with policy_sess.as_default(
): # create the Controller and build the internal policy network
controller = Controller(policy_sess,
NUM_LAYERS,
state_space,
reg_param=REGULARIZATION,
exploration=EXPLORATION,
controller_cells=CONTROLLER_CELLS,
embedding_dim=EMBEDDING_DIM,
restore_controller=RESTORE_CONTROLLER)
# get an initial random state space if controller needs to predict an
# action from the initial state
#随机初始化
# state = state_space.get_random_state_space(NUM_LAYERS)
state = state_space.get_state([3, 'relu', 3, 'linear'])
print("Initial Random State : ", state_space.parse_state_space_list(state))
print()
# clear the previous files
controller.remove_files()
isFirst = True
# train for number of trails
for trial in range(MAX_TRIALS):
if isFirst:
actions = state
isFirst = False
else:
with policy_sess.as_default():
K.set_session(policy_sess)
actions = controller.get_action(
with policy_sess.as_default():
# create the Controller and build the internal policy network
controller = Controller(policy_sess,
NUM_LAYERS,
state_space,
reg_param=REGULARIZATION,
exploration=EXPLORATION,
controller_cells=CONTROLLER_CELLS,
embedding_dim=EMBEDDING_DIM,
restore_controller=RESTORE_CONTROLLER)
# get an initial random state space if controller needs to predict an
# action from the initial state
#随机初始化
# state = state_space.get_random_state_space(NUM_LAYERS)
state = state_space.get_state([1, 'relu', 1, 'linear'])
print("Initial Random State : ", state_space.parse_state_space_list(state))
print()
# clear the previous files
controller.remove_files()
isFirst = True
# train for number of trails
for trial in range(MAX_TRIALS):
if isFirst:
actions = state
isFirst = False
else:
np.random.seed()
with policy_sess.as_default():
K.set_session(policy_sess)