什么是 Azure 机器学习工作室？What is Azure Machine Learning studio?
本文将介绍 Azure 机器学习工作室，这是 Azure 机器学习中面向数据科学家开发人员的 Web 门户。In this article, you learn about Azure Machine Learning studio, the web portal for data scientist developers in Azure Machine Learning. 此工作室将无代码和代码优先体验结合起来，打造包容的数据科学平台。The studio combines no-code and code-first experiences for an inclusive data science platform.
本文内容：In this article you learn:
- 如何在工作室中创作机器学习项目。How to author machine learning projects in the studio.
- 如何在工作室中管理资产和资源。How to manage assets and resources in the studio.
- Azure 机器学习工作室和机器学习工作室（经典）之间的差异。The differences between Azure Machine Learning studio and ML Studio (classic).
创作机器学习项目Author machine learning projects
该工作室提供多种创作体验，具体取决于类型项目和用户体验级别。The studio offers multiple authoring experiences depending on the type project and the level of user experience.
NotebookNotebooks
在直接集成到工作室中的托管 Jupyter Notebook 服务器中编写和运行自己的代码。Write and run your own code in managed Jupyter Notebook servers that are directly integrated in the studio.
Azure 机器学习设计器Azure Machine Learning designer
使用设计器可在不编写任何代码的情况下训练和部署机器学习模型。Use the designer to train and deploy machine learning models without writing any code. 拖放数据集和模块以创建 ML 管道。Drag and drop datasets and modules to create ML pipelines. 尝试设计器教程。Try out the designer tutorial.
自动化机器学习 UIAutomated machine learning UI
了解如何通过易于使用的界面创建自动化 ML 试验。Learn how to create automated ML experiments with an easy-to-use interface.
数据标记Data labeling
使用 Azure 机器学习数据标签来高效地协调数据标签项目。Use Azure Machine Learning data labeling to efficiently coordinate data labeling projects.
管理资产和资源Manage assets and resources
直接在浏览器中管理机器学习资产。Manage your machine learning assets directly in your browser. 在 SDK 和工作室之间的同一工作区中共享资产，以实现无缝体验。Assets are shared in the same workspace between the SDK and the studio for a seamless experience. 使用工作室管理：Use the studio to manage:
- 模型Models
- 数据集Datasets
- 数据存储Datastores
- 计算资源Compute resources
- 笔记本Notebooks
- 试验Experiments
- 运行日志Run logs
- 管道Pipelines
- 管道终结点Pipeline endpoints
即使你是经验丰富的开发人员，工作室也可以简化你管理工作区资源的方式。Even if you're an experienced developer, the studio can simplify how you manage workspace resources.
机器学习工作室（经典）与 Azure 机器学习工作室ML Studio (classic) vs Azure Machine Learning studio
机器学习工作室（经典）于 2015 年发布，是我们的第一个拖放式机器学习生成器。Released in 2015, ML Studio (classic) was our first drag-and-drop machine learning builder. 它是一个只提供视觉体验的独立服务。It is a standalone service that only offers a visual experience. 工作室（经典）不与 Azure 机器学习进行互操作。Studio (classic) does not interoperate with Azure Machine Learning.
Azure 机器学习是一种独立的新式服务，可提供完整的数据科学平台。Azure Machine Learning is a separate and modernized service that delivers a complete data science platform. 它同时支持代码优先和低代码体验。It supports both code-first and low-code experiences.
Azure 机器学习工作室是 Azure 机器学习中的 Web 门户，其中包含用于项目创作和资产管理的低代码和无代码选项。Azure Machine Learning studio is a web portal in Azure Machine Learning that contains low-code and no-code options for project authoring and asset management.
我们建议新用户选择 Azure 机器学习而不是机器学习工作室（经典），以使用最新的数据科学工具。We recommend that new users choose Azure Machine Learning, instead of ML Studio (classic), for the latest range of data science tools.
功能比较Feature comparison
下表总结了机器学习工作室（经典）和 Azure 机器学习之间的一些重要差异。The following table summarizes the key differences between ML Studio (classic) and Azure Machine Learning.
| 功能Feature | 机器学习工作室（经典版）ML Studio (classic) | Azure 机器学习Azure Machine Learning |
|---|---|---|
| 拖放界面Drag and drop interface | 经典体验Classic experience | 更新的体验 - Azure 机器学习设计器Updated experience - Azure Machine Learning designer |
| 代码 SDKCode SDKs | 不支持Unsupported | 与 Azure 机器学习 Python 和 R SDK 完全集成Fully integrated with Azure Machine Learning Python and R SDKs |
| 试验Experiment | 可缩放（10 GB 训练数据限制）Scalable (10-GB training data limit) | 使用计算目标进行缩放Scale with compute target |
| 训练计算目标Training compute targets | 专用计算目标，仅限 CPU 支持Proprietary compute target, CPU support only | 各种可自定义的训练计算目标。Wide range of customizable training compute targets. 包括 GPU 和 CPU 支持Includes GPU and CPU support |
| 部署计算目标Deployment compute targets | 专用 Web 服务格式，不可自定义Proprietary web service format, not customizable | 各种可自定义的部署计算目标。Wide range of customizable deployment compute targets. 包括 GPU 和 CPU 支持Includes GPU and CPU support |
| ML 管道ML Pipeline | 不支持Not supported | 生成灵活的模块化管道，用于自动完成工作流Build flexible, modular pipelines to automate workflows |
| MLOpsMLOps | 基本模型管理和部署；仅 CPU 部署Basic model management and deployment; CPU only deployments | 实体版本控制（模型、数据、工作流）、工作流自动化、与 CICD 工具集成、CPU 和 GPU 部署，等等Entity versioning (model, data, workflows), workflow automation, integration with CICD tooling, CPU and GPU deployments and more |
| 模型格式Model format | 专用格式，仅限工作室（经典）Proprietary format, Studio (classic) only | 多个受支持的格式，具体取决于训练作业类型Multiple supported formats depending on training job type |
| 自动化模型训练和超参数优化Automated model training and hyperparameter tuning | 不支持Not supported | 受支持。Supported. 代码优先和无代码选项。Code-first and no-code options. |
| 数据偏移检测Data drift detection | 不支持Not supported | 支持Supported |
| 数据标签项目Data labeling projects | 不支持Not supported | 支持Supported |
后续步骤Next steps
请访问工作室，或浏览以下教程中的不同创作选项：Visit the studio, or explore the different authoring options with these tutorials:
-
- 在自己的开发环境开始使用Get started in your own development environment
- 在计算实例上使用 Jupyter 笔记本来训练和部署模型Use Jupyter notebooks on a compute instance to train & deploy models
- 使用自动化机器学习训练和部署模型Use automated machine learning to train & deploy models
- 使用设计器训练和部署模型Use the designer to train & deploy models
- 在受保护的虚拟网络中使用工作室Use studio in a secured virtual network