Hi, when putting together the ResNet50 model, my tests seem to not passed even though code is correct and past tests were passed. I get the following error:
After debugging, I realized this is due to a mistake on my end when coding the convolutional block. I set the padding strategy of the third component to ¨same¨ instead of ¨valid¨. In reality, this makes no difference as they will create the same output since it is performing a 1x1 convolution (and so it passes the tests).
{moderator edit - solution code removed}
Thought I would share this in case it was useful for anyone.