train problem

[soledad-deng]

For i run the modle with tensorflow 1.1.0-rc2, i modify the input arguments of tf.concat in ops.py and tf.nn.sigmoid_cross_entropy_with_logits in FaceAging.py. To avoid the error[ValueError: Variable E_conv0/w/Adam/ does not exist], i try GradientDescentOptimizer instead of AdamOptimizer. It runs successfully. But, the jpgs in save/sample are all like this

[soledad-deng]

GradientDescentOptimizer doesn't work, i solved it with AdamOptimizer