Thank you for your work, I have a question about sequence embedding. The screenshot is from transformer.py
When you get sequence embedding, the position embedding has already been added to sequence embedding as fllows:
Why do you input the same position embedding into decoder layer ? After this operation, position embedding is added to sequence embedding twice.
Thank you for your work, I have a question about sequence embedding. The screenshot is from transformer.py When you get sequence embedding, the position embedding has already been added to sequence embedding as fllows:
Why do you input the same position embedding into decoder layer ? After this operation, position embedding is added to sequence embedding twice.
![image](https://user-images.githubusercontent.com/19609285/142346573-325d1a61-a73c-4472-b271-ff802f9164fa.png)