def make_env(name, action_repeat):
env = gym.make(name)
env = ActionRepeat(env, action_repeat)
env = RescaleAction(env, -1.0, 1.0)
train_env = ObservationNormalize(env)
test_env = TestObservationNormalize(train_env)
return train_env, test_env
def main():
parser = ArgumentParser()
parser.add_argument('--env', type=str, default='HalfCheetah-v2')
parser.add_argument('--seed', type=int, default=100)
parser.add_argument('--use_obs_filter',
dest='obs_filter',
action='store_true')
parser.add_argument('--update_every_n_steps', type=int, default=1)
parser.add_argument('--n_random_actions', type=int, default=10000)
parser.add_argument('--n_collect_steps', type=int, default=1000)
parser.add_argument('--n_evals', type=int, default=1)
parser.add_argument('--save_model', dest='save_model', action='store_true')
parser.set_defaults(obs_filter=False)
parser.set_defaults(save_model=False)
args = parser.parse_args()
params = vars(args)
seed = params['seed']
env = gym.make(params['env'])
env = RescaleAction(env, -1, 1)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
agent = SAC_Agent(seed, state_dim, action_dim)
train_agent_model_free(agent=agent, env=env, params=params)
def make_env(
env_name,
seed,
save_dir = None,
add_episode_monitor = True,
action_repeat = 1,
frame_stack = 1,
):
"""Env factory with wrapping.
Args:
env_name: The name of the environment.
seed: The RNG seed.
save_dir: Specifiy a save directory to wrap with `VideoRecorder`.
add_episode_monitor: Set to True to wrap with `EpisodeMonitor`.
action_repeat: A value > 1 will wrap with `ActionRepeat`.
frame_stack: A value > 1 will wrap with `FrameStack`.
Returns:
gym.Env object.
"""
# Check if the env is in x-magical.
xmagical.register_envs()
if env_name in xmagical.ALL_REGISTERED_ENVS:
env = gym.make(env_name)
else:
raise ValueError(f"{env_name} is not a valid environment name.")
if add_episode_monitor:
env = wrappers.EpisodeMonitor(env)
if action_repeat > 1:
env = wrappers.ActionRepeat(env, action_repeat)
env = RescaleAction(env, -1.0, 1.0)
if save_dir is not None:
env = wrappers.VideoRecorder(env, save_dir=save_dir)
if frame_stack > 1:
env = wrappers.FrameStack(env, frame_stack)
# Seed.
env.seed(seed)
env.action_space.seed(seed)
env.observation_space.seed(seed)
return env
def test_rescale_action():
env = gym.make("CartPole-v1")
with pytest.raises(AssertionError):
env = RescaleAction(env, -1, 1)
del env
env = gym.make("Pendulum-v1")
wrapped_env = RescaleAction(gym.make("Pendulum-v1"), -1, 1)
seed = 0
obs = env.reset(seed=seed)
wrapped_obs = wrapped_env.reset(seed=seed)
assert np.allclose(obs, wrapped_obs)
obs, reward, _, _ = env.step([1.5])
with pytest.raises(AssertionError):
wrapped_env.step([1.5])
wrapped_obs, wrapped_reward, _, _ = wrapped_env.step([0.75])
assert np.allclose(obs, wrapped_obs)
assert np.allclose(reward, wrapped_reward)
def main():
parser = ArgumentParser()
parser.add_argument('--env', type=str, default='HalfCheetah-v2')
parser.add_argument('--seed', type=int, default=100)
parser.add_argument('--use_obs_filter',
dest='obs_filter',
action='store_true')
parser.add_argument('--update_every_n_steps', type=int, default=1)
parser.add_argument('--n_random_actions', type=int, default=10000)
parser.add_argument('--n_collect_steps', type=int, default=1000)
parser.add_argument('--n_evals', type=int, default=1)
parser.add_argument('--experiment_name', type=str, default='')
parser.add_argument('--make_gif', dest='make_gif', action='store_true')
parser.add_argument('--save_model', dest='save_model', action='store_true')
