C5W4A1 - Encoder layer error

[Kavya_Prasad]

Hi,

I’m using the following:
self_mha_output = self.mha(query = x, key = x, value = x, attention_mask = mask)(x)

But I see the error below:

Any help is appreciated - been stuck for a long time!

Thanks!

[TMosh]

In EncoderLayer(), you’re only defining the layer. So you do not need to provide the (x) data argument. Just leave that off.

[Kavya_Prasad]

Thanks a lot, this worked!

I’m facing a different error in the cell after that though, for the Encoder layer itself:

I’m using tf.math.sqrt in
x *= tf.math.sqrt(self.embedding_dim)

But this seems to be causing an issue. I’ve tried using tf.cast for self.embedding_dim (this causes an error in the following line of code), tf.keras.backend.sqrt instead of tf.math.sqrt, and np.sqrt, none of which work. Could you give me some suggestions?

[TMosh]

Did you read this line from the instructions?

You can see how to use tf.cast() in other places in the notebook.