def test_unflattened_input(self):
env = GymEnv(DummyBoxEnv(obs_dim=(2, 2)))
cmb = ContinuousMLPBaseline(env_spec=env.spec)
env.reset()
es = env.step(1)
obs, rewards = es.observation, es.reward
train_paths = [{'observations': [obs], 'returns': [rewards]}]
cmb.fit(train_paths)
paths = {'observations': [obs]}
prediction = cmb.predict(paths)
assert np.allclose(0., prediction)
def test_fit(self, obs_dim):
box_env = GarageEnv(DummyBoxEnv(obs_dim=obs_dim))
with mock.patch(('garage.tf.baselines.'
'continuous_mlp_baseline.'
'ContinuousMLPRegressor'),
new=SimpleMLPRegressor):
cmb = ContinuousMLPBaseline(env_spec=box_env.spec)
paths = [{
'observations': [np.full(obs_dim, 1)],
'returns': [1]
}, {
'observations': [np.full(obs_dim, 2)],
'returns': [2]
}]
cmb.fit(paths)
obs = {'observations': [np.full(obs_dim, 1), np.full(obs_dim, 2)]}
prediction = cmb.predict(obs)
assert np.array_equal(prediction, [1, 2])
def test_fit_unnormalized(self):
box_env_spec = GarageEnv(DummyBoxEnv(obs_dim=(2, ))).spec
cmb = ContinuousMLPBaseline(env_spec=box_env_spec,
normalize_inputs=False)
train_paths, _, paths, expected = get_train_test_data()
for _ in range(20):
cmb.fit(train_paths)
prediction = cmb.predict(paths)
assert np.allclose(prediction, expected, rtol=0, atol=0.1)
x_mean = self.sess.run(cmb._x_mean)
x_mean_expected = np.zeros_like(x_mean)
x_std = self.sess.run(cmb._x_std)
x_std_expected = np.ones_like(x_std)
assert np.allclose(x_mean, x_mean_expected)
assert np.allclose(x_std, x_std_expected)
def test_fit_normalized(self):
box_env_spec = GarageEnv(DummyBoxEnv(obs_dim=(2, ))).spec
cmb = ContinuousMLPBaseline(env_spec=box_env_spec)
train_paths, observations, paths, expected = get_train_test_data()
for _ in range(20):
cmb.fit(train_paths)
prediction = cmb.predict(paths)
assert np.allclose(prediction, expected, rtol=0, atol=0.1)
x_mean = self.sess.run(cmb._x_mean)
x_mean_expected = np.mean(observations, axis=0, keepdims=True)
x_std = self.sess.run(cmb._x_std)
x_std_expected = np.std(observations, axis=0, keepdims=True)
assert np.allclose(x_mean, x_mean_expected)
assert np.allclose(x_std, x_std_expected)