如何使用tf.train训练使用tf.Keras模型创建的模型？

Question

I am creating a GAN using Tensorflow. I decided to make Generator and Discriminator using Keras as layers became complex (since defining layers in Keras is easier). I have loss and training code for GAN in Tensorflow but I don't know how to train Keras model using that.

D_loss_real = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=D_logit_real, labels=tf.ones_like(D_logit_real)))
D_loss_fake = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=D_logit_fake, labels=tf.zeros_like(D_logit_fake)))
D_loss = D_loss_real + D_loss_fake
G_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(logits=D_logit_fake, labels=tf.ones_like(D_logit_fake)))

D_solver = tf.train.AdamOptimizer().minimize(D_loss,var_list = ?)
G_solver = tf.train.AdamOptimizer().minimize(G_loss,var_list = ?)

I was passing tf.Variables (that were used in GAN when making with tf) in var_list. However, I don't know how to minimize the keras model since keras takes care of weights on its own.
I cannot use Keras's compile and fit function because Generator and Discriminator are too interdependent while training.

How do I train my keras with loss as shown above?

Answer 1

This article will help. Basically, you just need to set your Keras session same as your Tensorflow.