4f2535df9cb702854ef892c0d2a92ef068636ce0,examples/reinforcement_learning/baselines/algorithms/sac/sac.py,,learn,#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#Any#,242

Before Change

        rewards = []
        sac_trainer.load_weights()

        while frame_idx < test_timesteps:
            state = env.reset()
            state = state.astype(np.float32)
            episode_reward = 0
            if frame_idx < 1:
                _ = sac_trainer.policy_net([state])  // need an extra call to make inside functions be able to use forward

            for step in range(max_steps):
                action = sac_trainer.policy_net.get_action(state, deterministic=DETERMINISTIC)
                next_state, reward, done, _ = env.step(action)
                next_state = next_state.astype(np.float32)
                env.render()
                done = 1 if done ==True else 0

                state = next_state
                episode_reward += reward
                frame_idx += 1

                // if frame_idx % 50 == 0:
                //     plot(frame_idx, rewards)

                if done:
                    break
            print("Episode: ", frame_idx / max_steps, "| Episode Reward: ", episode_reward)
            rewards.append(episode_reward)

After Change

                plot(rewards, Algorithm_name="SAC", Env_name=env_id)
                sac_trainer.save_weights()
            print("Episode: {}/{}  | Episode Reward: {:.4f}  | Running Time: {:.4f}"\
            .format(eps, train_episodes, episode_reward, time.time()-t0 ))
            rewards.append(episode_reward)
        sac_trainer.save_weights()

Instances