def __init__(self,
env,
num_init_random_rollouts=10,
max_rollout_length=500,
num_onplicy_iters=10,
num_onpolicy_rollouts=10,
training_epochs=60,
training_batch_size=512,
render=False,
mpc_horizon=15,
num_random_action_selection=4096,
nn_layers=1):
self._env = env
self._max_rollout_length = max_rollout_length
self._num_onpolicy_iters = num_onplicy_iters
self._num_onpolicy_rollouts = num_onpolicy_rollouts
self._training_epochs = training_epochs
self._training_batch_size = training_batch_size
self._render = render
logger.info('Gathering random dataset')
self._random_dataset = self._gather_rollouts(utils.RandomPolicy(env),
num_init_random_rollouts)
logger.info('Creating policy')
self._policy = ModelBasedPolicy(
env,
self._random_dataset,
horizon=mpc_horizon,
num_random_action_selection=num_random_action_selection)
timeit.reset()
timeit.start('total')
def _log(self, dataset):
timeit.stop('total')
dataset.log()
logger.dump_tabular(print_func=logger.info)
logger.debug('')
for line in str(timeit).split('\n'):
logger.debug(line)
timeit.reset()
timeit.start('total')
def _log(self, dataset):
# stop timing
timeit.stop('total')
# print logging information
dataset.log()
logger.dump_tabular(print_func=logger.info)
logger.debug('')
for line in str(timeit).split('\n'):
logger.debug(line)
# reset timing
timeit.reset()
timeit.start('total')
