@misc{he2024fastdecode,
  title={FastDecode: High-Throughput GPU-Efficient LLM Serving using Heterogeneous Pipelines},
  author={Jiaao He and Jidong Zhai},
  year={2024},
  eprint={2403.11421},
  archivePrefix={arXiv},
  primaryClass={cs.DC},
  url={https://arxiv.org/abs/2403.11421}
}