def create_dyna(self,
env_spec=None,
model_free_algo=None,
dyanmics_model=None,
name='dyna'):
if not env_spec:
model_free_algo, local = self.create_ddpg()
dyanmics_model, _ = self.create_continuous_mlp_global_dynamics_model(
env_spec=local['env_spec'])
env_spec = local['env_spec']
env = local['env']
algo = Dyna(env_spec=env_spec,
name=name,
model_free_algo=model_free_algo,
dynamics_model=dyanmics_model,
config_or_config_dict=dict(dynamics_model_train_iter=1,
model_free_algo_train_iter=1))
algo.set_terminal_reward_function_for_dynamics_env(
terminal_func=RandomTerminalFunc(), reward_func=RandomRewardFunc())
return algo, locals()
def pendulum_task_fn():
GlobalConfig().set('DEFAULT_EXPERIMENT_END_POINT',
exp_config['DEFAULT_EXPERIMENT_END_POINT'])
env = make('Pendulum-v0')
name = 'benchmark'
env_spec = EnvSpec(obs_space=env.observation_space,
action_space=env.action_space)
mlp_q = MLPQValueFunction(env_spec=env_spec,
name_scope=name + '_mlp_q',
name=name + '_mlp_q',
**exp_config['MLPQValueFunction'])
policy = DeterministicMLPPolicy(env_spec=env_spec,
name_scope=name + '_mlp_policy',
name=name + '_mlp_policy',
output_low=env_spec.action_space.low,
output_high=env_spec.action_space.high,
**exp_config['DeterministicMLPPolicy'],
reuse=False)
ddpg = DDPG(env_spec=env_spec,
policy=policy,
value_func=mlp_q,
name=name + '_ddpg',
**exp_config['DDPG'])
mlp_dyna = ContinuousMLPGlobalDynamicsModel(env_spec=env_spec,
name_scope=name + '_mlp_dyna',
name=name + '_mlp_dyna',
**exp_config['DynamicsModel'])
algo = Dyna(env_spec=env_spec,
name=name + '_dyna_algo',
model_free_algo=ddpg,
dynamics_model=mlp_dyna,
config_or_config_dict=dict(dynamics_model_train_iter=10,
model_free_algo_train_iter=10))
algo.set_terminal_reward_function_for_dynamics_env(
terminal_func=FixedEpisodeLengthTerminalFunc(
max_step_length=env.unwrapped._max_episode_steps,
step_count_fn=algo.dynamics_env.total_step_count_fn),
reward_func=REWARD_FUNC_DICT['Pendulum-v0']())
agent = Agent(env=env,
env_spec=env_spec,
algo=algo,
exploration_strategy=None,
noise_adder=AgentActionNoiseWrapper(
noise=NormalActionNoise(),
noise_weight_scheduler=ConstantSchedule(value=0.3),
action_weight_scheduler=ConstantSchedule(value=1.0)),
name=name + '_agent')
flow = DynaFlow(
train_sample_count_func=lambda: get_global_status_collect()
('TOTAL_AGENT_TRAIN_SAMPLE_COUNT'),
config_or_config_dict=exp_config['DynaFlow'],
func_dict={
'train_algo': {
'func': agent.train,
'args': list(),
'kwargs': dict(state='state_agent_training')
},
'train_algo_from_synthesized_data': {
'func': agent.train,
'args': list(),
'kwargs': dict(state='state_agent_training', train_iter=1)
},
'train_dynamics': {
'func': agent.train,
'args': list(),
'kwargs': dict(state='state_dynamics_training')
},
'test_algo': {
'func': agent.test,
'args': list(),
'kwargs': dict(sample_count=1, sample_trajectory_flag=True)
},
'test_dynamics': {
'func': agent.algo.test_dynamics,
'args': list(),
'kwargs': dict(sample_count=10, env=env)
},
'sample_from_real_env': {
'func':
agent.sample,
'args':
list(),
'kwargs':
dict(sample_count=10,
env=agent.env,
in_which_status='TRAIN',
store_flag=True)
},
'sample_from_dynamics_env': {
'func':
agent.sample,
'args':
list(),
'kwargs':
dict(sample_count=50,
sample_type='transition',
env=agent.algo.dynamics_env,
in_which_status='TRAIN',
store_flag=False)
}
})
experiment = Experiment(tuner=None,
env=env,
agent=agent,
flow=flow,
name=name)
experiment.run()
def task_fn():
env = make('Pendulum-v0')
name = 'demo_exp'
env_spec = EnvSpec(obs_space=env.observation_space,
action_space=env.action_space)
mlp_q = MLPQValueFunction(env_spec=env_spec,
