Attend to Not Attended: Structure-then-Detail Token Merging for Post-training DiT Acceleration

Documenti analoghi

• HeadRouter: A Training-free Image Editing Framework for MM-DiTs by Adaptively Routing Attention Heads
  di: Xu, Yu, et al.
  Pubblicazione: (2024)
• PTQ4DiT: Post-training Quantization for Diffusion Transformers
  di: Wu, Junyi, et al.
  Pubblicazione: (2024)
• Where MLLMs Attend and What They Rely On: Explaining Autoregressive Token Generation
  di: Chen, Ruoyu, et al.
  Pubblicazione: (2025)
• Q-DiT: Accurate Post-Training Quantization for Diffusion Transformers
  di: Chen, Lei, et al.
  Pubblicazione: (2024)
• Reasoning to Attend: Try to Understand How <SEG> Token Works
  di: Qian, Rui, et al.
  Pubblicazione: (2024)