Medium
How Self-Attention with Relative Position Representations works
An explanation of how Relative Position Representation embeddings allow Transformers to encode sequential information in an input sequence.
See this content immediately after install