Hypernova 60B

New technique slashes knowledge distillation costs for LLMs

Researchers have developed a more efficient method for knowledge distillation in large language models, significantly reducing the computational cost and memory requirements. This new technique involves caching the teac…