|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/nuts-and-bolts-of-mlflow/
课程评论：没有评论
课程名称：MLFlow的基础知识 课程概述：本课程聚焦于MLOps。选择MLOps工具栈时，MLFlow可能是追踪实验、注册和服务模型的最流行解决方案。本课程将深入探讨MLFlow的工作原理，以及如何利用MLFlow与亚马逊网络服务（AWS）构建自己的MLOps工具栈。 课程将从MLFlow的整体概述开始，介绍其在机器学习和数据科学中的必要性。接下来，我们将详细探索MLFlow中最重要的组件——MLflow追踪，讲解追踪的工作原理及其跟踪的内容。 随后，我们将学习MLflow模型注册，涵盖如何在MLflow中注册模型及管理其生命周期。同时，我们还将学习如何从注册表中检索模型以进行预测。 接下来是MLFlow模型部分，探讨模型的工作原理以及保存模型的不同类型（flavours）。我们将对一些模型进行服务以进行预测。 最后一部分为可选环节，将逐步讲解如何基于MLFlow利用亚马逊网络服务（如Amazon EC2、Amazon S3和Amazon RDS）构建MLOps架构。 需要注意的是，本课程不会专注于数据科学和机器学习，因此不要期待学习机器学习模型的详细内容。我们将使用一个简单的聚类模型作为例子，以说明任何机器学习模型的应用。祝您好运！
This course is about MLOps.When choosing your MLOps stack, MLFlow is probably the most populat solution for tracking experiments, registering and serving models.This course will give you a deep dive on how MLFlow works and how you can build your own MLOps stack with mlflow using Amazon Web Services (AWS).We will start the course by giving an overall overvew of what mlflow is and why it is necessary for Machine Learning and Data Science. Next we will explore in detail the most important component of MLFlow which is mlflow tracking where we will have a look at how tracking works and how you what can be tracked.Next, we will move to MLflow model registry where we will cover how to register a model in a mlflow and how to manage its lifecycle. We will also learn how to retrieve a model from the registry in order to make predictions.The next topic is MLFlow models. Here, we will explore how models work as well as the different types (flavours) of a saved model. We will also, serve some of the models in order to make predictions.The last section is optional and will cover how to build, step by step, an MLOps architecture based on MLFlow using Amazon Web Services such as Amazon EC2, Amazon S3 and Amazon RDS.This course will not focus on data science and machine learning, so do not except to learn the details of Machine Learning models. We will take a simple clustering model as an example that will illustrate any Machine Learning Model.Good luck.
