The authors first reparameterize causal Conv3D layers into a spatial component and a temporal correction term.
LightVAE speeds up video autoencoders by exploiting temporal and depth redundancy
The authors achieve up to 2.97× decoding throughput on MiniMax-H3 while retaining 99.51% of its VBench-2.0 score.
Chinese Tech
Shiqiao Gu · Yang Yong · Fengjun Zhong · Ruihao Gong · Chengtao Lv · Xiangyu Fan · +2 more
SenseTime · Beihang University · Nanyang Technological University
Research Digest··2 min read
The authors present LightVAE, a compression framework for pretrained video VAEs that accelerates decoding with minimal quality loss.
Why this paper
From SenseTime and 2 others
In one line
LightVAE accelerates video VAE decoding by up to 2.24x through temporal low rank compression and block pruning.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ✓Compute or model size stated (hardware NVIDIA H100)
- ·No stated limitations found
- ✓Reports numbers on named benchmarks (4 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§