https://research.nvidia.com/labs/eai/author/luozhou-wang/Luozhou WangHugo Blox Builder (https://hugoblox.com)en-usWed, 03 Jun 2026 00:00:00 +0000https://research.nvidia.com/labs/eai/media/icon_hu_67e9a0fea4325d02.pnghttps://research.nvidia.com/labs/eai/author/luozhou-wang/https://research.nvidia.com/labs/eai/post/scaling-video-training-with-parallelism/Wed, 03 Jun 2026 00:00:00 +0000https://research.nvidia.com/labs/eai/post/scaling-video-training-with-parallelism/<p>We published a new blog post: <a href="https://research.nvidia.com/labs/eai/blogs/scaling-video-training-with-parallelism/">Scaling Video Training with Parallelism</a>.</p> <p>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.</p>https://research.nvidia.com/labs/eai/post/video-gen-is-an-infra-problem/Tue, 26 May 2026 00:00:00 +0000https://research.nvidia.com/labs/eai/post/video-gen-is-an-infra-problem/<p>We published a new blog post: <a href="https://research.nvidia.com/labs/eai/blogs/video-gen-is-an-infra-problem/">Why Video Gen Is an Infra Problem</a>.</p> <p>A good video model is no longer the whole story; practical video generation also requires memory, decoding, scheduling, compression, parallel execution, and deployment-aware design.</p>