Abstract:The advent of larger machine learning (ML) models have improved state-of-the-art (SOTA) performance in various modeling tasks, ranging from computer vision to natural language. As ML models continue increasing in size, so does their respective energy consumption and computational requirements. However, the methods for tracking, reporting, and comparing energy consumption remain limited. We presentEnergyVis, an interactive energy consumption tracker for ML models. Consisting of multiple coordinated views, EnergyVis enables researchers to interactively track, visualize and compare model energy consumption across key energy consumption and carbon footprint metrics (kWh and CO2), helping users explore alternative deployment locations and hardware that may reduce carbon footprints. EnergyVis aims to raise awareness concerning computational sustainability by interactively highlighting excessive energy usage during model training; and by providing alternative training options to reduce energy usage.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the ever - increasing energy consumption and computational requirements in the training and evaluation of machine learning (ML) models, as well as the current lack of effective tools to track, report, and compare the energy consumption of these models. As the scale of ML models continues to expand, their energy consumption and carbon footprint are also increasing day by day, which has a negative impact on the environment and makes it difficult for some small research teams and companies to bear the high computational costs. Specifically, the paper points out: 1. **Energy Consumption Problem**: The number of parameters in large - scale ML models has increased by more than 300,000 times from 2012 to 2017, resulting in huge energy consumption. For example, the carbon emissions generated by training a state - of - the - art natural language processing (NLP) model (such as BERT) are equivalent to the carbon emissions of a trans - Atlantic flight. 2. **Inequality in Computational Resources**: As the scale of models expands, the demand for computational resources also increases, which makes it impossible for many small research teams and companies to replicate and use these models, thus exacerbating the inequality in computational resources. 3. **Lack of Effective Tracking Tools**: Most of the existing tools track and report energy consumption through the command - line interface or static charts, lacking interactivity and real - time performance, and unable to provide users with sufficient information to optimize energy use during the model training process. To solve these problems, the authors propose EnergyVis, an interactive energy consumption tracking system, aiming to help researchers and practitioners: - **Real - Time Tracking and Visualization**: Through multiple coordinated views, users can track and visualize the energy consumption and carbon footprint of ML models in real - time. - **Explore Alternatives**: Provide choices of different hardware and geographical locations to help users find more energy - efficient training methods. - **Raise Awareness of Computational Sustainability**: By highlighting excessive energy use, prompt users to consider more environmentally - friendly training options. In conclusion, the goal of EnergyVis is to provide an interactive platform, enabling users to better understand and manage the energy consumption of ML models, thereby reducing the carbon footprint and promoting the fair distribution of computational resources.

EnergyVis: Interactively Tracking and Exploring Energy Consumption for ML Models

Towards Low-Budget Energy Efficiency Design in Additive Manufacturing Based on Variational Scale-Aware Transformer

Computing Within Limits: An Empirical Study of Energy Consumption in ML Training and Inference

Full-Cycle Energy Consumption Benchmark for Low-Carbon Computer Vision

Carbon Emissions and Large Neural Network Training

Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models

Accuracy is not the only Metric that matters: Estimating the Energy Consumption of Deep Learning Models

Impact of ML Optimization Tactics on Greener Pre-Trained ML Models

eco2AI: Carbon Emissions Tracking of Machine Learning Models as the First Step Towards Sustainable AI

Towards energy-efficient Deep Learning: An overview of energy-efficient approaches along the Deep Learning Lifecycle

Watt For What: Rethinking Deep Learning's Energy-Performance Relationship

Unveiling Energy Efficiency in Deep Learning: Measurement, Prediction, and Scoring across Edge Devices

Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning

Normalizing Energy Consumption for Hardware-Independent Evaluation

How to use model architecture and training environment to estimate the energy consumption of DL training

How to estimate carbon footprint when training deep learning models? A guide and review

Great Power, Great Responsibility: Recommendations for Reducing Energy for Training Language Models

How green is continual learning, really? Analyzing the energy consumption in continual training of vision foundation models

Towards a Catalog of Energy Patterns in Deep Learning Development

How to Sustainably Monitor ML-Enabled Systems? Accuracy and Energy Efficiency Tradeoffs in Concept Drift Detection

EMERS: Energy Meter for Recommender Systems