possible opposite explanation of hidden compared to output in notebook #3

[Hadar933]

In your forward method, inside Encoder, you specify that

    #outputs = [src len, batch size, hid dim * num directions]
    #hidden = [n layers * num directions, batch size, hid dim]
    
    #hidden is stacked [forward_1, backward_1, forward_2, backward_2, ...]
    #outputs are always from the last layer

While the dimensions are correct, I think that it is the outputs that are stacked, and the hidden is from the last layer (see GRU docs)