parser.add_argument('--total_steps', type=int, default=int(1e7))
parser.set_defaults(obs_filter=False)
parser.set_defaults(save_model=False)
args = parser.parse_args()
params = vars(args)
seed = params['seed']
all_envs = gym.envs.registry.all()
available_envs = [env_spec.id for env_spec in all_envs]
env_name = params['env']
if env_name in available_envs:
env = gym.make(params['env'])
elif env_name == 'cartpole-swingup_sparse':
env = dmc2gym.make(domain_name='cartpole',
task_name='swingup_sparse',
seed=0)
else:
raise Exception("Invalid environment name")
env = RescaleAction(env, -1, 1)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
agent = SAC_Agent(seed, state_dim, action_dim)
train_agent_model_free(agent=agent, env=env, params=params)
def main():
parser = ArgumentParser()
parser.add_argument('--env', type=str, default='HalfCheetah-v2')
parser.add_argument('--alg',
type=str,
default='td3',
choices={'td3', 'sac', 'tds', 'mepg'})
parser.add_argument('--yaml_config', type=str, default=None)
parser.add_argument('--seed', type=int, default=100)
parser.add_argument('--use_obs_filter',
dest='obs_filter',
action='store_true')
parser.add_argument('--update_every_n_steps', type=int, default=None)
parser.add_argument('--n_random_actions', type=int, default=None)
parser.add_argument('--n_collect_steps', type=int, default=None)
parser.add_argument('--n_evals', type=int, default=1)
parser.add_argument('--total_timesteps', type=int, default=1e7)
parser.add_argument('--save_model', dest='save_model', action='store_true')
parser.add_argument('--experiment_name', type=str, default=None)
parser.add_argument('--make_gif', dest='make_gif', action='store_true')
parser.add_argument('--checkpoint_interval', type=int, default=500000)
parser.add_argument('--save_replay_pool', type=bool, default=False)
parser.add_argument('--load_model_path', type=str, default=None)
parser.set_defaults(obs_filter=False)
parser.set_defaults(save_model=False)
args = parser.parse_args()
params = vars(args)
seed = params['seed']
env = gym.make(params['env'])
env = RescaleAction(env, -1, 1)
state_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
params, agent = get_agent_and_update_params(seed, state_dim, action_dim,
params)
train_agent_model_free(agent=agent, env=env, params=params)
def main():
envs = {
0: ['Walker2d-v2', 5],
1: ['Hopper-v2', 5],
2: ['HalfCheetah-v2', 1]
}
ind = 1
env_name = envs[ind][0]
env = gym.make(env_name)
env = RescaleAction(env, -1, 1)
obs_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
print(action_dim, env.action_space.low, env.action_space.high)
critic_net = DoubleQFunc(obs_dim, action_dim)
target_net = copy.deepcopy(critic_net)
target_net.eval()
policy = Policy(obs_dim, action_dim)
train(env, critic_net, target_net, policy)
reward_sum = 0
for _ in range(n_starts):
done = False
state = env.reset()
while (not done):
action = get_action(state, policy, deterministic=True)
nextstate, reward, done, _ = env.step(action)
reward_sum += reward
state = nextstate
env.render()
return reward_sum / n_starts
envs = {0: ['Walker2d-v2', 5], 1: ['Hopper-v2', 5], 2: ['HalfCheetah-v2', 1]}
ind = 1
env_name = envs[ind][0]
env = gym.make(env_name)
env = RescaleAction(env, -1, 1)
obs_dim = env.observation_space.shape[0]
action_dim = env.action_space.shape[0]
print(action_dim, env.action_space.low, env.action_space.high)
save_path = "Hopper/Policy200000pt"
# policy = Policy(obs_dim, action_dim)
policy = torch.load(save_path)
evaluate_agent(env, policy, False, n_starts=10000)