DLS course 2 week 2 excersise 6 issue with code

Course Q&A Deep Learning Specialization Improving Deep Neural Networks: Hyperparameter tun
1

KeyError Traceback (most recent call last)
in
1 # train 3-layer model
2 layers_dims = [train_X.shape[0], 5, 2, 1]
----> 3 parameters = model(train_X, train_Y, layers_dims, optimizer = “adam”)
4
5 # Predict

in model(X, Y, layers_dims, optimizer, learning_rate, mini_batch_size, beta, beta1, beta2, epsilon, num_epochs, print_cost)
37 v = initialize_velocity(parameters)
38 elif optimizer == “adam”:
—> 39 v, s = initialize_adam(parameters)
40
41 # Optimization loop

in initialize_adam(parameters)
34 # s[“db” + str(l)] = …
35 # YOUR CODE STARTS HERE
—> 36 v[“dW” + str(l)] = np.zeros(grads[“dW” + str(l)].shape)
37 v[“db” + str(l)] = np.zeros(grads[“db” + str(l)].shape)
38 s[“dW” + str(l)] = np.zeros(grads[“dW” + str(l)].shape)

KeyError: ‘dW3’

got this error in the exercise 6 adam section even though you don’t need to write any of your own code. every cell before that ra perfectly and passed all tests, don’t know what to do!

2

But you did write the code for initialize_adam, which you are calling there, right? It looks like that code is incorrect: it is a mistake to reference the variable grads within initialize_adam, since that is not a parameter of that function, right? Maybe that works in that one test case for that function because there happens to be a global variable grads with the right elements, but it is a mistake. That’s a bug in the test case if it doesn’t catch that error.

3

Yes, it turns out that the tests do not catch this error for initialize_adam: you can reference the global dictionary grads and it still passes the test. The same bug exists for initialize_velocity as well. Eeeeek!

I will file this bug immediately against the tests in the notebook, but in the meantime you just need to avoid stepping on this landmine.