I am not sure the dimension for the final two Dense layers, how much units should I create?
I presume you’re referring to the last part of the Transformer class “call” method.
All you need to use is the final_layer() method and pass it the decoder output.
Yes, I do, but since it mentioned we need two fully connected layers, thus I also add a Dense layer with relu activation, But I need to add units for it, That’s where I got stuck.
Ok, I just try to remove the Dense layer that I added… well, it turns out, I just need to add the final layer, and it work.
That’s a bit confusing, I mean, based on previous text, We need two layers, but the final layer is just one layer.
That’s a mistake in the notebook instructions. I’ll report it to the staff.