Wrong comments in the assignment of C4W2

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

Hi guys,

In the call function of class EncoderLayer. the comments says the shape of x is (batch_size, input_seq_len, fully_connected_dim). I think this is not correct, it should be (batch_size, input_seq_len, embedding_dim). fully_connected_dim displays several times in coments, I think they are almost all wrong. Please correct me If I am wrong. Please correct the comments If I am right.

Thank you very much.

def call(self, x, training, mask):
        """
        Forward pass for the Encoder Layer
        
        Arguments:
            x (tf.Tensor): Tensor of shape (batch_size, input_seq_len, fully_connected_dim)
1 Like
2

Hi @duanxiugang

I agree! I think it should be the dimensionality of the embedding space (embedding_dim) unless you are use a fully connected (dense) layer as part of the embedding process.

3

Hi @duanxiugang

This has been already reported and it’s been worked upon, I am still tagging a staff member, for them to know that the correction is not done yet.

@lucas.coutinho this issue was raised by previous learner where it was told this need to be corrected. Can you check if this was reported in the repo.

Regards
DP

4

Thanks @duanxiugang for reporting this! I’ll have a look.

Also, thanks @Deepti_Prasad for tagging me.

Thanks,
Lucas

1 Like