from scipy import linalg as la
from scipy.sparse import linalg as sla
from scipy.linalg import eigh
from scipy.sparse.linalg import eigsh
from scipy.linalg import svd
from scipy.sparse.linalg import svds
import os
import numpy as np
import pickle
sim_type = 'ex_ra'
method = 'smf'
decom_method = 'svd'
args = argument.parse_args()
net_file = args.net_file
print('netfile: {}'.format(net_file))
emb_file = args.emb_file
print('emb_file: {}'.format(emb_file))
net_name = args.net_name
print('net_name: {}'.format(net_name))
emb_dim = args.emb_dim
print('emb_dim: {}'.format(emb_dim))
alpha = args.alpha
print('alpha: {}'.format(alpha))
import os
import shutil
import random
import Preproccessor
import tensorflow as tf
import numpy as np
from argument import parse_args
from model import DCGAN as GAN
from scipy import misc
from tensorflow.examples.tutorials.mnist import input_data
from keras.utils import to_categorical
from utils import save_image_train, save_image_train_by_digit
LOAD_FROM_MNIST = False
args = parse_args()
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
if LOAD_FROM_MNIST:
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
else:
prep = Preproccessor.Preprocessor(
image_shape=[args.img_width, args.img_height, 3])
def process_img(img):
if args.prep:
return (img - 0.5) / 0.5
else:
return img
tf.set_random_seed(params.validseed)
saved_state = brain.save_state(sess)
total_loss = 0.0
try:
for eval_i in range(num_samples):
loss, _ = sess.run([brain._valid_loss_op, brain._update_state_op])
total_loss += loss
print("Validation loss = %f" % (total_loss / num_samples))
except tf.errors.OutOfRangeError:
print("No more samples for evaluation. This should not happen")
raise
brain.load_state(sess, saved_state)
# restore seed
if fix_seed:
np.random.set_state(np_random_state)
tf.set_random_seed(np.random.randint(
999999)) # cannot save tf seed, so generate random one from numpy
return total_loss
if __name__ == '__main__':
params = parse_args()
run_training(params)
