def test_sac(
observation_shape, action_size, q_func_factory, scaler, action_scaler
):
sac = SAC(
q_func_factory=q_func_factory,
scaler=scaler,
action_scaler=action_scaler,
)
algo_tester(sac, observation_shape)
algo_update_tester(sac, observation_shape, action_size)
def test_sac(
observation_shape,
action_size,
q_func_factory,
scalers,
target_reduction_type,
):
scaler, action_scaler, reward_scaler = scalers
sac = SAC(
q_func_factory=q_func_factory,
scaler=scaler,
action_scaler=action_scaler,
reward_scaler=reward_scaler,
target_reduction_type=target_reduction_type,
)
algo_tester(sac,
observation_shape,
test_policy_copy=True,
test_q_function_copy=True)
algo_update_tester(sac, observation_shape, action_size)
def test_sac_performance(q_func_factory):
if q_func_factory == "iqn" or q_func_factory == "fqf":
pytest.skip("IQN is computationally expensive")
sac = SAC(q_func_factory=q_func_factory)
algo_pendulum_tester(sac, n_trials=3)
def test_sac_performance(q_func_type):
if q_func_type == 'iqn' or q_func_type == 'fqf':
pytest.skip('IQN is computationally expensive')
sac = SAC(q_func_type=q_func_type)
algo_pendulum_tester(sac, n_trials=3)
def test_sac(observation_shape, action_size, q_func_type, scaler):
sac = SAC(q_func_type=q_func_type, scaler=scaler)
algo_tester(sac, observation_shape)
algo_update_tester(sac, observation_shape, action_size)