Whether it is right about KL_loss

[yangzhikai]

I find the calculation of KL-loss is different from VAE.
Because:
latent_loss = 0.5 * tf.reduce_sum(　tf.square(z_mean) + tf.square(z_stddev)
　　　　　　　　　　　　　　　　　　　　　　- tf.log(tf.square(z_stddev)) - 1,　1　)
I think the "self.mu.pow(2) " of line-154 in bicycle_gan_model.py should be "self.mu.pow(-2)".

[junyanz]

@pathak22 I think it should be pow(2) according to Equation (10) in the VAE paper.