1 paper · 1 filter
Sophie Ostmeier, Brian Axelrod, Maya Varma +3
Transformer architectures rely on position encodings to model the spatial structure of input data. Rotary Position Encoding (RoPE) is a widely used method in language models that e…