# discriminator loss
loss_d_r = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_real, labels=y_real))
loss_d_f = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_fake, labels=y_fake))
loss_d = (loss_d_r + loss_d_f) / 2
print 'loss_d', loss_d.get_shape()
# generator loss
loss_g = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_fake, labels=y_real))
# categorical factor loss
loss_c = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits=cat_fake, labels=z_cat))
# continuous factor loss
loss_con =tf.reduce_mean(tf.square(con_fake-z_con))
train_disc, disc_global_step = optim(loss_d + loss_c + loss_con, lr=0.0001, optim = 'Adm', category='discriminator')
train_gen, gen_global_step = optim(loss_g + loss_c + loss_con, lr=0.001, optim = 'Adm', category='generator')
init = tf.global_variables_initializer()
saver = tf.train.Saver()
print train_gen
cur_epoch = 0
cur_step = 0
with tf.Session() as sess:
sess.run(init)
coord, threads = queue_context(sess)
try:
while not coord.should_stop():
cur_step += 1
loss_d = (loss_d_r + loss_d_f) / 2
print 'loss_d', loss_d.get_shape()
# generator loss
loss_g = tf.reduce_mean(
tf.nn.sigmoid_cross_entropy_with_logits(logits=disc_fake, labels=y_real))
# categorical factor loss
loss_c = tf.reduce_mean(
tf.nn.sparse_softmax_cross_entropy_with_logits(logits=cat_fake,
labels=z_cat))
# continuous factor loss
loss_con = tf.reduce_mean(tf.square(con_fake - z_con))
train_disc, disc_global_step = optim(loss_d + loss_c + loss_con,
lr=0.0001,
optim='Adm',
category='discriminator')
train_gen, gen_global_step = optim(loss_g + loss_c + loss_con,
lr=0.001,
optim='Adm',
category='generator')
init = tf.global_variables_initializer()
saver = tf.train.Saver()
print train_gen
cur_epoch = 0
cur_step = 0
with tf.Session() as sess:
sess.run(init)
coord, threads = queue_context(sess)