Entity

KV Cache

Key-Value caching mechanism in transformer models that stores previous attention computations to accelerate sequential generation

术语属性

  • type:Optimization
  • purpose:Inference acceleration
  • impact:Memory efficiency

相关文章