0bb6982bd06bf21de58e61f021626ade1c9b6101,ch14/04_train_ddpg.py,,,#,46

Before Change

                actor_loss_v = -net.critic(states_v, cur_actions_v)
                actor_loss_v = actor_loss_v.mean()
                actor_loss_v.backward()
                net.n_critic.zero_grad()
                optimizer.step()
                tb_tracker.track("loss_actor", actor_loss_v, frame_idx)

                tgt_net.alpha_sync(alpha=1-1e-3)

After Change

    test_env = gym.make(ENV_ID)

    act_net = model.DDPGActor(env.observation_space.shape[0], env.action_space.shape[0])
    crt_net = model.DDPGCritic(env.observation_space.shape[0], env.action_space.shape[0])
    if args.cuda:
        act_net.cuda()
        crt_net.cuda()

Instances