5 papers
G-Power: Architecture-level GPU Power Modeling with Aggregated Knowledge Foundations from Known GPUs
Qijun Zhang, Yao Lu, Shang Liu +4
Graphics Processing Units (GPUs) have been serving as critical computation resources for large-scale parallel computations. With increasing chip complexity, power efficiency has be…
Full-Pipeline Inference Optimization for MiMo-V2.5 Series: Pushing Hybrid SWA Efficiency to the Limit
Xiaomi MiMo Team, Anqi Liu, Aoxin Ma +28
We present a full-pipeline inference optimization for the MiMo-V2.5 model family, which combines Hybrid Sliding Window Attention (Hybrid SWA), sparse Mixture-of-Experts (MoE), and…
MoE-Hub: Taming Software Complexity for Seamless MoE Overlap with Hardware-Accelerated Communication on Multi-GPU Systems
Zhuoshan Zhou, Chen Zhang, Shuyi Zhang +10
The Mixture-of-Experts (MoE) architecture is crucial for scaling large language models, but its scalability is severely limited by inter-GPU communication bottlenecks in multi-GPU…
Towards Compute-Aware In-Switch Computing for LLMs Tensor-Parallelism on Multi-GPU Systems
Chen Zhang, Qijun Zhang, Zhuoshan Zhou +10
Tensor parallelism (TP) in large-scale LLM inference and training introduces frequent collective operations that dominate inter-GPU communication. While in-switch computing, exempl…
Accelerating MoE with Dynamic In-Switch Computing on Multi-GPUs
Qijun Zhang, Chen Zhang, Zhuoshan Zhou +10
Mixture-of-Experts (MoE) has been adopted by many leading large models to reduce computational requirements. However, frequent inter-GPU communication in MoE expert parallelism (EP…