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

Abstract

Accurate reporting of energy and carbon usage is essential for understanding the potential climate impacts of machine learning research. We introduce a framework that makes this easier by providing a simple interface for tracking realtime energy consumption and carbon emissions, as well as generating standardized online appendices. Utilizing this framework, we create a leaderboard for energy efficient reinforcement learning algorithms to incentivize responsible research in this area as an example for other areas of machine learning. Finally, based on case studies using our framework, we propose strategies for mitigation of carbon emissions and reduction of energy consumption. By making accounting easier, we hope to further the sustainable development of machine learning experiments and spur more research into energy efficient algorithms.

Published in JMLR: https://jmlr.org/papers/v21/20-312.html

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Quantifying the Carbon Emissions of Machine Learning
• Single Headed Attention RNN: Stop Thinking With Your Head
• Accelerating Deep Learning by Focusing on the Biggest Losers

Cited by in corpus (25)

• How I failed machine learning in medical imaging -- shortcomings and recommendations
• Sustainable AI: Environmental Implications, Challenges and Opportunities
• Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models
• Impact of Tokenization on Language Models: An Analysis for Turkish
• The climate impact of ICT: A review of estimates, trends and regulations
• A first look into the carbon footprint of federated learning
• Multi-Mask Self-Supervised Learning for Physics-Guided Neural Networks in Highly Accelerated MRI
• Towards Sustainable Artificial Intelligence: An Overview of Environmental Protection Uses and Issues
• An astronomical institute's perspective on meeting the challenges of the climate crisis
• From Clicks to Carbon: The Environmental Toll of Recommender Systems
• When BERT Plays the Lottery, All Tickets Are Winning
• A Synthesis of Green Architectural Tactics for ML-Enabled Systems
• Efficiency is Not Enough: A Critical Perspective of Environmentally Sustainable AI
• Greening AI-enabled Systems with Software Engineering: A Research Agenda for Environmentally Sustainable AI Practices
• Length-Adaptive Transformer: Train Once with Length Drop, Use Anytime with Search
• Streamlining Tensor and Network Pruning in PyTorch
• Power Consumption Variation over Activation Functions
• adaptNMT: an open-source, language-agnostic development environment for Neural Machine Translation
• IrEne: Interpretable Energy Prediction for Transformers
• Towards Accurate and Reliable Energy Measurement of NLP Models
• TDprop: Does Jacobi Preconditioning Help Temporal Difference Learning?
• Supervision Accelerates Pre-training in Contrastive Semi-Supervised Learning of Visual Representations
• Highly Efficient Knowledge Graph Embedding Learning with Orthogonal Procrustes Analysis
• NeurIPS 2020 NLC2CMD Competition: Translating Natural Language to Bash Commands
• Can Federated Learning Save The Planet?