def test_ddpg_performance(q_func_factory):
# not good enough for batch RL, but check if it works without errors.
try:
ddpg = DDPG(q_func_type=q_func_factory)
algo_pendulum_tester(ddpg, n_trials=1)
except AssertionError:
pass
def test_ddpg(observation_shape, action_size, q_func_factory, scaler,
action_scaler):
ddpg = DDPG(
q_func_factory=q_func_factory,
scaler=scaler,
action_scaler=action_scaler,
)
algo_tester(ddpg, observation_shape)
algo_update_tester(ddpg, observation_shape, action_size)
def test_ddpg(
observation_shape,
action_size,
q_func_factory,
scalers,
target_reduction_type,
):
scaler, action_scaler, reward_scaler = scalers
ddpg = DDPG(
q_func_factory=q_func_factory,
scaler=scaler,
action_scaler=action_scaler,
reward_scaler=reward_scaler,
target_reduction_type=target_reduction_type,
)
algo_tester(
ddpg,
observation_shape,
test_policy_copy=True,
test_q_function_copy=True,
)
algo_update_tester(ddpg, observation_shape, action_size)
def test_ddpg(observation_shape, action_size, q_func_type, scaler):
ddpg = DDPG(q_func_type=q_func_type, scaler=scaler)
algo_tester(ddpg, observation_shape)
algo_update_tester(ddpg, observation_shape, action_size)