Yukang Chen

@yukangchen_

Research Scientist @NVIDIA, work in Efficient and Long AI.

Boston USA

yukangchen.com

Replies Reposts Media

Pinned
Yukang Chen
@yukangchen_
May 19
🚀 Excited to release LongLive 2.0! 🎬 An end-to-end infrastructure for long video generation, with FP4 and parallelism at the core of both training and inference. ⚡45.7 FPS generation speed on 5B model⚡ ✨ LongLive 2.0 supports real-video training, few-step distillation,
00:00
7
Yukang Chen
@yukangchen_
Aug 4
🚀 Very excited that TriAttention has been integrated into NVIDIA TensorRT-LLM! ⚡️ TriAttention is an agent-friendly and infra-aware KV cache compression method, for efficient LLM inference. 🔗 Link:
TensorRT-LLM/examples/kv_cache_compression/triattention.md at main · NVIDIA/TensorRT-LLM
From github.com
4
Yukang Chen
@yukangchen_
Jul 15
We are excited to share a new blog. Hope you enjoy reading.
Shuai Yang
@ShuaiYa68505475
Jul 15
Our blog surveys the evolving landscape of autoregressive video generation—how causal models are turning video diffusion into a practical foundation for real-time, interactive, and long generation research.nvidia.com/labs/eai/blogs… @yukangchen_ @AaronWeiHuang @WeianMaoX @songhan_mit
00:00
Yukang Chen
@yukangchen_
Jun 30
We are excited to share a new blog “Pushing Intelligence to 4-bit.” research.nvidia.com/labs/eai/blogs… This blog discusses how FP4 and Blackwell hardware make 4-bit floating point practical for training and inference, covering LLMs, KV cache, attention, and Video Gen.
Aaron Huang
@AaronWeiHuang
Jun 30
🔗 Our new blog looks at how FP4 is moving beyond compression into a practical primitive for training and inference across both LLMs and diffusion models: research.nvidia.com/labs/eai/blogs… 1. Why Four Bits Is Hard: Only 15 values make scaling critical. 2. NVFP4: Smaller blocks and finer
00:00
Yukang Chen
@yukangchen_
Jun 24
Birthday today, and my citations crossed 10,000 — couldn’t ask for a better gift. Grateful to all my collaborators, and excited to keep pushing forward! 🎂📚✨
10