Quantifying the Carbon Emissions of Machine Learning

Abstract

From an environmental standpoint, there are a few crucial aspects of training a neural network that have a major impact on the quantity of carbon that it emits. These factors include: the location of the server used for training and the energy grid that it uses, the length of the training procedure, and even the make and model of hardware on which the training takes place. In order to approximate these emissions, we present our Machine Learning Emissions Calculator, a tool for our community to better understand the environmental impact of training ML models. We accompany this tool with an explanation of the factors cited above, as well as concrete actions that individual practitioners and organizations can take to mitigate their carbon emissions.

Machine Learning Emissions Calculator: https://mlco2.github.io/impact/

References in corpus (4)

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Convolutional Neural Networks for Medical Image Analysis: Full Training or Fine Tuning?
• NeuralPower: Predict and Deploy Energy-Efficient Convolutional Neural Networks
• Computational Physics on Graphics Processing Units

Cited by in corpus (49)

• On the Opportunities and Risks of Foundation Models
• Multitask Prompted Training Enables Zero-Shot Task Generalization
• ChemBERTa: Large-Scale Self-Supervised Pretraining for Molecular Property Prediction
• Sustainable AI: Environmental Implications, Challenges and Opportunities
• Towards the Systematic Reporting of the Energy and Carbon Footprints of Machine Learning
• Carbon Emissions and Large Neural Network Training
• Carbontracker: Tracking and Predicting the Carbon Footprint of Training Deep Learning Models
• FMix: Enhancing Mixed Sample Data Augmentation
• Data augmentation instead of explicit regularization
• Robust deep learning-based semantic organ segmentation in hyperspectral images
• A first look into the carbon footprint of federated learning
• Towards Green Automated Machine Learning: Status Quo and Future Directions
• Analyzing Sustainability Reports Using Natural Language Processing
• Large-scale, Language-agnostic Discourse Classification of Tweets During COVID-19
• CCVS: Context-aware Controllable Video Synthesis
• Localising In-Domain Adaptation of Transformer-Based Biomedical Language Models
• Investigating Cosmological GAN Emulators Using Latent Space Interpolation
• Dissecting FLOPs along input dimensions for GreenAI cost estimations
• The MultiBERTs: BERT Reproductions for Robustness Analysis
• C5T5: Controllable Generation of Organic Molecules with Transformers
• RETRIEVE: Coreset Selection for Efficient and Robust Semi-Supervised Learning
• Graph Neural Networks with Local Graph Parameters
• DASS: Differentiable Architecture Search for Sparse neural networks
• Densely connected normalizing flows
• TöRF: Time-of-Flight Radiance Fields for Dynamic Scene View Synthesis
• Transformers for Low-Resource Languages: Is Féidir Linn!
• Implicit Deep Adaptive Design: Policy-Based Experimental Design without Likelihoods
• Supervising the Transfer of Reasoning Patterns in VQA
• Curb Your Carbon Emissions: Benchmarking Carbon Emissions in Machine Translation
• Data augmentation and image understanding
• Low Distortion Block-Resampling with Spatially Stochastic Networks
• Greenformer: Factorization Toolkit for Efficient Deep Neural Networks
• Optimizing Reusable Knowledge for Continual Learning via Metalearning
• Robust Embeddings Via Distributions
• ClimateGAN: Raising Climate Change Awareness by Generating Images of Floods
• An energy-based comparative analysis of common approaches to text classification in the Legal domain
• On Contrastive Representations of Stochastic Processes
• PCA-based Multi Task Learning: a Random Matrix Approach
• Statistical discrimination in learning agents
• Toward Foundation Models for Earth Monitoring: Proposal for a Climate Change Benchmark
• Independent Ethical Assessment of Text Classification Models: A Hate Speech Detection Case Study
• TABBIE: Pretrained Representations of Tabular Data
• NeurIPS 2020 NLC2CMD Competition: Translating Natural Language to Bash Commands
• CARMS: Categorical-Antithetic-REINFORCE Multi-Sample Gradient Estimator
• Can Federated Learning Save The Planet?
• Genealogical Population-Based Training for Hyperparameter Optimization
• Self-Supervised Visual Representation Learning Using Lightweight Architectures
• Distributed Reconstruction Algorithm for Electron Tomography with Multiple-scattering Samples
• A Survey on Green Deep Learning