694b2e53a79a020db8d48bb2ced4ab80d0cb962b,tf_agents/drivers/dynamic_episode_driver_test.py,DynamicEpisodeDriverTest,testOneStepUpdatesObservers,#DynamicEpisodeDriverTest#,73

Before Change

    self.assertEqual(policy_state, [3])

  def testOneStepUpdatesObservers(self):
    if tf.executing_eagerly():
      self.skipTest("b/123880410")
    env = tf_py_environment.TFPyEnvironment(
        driver_test_utils.PyEnvironmentMock())
    policy = driver_test_utils.TFPolicyMock(
        env.time_step_spec(), env.action_spec())

After Change

    self.assertEqual(self.evaluate(num_episodes_observer.num_episodes), 5)
    // Two steps per episode.
    self.assertEqual(self.evaluate(num_steps_observer.num_steps), 10)
    self.assertEqual(self.evaluate(num_steps_transition_observer.num_steps), 10)

  def testMultiStepUpdatesObservers(self):
    env = tf_py_environment.TFPyEnvironment(
        driver_test_utils.PyEnvironmentMock())

Instances