Visão Geral
O curso "Mastering Kubeflow for Machine Learning Workflows" oferece um treinamento abrangente para dominar o Kubeflow, uma poderosa plataforma de código aberto voltada para o desenvolvimento, treinamento e implantação de modelos de machine learning em escala. Ao longo deste curso, você aprenderá a construir pipelines eficientes, automatizar fluxos de trabalho, monitorar e gerenciar seus modelos de machine learning, além de utilizar o Kubeflow em ambientes de produção.
Conteúdo Programatico
Part 1: Introduction to Kubeflow
Module 1.1: What is Kubeflow?
- Understanding Kubeflow and its components
- Kubeflow vs. other ML platforms
Module 1.2: Setting up Kubeflow
- Installation on Kubernetes
- Exploring the Kubeflow UI and dashboards
Part 2: Building ML Pipelines
Module 2.1: Introduction to Kubeflow Pipelines
- Designing and constructing ML pipelines
- Using Jupyter notebooks in Kubeflow
Module 2.2: Managing Data in Kubeflow
- Data ingestion, processing, and versioning
- Integrating external data sources
Part 3: Advanced Pipelines and Hyperparameter Tuning
Module 3.1: Advanced ML Pipelines
- Using pre-built components and containers
- Creating custom pipeline components
Module 3.2: Hyperparameter Tuning with Katib
- Overview of Katib
- Setting up and running hyperparameter tuning experiments
Part 4: Model Serving and Scaling
Module 4.1: Serving Models with Kubeflow
- Introduction to KFServing
- Deploying and managing model servers
Module 4.2: Autoscaling ML Models
- Scaling model inference
- Performance optimization strategies
Part 5: MLOps with Kubeflow
Module 5.1: Kubeflow for MLOps
- Continuous integration and delivery (CI/CD) for ML models
- Best practices for ML lifecycle management
Module 5.2: Course Capstone Project
- Applying learned skills to a real-world problem
- Group presentations of projects