The graph attached is my work with Keynote to focus on convolutional layers, but you can get a similar one with this.
from keras.utils.vis_utils import plot_model
plot_model(model, show_shapes=True, show_layer_names=True)
1 Like
Thanks so much! That’ll be useful. It looks like it works at the model level, so in order to get a graph of my identity/convolution layers, I need to embed them in a model. I just tried it and it seems to work. I’d send the output, but it’s really long, but it does have the skip connections, so, at least seems to be working correctly.
Anyway, another useful tool. Thanks, Nobu.
1 Like
Could you tell what was the source of error…
I have been going through it since last few days…
1 Like
Hi, ,Priyanshu,
The resolution was in Course 4, which I no longer seem to have access to. Is there a policy that students are barred from review earlier course material?
1 Like
As near as I can remember, the error was in this line in the Residual Nets:
X_shortcut = Conv2D(filters = F3, kernel_size = 1, strides = (s,s), padding='valid', kernel_initializer = initializer(seed=0))(X_shortcut)
I had X instead of X_shortcut as the final argument for Conv2D.
Hope that helps.
4 Likes
AHHHH, and that is true for 2 lines. Not just one !!!
Thanks guys !
2 Likes
I’m having a different problem which says AssertionError:Looks like the BatchNormalization units are not working
Wt shall I do ?
1 Like
Thank you Marshall, and I think this problem is related to original code because in the shortcut path has been wrote “X” that is confusing, or maybe they want to test our understanding. 
SHORTCUT PATH
X = None
X = None
1 Like
Thanks to everyone I managed to correct my code. But, my laptop is not resilient enough, the kernel dies whenever I run the model, its not able to handle the load. Can I still submit without running the model.fit() and the test_accuracy?
1 Like
@valentina sorry to bother u but how did u actually solve the problem, i literally tried everything and everything seems fine to me, still i have that error. i assume that’s similar to yours as well.
Your post is a duplicate, I’ve answered on your other thread.
I got the problem. In all 3 Conv2D layers check the "filters = " (all 3 should be different) and “kernel_size =” (second is different). Check the instructions above 
I have the same problem as you. I have reviewed the suggestions you provided above, but still got an error.