Getting started（入门指南）¶
Prior to starting modeling workflows, ensure you have a high-level understanding of core Foundry concepts, Code Repositories, Python transforms, Ontology, and Functions.
If you are new to Foundry, the supervised learning tutorial provides a walkthrough from setting up a new project to deploying a model into production.
Review the dedicated page to learn more about the considerations involved in choosing the right tool.
Common model creation workflows¶
- Train a model with data from the Palantir platform in a Jupyter® code workspace, code repository, or in a model studio for no-code model training
- Import model weights and register them as a model
- Import and register a container as a model
- Register an externally hosted model
Deploying a model for inference and consumption¶
- Using a model as an input to Python transforms
- Using a model in Pipeline Builder
- Creating a batch deployment in Modeling Objectives and evaluating model performance
- Configuring a live deployment to serve the model as a REST endpoint
Registering a model for use in the Ontology¶
- Register a live deployment to be used as a function
- Use the model function in another function to write custom business logic using model predictions
中文翻译¶
入门指南¶
在开始建模工作流之前，请确保您对 Foundry 核心概念、代码仓库、Python 转换、本体和函数有高层级的理解。
如果您是 Foundry 新手，监督学习教程提供了从设置新项目到将模型部署到生产环境的完整操作指南。
请查阅专用页面，了解选择合适工具时需考虑的相关因素。
常见模型创建工作流¶
- 在 Jupyter® 代码工作区、代码仓库或模型工作室中，使用 Palantir 平台的数据训练模型（支持无代码模型训练）
- 导入模型权重并将其注册为模型
- 导入并注册容器作为模型
- 注册外部托管的模型