def test_double_dqn_impl(
observation_shape,
action_size,
learning_rate,
optim_factory,
encoder_factory,
q_func_factory,
gamma,
n_critics,
target_reduction_type,
scaler,
augmentation,
):
impl = DoubleDQNImpl(
observation_shape=observation_shape,
action_size=action_size,
learning_rate=learning_rate,
optim_factory=optim_factory,
encoder_factory=encoder_factory,
q_func_factory=create_q_func_factory(q_func_factory),
gamma=gamma,
n_critics=n_critics,
target_reduction_type=target_reduction_type,
use_gpu=None,
scaler=scaler,
augmentation=augmentation,
)
torch_impl_tester(impl,
discrete=True,
deterministic_best_action=q_func_factory != "iqn")
def test_double_dqn_impl(
observation_shape,
action_size,
learning_rate,
optim_factory,
encoder_factory,
q_func_factory,
gamma,
n_critics,
bootstrap,
share_encoder,
scaler,
augmentation,
):
impl = DoubleDQNImpl(
observation_shape,
action_size,
learning_rate,
optim_factory,
encoder_factory,
create_q_func_factory(q_func_factory),
gamma,
n_critics,
bootstrap,
share_encoder,
use_gpu=False,
scaler=scaler,
augmentation=augmentation,
)
torch_impl_tester(
impl, discrete=True, deterministic_best_action=q_func_factory != "iqn"
)
def test_double_dqn_impl(observation_shape, action_size, learning_rate, gamma,
n_critics, bootstrap, share_encoder, eps,
use_batch_norm, q_func_type, scaler, augmentation,
n_augmentations, encoder_params):
impl = DoubleDQNImpl(observation_shape,
action_size,
learning_rate,
gamma,
n_critics,
bootstrap,
share_encoder,
eps,
use_batch_norm,
q_func_type=q_func_type,
use_gpu=False,
scaler=scaler,
augmentation=augmentation,
n_augmentations=n_augmentations,
encoder_params=encoder_params)
torch_impl_tester(impl,
discrete=True,
deterministic_best_action=q_func_type != 'iqn')