GAN因TypeError而无法工作：只能将列表（而不是“浮点”）连接到列表

Question

GANs模型代码中存在问题的函数：

def generate_image(iteration):
    for i in range(CLASSES):
        curLabel = genRandomLabels(BATCH_SIZE, CLASSES, condition=i)
        samples = session.run(
            all_fixed_noise_samples, feed_dict={sample_labels: curLabel}
        )
        
        samples = (samples + 1.0) * (255.99 / 2).astype("int32")
        lib.save_images.save_images(
            samples.reshape((BATCH_SIZE, 3, DIM, DIM)),
            "generated/samples_{}_{}.png".format(str(i), iteration),
        )

我明白了：

samples = (samples + 1.0) * (255.99 / 2).astype("int32")

TypeError: can only concatenate list (not "float") to list

如何解决这个问题

Answer 1

似乎您正在尝试执行阵列广播，但它不适用于Python列表：

samples = [0, 1, 2]
samples = samples + 1.0

                                     -

TypeError                                 Traceback (most recent call last)

<ipython-input-1-bfd1747a1a63> in <module>()
      1 samples = [0, 1, 2]
      2 
  > 3 samples = samples + 1.0

TypeError: can only concatenate list (not "float") to list

但是，它确实适用于Numpy阵列：

import numpy as np

samples = np.array(samples) + 1.0
samples
# array([1., 2., 3.])