Python MujocoInputFilter 예제들

예제 #1

schedule_params.heatup_steps = EnvironmentSteps(1000)

#########
# Agent #
#########
agent_params = DDPGAgentParameters()
agent_params.network_wrappers['actor'].input_embedders_parameters['measurements'] = \
    agent_params.network_wrappers['actor'].input_embedders_parameters.pop('observation')
agent_params.network_wrappers['critic'].input_embedders_parameters['measurements'] = \
    agent_params.network_wrappers['critic'].input_embedders_parameters.pop('observation')
agent_params.network_wrappers['actor'].input_embedders_parameters['measurements'].scheme = [Dense([300])]
agent_params.network_wrappers['actor'].middleware_parameters.scheme = [Dense([200])]
agent_params.network_wrappers['critic'].input_embedders_parameters['measurements'].scheme = [Dense([400])]
agent_params.network_wrappers['critic'].middleware_parameters.scheme = [Dense([300])]
agent_params.network_wrappers['critic'].input_embedders_parameters['action'].scheme = EmbedderScheme.Empty
agent_params.input_filter = MujocoInputFilter()
agent_params.input_filter.add_reward_filter("rescale", RewardRescaleFilter(1/10.))

###############
# Environment #
###############
env_params = ControlSuiteEnvironmentParameters()
env_params.level = SingleLevelSelection(control_suite_envs)

vis_params = VisualizationParameters()
vis_params.video_dump_methods = [SelectedPhaseOnlyDumpMethod(RunPhase.TEST), MaxDumpMethod()]
vis_params.dump_mp4 = False

########
# Test #
########

예제 #2

파일: Mujoco_ClippedPPO.py 프로젝트: mdavala/coach

agent_params.network_wrappers[
    'main'].middleware_parameters.activation_function = 'tanh'
agent_params.network_wrappers['main'].batch_size = 64
agent_params.network_wrappers['main'].optimizer_epsilon = 1e-5
agent_params.network_wrappers['main'].adam_optimizer_beta2 = 0.999

agent_params.algorithm.clip_likelihood_ratio_using_epsilon = 0.2
agent_params.algorithm.clipping_decay_schedule = LinearSchedule(
    1.0, 0, 1000000)
agent_params.algorithm.beta_entropy = 0
agent_params.algorithm.gae_lambda = 0.95
agent_params.algorithm.discount = 0.99
agent_params.algorithm.optimization_epochs = 10
agent_params.algorithm.estimate_state_value_using_gae = True

agent_params.input_filter = MujocoInputFilter()
agent_params.exploration = AdditiveNoiseParameters()
agent_params.pre_network_filter = MujocoInputFilter()
agent_params.pre_network_filter.add_observation_filter(
    'observation', 'normalize_observation',
    ObservationNormalizationFilter(name='normalize_observation'))

###############
# Environment #
###############
env_params = Mujoco()
env_params.level = SingleLevelSelection(mujoco_v2)

vis_params = VisualizationParameters()
vis_params.video_dump_methods = [
    SelectedPhaseOnlyDumpMethod(RunPhase.TEST),