name_scope=name + '_mlp_q',
name=name + '_mlp_q',
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"N_UNITS": 1,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}])
policy = DeterministicMLPPolicy(env_spec=env_spec,
name_scope=name + '_mlp_policy',
name=name + '_mlp_policy',
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"N_UNITS": env_spec.flat_action_dim,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}],
reuse=False)
ddpg = DDPG(env_spec=env_spec,
config_or_config_dict={
"REPLAY_BUFFER_SIZE": 10000,
"GAMMA": 0.999,
"CRITIC_LEARNING_RATE": 0.001,
"ACTOR_LEARNING_RATE": 0.001,
"DECAY": 0.5,
"BATCH_SIZE": 50,
"TRAIN_ITERATION": 1,
"critic_clip_norm": 0.1,
"actor_clip_norm": 0.1,
},
value_func=mlp_q,
policy=policy,
name=name + '_ddpg',
replay_buffer=None)
mlp_dyna = ContinuousMLPGlobalDynamicsModel(
env_spec=env_spec,
name_scope=name + '_mlp_dyna',
name=name + '_mlp_dyna',
learning_rate=0.01,
state_input_scaler=RunningStandardScaler(dims=env_spec.flat_obs_dim),
action_input_scaler=RunningStandardScaler(
dims=env_spec.flat_action_dim),
output_delta_state_scaler=RunningStandardScaler(
dims=env_spec.flat_obs_dim),
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"L1_NORM": 0.0,
"L2_NORM": 0.0,
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"L1_NORM": 0.0,
"L2_NORM": 0.0,
"N_UNITS": env_spec.flat_obs_dim,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}])
algo = Dyna(env_spec=env_spec,
name=name + '_dyna_algo',
model_free_algo=ddpg,
dynamics_model=mlp_dyna,
config_or_config_dict=dict(dynamics_model_train_iter=10,
model_free_algo_train_iter=10))
# For examples only, we use random reward function and terminal function with fixed episode length.
algo.set_terminal_reward_function_for_dynamics_env(
terminal_func=FixedEpisodeLengthTerminalFunc(
max_step_length=env.unwrapped._max_episode_steps,
step_count_fn=algo.dynamics_env.total_step_count_fn),
reward_func=RandomRewardFunc())
agent = Agent(
env=env,
env_spec=env_spec,
algo=algo,
algo_saving_scheduler=PeriodicalEventSchedule(
t_fn=lambda: get_global_status_collect()
('TOTAL_AGENT_TRAIN_SAMPLE_COUNT'),
trigger_every_step=20,
after_t=10),
name=name + '_agent',
exploration_strategy=EpsilonGreedy(action_space=env_spec.action_space,
init_random_prob=0.5))
flow = create_dyna_flow(
train_algo_func=(agent.train, (), dict(state='state_agent_training')),
train_algo_from_synthesized_data_func=(
agent.train, (), dict(state='state_agent_training')),
train_dynamics_func=(agent.train, (),
dict(state='state_dynamics_training')),
test_algo_func=(agent.test, (), dict(sample_count=10)),
test_dynamics_func=(agent.algo.test_dynamics, (),
dict(sample_count=10, env=env)),
sample_from_real_env_func=(agent.sample, (),
dict(sample_count=10,
env=agent.env,
in_which_status='TRAIN',
store_flag=True)),
sample_from_dynamics_env_func=(agent.sample, (),
dict(sample_count=10,
env=agent.algo.dynamics_env,
in_which_status='TRAIN',
store_flag=True)),
train_algo_every_real_sample_count_by_data_from_real_env=10,
train_algo_every_real_sample_count_by_data_from_dynamics_env=10,
test_algo_every_real_sample_count=10,
test_dynamics_every_real_sample_count=10,
train_dynamics_ever_real_sample_count=10,
start_train_algo_after_sample_count=1,
start_train_dynamics_after_sample_count=1,
start_test_algo_after_sample_count=1,
start_test_dynamics_after_sample_count=1,
warm_up_dynamics_samples=1)
experiment = Experiment(tuner=None,
env=env,
agent=agent,
flow=flow,
name=name + '_exp')
experiment.run()
def task_fn():
# create the gym environment by make function
env = make('Pendulum-v0')
