Scaling Video Training with Parallelism

Yukang Chen, Luozhou Wang, Wei Huang, Shuai Yang, Weian Mao, Song Han
Jun 3, 2026 1 min read

We published a new blog post: Scaling Video Training with Parallelism.

Long-video training changes the unit of distributed computation: instead of only splitting across samples, the system must split inside one long video sample. The post discusses sequence parallelism for long-video understanding and generation, including LongVILA MM-SP and LongLive-2.0 Balanced SP.

Video Generation Long Video Training Sequence Parallelism Infrastructure
Yukang Chen
Yukang Chen
Senior Research Scientist

Senior Research Scientist at NVIDIA Research.

Song Han
Song Han
Associate Professor

Song Han is an associate professor at MIT EECS.