English-to-Spanish translation with a sequence-to-sequence Transformer: TransformerDecoder second attention_mask maybe wrong

[xwklwwlf]

Issue Type

Bug

Source

source

Keras Version

keras 2.15

Custom Code

Yes

OS Platform and Distribution

mac os

Python version

3.10

GPU model and memory

No response

Current Behavior?

https://github.com/keras-team/keras-io/blob/master/examples/nlp/neural_machine_translation_with_transformer.py
line 362 attention_mask=padding_mask,
if the sequence_length of encoder_inputs and decoder_inputs is different, the code will report an error.
I think attention_mask should be set to padding_mask of encoder_inputs, not padding_mask of decoder_inputs.

Standalone code to reproduce the issue or tutorial link

https://keras.io/examples/nlp/neural_machine_translation_with_transformer/
https://github.com/keras-team/keras-io/blob/master/examples/nlp/neural_machine_translation_with_transformer.py

Relevant log output

No response