class RNDNetwork(torch.nn.Module):
EPSILON = 1e-10
def __init__(self, specs: BehaviorSpec, settings: RNDSettings) -> None:
super().__init__()
state_encoder_settings = NetworkSettings(
normalize=True,
hidden_units=settings.encoding_size,
num_layers=3,
vis_encode_type=EncoderType.SIMPLE,
memory=None,
)
self._encoder = NetworkBody(specs.observation_shapes,
state_encoder_settings)
def forward(self, mini_batch: AgentBuffer) -> torch.Tensor:
n_vis = len(self._encoder.visual_processors)
hidden, _ = self._encoder.forward(
vec_inputs=[
ModelUtils.list_to_tensor(mini_batch["vector_obs"],
dtype=torch.float)
],
vis_inputs=[
ModelUtils.list_to_tensor(mini_batch["visual_obs%d" % i],
dtype=torch.float)
for i in range(n_vis)
],
)
self._encoder.update_normalization(
torch.tensor(mini_batch["vector_obs"]))
return hidden
class RNDNetwork(torch.nn.Module):
EPSILON = 1e-10
def __init__(self, specs: BehaviorSpec, settings: RNDSettings) -> None:
super().__init__()
state_encoder_settings = NetworkSettings(
normalize=True,
hidden_units=settings.encoding_size,
num_layers=3,
vis_encode_type=EncoderType.SIMPLE,
memory=None,
)
self._encoder = NetworkBody(specs.observation_specs, state_encoder_settings)
def forward(self, mini_batch: AgentBuffer) -> torch.Tensor:
n_obs = len(self._encoder.processors)
np_obs = ObsUtil.from_buffer(mini_batch, n_obs)
# Convert to tensors
tensor_obs = [ModelUtils.list_to_tensor(obs) for obs in np_obs]
hidden, _ = self._encoder.forward(tensor_obs)
self._encoder.update_normalization(mini_batch)
return hidden
class RNDNetwork(torch.nn.Module):
EPSILON = 1e-10
def __init__(self, specs: BehaviorSpec, settings: RNDSettings) -> None:
super().__init__()
state_encoder_settings = settings.network_settings
if state_encoder_settings.memory is not None:
state_encoder_settings.memory = None
logger.warning(
"memory was specified in network_settings but is not supported by RND. It is being ignored."
)
self._encoder = NetworkBody(specs.observation_specs,
state_encoder_settings)
def forward(self, mini_batch: AgentBuffer) -> torch.Tensor:
n_obs = len(self._encoder.processors)
np_obs = ObsUtil.from_buffer(mini_batch, n_obs)
# Convert to tensors
tensor_obs = [ModelUtils.list_to_tensor(obs) for obs in np_obs]
hidden, _ = self._encoder.forward(tensor_obs)
self._encoder.update_normalization(mini_batch)
return hidden
