if __name__ == '__main__':
param = argparse.ArgumentParser(usage='param', description='for main')
param.add_argument('--model',
type=str,
default='Single',
help='choose one model')
param.add_argument('--lr', type=float, default=0.001, help='learning rate')
param.add_argument('--optimizer',
type=str,
default='adam',
help='you can also choose sgd, adagrad, rmsprop')
param.add_argument('--shell',
type=bool,
default=False,
help='learning rate')
temparam = param.parse_args()
config = configure()
config.lr = temparam.lr
config.lr_method = temparam.optimizer
model_ = getattr(models, temparam.model)
model = model_(config)
print 'Present used model %s\n' % model
model.train()
if temparam.shell:
print('\n######################\n open ipython for eval')
_start_shell(locals())
# end_point = Point(x2, y2, z2, 1)
## experiment for group 2: 5*20*20 grid scale
# 高度
x1 = 0
x2 = 0
# 长度
y1 = 0
y2 = 18
# 宽度
z1 = 9
z2 = 9
starting_point = Point(x1, y1, z1, 1)
end_point = Point(x2, y2, z2, 1)
config = configure(grid_x, grid_y, grid_z, safety_threshold, privacy_threshold,
privacy_radius, starting_point, end_point, delay)
T_budget = config.T_budget
T_optimal = config.T_optimal
# print(occ_grid)
occ_grid_known = copy.deepcopy(occ_grid)
for i in range(occ_grid.shape[0]):
for j in range(occ_grid.shape[1]):
for k in range(occ_grid.shape[2]):
if occ_grid_known[i][j][k] > 1:
occ_grid_known[i][j][k] = 1
## ref == 1 to show the initial results without self-adaptive motion planning
ref = 0
if ref == 1:
reference_path = np.load('../data_raw/reference_path.npy')
end_point = Point(x2, y2, z2, 1)
# alpha ,beta 固定
# alpha_list = [4/2, 5/3, 6/4, 7/5, 8/6, 9/7, 10/8, 11/9, 12/10, 13/11]
# alpha = alpha_list[i % 10]
# beta_list = [3/2, 4/3, 5/4, 6/5, 7/6, 8/7, 9/8, 10/9, 11/10, 12/11]
# beta = beta_list[i % 10]
alpha = 5 / 3
beta = 4 / 3
# alpha = 10
# beta = 10
Kca = 10
config = configure(grid_x, grid_y, grid_z, safety_threshold,
privacy_threshold, privacy_radius,
starting_point, end_point, viewradius, alpha,
beta, exploration_rate, preference)
T_budget = alpha * (abs(x1 - x2) + abs(y1 - y2) + abs(z1 - z2))
T_optimal = beta * (abs(x1 - x2) + abs(y1 - y2) + abs(z1 - z2))
log.info(
"Iteration: %d; Configuration: grid: %d, safety_threshold: %f, privacy_threshold: %f, the starting point: [%d, %d, %d]; the end point: [%d, %d, %d]; T_budget(alpha): "
"%f (%f); "
"T_optimal(beta): %f (%f); Exploration_rate: %f; Preference: %f; View_radius: %f"
% (iteration, grid_x, safety_threshold, privacy_threshold, x1,
y1, z1, x2, y2, z2, T_budget, alpha, T_optimal, beta,
exploration_rate, preference, viewradius))
SaveMap(config, iteration, exploration_rate, num)
reinitial_flag = 1
refpath = []
planpath = []
unk_id = self.vocab.word2id('<unk>')
with open(self.config.eval_data, "rb") as analogy_f:
for line in analogy_f:
if line.startswith(b":"): # Skip comments.
continue
words = line.strip().lower().split(b" ")
ids = self.vocab.word2id(words)
if unk_id in ids or len(ids) != 4:
questions_skipped += 1
else:
questions.append(np.array(ids))
print("Eval analogy file: ", self.config.eval_data)
print("Questions: ", len(questions))
print("Skipped: ", questions_skipped)
self._analogy_questions = np.array(questions, dtype=np.int32)
def _start_shell(local_ns=None):
''''''
import IPython
user_ns = {}
if local_ns:
user_ns.update(local_ns)
user_ns.update(globals())
IPython.start_ipython(argv=[], user_ns=user_ns)
if __name__ == '__main__':
param = configure()
train = emTrainer(param)
