ecbe066e40882b166401b70ae9c4f1d535c93b12,texar/losses/adv_losses.py,,binary_adversarial_losses,#Any#Any#Any#Any#,12

Before Change

    num_real_data = tf.shape(real_loss)[0]
    ave_real_loss = tf.reduce_sum(real_loss) / tf.to_float(num_real_data)
    fake_logits = discriminator_fn(fake_data)
    fake_loss = tf.nn.sigmoid_cross_entropy_with_logits(
        logits=fake_logits, labels=tf.zeros_like(fake_logits))
    num_fake_data = tf.shape(fake_loss)[0]
    ave_fake_loss = tf.reduce_sum(fake_loss) / tf.to_float(num_fake_data)
    disc_loss = ave_real_loss + ave_fake_loss
    if mode == "min_fake":
        gen_loss = - ave_fake_loss

After Change

    real_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
        logits=real_logits, labels=tf.ones_like(real_logits)))
    fake_logits = discriminator_fn(fake_data)
    fake_loss = tf.reduce_mean(tf.nn.sigmoid_cross_entropy_with_logits(
        logits=fake_logits, labels=tf.zeros_like(fake_logits)))
    d_loss = real_loss + fake_loss
    if mode == "min_fake":
        g_loss = - fake_loss

Instances