4 papers indexed
Xiaohan Zhang, Yuqing Wen, Junlin Chen, Yuqi Tang, Yiting He, Lizhuo Shao, et al.
Multi-reference-to-audio-video (MR2AV) generation aims to generate coherent audio-video content conditioned on multiple references and textual instructions. Existing benchmarks mainly focus on text-driven generation, single-reference subject preservation, or isolated audio-video…
Xiang Zhao, Ronghui Quan, Yaqi Xiao, Junlin Chen
Parachutes are widely utilized in aviation, aerospace and lifesaving missions. As the initial stage of parachute deployment, suspension line extraction and straightening directly determines the smooth implementation of subsequent inflation procedures. This ultra-short process inv…
Mingkai Zheng, Junlin Chen, Haotian Xie, Zhao Zhang
Communication increasingly dominates the cost of Large Language Model (LLM) pre-training, especially under data-parallel and sharded training schemes, where gradient synchronization and parameter reconstruction overhead increase with model size and system scale. Existing communic…
Haotian Xie, Junlin Chen, Mingkai Zheng, Lishan Yang, Zhao Zhang
State-of-the-art large language model (LLM) training takes tens of thousands of graphics processing units (GPUs) for months and encounters failures across the software and hardware stack. Existing fault-tolerance mechanisms either impose non-trivial overhead during failure-free e…