|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/mastering-advanced-mlops-on-gcp-cicd-kubernetes-kubeflow/
课程评论：没有评论
课程名称：从初学者到高级 MLOps：GCP-CI/CD、Kubernetes Jenkins 课程概述： 本课程涵盖初学者到高级水平的 MLOps，介绍构建、部署和自动化机器学习（ML）模型在生产环境中所需的各种技术和工具。课程内容广泛，包括实验跟踪、模型管理、数据与代码版本控制、CI/CD 管道、云基础设施、部署与容器化等多个方面。 使用的技术与工具： 1. 实验跟踪与模型管理：MLFlow、Comet-ML、TensorBoard 2. 数据与代码版本控制：DVC、Git、GitHub、GitLab 3. CI/CD 管道与自动化：Jenkins、ArgoCD、GitHub Actions、GitLab CI/CD、CircleCI 4. 云与基础设施：GCP（Google Cloud Platform）、Minikube、Google Cloud Run、Kubernetes 5. 部署与容器化：Docker、Kubernetes、FastAPI、Flask 6. 数据工程与特征存储：PostgreSQL、Redis、Astro Airflow、PSYCOPG2 7. ML 监控与漂移检测：Prometheus、Grafana、Alibi-Detect 8. API 与 Web 应用开发：FastAPI、Flask、ChatGPT、Postman、SwaggerUI 这些工具与技术如何提供帮助： - 实验跟踪与模型管理帮助记录、比较和跟踪不同 ML 模型的实验，提升模型的可追溯性。 - 数据与代码版本控制确保通过跟踪数据变更来实现可再现性，DVC 处理大数据集，GitHub/GitLab 进行代码和流水线的版本控制。 - CI/CD 管道与自动化简化了从模型训练到部署的整个 ML 工作流程，提高了效率。 - Cloud & Infrastructure 提供可扩展的数据存储、模型训练和部署基础设施，Minikube 支持在本地环境中测试 Kubernetes。 - 部署与容器化使用 Docker 使应用程序可移植和可扩展，而 Kubernetes 管理 ML 部署以确保高可用性。 - 数据工程与特征存储通过 PostgreSQL 和 Redis 存储结构化和实时 ML 特征，Airflow 自动化 ETL 流程。 - ML 监控与漂移检测使用 Prometheus 和 Grafana 进行实时性能可视化，Alibi-Detect 有助于识别数据漂移和模型减效。 - API 与 Web 应用开发使用 FastAPI 与 Flask 构建实时模型推断 API，ChatGPT 集成提升聊天机器人应用。 课程确保提供全面的实践方法，通过涵盖数据摄取、模型训练、版本控制、部署、监控和 CI/CD 自动化等所有方面，使 ML 项目达到可生产和可扩展的标准。
This Beginner to Advanced MLOps Course covers a wide range of technologies and tools essential for building, deploying, and automating ML models in production.Technologies & Tools Used Throughout the CourseExperiment Tracking & Model Management: MLFlow, Comet-ML, TensorBoardData & Code Versioning: DVC, Git, GitHub, GitLabCI/CD Pipelines & Automation: Jenkins, ArgoCD, GitHub Actions, GitLab CI/CD, CircleCICloud & Infrastructure: GCP (Google Cloud Platform), Minikube, Google Cloud Run, KubernetesDeployment & Containerization: Docker, Kubernetes, FastAPI, FlaskData Engineering & Feature Storage: PostgreSQL, Redis, Astro Airflow, PSYCOPG2ML Monitoring & Drift Detection: Prometheus, Grafana, Alibi-DetectAPI & Web App Development: FastAPI, Flask, ChatGPT, Postman, SwaggerUIHow These Tools & Techniques HelpExperiment Tracking & Model ManagementHelps in logging, comparing, and tracking different ML model experiments.MLFlow & Comet-ML provide centralized tracking of hyperparameters and performance metrics.Data & Code VersioningEnsures reproducibility by tracking data changes over time.DVC manages large datasets, and GitHub/GitLab maintains version control for code and pipelines.CI/CD Pipelines & AutomationAutomates ML workflows from model training to deployment.Jenkins, GitHub Actions, GitLab CI/CD, and ArgoCD handle continuous integration & deployment.Cloud & InfrastructureGCP provides scalable infrastructure for data storage, model training, and deployment.Minikube enables Kubernetes testing on local machines before deploying to cloud environments.Deployment & ContainerizationDocker containerizes applications, making them portable and scalable.Kubernetes manages ML deployments for high availability and scalability.Data Engineering & Feature StoragePostgreSQL & Redis store structured and real-time ML features.Airflow automates ETL pipelines for seamless data processing.ML Monitoring & Drift DetectionPrometheus & Grafana visualize ML model performance in real-time.Alibi-Detect helps in identifying data drift and model degradation.API & Web App DevelopmentFastAPI & Flask create APIs for real-time model inference.ChatGPT integration enhances chatbot-based ML applications.SwaggerUI & Postman assist in API documentation & testing.This course ensures a complete hands-on approach to MLOps, covering everything from data ingestion, model training, versioning, deployment, monitoring, and CI/CD automation to make ML projects production-ready and scalable.
