optimize
we get the firstagent.state_size
frames as state and lastagent.state_size
frames fornext_state