在急切模式下实例化InceptionResNetV2时出错

Question

tensorFlow version (use command below): 1.10
Python version:3.6
OS Platform: window 10

描述问题：

我使用了tf.enable_eager_execution()和create InceptionResNetV2模式从here将是错误的，只有InceptionResNetV2。我需要使用tf.enable_eager_execution()不能没有的东西。

如何修复它？

创建模式

import tensorflow as tf
tf.enable_eager_execution()
image_model =tf.keras.applications.InceptionResNetV2(include_top=False, weights='imagenet')

错误

TypeError Traceback (most recent call last)
in ()
----> 1 image_model =tf.keras.applications.InceptionResNetV2(include_top=False, weights='imagenet')

D:\anaconda\lib\site-packages\tensorflow\python\keras\applications\inception_resnet_v2.py in InceptionResNetV2(include_top, weights, input_tensor, input_shape, pooling, classes)
304 for block_idx in range(1, 11):
305 x = inception_resnet_block(
--> 306 x, scale=0.17, block_type='block35', block_idx=block_idx)
307
308 # Mixed 6a (Reduction-A block): 17 x 17 x 1088

D:\anaconda\lib\site-packages\tensorflow\python\keras\applications\inception_resnet_v2.py in inception_resnet_block(x, scale, block_type, block_idx, activation)
187 output_shape=K.int_shape(x)[1:],
188 arguments={'scale': scale},
--> 189 name=block_name)([x, up])
190 if activation is not None:
191 x = Activation(activation, name=block_name + '_ac')(x)

D:\anaconda\lib\site-packages\tensorflow\python\keras\engine\base_layer.py in call(self, inputs, *args, **kwargs)
745 input_shapes = nest.map_structure(lambda x: x.shape, inputs)
746
--> 747 output_shapes = self.compute_output_shape(input_shapes)
748 output_shapes = nest.flatten(output_shapes)
749 outputs = [

D:\anaconda\lib\site-packages\tensorflow\python\keras\layers\core.py in compute_output_shape(self, input_shape)
678
679 def compute_output_shape(self, input_shape):
--> 680 input_shape = tuple(tensor_shape.TensorShape(input_shape).as_list())
681
682 if self._output_shape is None:

D:\anaconda\lib\site-packages\tensorflow\python\framework\tensor_shape.py in init(self, dims)
539 else:
540 # Got a list of dimensions
--> 541 self._dims = [as_dimension(d) for d in dims_iter]
542 self._ndims = None
543

D:\anaconda\lib\site-packages\tensorflow\python\framework\tensor_shape.py in (.0)
539 else:
540 # Got a list of dimensions
--> 541 self._dims = [as_dimension(d) for d in dims_iter]
542 self._ndims = None
543

D:\anaconda\lib\site-packages\tensorflow\python\framework\tensor_shape.py in as_dimension(value)
480 return value
481 else:
--> 482 return Dimension(value)
483
484

D:\anaconda\lib\site-packages\tensorflow\python\framework\tensor_shape.py in init(self, value)
35 raise TypeError("Cannot convert %s to Dimension" % value)
36 else:
---> 37 self._value = int(value)
38 if (not isinstance(value, compat.bytes_or_text_types) and
39 self._value != value):

TypeError: int() argument must be a string, a bytes-like object or a number, not 'TensorShape'

Answer 1

实际上，在启用tf.enable_eager_execution()时，运行tf.keras.applications.InceptionResNetV2()存在一个错误。我看到一个社区成员已经在TensorFlow Github page上提交了一个bug问题。