Compute total loss error

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
6

The inputs to the TF cost function need to be in “samples first” orientation. If the question is why it needs to be this way, please see this thread.