|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/mlflow-kubernetes-mlops/
课程评论：没有评论
**课程名称：** MLflow for Kubernetes: Deploy and Manage ML Models at Scale **课程概述：** 本课程是一门全面且实用的课程，旨在帮助学员掌握使用 MLflow、Kubernetes、Docker 和 KServe 将机器学习模型从实验阶段部署到可扩展、生产就绪的 AI 服务。课程将从介绍 Kubernetes 和 MLflow 在现代 AI 可扩展性方面的重要性入手，阐述它们如何简化从实验跟踪到生产环境模型服务的整个 ML 生命周期。 **核心学习内容：** * **环境搭建：** 学习安装 MLflow、配置 Minikube 以及在 Kubernetes 上部署 KServe。 * **模型训练与跟踪：** 利用 MLflow Autologging 和 UI 可视化功能监控机器学习实验。 * **超参数调优与模型选择：** 进行随机搜索实验，并在 MLflow 中直接比较模型性能。 * **模型本地打包与服务：** 构建 Docker 镜像，并使用 MLServer 模型进行快速本地测试。 * **模型大规模部署到 Kubernetes：** 创建 KServe InferenceService YAML 文件，使用 kubectl 部署模型，并掌握故障排除的最佳实践。 * **推理与服务监控：** 发送请求，解读结果，并监控 Kubernetes Pods 和日志以确保服务健康运行。 * **生产级实践：** 探索自动扩展、金丝雀部署、A/B 测试，并使用 MLflow Model Registry 进行版本控制和治理。 **课程目标：** 完成本课程后，学员将能够自信地将 ML 模型大规模投入生产，利用 CI/CD 概念自动化部署工作流，并管理从训练到生产推理的完整生命周期。 **适合人群：** 机器学习工程师、MLOps 专家以及希望从 Notebook 进阶、构建真实世界、可扩展 ML 系统的数据科学家。
Deploying machine learning models to production doesn't have to be painful.This comprehensive, hands-on course will teach you step-by-step how to make the leap from experiments to scalable, production-ready AI services using MLflow, Kubernetes, Docker, and KServe.You will start by learning why Kubernetes and MLflow are essential for modern AI scalability, and how they can streamline the entire ML lifecycle - from tracking experiments to serving models in production environments. Through carefully designed lessons and real-world projects, you will build deep practical knowledge in:Setting up your environment - Install MLflow, configure Minikube, and deploy KServe on Kubernetes.Training and tracking models - Use MLflow Autologging and UI visualization to monitor your machine learning experiments.Hyperparameter tuning and model selection - Run randomized search experiments and compare model performance directly in MLflow.Packaging and serving models locally - Build Docker images and serve models with MLServer for quick local testing.Deploying models to Kubernetes at scale - Create KServe InferenceService YAML files and deploy models using kubectl with troubleshooting best practices.Performing inference and monitoring services - Send requests, interpret results, and monitor Kubernetes pods and logs for healthy service operations.Implementing production-level practices - Explore autoscaling, canary deployments, A/B testing, and use MLflow Model Registry for versioning and governance.By the end of the course, you will be able to confidently operationalize ML models at scale, automate deployment workflows using CI/CD concepts, and manage the full lifecycle from training to production inference.This course is ideal for ML engineers, MLOps specialists, and data scientists ready to move beyond notebooks and start building real-world, scalable ML systems.
