Week 4 - decoder Exercise 7

Course Q&A Deep Learning Specialization Sequence Models
1

I have completed the decoderLayer and all tests passed.

However, I got this error for ### Exercise 7 - Decoder

AssertionError Traceback (most recent call last)
in
1 # UNIT TEST
----> 2 Decoder_test(Decoder, create_look_ahead_mask, create_padding_mask)

~/work/W4A1/public_tests.py in Decoder_test(target, create_look_ahead_mask, create_padding_mask)
221 assert tf.is_tensor(x), “Wrong type for x. It must be a dict”
222 assert np.allclose(tf.shape(x), tf.shape(encoderq_output)), f"Wrong shape. We expected { tf.shape(encoderq_output)}"
→ 223 assert np.allclose(x[1, 1], [-0.2715261, -0.5606001, -0.861783, 1.69390933]), “Wrong values in x”
224
225 keys = list(attention_weights.keys())

AssertionError: Wrong values in x

It seems very straightforward, so I have no idea how to look for the problem. What is the hard part that is likely to miss here?

2

Passing the tests for DecoderLayer() does not prove your code is perfect.

So I recommend you go over your DecoderLayer() code in detail and be sure you’ve followed all of the guidance in the notebook.

Particularly: watch where you use training=training. This is a common issue.

Incorrect use of training=training can also cause issues in Decoder()

3

Thanks for the hint. I would never have thought the problem is in the decoderLayer.

4

I used training= 3 times. Two times I set it to be True as I see # Dropout will be applied during training. The last time I set it to =training as suggested? Does this seem to be correct?

5

I set them all equal to train and it worked.

6

Thanks so much

7

In DecoderLayer(), you only use training=training once.