Error in C5_W4_A1 Part4(EncoderLayer) instructions

Course Q&A Deep Learning Specialization Sequence Models
,
1

In version “v1.6” of the C5_W4_A1_Transformer_Subclass_v1 there is an error in the comments of Part 4’s coding assignment that is bound to trip people up.

Where it says “pass the output of the multi-head attention layer through a ffn (~1 line)” in exercise 4 it should say “pass the output of the normalized multi-head attention layer through a ffn (~1 line)”

2

Thanks for your report.

1 Like
3

Maybe that would make it clearer, but here’s the comment in the template code for the previous line of code:

# skip connection
# apply layer normalization on sum of the input and the attention output to get the  
# output of the multi-head attention layer (~1 line)

In other words what they mean by “the output of the multi-head attention layer” is only ambiguous if you missed the point of the previous comment. And the diagram in Figure 2a.

But more clarity is never a bad thing :nerd_face: …