mlsanityData to 14 Oct 2026

EnergyLens: Predictive Energy-Aware Exploration for Multi-GPU LLM Inference Optimization

arXiv 2605.14249 ↗First posted 14 May 2026cs.LGcited by 0

On the map: Mixture-of-experts → Efficient MoE training and inference systems

Abstract (arXiv)

We present EnergyLens, an end-to-end framework for energy-aware large language model (LLM) inference optimization. As LLMs scale, predicting and reducing their energy footprint has become critical for sustainability and datacenter operations, yet existing approaches either require production-level code and expensive profiling or fail to accurately capture multi-GPU energy behavior. As a result, practitioners lack tools for deciding which optimizations to prioritize and for selecting among existing deployment configurations when exhaustive profiling is impractical. EnergyLens addresses this gap with an intuitive einsum-based interface that captures LLM specifications including fusion, parallelism, and compute-communication overlap, combined with load-imbalance-aware MoE modeling and an empirically driven communication energy model for multi-GPU settings. We validate EnergyLens on Llama3 and Qwen3-MoE across tensor-parallel and expert-parallel configurations, achieving mean absolute percentage errors (MAPEs) between 9.25% and 13.19% for multi-GPU prefill and decode energy, and 12.97% across SM allocations for Megatron-style overlap. Our energy-driven exploration reveals up to 1.47x and 52.9x energy variation across configurations in prefill and decode efficiency and motivates distributed serving. We further show that compute-communication overlap is difficult to optimize with intuition alone, but EnergyLens correctly identifies Pareto-optimal overlap configurations.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Efficient MoE training and inference systems, one of 12 directions in Mixture-of-experts. Builds on 18 of its 26 references that are in the map. First posted on arXiv 14 May 2026.

Builds on

18 of 26 references.

References · 18

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2019Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelismcited by 1,868
  3. 2024DistServe: Disaggregating Prefill and Decoding for Goodput-optimized Large Language Model Servingcited by 617
  4. 2019Quantifying the Carbon Emissions of Machine Learningcited by 521
  5. 2022DeepSpeed Inference: Enabling Efficient Inference of Transformer Models at Unprecedented Scalecited by 407
  6. 2022Estimating the Carbon Footprint of BLOOM, a 176B Parameter Language Modelcited by 265
  7. 2020Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learningcited by 249
  8. 2023DeepSpeed Ulysses: System Optimizations for Enabling Training of Extreme Long Sequence Transformer Modelscited by 245
  9. 2024Vidur: A Large-Scale Simulation Framework For LLM Inferencecited by 117
  10. 2023ASTRA-sim2.0: Modeling Hierarchical Networks and Disaggregated Systems for Large-model Training at Scalecited by 47
  11. 2024LLMServingSim: A HW/SW Co-Simulation Infrastructure for LLM Inference Serving at Scalecited by 31
  12. 2024LLMCO2: Advancing Accurate Carbon Footprint Prediction for LLM Inferencescited by 19
6 more
  1. 2024Forecasting GPU Performance for Deep Learning Training and Inferencecited by 15
  2. 2025Efficient and Adaptable Overlapping for Computation and Communication via Signaling and Reorderingcited by 10
  3. 2026LLMServingSim 2.0: A Unified Simulator for Heterogeneous and Disaggregated LLM Serving Infrastructurecited by 9
  4. 2025Lumos: Efficient Performance Modeling and Estimation for Large-scale LLM Trainingcited by 7
  5. 2025Characterizing Compute-Communication Overlap in GPU-Accelerated Distributed Deep Learning: Performance and Power Implicationscited by 3
  6. 2026EnergAIzer: Fast and Accurate GPU Power Estimation Framework for AI Workloadscited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Watt Counts: Energy-Aware Benchmark for Sustainable LLM Inference on Heterogeneous GPU Architectures
  2. 2026Understanding Efficiency: Quantization, Batching, and Serving Strategies in LLM Energy Use
  3. 2026EnergyLens: Interpretable Closed-Form Energy Models for Multimodal LLM Inference Serving
  4. 2026PALS: Power-Aware LLM Serving for Mixture-of-Experts Models
  5. 2024SLO-aware GPU Frequency Scaling for Energy Efficient LLM Inference Serving
  6. 2025Characterizing LLM Inference Energy-Performance Tradeoffs across Workloads and GPU Scaling
  7. 2024Towards Greener LLMs: Bringing Energy-Efficiency to the Forefront of LLM Inference
  8. 2026OmniPilot: An Uncertainty-Aware LLM Inference Advisor for Heterogeneous GPU Clusters ↗