[論文レビュー] MLOps: A Step Forward to Enterprise Machine Learning

Aliza Tabassam|arXiv (Cornell University)|May 27, 2023

Big Data and Business Intelligence被引用数 8

ひとこと要約

要約は一文または二文の直接的な回答

ABSTRACT

Machine Learning Operations (MLOps) is becoming a highly crucial part of businesses looking to capitalize on the benefits of AI and ML models. This research presents a detailed review of MLOps, its benefits, difficulties, evolutions, and important underlying technologies such as MLOps frameworks, Docker, GitHub actions, and Kubernetes. The MLOps workflow, which includes model design, deployment, and operations, is explained in detail along with the various tools necessary for both model and data exploration and deployment. This article also puts light on the end-to-end production of ML projects using various maturity levels of automated pipelines, with the least at no automation at all and the highest with complete CI/CD and CT capabilities. Furthermore, a detailed example of an enterprise-level MLOps project for an object detection service is used to explain the workflow of the technology in a real-world scenario. For this purpose, a web application hosting a pre-trained model from TensorFlow 2 Model Zoo is packaged and deployed to the internet making sure that the system is scalable, reliable, and optimized for deployment at an enterprise level.

研究の動機と目的

• Motivate the importance of MLOps for leveraging AI/ML in enterprises.
• Summarize the evolution, benefits, and challenges of MLOps.
• Identify core technologies, tools, and frameworks used in MLOps.
• Describe end-to-end AI project production using automated pipelines at varying maturity levels.

提案手法

• Survey of MLOps concepts, workflows, and lifecycle stages (design, deployment, operations).
• Discussion of underlying technologies and tools such as MLOps frameworks, Docker, GitHub Actions, and Kubernetes.
• Illustration of end-to-end production of ML projects with progression from low to high automation (CI/CD and CT).
• Provision of a concrete enterprise example involving packaging and deploying a pre-trained TensorFlow 2 Model Zoo model for object detection.
• Description of considerations for scalability, reliability, and enterprise deployment.

実験結果

リサーチクエスチョン

• RQ1 What are the benefits and difficulties of adopting MLOps in enterprise settings?
• RQ2 What are the essential technologies and frameworks that support MLOps workflows?
• RQ3 How can MLOps enable end-to-end production pipelines with varying levels of automation?
• RQ4 What does a real-world enterprise MLOps deployment look like for an object detection service?

主な発見

• MLOps frameworks, Docker, GitHub Actions, and Kubernetes are central to modern MLOps ecosystems.
• End-to-end ML production can be achieved through automated pipelines with different maturity levels, from minimal automation to full CI/CD and CT.
• An enterprise-level object detection project demonstrates a scalable and reliable deployment workflow.
• Packaging and deploying a pre-trained TensorFlow 2 Model Zoo model can be made accessible via internet-hosted web applications.