230503ac56636d91d53d728c247c6ea3d096eeb9,tf_agents/environments/tf_py_environment_test.py,TFPYEnvironmentTest,testTwoStepsDependenceOnTheFirst,#TFPYEnvironmentTest#Any#,250

Before Change

    self.assertEqual([2], time_step.observation)
    self.assertEqual(1., time_step.reward)
    self.assertEqual(0., time_step.discount)
    self.assertEqual(np.array([0]), step_data)
    self.assertEqual([1, 2], py_env.actions_taken)

  @parameterized.parameters({"batch_py_env": True}, {"batch_py_env": False})
  def testFirstObservationIsPreservedAfterTwoSteps(self, batch_py_env):

After Change

    else:
      tf_env = tf_py_environment.TFPyEnvironment(py_env)
    time_step = tf_env.current_time_step()
    with tf.control_dependencies([time_step.step_type]):
      action = tf.constant([1])
    time_step = tf_env.step(action)
    with tf.control_dependencies([time_step.step_type]):
      action = tf.constant([2])
    time_step = self.evaluate(tf_env.step(action))

Instances