Towards Sustainable NLP: Insights from Benchmarking Inference Energy in Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Poddar, Soham, Koley, Paramita, Misra, Janardan, Podder, Sanjay, Ganguly, Niloy, Ghosh, Saptarshi
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910876518318080
author Poddar, Soham
Koley, Paramita
Misra, Janardan
Podder, Sanjay
Ganguly, Niloy
Ghosh, Saptarshi
author_facet Poddar, Soham
Koley, Paramita
Misra, Janardan
Podder, Sanjay
Ganguly, Niloy
Ghosh, Saptarshi
contents Large language models (LLMs) are increasingly recognized for their exceptional generative capabilities and versatility across various tasks. However, the high inference costs associated with these models have not received adequate attention, particularly when compared to the focus on training costs in existing research. In response to this gap, our study conducts a comprehensive benchmarking of LLM inference energy across a wide range of NLP tasks, where we analyze the impact of different models, tasks, prompts, and system-related factors on inference energy. Specifically, our experiments reveal several interesting insights, including strong correlation of inference energy with output token length and response time. Also, we find that quantization and optimal batch sizes, along with targeted prompt phrases, can significantly reduce energy usage. This study is the first to thoroughly benchmark LLM inference across such a diverse range of aspects, providing insights and offering several recommendations for improving energy efficiency in model deployment.
format Preprint
id arxiv_https___arxiv_org_abs_2502_05610
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Towards Sustainable NLP: Insights from Benchmarking Inference Energy in Large Language Models
Poddar, Soham
Koley, Paramita
Misra, Janardan
Podder, Sanjay
Ganguly, Niloy
Ghosh, Saptarshi
Computation and Language
Large language models (LLMs) are increasingly recognized for their exceptional generative capabilities and versatility across various tasks. However, the high inference costs associated with these models have not received adequate attention, particularly when compared to the focus on training costs in existing research. In response to this gap, our study conducts a comprehensive benchmarking of LLM inference energy across a wide range of NLP tasks, where we analyze the impact of different models, tasks, prompts, and system-related factors on inference energy. Specifically, our experiments reveal several interesting insights, including strong correlation of inference energy with output token length and response time. Also, we find that quantization and optimal batch sizes, along with targeted prompt phrases, can significantly reduce energy usage. This study is the first to thoroughly benchmark LLM inference across such a diverse range of aspects, providing insights and offering several recommendations for improving energy efficiency in model deployment.
title Towards Sustainable NLP: Insights from Benchmarking Inference Energy in Large Language Models
topic Computation and Language
url https://arxiv.org/abs/2502.05610