Running GAN example with placeholder for data.train, data.train.label instead of queue runner

[mjdietzx]

@tf.sg_train_func
def alt_train(sess, opt):
    x_batch, x_label_batch = train_generator.next()
    l_disc = sess.run([loss_disc, train_disc], {x: x_batch, x_label: x_label_batch})[0]  # training discriminator
    l_gen = sess.run([loss_gen, train_gen], {x: x_batch, x_label: x_label_batch})[0]  # training generator
    return np.mean(l_disc) + np.mean(l_gen)

This causes some problems in the function: sugartensor.sg_train.sg_train_func(func)

Specifically, because it is required for x_label and x_label_batch to be passed when calling sess.run() calls to sess.run() like this: sugartensor/sg_train.py#L331 fail.

The error thrown: InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'Placeholder' with dtype float [[Node: Placeholder = Placeholder[dtype=DT_FLOAT, shape=[], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

My temporary workaround is just to comment out sugartensor/sg_train.py#L331

[buriburisuri]

@mjdietzx Thanks for your information. I'll check this and think about placeholder instead of queue runner.