C4W1 Exercise4 Translator

Course Q&A Natural Language Processing NLP with Attention Models
1

I passed the quick check but the unit test gave me the error as below. I wonder what could be done wrong?

image
image3486×1548 487 KB

2

Hi @Yizhuo_Han

You probably hard coded 12000 value somewhere where you should have used vocab_size parameter (most probably in the decoder’s output layer).

Cheers

1 Like
4

Hi, I have not hard coded 12000… for the decoder output layer I have self.output_layer = tf.keras.layers.Dense(
units=vocab_size,
activation=‘log_softmax’
)
I wonder if i used the wrong log_softmax activation function? dont know if it matters

5

From the instructions:

Make sure to use a logsoftmax activation function for this one, which you can get as tf.nn.log_softmax.

It’s a shame, but in my experience in TensorFlow it matters (whether “log_softmax” or tf.nn.log_softmax).