Module tests.moog.env_wrappers.test_gym_wrapper

Tests for moog/env_wrappers/gym_wrapper.py.

To run this test, navigate to this directory and run

$ pytest test_gym_wrapper.py --capture=tee-sys

Note: The –capture=tee-sys routes print statements to stdout, which is useful for debugging.

Alternatively, to run this test and any others, navigate to any parent directory and simply run

$ pytest --capture=tee-sys

This will run all test_* files in children directories.

Classes

class TestGymWrapper

Expand source code

class TestGymWrapper():

    def testJoystickActions(self):
        def _state_initializer():
            agent = sprite.Sprite(x=0.5, y=0.5, scale=0.1, c0=128)
            return collections.OrderedDict([('agent', [agent])])

        max_episode_length = 5
        task = tasks.Reset(
            lambda _: True, steps_after_condition=max_episode_length - 1)

        env = environment.Environment(
            state_initializer=_state_initializer,
            physics=physics_lib.Physics(),
            task=task,
            action_space=action_spaces.Joystick(),
            observers={'image': observers.PILRenderer(image_size=(64, 64))},
        )
        
        gym_env = gym_wrapper.GymWrapper(env)

        assert (
            gym_env.observation_space ==
            spaces.Dict({
                'image': spaces.Box(0, 255, shape=(64, 64, 3), dtype=np.uint8)
            })
        )
        assert (
            gym_env.action_space ==
            spaces.Box(-1., 1., shape=(2,), dtype=np.float32)
        )

        for _ in range(3):
            gym_env.reset()
            for _ in range(max_episode_length - 1):
                action = gym_env.action_space.sample()
                obs, reward, done, _ = gym_env.step(action)
                assert (obs['image'].dtype == np.uint8)
                assert not done
                assert (reward == 0.)
            action = gym_env.action_space.sample()
            _, _, done, _ = gym_env.step(action)
            assert done
            _, _, done, _ = gym_env.step(action)
            assert not done

    def testGridActions(self):
        def _state_initializer():
            agent = sprite.Sprite(x=0.5, y=0.5, scale=0.1, c0=128)
            return collections.OrderedDict([('agent', [agent])])
        
        max_episode_length = 5
        task = tasks.Reset(
            lambda _: True, steps_after_condition=max_episode_length - 1)
        env = environment.Environment(
            state_initializer=_state_initializer,
            physics=physics_lib.Physics(),
            task=task,
            action_space=action_spaces.Grid(),
            observers={'image': observers.PILRenderer(image_size=(64, 64))})
        gym_env = gym_wrapper.GymWrapper(env)

        assert (
            gym_env.observation_space ==
            spaces.Dict({
                'image': spaces.Box(0, 255, shape=(64, 64, 3), dtype=np.uint8)
            })
        )
        assert (gym_env.action_space == spaces.Discrete(5))

        for _ in range(3):
            gym_env.reset()
            for _ in range(max_episode_length - 1):
                action = gym_env.action_space.sample()
                obs, reward, done, _ = gym_env.step(action)
                assert (obs['image'].dtype == np.uint8)
                assert not done
                assert reward == 0.
            action = gym_env.action_space.sample()
            _, _, done, _ = gym_env.step(action)
            assert done
            _, _, done, _ = gym_env.step(action)
            assert not done

Methods

def testGridActions(self)

Expand source code

def testGridActions(self):
    def _state_initializer():
        agent = sprite.Sprite(x=0.5, y=0.5, scale=0.1, c0=128)
        return collections.OrderedDict([('agent', [agent])])
    
    max_episode_length = 5
    task = tasks.Reset(
        lambda _: True, steps_after_condition=max_episode_length - 1)
    env = environment.Environment(
        state_initializer=_state_initializer,
        physics=physics_lib.Physics(),
        task=task,
        action_space=action_spaces.Grid(),
        observers={'image': observers.PILRenderer(image_size=(64, 64))})
    gym_env = gym_wrapper.GymWrapper(env)

    assert (
        gym_env.observation_space ==
        spaces.Dict({
            'image': spaces.Box(0, 255, shape=(64, 64, 3), dtype=np.uint8)
        })
    )
    assert (gym_env.action_space == spaces.Discrete(5))

    for _ in range(3):
        gym_env.reset()
        for _ in range(max_episode_length - 1):
            action = gym_env.action_space.sample()
            obs, reward, done, _ = gym_env.step(action)
            assert (obs['image'].dtype == np.uint8)
            assert not done
            assert reward == 0.
        action = gym_env.action_space.sample()
        _, _, done, _ = gym_env.step(action)
        assert done
        _, _, done, _ = gym_env.step(action)
        assert not done

def testJoystickActions(self)

Expand source code

def testJoystickActions(self):
    def _state_initializer():
        agent = sprite.Sprite(x=0.5, y=0.5, scale=0.1, c0=128)
        return collections.OrderedDict([('agent', [agent])])

    max_episode_length = 5
    task = tasks.Reset(
        lambda _: True, steps_after_condition=max_episode_length - 1)

    env = environment.Environment(
        state_initializer=_state_initializer,
        physics=physics_lib.Physics(),
        task=task,
        action_space=action_spaces.Joystick(),
        observers={'image': observers.PILRenderer(image_size=(64, 64))},
    )
    
    gym_env = gym_wrapper.GymWrapper(env)

    assert (
        gym_env.observation_space ==
        spaces.Dict({
            'image': spaces.Box(0, 255, shape=(64, 64, 3), dtype=np.uint8)
        })
    )
    assert (
        gym_env.action_space ==
        spaces.Box(-1., 1., shape=(2,), dtype=np.float32)
    )

    for _ in range(3):
        gym_env.reset()
        for _ in range(max_episode_length - 1):
            action = gym_env.action_space.sample()
            obs, reward, done, _ = gym_env.step(action)
            assert (obs['image'].dtype == np.uint8)
            assert not done
            assert (reward == 0.)
        action = gym_env.action_space.sample()
        _, _, done, _ = gym_env.step(action)
        assert done
        _, _, done, _ = gym_env.step(action)
        assert not done