def env_contr():
#env = gym.make("CartPole-v0")#
env = simple_world_comm.env()
env = pad_observations(env)
env = pad_action_space(env)
markov_env = aec_to_markov(env)
venv = MarkovVectorEnv(markov_env)
return venv
def env_fn():
#env = gym.make("CartPole-v0")#
env = simple_world_comm_v0.env()
# print(env.action_spaces.values())
# exit(0)
env = pad_observations(env)
env = pad_action_space(env)
markov_env = aec_to_markov(env)
venv = MarkovVectorEnv(markov_env)
return venv
def test_dehomogenize():
base_act_spaces = {
"a_{}".format(idx): Discrete(5 + idx)
for idx in range(2)
}
base_env = DummyEnv(base_obs, base_obs_space, base_act_spaces)
env = pad_action_space(base_env)
env.reset()
assert all([s.n == 6 for s in env.action_spaces.values()])
env.step(5)
}
base_act_spaces = {"a_{}".format(idx): Discrete(5) for idx in range(2)}
return DummyEnv(base_obs, base_obs_space, base_act_spaces)
wrappers = [
aec_wrappers.color_reduction(new_dummy(), "R"),
aec_wrappers.down_scale(new_dummy(), x_scale=5, y_scale=10),
aec_wrappers.dtype(new_dummy(), np.int32),
aec_wrappers.flatten(new_dummy()),
aec_wrappers.reshape(new_dummy(), (64, 3)),
aec_wrappers.normalize_obs(new_dummy(), env_min=-1, env_max=5.),
aec_wrappers.frame_stack(new_dummy(), 8),
aec_wrappers.pad_observations(new_dummy()),
aec_wrappers.pad_action_space(new_dummy()),
aec_wrappers.continuous_actions(new_dummy()),
aec_wrappers.agent_indicator(new_dummy(), True),
aec_wrappers.agent_indicator(new_dummy(), False),
]
@pytest.mark.parametrize("env", wrappers)
def test_basic_wrappers(env):
obs = env.reset()
act_space = env.action_spaces[env.agent_selection]
obs_space = env.observation_spaces[env.agent_selection]
first_obs = env.observe("a_0")
print(obs_space.low.shape)
print(obs_space.high.shape)
print(first_obs.shape)