# give your experiment a name which is used to generate the log path etc.
name = 'demo_exp'
# construct the environment specification
env_spec = EnvSpec(obs_space=env.observation_space,
action_space=env.action_space)
# construct the neural network to approximate q function of DDPG
mlp_q = MLPQValueFunction(env_spec=env_spec,
name_scope=name + '_mlp_q',
name=name + '_mlp_q',
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"N_UNITS": 1,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}])
# construct the neural network to approximate policy for DDPG
policy = DeterministicMLPPolicy(env_spec=env_spec,
name_scope=name + '_mlp_policy',
name=name + '_mlp_policy',
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"N_UNITS": env_spec.flat_action_dim,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}],
reuse=False)
# construct the DDPG algorithms
ddpg = DDPG(env_spec=env_spec,
config_or_config_dict={
"REPLAY_BUFFER_SIZE": 10000,
"GAMMA": 0.999,
"CRITIC_LEARNING_RATE": 0.001,
"ACTOR_LEARNING_RATE": 0.001,
"DECAY": 0.5,
"BATCH_SIZE": 50,
"TRAIN_ITERATION": 1,
"critic_clip_norm": 0.1,
"actor_clip_norm": 0.1,
},
value_func=mlp_q,
policy=policy,
name=name + '_ddpg',
replay_buffer=None)
# construct a neural network based global dynamics model to approximate the state transition of environment
mlp_dyna = ContinuousMLPGlobalDynamicsModel(
env_spec=env_spec,
name_scope=name + '_mlp_dyna',
name=name + '_mlp_dyna',
learning_rate=0.01,
state_input_scaler=RunningStandardScaler(dims=env_spec.flat_obs_dim),
action_input_scaler=RunningStandardScaler(
dims=env_spec.flat_action_dim),
output_delta_state_scaler=RunningStandardScaler(
dims=env_spec.flat_obs_dim),
mlp_config=[{
"ACT": "RELU",
"B_INIT_VALUE": 0.0,
"NAME": "1",
"L1_NORM": 0.0,
"L2_NORM": 0.0,
"N_UNITS": 16,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}, {
"ACT": "LINEAR",
"B_INIT_VALUE": 0.0,
"NAME": "OUPTUT",
"L1_NORM": 0.0,
"L2_NORM": 0.0,
"N_UNITS": env_spec.flat_obs_dim,
"TYPE": "DENSE",
"W_NORMAL_STDDEV": 0.03
}])
# finally, construct the Dyna algorithms with a model free algorithm DDGP, and a NN model.
algo = Dyna(env_spec=env_spec,
name=name + '_dyna_algo',
model_free_algo=ddpg,
dynamics_model=mlp_dyna,
config_or_config_dict=dict(dynamics_model_train_iter=10,
model_free_algo_train_iter=10))
# To make the NN based dynamics model a proper environment so be a sampling source for DDPG, reward function and
# terminal function need to be set.
# For examples only, we use random reward function and terminal function with fixed episode length.
algo.set_terminal_reward_function_for_dynamics_env(
terminal_func=FixedEpisodeLengthTerminalFunc(
max_step_length=env.unwrapped._max_episode_steps,
step_count_fn=algo.dynamics_env.total_step_count_fn),
reward_func=RandomRewardFunc())
# construct agent with additional exploration strategy if needed.
agent = Agent(
env=env,
env_spec=env_spec,
algo=algo,
algo_saving_scheduler=PeriodicalEventSchedule(
t_fn=lambda: get_global_status_collect()
('TOTAL_AGENT_TRAIN_SAMPLE_COUNT'),
trigger_every_step=20,
after_t=10),
name=name + '_agent',
exploration_strategy=EpsilonGreedy(action_space=env_spec.action_space,
init_random_prob=0.5))
# construct the training flow, called Dyna flow. It defines how the training proceed, and the terminal condition
flow = create_dyna_flow(
train_algo_func=(agent.train, (), dict(state='state_agent_training')),
train_algo_from_synthesized_data_func=(
agent.train, (), dict(state='state_agent_training')),
train_dynamics_func=(agent.train, (),
dict(state='state_dynamics_training')),
test_algo_func=(agent.test, (), dict(sample_count=1)),
test_dynamics_func=(agent.algo.test_dynamics, (),
dict(sample_count=10, env=env)),
sample_from_real_env_func=(agent.sample, (),
dict(sample_count=10,
env=agent.env,
store_flag=True)),
sample_from_dynamics_env_func=(agent.sample, (),
dict(sample_count=10,
env=agent.algo.dynamics_env,
store_flag=True)),
train_algo_every_real_sample_count_by_data_from_real_env=40,
train_algo_every_real_sample_count_by_data_from_dynamics_env=40,
test_algo_every_real_sample_count=40,
test_dynamics_every_real_sample_count=40,
train_dynamics_ever_real_sample_count=20,
start_train_algo_after_sample_count=1,
start_train_dynamics_after_sample_count=1,
start_test_algo_after_sample_count=1,
start_test_dynamics_after_sample_count=1,
warm_up_dynamics_samples=1)
# construct the experiment
experiment = Experiment(tuner=None,
env=env,
agent=agent,
flow=flow,
name=name + '_exp')
# run!
experiment.run()