def main():
ap = argparse.ArgumentParser()
ap.add_argument('--arch', required=True, nargs='+', type=int)
ap.add_argument('--lr', required=False, type=int, default=.01)
ap.add_argument('--epochs', required=False, type=int, default=100)
ap.add_argument('--iters', required=True, type=int)
ap.add_argument('--trials', required=True, type=int)
ap.add_argument('--env', required=True)
ap.add_argument('--t', required=True, type=int)
args = vars(ap.parse_args())
opt = Options()
opt.envname = args['env']
opt.trials = args['trials']
opt.iters = args['iters']
opt.epochs = args['epochs']
opt.lr = args['lr']
opt.arch = args['arch']
opt.t = args['t']
opt.filename = '/Users/JonathanLee/experts/' + opt.envname + '.pkl'
opt.env = gym.envs.make(opt.envname).env
opt.sim = gym.envs.make(opt.envname).env
opt.pi = load_policy.load_policy(opt.filename)
opt.sess = tf.Session()
opt.sup = NetSupervisor(opt.pi, opt.sess)
run_trial(opt)
def main():
# ------------------------------------------------ Training Phase ------------------------------------------------
# image_files = random.sample(glob.glob('E:\\work\\pedestrian_crop_python_process\\Pedestrain_cropDB\\train\\0\\*.bmp'), 10)
# image_files = random.sample(glob.glob('data/0.normal/*.bmp'), 10)
# data_in = data_read(image_files)
opt = Options().parse()
opt.iwidth = map_x_size
opt.iheight = map_y_size
#---new--- depth for size
ctinit = map_x_size
while ctinit > 4:
ctinit = ctinit / 2
opt.ctinit = int(ctinit)
#---new---
opt.batchsize = 64
opt.epochs = 1000
opt.mask = 0 # 1: masking for simulation map
opt.time = datetime.now()
train_dataloader = load_data(
'./data/unsupervised/train/') # path to trainset
result_path = './results/{0}/'.format(
opt.time) # reconstructions durnig the training
if not os.path.isdir(result_path):
os.mkdir(result_path)
# dataloader = load_data(opt, data_in)
model = AAE_basic(opt, train_dataloader)
model.train()