7 papers
SAKI: Score-Aware Low-Rank Key Indexing with Random-Matrix Noise Correction for KV Retrieval
Lin Zhang
Existing low rank KV cache methods preserve either model weights or key variance, neither of which directly reflects the attention scores used during inference. We derive the expec…
Zellige: Moldable Sequence Placement for Mixed Image-Video DiT Training
Guangyu Xiang, Xueze Kang, Minwei Zhao +4
High-quality video generation requires training Diffusion Transformers (DiTs) jointly on image and video data, posing a mixed-length sequence training problem across GPUs. Existing…
Xema: Efficient Diffusion Serving through Fine-Grained Memory Management and Auto-Configuration
Xueze Kang, Guangyu Xiang, Suyi Li +4
Xema is a system that reduces GPU memory usage for diffusion model serving by analyzing tensor lifetimes to apply targeted memory mitigation and by planning parallelism and concurr…
KernelFlume: Elastic Core-Attention Scaling for Agentic Long-Context Decoding
Guangyu Xiang, Xueze Kang, Lin Zhang +4
LLM serving is increasingly dominated by long and dynamic decode workloads from agents, reasoning models, and extended conversations. When bursty long-context demand exceeds deploy…
HyLRA: Hybrid Layer Reuse Attention for Efficient Long-Context Inference
Xuan Ai, Qingqing Yang, Peng Wang +4
Long-context inference in Large Language Models (LLMs) is bottlenecked by the quadratic computation complexity of attention and the substantial memory footprint of Key-Value (KV) c…
HierMoE: Accelerating MoE Training with Hierarchical Token Deduplication and Expert Swap
Wenxiang Lin, Xinglin Pan, Lin Zhang +3
The sparsely activated mixture-of-experts (MoE) transformer has become a common architecture for large language models (LLMs) due to its sparsity, which requires fewer computationa…