OmniCache: A Trajectory-Oriented Global Perspective on Training-Free Cache Reuse for Diffusion Transformer Models

Documenti analoghi

• QNCD: Quantization Noise Correction for Diffusion Models
  di: Chu, Huanpeng, et al.
  Pubblicazione: (2024)
• AdaCorrection: Adaptive Offset Cache Correction for Accurate Diffusion Transformers
  di: Liu, Dong, et al.
  Pubblicazione: (2026)
• Rethinking Token-wise Feature Caching: Accelerating Diffusion Transformers with Dual Feature Caching
  di: Zou, Chang, et al.
  Pubblicazione: (2024)
• FastCache: Fast Caching for Diffusion Transformer Through Learnable Linear Approximation
  di: Liu, Dong, et al.
  Pubblicazione: (2025)
• CacheQuant: Comprehensively Accelerated Diffusion Models
  di: Liu, Xuewen, et al.
  Pubblicazione: (2025)