tl.AttentionQKV layer is by default nested inside the Serial Combinator

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

As instructed, I am trying to use tl.AttentionQKV inside NMTAttn function, still getting wrong model as tl.AttentionQKV layer is by default nested inside the Serial Combinator.What to do? Anyone can please help.
pip install trax installed trax 1.4.1

Screenshot (726)
Screenshot (726)898×446 23.1 KB

2

Hi shubhsr,

From the screenshot I cannot see anything wrong. Did you manage to resolve this? If not, feel free to send me your notebook as an attachment to a direct message, and I can have a look.