1a89404a84393211c7b370077fafe536a8f44762,ch17/lib/common.py,,iterate_batches,#Any#Any#Any#,64

Before Change

        _, values_v = net(obs_v)
        values_last = values_v.squeeze().data.cpu().numpy()
        // prepare before rollouts calculation
        mb_dones = np.roll(mb_dones, -1, axis=1)
        mb_dones[:, -1] = cur_dones

        for e_idx, (rewards, dones, value) in enumerate(zip(mb_rewards, mb_dones, values_last)):
            rewards = rewards.tolist()

After Change

    mb_actions = np.zeros((NUM_ENVS, REWARD_STEPS), dtype=np.int32)

    while True:
        batch_dones = [[dones[-1]] for dones in batch_dones]
        done_rewards = []
        done_steps = []
        for n in range(REWARD_STEPS):
            obs_v = ptan.agent.default_states_preprocessor(obs)

Instances