Skip to content
Synced · WeChat

FlashAR speeds up pretrained autoregressive image models by 22.9x using 0.05% data

22.9倍加速!FlashAR:仅用0.05%数据,让预训练好的自回归图像模型飞起来

Zhejiang University and the University of Adelaide introduced FlashAR, using 0.05% of the original training data to reduce Emu3.5-Image-34B 512×512 generation latency from 130.10 seconds to 5.68 seconds, while GenEval changed from 80.48 to 80.29.

Why it matters: Zhejiang University and the University of Adelaide propose FlashAR, which fine-tunes on just 0.05% of the original training data and cuts Emu3.5-Image-34B 512x512 generation from 130.10 seconds to 5.68 seconds.

Read the original ↗Export Markdown