if __name__ == '__main__':
config1 = config()
# make env
args = xworld_args.parser().parse_args()
args.visible_radius_unit_side = config1.visible_radius_unit_side
args.visible_radius_unit_front = config1.visible_radius_unit_front
args.ego_centric = config1.ego_centric
args.map_config = config1.map_config_file
args.goal_id = 0
args.israndom_goal = False
env = xworld_navi_goal_obs_crop.XWorldNaviGoal(args)
# exploration strategy
exp_schedule = LinearExploration(env.agent.num_actions, config1.eps_begin,
config1.eps_end, config1.eps_nsteps)
# learning rate schedule
lr_schedule = LinearSchedule(config1.lr_begin, config1.lr_end,
config1.lr_nsteps)
# train model
g1 = tf.Graph()
model = DRQN(env, config1, g1)
g2 = tf.Graph()
model2 = DRQN(env, config1, g2)
#with g1.as_default():
# params = tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES, scope='q')
#print(params)
self.evaluate()
if __name__ == '__main__':
# make env
args = xworld_args.parser().parse_args()
args.visible_radius_unit_side = config.visible_radius_unit_side
args.visible_radius_unit_front = config.visible_radius_unit_front
args.ego_centric = config.ego_centric
args.map_config = config.map_config_file
env = xworld_navi_goal.XWorldNaviGoal(args)
env.teacher.israndom_goal = False
env.teacher.goal_id = 0
# exploration strategy
exp_schedule = LinearExploration(env, config.eps_begin,
config.eps_end, config.eps_nsteps)
# learning rate schedule
lr_schedule = LinearSchedule(config.lr_begin, config.lr_end,
config.lr_nsteps)
# train model
model = DRQN(env, config)
shutil.copyfile('./configs/drqn_xworld.py', config.output_path+'config.py')
shutil.copy(os.path.realpath(__file__), config.output_path)
shutil.copy(config.map_config_file, config.output_path)
if config.deploy_only:
model.deploy()
else:
model.run(exp_schedule, lr_schedule)
args = xworld_args.parser().parse_args()
args.visible_radius_unit_side = config_i.visible_radius_unit_side
args.visible_radius_unit_front = config_i.visible_radius_unit_front
args.ego_centric = config_i.ego_centric
args.map_config = config_i.map_config_file
args.goal_id = 0
args.israndom_goal = False
env_i = xworld_navi_goal_gt_dnc.XWorldNaviGoal(args)
# load model
g_a = tf.Graph()
model_a = DRQN(env_a, config_a, g_a)
model_a.initialize()
# exploration strategy
exp_schedule = LinearExploration(2, config_i.eps_begin,
config_i.eps_end, config_i.eps_nsteps)
# learning rate schedule
lr_schedule = LinearSchedule(config_i.lr_begin, config_i.lr_end,
config_i.lr_nsteps)
# train model
g_i = tf.Graph()
model_i = DRQN_planner(env_i, config_i, g_i)
'''
model_i.initialize()
nldms = []
nins = 2
nway = 5
ndigits = 2