module 'tensorflow' has no attribute 'random_normal'

Question

i was running this step and it had error

from efficientnet import EfficientNetB0 as Net
from efficientnet import center_crop_and_resize, preprocess_input
# loading pretrained conv base model
conv_base = Net(weights='imagenet', include_top=False, 
input_shape=input_shape)   

error:

AttributeError Traceback (most recent call last)
<ipython-input-13-34d286b24b60> in <module>()
      2 from efficientnet import center_crop_and_resize, preprocess_input
      3 # loading pretrained conv base model
----> 4 conv_base = Net(weights='imagenet', include_top=False, input_shape=input_shape)

4 frames
/content/efficientnet_keras_transfer_learning/efficientnet/model.py in __call__(***failed resolving arguments***)
     65         kernel_height, kernel_width, _, out_filters = shape
     66         fan_out = int(kernel_height * kernel_width * out_filters)
---> 67         return tf.random_normal(
     68             shape, mean=0.0, stddev=np.sqrt(2.0 / fan_out), dtype=dtype)
     69 

AttributeError: module 'tensorflow' has no attribute 'random_normal'

code ref: https://colab.research.google.com/github/Tony607/efficientnet_keras_transfer_learning/blob/master/Keras_efficientnet_transfer_learning.ipynb

Thank you so much

Answer 1

Google Colab by default uses Tensorflow 2.x, where this function is called tensorflow.random.normal. The repository you link to seems to be using Tensorflow 1.x, where it was called tensorflow.random_normal. You can either:

• clone the repository and update its code to run correctly on the newer version, or

• instruct Google Colab to use the old Tensorflow using this instruction before import tensorflow:

    %tensorflow_version 1.x