Transformers Without the RNN

The previous post walked through attention as an add-on to RNNs — a way to let the decoder look back at encoder hidden states instead of… Continue reading on Towards AI »