def setUpClass(cls):
cls.parser = apex_argument()
cls.parser.set_defaults(n_training=10)
cls.parser.set_defaults(param_update_freq=1)
cls.parser.set_defaults(test_freq=10)
cls.parser.set_defaults(n_explorer=2)
cls.parser.set_defaults(n_env=4)
cls.parser.set_defaults(local_buffer_size=64)
def test_run(parser):
parser = apex_argument()
parser.set_defaults(n_training=10)
parser.set_defaults(param_update_freq=1)
parser.set_defaults(test_freq=10)
parser.set_defaults(n_env=64)
parser.set_defaults(local_buffer_size=64)
_test_run_discrete(parser)
_test_run_continuous(parser)
critic_units=[400, 300],
memory_capacity=memory_capacity)
def get_weights_fn(policy):
# TODO: Check if following needed
import tensorflow as tf
with tf.device(policy.device):
return [
policy.actor.weights, policy.critic.weights,
policy.critic_target.weights
]
def set_weights_fn(policy, weights):
actor_weights, critic_weights, critic_target_weights = weights
update_target_variables(policy.actor.weights, actor_weights, tau=1.)
update_target_variables(policy.critic.weights, critic_weights, tau=1.)
update_target_variables(policy.critic_target.weights,
critic_target_weights,
tau=1.)
if __name__ == '__main__':
parser = apex_argument()
parser.add_argument('--env-name', type=str, default="Pendulum-v0")
parser = DDPG.get_argument(parser)
args = parser.parse_args()
run(args, env_fn(args.env_name), policy_fn, get_weights_fn, set_weights_fn)