Unlike in RNNs, inputs into a transformer need to be encoded with positions. In this video, I showed how positional encoding are computed using a simple numerical example.