5bb63d23a7722d060f7df345a9d2b51dc98c035a,ch09/02_cartpole_reinforce.py,,,#,34

Before Change

        step_rewards = step_rewards[-1000:]

        baseline = np.mean(step_rewards)
        writer.add_scalar("baseline", baseline, step_idx)
        batch_states.append(exp.state)
        batch_actions.append(int(exp.action))
        batch_scales.append(exp.reward - baseline)

        // handle new rewards
        new_rewards = exp_source.pop_total_rewards()

After Change

    batch_episodes = 0
    batch_states, batch_actions, batch_qvals = [], [], []
    cur_states, cur_actions, cur_rewards = [], [], []

    for step_idx, exp in enumerate(exp_source):
        cur_states.append(exp.state)

Instances