How does Centralpoint handle Multi-Head Attention?

Multi-head-attention models in Centralpoint: Centralpoint operates above whatever attention variant powers your models — full MHA, GQA, MQA — in a model-agnostic platform. The pla