如何解决Inception Score代码中的AttributeError错误？

运行用于评估生成图片的Inception Score代码时持续报错，相关函数代码如下：

# Code adapted from
# https://github.com/openai/improved-gan/blob/master/inception_score/model.py
# which was in turn derived from
# tensorflow/tensorflow/models/image/imagenet/classify_image.py
# ...
def _init_inception():
    global softmax
    if not os.path.exists(MODEL_DIR):
        os.makedirs(MODEL_DIR)
    filename = DATA_URL.split('/')[-1]
    filepath = os.path.join(MODEL_DIR, filename)
    if not os.path.exists(filepath):
        def _progress(count, block_size, total_size):
            sys.stdout.write('\r>> Downloading %s %.1f%%' % (
                filename, float(count * block_size) / float(total_size) * 100.0))
            sys.stdout.flush()

        filepath, _ = urllib.request.urlretrieve(DATA_URL, filepath, _progress)
        print()
        statinfo = os.stat(filepath)
        print('Succesfully downloaded', filename, statinfo.st_size, 'bytes.')
    tarfile.open(filepath, 'r:gz').extractall(MODEL_DIR)
    #with tf.gfile.FastGFile(os.path.join(
    with tf.io.gfile.GFile(os.path.join(
            MODEL_DIR, 'classify_image_graph_def.pb'), 'rb') as f:
        graph_def = tf.compat.v1.GraphDef()
        graph_def.ParseFromString(f.read())
        _ = tf.import_graph_def(graph_def, name='')
    # Works with an arbitrary minibatch size.
    with tf.compat.v1.Session() as sess:
        pool3 = sess.graph.get_tensor_by_name('pool_3:0')
        ops = pool3.graph.get_operations()
        for op_idx, op in enumerate(ops):
            for o in op.outputs:
                shape = o.get_shape()
                shape = [s.value for s in shape]
                new_shape = []
                for j, s in enumerate(shape):
                    if s == 1 and j == 0:
                        new_shape.append(None)
                    else:
                        new_shape.append(s)
                o.set_shape(tf.TensorShape(new_shape))
        w = sess.graph.get_operation_by_name("softmax/logits/MatMul").inputs[1]
        logits = tf.matmul(tf.squeeze(pool3, [1, 2]), w)
        softmax = tf.nn.softmax(logits)

错误出现在标注# Works with an arbitrary minibatch size.的代码段，错误信息如下：

Traceback (most recent call last):
  File "C:\Users\k\anaconda3\envs\tf\lib\contextlib.py", line 130, in __exit__
    self.gen.throw(type, value, traceback)
  File "C:\Users\k\anaconda3\envs\tf\lib\site-packages\tensorflow_core\python\framework\ops.py", line 5385, in get_controller
    yield g
  File "D:/Git/Graduation/mytext2image-main/IS.py", line 182, in _init_inception
    shape = [s.value for s in shape]
  File "D:/Git/Graduation/mytext2image-main/IS.py", line 182, in <listcomp>
    shape = [s.value for s in shape]
AttributeError: 'NoneType' object has no attribute 'value'

疑问：该问题是否与TensorFlow版本有关？已将所有tf替换为tf.compat.v1，求解决帮助。

解决方案 #

这个错误确实和TensorFlow版本兼容有关：旧代码基于TF1.x原生API编写，即使替换为tf.compat.v1，部分API的行为仍有变化。当TensorShape的维度为未知（动态维度，值为None）时，直接访问s.value会返回None，触发AttributeError。

具体修改步骤 #

将报错的维度遍历代码替换为更安全的标准写法：
原代码中：

shape = o.get_shape()
shape = [s.value for s in shape]

替换为：

shape = o.get_shape().as_list()

as_list()是TensorFlow中获取TensorShape维度列表的标准方法，会直接返回包含None的维度列表，无需手动遍历访问.value，同时兼容TF1.x和tf.compat.v1模式。

修改后的完整代码片段 #

# Works with an arbitrary minibatch size.
with tf.compat.v1.Session() as sess:
    pool3 = sess.graph.get_tensor_by_name('pool_3:0')
    ops = pool3.graph.get_operations()
    for op_idx, op in enumerate(ops):
        for o in op.outputs:
            shape = o.get_shape().as_list()  # 核心修改处
            new_shape = []
            for j, s in enumerate(shape):
                if s == 1 and j == 0:
                    new_shape.append(None)
                else:
                    new_shape.append(s)
            o.set_shape(tf.TensorShape(new_shape))
    w = sess.graph.get_operation_by_name("softmax/logits/MatMul").inputs[1]
    logits = tf.matmul(tf.squeeze(pool3, [1, 2]), w)
    softmax = tf.nn.softmax(logits)

补充说明 #

你已替换的tf.io.gfile.GFile、tf.compat.v1.Session等API都是正确的兼容写法，无需调整。修改后即可解决动态维度导致的NoneType属性访问错误。

内容的提问来源于stack exchange，提问作者zli2001