Transformer-model

built on the attention mechanism, which was initially designed as an improvement for encoder-decoder RNNs in sequence-to-sequence tasks like machine translation