'KerasTensor' object has no attribute '_keras_shape

[Vidhya-Vidhya]

Respected sir, I am getting the following error while using your code.

**depth = 20 base_model = 'inception_resnet_v2'

Choose what attention_module to use: cbam_block / se_block / None

attention_module = 'cbam_block' model_type = base_model if attention_module=='cbam_block' else basemodel+''+attention_module model = InceptionResNetV2(input_shape=(299,299,3), classes=num_classes, attention_module=attention_module)

AttributeError Traceback (most recent call last) Input In [7], in <cell line: 7>() ----> 7 model = InceptionResNetV2(input_shape=(299,299,3), classes=num_classes, attention_module=attention_module)

Input In [3], in InceptionResNetV2(include_top, weights, input_tensor, input_shape, pooling, classes, attention_module) 144 x = Concatenate(axis=channel_axis, name='mixed_5b')(branches) 146 # squeeze and excite block --> 147 x = se_block(x) 149 # 10x block35 (Inception-ResNet-A block): 35 x 35 x 320 150 for block_idx in range(1, 11):

Input In [1], in se_block(input_feature, ratio) 15 def se_block(input_feature, ratio=8): 18 channel_axis = 1 if K.image_data_format() == "channels_first" else -1 ---> 19 channel = input_feature._keras_shape[channel_axis] 21 se_feature = GlobalAveragePooling2D()(input_feature) 22 se_feature = Reshape((1, 1, channel))(se_feature)

AttributeError: 'KerasTensor' object has no attribute '_keras_shape'

I am using jupyter command prompt in anaconda. my python version is 3.6, tensorflow =2.2.0, keras=2.3.1 while using tensorflow =1.13.1 keras=2.2.0 in colab the code is running. but in jupyter the same version and the above version is not running. so kindly guide me to fix the issues as soon as possible

[junjie2008v]

try replace '_keras_shape' to 'shape'

[bolgoff]

try replace '_keras_shape' to 'shape'

Yeah. That's worked