Diffusion Model의 Reverse sampling을 더 빠르게 하는 새로운 방식으로 Distillation Approach가 제안되었다. 아래 2가지 논문을 살펴보자.

  1. Luhman & Luhman (2021). Knowledge distillation in iterative generative models for improved sampling speed.
  2. Salimans & Ho (2022). Progressive distillation for fast sampling of diffusion models.

Summary

Reference

Websites

[사이트 출처 1]

Papers

[1] Luhman, E., & Luhman, T. (2021). Knowledge distillation in iterative generative models for improved sampling speed. arXiv preprint arXiv:2101.02388.

[2] Salimans, T., & Ho, J. (2022). Progressive distillation for fast sampling of diffusion models. arXiv preprint arXiv:2202.00512.

댓글남기기