General Debugging

Course Q&A Deep Learning Specialization Sequence Models
1

When faced with an error like this:

IndexError Traceback (most recent call last)
in
----> 1 parameters, last_name = model(data.split(“\n”), ix_to_char, char_to_ix, 22001, verbose = True)
2
3 assert last_name == ‘Trodonosaurus\n’, “Wrong expected output”
4 print(“\033[92mAll tests passed!”)

in model(data_x, ix_to_char, char_to_ix, num_iterations, n_a, dino_names, vocab_size, verbose)
61 # Perform one optimization step: Forward-prop → Backward-prop → Clip → Update parameters
62 # Choose a learning rate of 0.01
—> 63 curr_loss, gradients, a_prev = optimize(X, Y, a_prev, parameters, learning_rate = 0.01)
64
65 ### END CODE HERE ###

in optimize(X, Y, a_prev, parameters, learning_rate)
32
33 # Forward propagate through time (≈1 line)
—> 34 loss, cache = rnn_forward(X, Y, a_prev, parameters)
35
36 # Backpropagate through time (≈1 line)

~/work/W1A2/utils.py in rnn_forward(X, Y, a0, parameters, vocab_size)
94 x[t] = np.zeros((vocab_size,1))
95 if (X[t] != None):
—> 96 x[t][X[t]] = 1
97
98 # Run one step forward of the RNN

IndexError: only integers, slices (:), ellipsis (...), numpy.newaxis (None) and integer or boolean arrays are valid indices

If the most recent call is last, should I start from the beginning (oldest/first call) when looking for the bug that is causing the issues. Or should I start at the last most recent call… in this case ‘rnn_forward.’

I’ve mostly been using the last most recent parts of errors to debug thus far in the DLS courses. I’m wondering if theres any more information I can squeeze out of this error message.
Thanks

2

Start from the bottom. The “IndexError:…” message refers to the line of code marked with line 96, where the arrow points.

The text just above that says that locates the error in the rnn_forward() function.

2 Likes
3

Yes, you can use bottom to top approach. Check the bottom most arrow, then the 2nd last arrow and so on. Moving from bottom to top and check where bug is…

Best,
Saif.

1 Like
5

Ok thanks. So its telling me I’m incorrectly indexing and points to an index that I don’t do inside of a function that I only reference, didn’t write myself. So if I didn’t write the function, and I didn’t write how they’re using it. The only thing I can think of is it doesn’t like my initialization of x with np.zeros((vocab_size, 1)). But that wouldn’t explain an index problem would it?
also worth noting I think that whole rnn_forwards and x initialization happened in a different cell that I already passed the tests for.

6

Sorry, i didn’t review the error stack far enough, as I was thinking of a different assignment.

If you go further toward the top by one more error, you see that the error is thrown by your code in optimize(), where it calls rnn_forward().

Then going back one error further, you’re calling optimize() from model().

So, I’d say there’s a problem in your model() function where it computes one of the parameters that is passed to optimize() - that’s either X, Y, or a_prev.

1 Like
7

Thank you.

x is none + single example ix, none in square brackets
y is just the end of y appended to the new line. so since y = x+1 Im just doing x[1:] plus [ixnewline] which is just the \n.
the problem could be a_prev, I’m initializing it to zeros shape na,1 but then I dont touch it again before it enters the optimization loop and is used by optimize…

Thanks