We present GreenAuto, an end-to-end automated platform designed for sustainable AI model exploration, generation, deployment, and evaluation. GreenAuto employs a Pareto front-based search method within an expanded neural architecture search (NAS) space, guided by gradient descent to optimize model exploration. Pre-trained kernel-level energy predictors estimate energy consumption across all models, providing a global view that directs the search toward more sustainable solutions. By automating performance measurements and iteratively refining the search process, GreenAuto demonstrates the efficient identification of sustainable AI models without the need for human intervention. please check our paper for more details: https://arxiv.org/abs/2501.14995
GreenAuto contains three essential components: (1) energy efficiency driven model search space that includes over 900,000 DNN models tailored for energy efficiency, (2) Pareto front-based model search algorithm that, combined with gradient descent-based model sampling, efficiently identifies promising model candidates from the search space, and (3) automated energy measurement function that leverages an external power monitor to capture accurate power consumption data during model execution.
The main function for the model search. SVGD_Pareto.py performs Pareto model search based on gradient descent.
Functions for power measurement using the Monsoon Power Monitor.
Outputs generated from the model search.
Estimated model accuracy and power consumption.
