That was what I was trying to ask… Now, I fully understand the reason.
In your local environment, you probably use newer version of packages including tensorflow and others. In that case, behavior of GlorotUniform, an initializer for convolutional_block(), is different.
Of course, a recommended way is to downgrade several packages to be consistent to the Coursera platform.
But, this case is relatively easy.
At first, you need to remove glorot_uniform from the import list. Then, add this.
from tensorflow.python.keras.initializers.initializers_v2 import GlorotUniform as glorot_uniform
Please try this.
And, if you keep going on a local environment, similar problems are sure to occur. I have all assignments in my local PC, but also have several conda environments.
So, if you think everything should be correctly implemented but get an error, please try on the Coursera environment. Then, you can see what you should do next.
Hope this helps.