I-max: Maximize the resolution potential of pre-trained rectified flow transformers with...

[Free GPT-4]
[PDF] arxiv.org

CLEAR: Conv-Like Linearization Revs Pre-Trained Diffusion Transformers Up

S Liu, Z Tan, X Wang - arxiv preprint arxiv:2412.16112, 2024 - arxiv.org
Diffusion Transformers (DiT) have become a leading architecture in image generation.
However, the quadratic complexity of attention mechanisms, which are responsible for …
保存 引用 相关文章 HTML 版