(s, noise_scale)
| 183 | sess.run(target_init) |
| 184 | |
| 185 | def get_action(s, noise_scale): |
| 186 | a = sess.run(mu, feed_dict={X: s.reshape(1,-1)})[0] |
| 187 | a += noise_scale * np.random.randn(num_actions) |
| 188 | return np.clip(a, -action_max, action_max) |
| 189 | |
| 190 | test_returns = [] |
| 191 | def test_agent(num_episodes=5): |