edce51a1976684e5a0cc1eef791d2c7aee72fc0d,hypergan/discriminators/densenet_discriminator.py,,discriminator,#Any#Any#Any#Any#Any#,6

Before Change

          if batch_norm is not None:
            net_dense = batch_norm(config["batch_size"]*2, name="d_expand_bna_"+str(i*10+j))(net_dense)
          net_dense = activation(net_dense)
          nets.append(conv2d(net_dense, size_dense, name="d_expand_layear"+str(i*10+j), k_w=3, k_h=3, d_h=1, d_w=1))
      net = tf.concat(3, [net]+nets)

      net = tf.nn.avg_pool(net, ksize=filter, strides=stride, padding="SAME")

After Change

              net_dense = activation(net_dense)
              if batch_norm is not None:
                net_dense = batch_norm(config["batch_size"]*2, name="d_expand_bna_"+str(i*10+j))(net_dense)
          newnet = conv2d(net_dense, size_dense, name="d_expand_layear"+str(i*10+j), k_w=3, k_h=3, d_h=1, d_w=1)
          net = tf.concat(3, [net, newnet])

      net = tf.nn.avg_pool(net, ksize=filter, strides=stride, padding="SAME")

      net = activation(net)
      if batch_norm is not None:
        net = batch_norm(config["batch_size"]*2, name="d_expand_b2na_"+str(i*10+j))(net)
      net = conv2d(net, int(net.get_shape()[3]), name="d_expand_la4year"+str(i*10+j), k_w=1, k_h=1, d_h=1, d_w=1)

      print("[discriminator] layer", net)

    k=-1

Instances