Long-video generation with diffusion transformers (DiTs) produces extremely long token sequences, making attention a dominant inference bottleneck. Dynamic sparse attention reduces computation, but its realized speedup remains limited because irregular query-row execution degrades L2 cache locality and increases HBM tr...
Zi-Jian Dai, Sen-Khoo Han, You-Hui Bai et al.· 0 citations
PEARL is an asynchronous agentic RL system that coordinates external resource elasticity, temporary reuse of idle training GPUs, and adaptive PD execution, and maintains a unified GPU--worker--role state and uses runtime profiles to predict rollout batch completion time.
Ji-Aan Zhu, Wei Gao, You-Hui Bai et al.· 0 citations
DeaMoE is proposed, a decoding-efficient MoE architecture, in which the experts are grouped into several departments, and the experts belonging to the same department share most parameters since they come from the same professional field, and additionally each expert contains a few private parameters to reflect its uni...
Ze-Wen Jin, Sheng-Yu Fu, Zeping Duan et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.