Exercise 3 - unet_model_value_error

Course Q&A Deep Learning Specialization Convolutional Neural Networks
1

I got this error

image
image757×501 24.3 KB

it seem the problem is in the number of filters so i have checked them but i didn’t figured it i used 2 * n_filters than 4 * , 8* , 16* for conv_block
and 8* n_filters than 4 * , 2* , 1* for upsampling_block

2

The instructions tell you what to do. The point is not to hard-code anything, but use the parameters that are passed and then multiply by the appropriate factors at each level.

3

yh this is what I did, I multiplied it by 2

4

The first conv_block uses just n_filters and then you double it every time after that, right?

Also make sure you do “Kernel → Restart and Clear Output”, “Cell → Run All” to make sure you are actually running the code that you are looking at.

5

If you look at the full “summary” of your model or the correct one, you can see that your error is on the very last conv_block call. Note that you have 368 filters, but 368 is not divisible by 32, right? And the value of n_filters in this test case is 32. So how could that happen?

2 Likes
6

Sorry I actually used n_filter for the first conv_block

7

the number of filter in conv_block 5 must be 16 * n_filter as the fourth time I doubled it

8

thank you :heart_eyes: , the error is actually in conv_block5 I have used 16n_classes* instead of n_fillters

1 Like