Do Developers Adopt Green Architectural Tactics for ML-Enabled Systems? A Mining Software Repository Study

2024-10-09

arXiv.org

AI Summary

This study investigates the real-world adoption of green software engineering practices by machine learning developers in open-source systems, focusing on the implementation status and barriers to computational resource optimization. Method: Leveraging 168 GitHub projects, we propose the first LLM-enhanced framework for automated identification of green architectural tactics, integrating static code analysis, documentation parsing, and natural language processing. Rigorous qualitative coding and statistical validation ensure analytical reliability. Contribution/Results: We provide the first empirical evidence on the actual adoption rates, contextual characteristics, and implementation bottlenecks of high-impact green tactics—including model pruning, quantized deployment, and hardware-aware training. We quantify their frequency of use and scenario-specific applicability, establishing a reusable empirical dataset, evidence-based guidelines, and foundational support for automation in green ML engineering.

Technology Category

• Machine Learning: Efficient ML / Green AI
• Search and Optimization: Metareasoning and Metaheuristics
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

As machine learning (ML) and artificial intelligence (AI) technologies become more widespread, concerns about their environmental impact are increasing due to the resource-intensive nature of training and inference processes. Green AI advocates for reducing computational demands while still maintaining accuracy. Although various strategies for creating sustainable ML systems have been identified, their real-world implementation is still underexplored. This paper addresses this gap by studying 168 open-source ML projects on GitHub. It employs a novel large language model (LLM)-based mining mechanism to identify and analyze green strategies. The findings reveal the adoption of established tactics that offer significant environmental benefits. This provides practical insights for developers and paves the way for future automation of sustainable practices in ML systems.

Problem

• Machine Learning
• Environmental Impact
• Resource Consumption

Innovation

• Green AI
• Large Language Model
• Environmental Impact Reduction