7 papers
ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
Sangjin Choi, Sukmin Cho, Yifan Xiong +3
In prefill-decode (PD) disaggregated LLM serving, each request is assigned to a decode worker after prefill. Existing decode routers balance only load; for mixture-of-experts (MoE)…
MSCCL++: Rethinking GPU Communication Abstractions for AI Inference
Changho Hwang, Peng Cheng, Roshan Dathathri +12
AI applications increasingly run on fast-evolving, heterogeneous hardware to maximize performance, but general-purpose libraries lag in supporting these features. Performance-minde…
Sigma-MoE-Tiny Technical Report
Qingguo Hu, Zhenghao Lin, Ziyue Yang +12
Mixture-of-Experts (MoE) has emerged as a promising paradigm for foundation models due to its efficient and powerful scalability. In this work, we present Sigma-MoE-Tiny, an MoE la…
SIGMA: An AI-Empowered Training Stack on Early-Life Hardware
Lei Qu, Lianhai Ren, Peng Cheng +12
An increasing variety of AI accelerators is being considered for large-scale training. However, enabling large-scale training on early-life AI accelerators faces three core challen…
ForestColl: Throughput-Optimal Collective Communications on Heterogeneous Network Fabrics
Liangyu Zhao, Saeed Maleki, Yuanhong Wang +4
As modern DNN models grow ever larger, collective communications between the accelerators (allreduce, etc.) emerge as a significant performance bottleneck. Designing efficient comm…
Optimizing Large Language Model Training Using FP4 Quantization
Ruizhe Wang, Yeyun Gong, Xiao Liu +5
The growing computational demands of training large language models (LLMs) necessitate more efficient methods. Quantized training presents a promising solution by enabling low-bit…