您现在访问的是微软AZURE全球版技术文档网站,若需要访问由世纪互联运营的MICROSOFT AZURE中国区技术文档网站,请访问 https://docs.azure.cn.

使用 Azure 机器学习的 MLOpsMLOps with Azure Machine Learning

MLOps (机器学习操作) 基于可提高工作效率的 DevOps 原则和实践,如持续集成、交付和部署。MLOps (machine learning operations) is based on DevOps principles and practices that increase workflow efficiencies like continuous integration, delivery, and deployment. MLOps 会将这些原则应用到机器学习过程,以便:MLOps applies these principles to the machine learning process in order to:

  • 更快地试验和开发模型。Experiment and develop models more quickly.
  • 更快地将模型部署到生产环境。Deploy models to production more quickly.
  • 实践和优化质量保证。Practice and refine quality assurance.

Azure 机器学习提供以下 MLOps 功能:Azure Machine Learning provides the following MLOps capabilities:

  • 创建可重复的管道。Create reproducible pipelines. 机器学习管道使你可以为数据准备、定型和评分过程定义可重复和可重复使用的步骤。Machine Learning pipelines enable you to define repeatable and reusable steps for your data preparation, training, and scoring processes.
  • 创建可重用的软件环境用于训练和部署模型。Create reusable software environments for training and deploying models.
  • 从任何位置注册、打包和部署模型。Register, package, and deploy models from anywhere. 您可以跟踪使用该模型所需的关联元数据。You can track the associated metadata required to use the model.
  • 捕获用于端到端生命周期的管理数据。Capture the governance data for the end-to-end lifecycle. 记录的信息可以包括模型的发布者、做出更改的原因,以及在生产环境中部署或使用模型的时间。The logged information can include who is publishing models, why changes were made, and when models were deployed or used in production.
  • 通知并向生命周期中的事件发出警报。Notify and alert on events in the lifecycle. 例如,你可以获得试验完成、模型注册、模型部署和数据偏差检测的警报。For example, you can get alerts for experiment completion, model registration, model deployment, and data drift detection.
  • 监视应用程序的操作和机器学习相关的问题。Monitor applications for operational and machine learning-related issues. 比较定型与推理之间的模型输入,探索特定于模型的指标,并在机器学习基础结构上提供监视和警报。Compare model inputs between training and inference, explore model-specific metrics, and provide monitoring and alerts on your machine learning infrastructure.
  • 通过 Azure 机器学习和 Azure Pipelines 自动化端到端机器学习生命周期。Automate the end-to-end machine learning lifecycle with Azure Machine Learning and Azure Pipelines. 通过管道,你可以经常更新模型、测试新模型,并与其他应用程序和服务一起不断推出新的机器学习模型。With pipelines, you can frequently update models, test new models, and continuously roll out new machine learning models alongside your other applications and services.

MLOps 与 Azure 机器学习的最佳实践Best practices for MLOps with Azure Machine Learning

模型与代码的不同之处在于,它们具有一种极差的保质期,并将降低,除非进行维护。Models differ from code because they have an organic shelf life and will deteriorate unless maintained. 部署完成后,他们可以增加真实的商业价值,当数据科学家获得采用标准工程实践的工具时,这会变得更容易。After they're deployed, they can add real business value, and this gets easier when data scientists are given the tools to adopt standard engineering practices.

MLOps with Azure 可帮助你:MLOps with Azure helps you:

  • 创建可重复的模型和可重复使用的定型管道。Create reproducible models and reusable training pipelines.
  • 简化对质量控制和 a/B 测试的模型打包、验证和部署。Simplify model packaging, validation, and deployment for quality control and a/B testing.
  • 说明并观察模型行为,并自动执行重新训练过程。Explain and observe model behavior, and automate the retraining process.

MLOps 改善了机器学习解决方案的质量和一致性。MLOps improves the quality and consistency of your machine learning solutions. 若要详细了解如何使用 Azure 机器学习来管理模型的生命周期,请参阅 MLOps:模型管理、部署和监视 Azure 机器学习To learn more about how to use Azure Machine Learning to manage the lifecycle of your models, see MLOps: model management, deployment, and monitoring with Azure Machine Learning.

后续步骤Next steps

阅读并探索以下资源来了解详细信息:Learn more by reading and exploring the following resources: