def _generate_reset_input(self, training, config: ArenaConfig) -> UnityRLInput:
rl_in = UnityRLInput()
rl_in.is_training = training
rl_in.command = 1
rl_reset = UnityRLResetInput()
if (config is not None):
rl_reset.CopyFrom(config.dict_to_arena_config())
result = UnityInput()
result.rl_input.CopyFrom(rl_in)
result.rl_reset_input.CopyFrom(rl_reset)
return result
def dict_to_arena_config(self) -> UnityRLResetInput:
config_out = UnityRLResetInput()
for k in self.arenas:
config_out.arenas[k].CopyFrom(ArenaParametersProto())
config_out.arenas[k].t = self.arenas[k].t
config_out.arenas[k].blackouts.extend(self.arenas[k].blackouts)
for item in self.arenas[k].items:
to_spawn = config_out.arenas[k].items.add()
to_spawn.name = item.name
to_spawn.positions.extend(
[v.to_proto() for v in item.positions])
to_spawn.rotations.extend(item.rotations)
to_spawn.sizes.extend([v.to_proto() for v in item.sizes])
to_spawn.colors.extend([v.to_proto() for v in item.colors])
return config_out