def _generate_reset_input(self) -> UnityInputProto:
rl_in = UnityRLInputProto()
rl_in.command = RESET
rl_in.side_channel = bytes(
self._side_channel_manager.generate_side_channel_messages()
)
return self._wrap_unity_input(rl_in)
def _generate_step_input(
self, vector_action: Dict[str, ActionTuple]
) -> UnityInputProto:
rl_in = UnityRLInputProto()
for b in vector_action:
n_agents = len(self._env_state[b][0])
if n_agents == 0:
continue
for i in range(n_agents):
action = AgentActionProto()
if vector_action[b].continuous is not None:
action.vector_actions_deprecated.extend(
vector_action[b].continuous[i]
)
action.continuous_actions.extend(vector_action[b].continuous[i])
if vector_action[b].discrete is not None:
action.vector_actions_deprecated.extend(
vector_action[b].discrete[i]
)
action.discrete_actions.extend(vector_action[b].discrete[i])
rl_in.agent_actions[b].value.extend([action])
rl_in.command = STEP
rl_in.side_channel = bytes(
self._side_channel_manager.generate_side_channel_messages()
)
return self._wrap_unity_input(rl_in)
def _generate_step_input(
self, vector_action: Dict[str, np.ndarray]) -> UnityInputProto:
rl_in = UnityRLInputProto()
for b in vector_action:
n_agents = len(self._env_state[b][0])
if n_agents == 0:
continue
for i in range(n_agents):
action = AgentActionProto(vector_actions=vector_action[b][i])
rl_in.agent_actions[b].value.extend([action])
rl_in.command = STEP
rl_in.side_channel = bytes(
self._generate_side_channel_data(self.side_channels))
return self.wrap_unity_input(rl_in)
def _generate_reset_input(self) -> UnityInputProto:
rl_in = UnityRLInputProto()
rl_in.command = RESET
rl_in.side_channel = bytes(
self._generate_side_channel_data(self.side_channels))
return self.wrap_unity_input(rl_in)