def _build_actor(self):
self.policy = create_deterministic_residual_policy(
self.observation_shape,
self.action_size,
self.action_flexibility,
self.use_batch_norm,
encoder_params=self.encoder_params)
def test_create_deterministic_residual_policy(observation_shape, action_size,
scale, batch_size,
use_batch_norm):
policy = create_deterministic_residual_policy(observation_shape,
action_size, scale,
use_batch_norm)
assert isinstance(policy, DeterministicResidualPolicy)
x = torch.rand((batch_size, ) + observation_shape)
action = torch.rand(batch_size, action_size)
y = policy(x, action)
assert y.shape == (batch_size, action_size)
def _build_actor(self):
self.policy = create_deterministic_residual_policy(
self.observation_shape, self.action_size, self.action_flexibility,
self.actor_encoder_factory)