Container models（容器模型（Container models））

:::callout{theme="warning" title="Prerequisites"} The following documentation assumes working knowledge of containerized infrastructure and concepts like container images. If you are unfamiliar with these topics, we recommend reviewing the Docker overview documentation ↗ :::

Foundry allows users to package container images along with a model adapter to create a container-backed Model. The model is then consumable in Foundry as with non-container-backed models. Container-backed models are particularly useful if the models are especially large, pre-trained, written in a language Foundry does not natively support (such as R), or already containerized for other usage.

Control Panel houses all features around container governance, including enabling container usage and managing vulnerabilities. If the container setting is not enabled, all Foundry jobs and deployments relying on imported containers will fail.

Once container workflows are enabled, the first step is to create a model and begin pushing images into the platform.

---

中文翻译

---

容器模型（Container models）

:::callout{theme="warning" title="前提条件"} 以下文档假定您已具备容器化基础设施和容器镜像（container images）等概念的实际操作知识。如果您不熟悉这些主题，建议先查阅 Docker 概述文档 ↗。 :::

Foundry 允许用户将容器镜像（container images）与模型适配器（model adapter）打包，从而创建基于容器的模型（Model）。该模型在 Foundry 中的使用方式与非容器模型相同。基于容器的模型特别适用于以下场景：模型体积特别庞大、已预训练、使用 Foundry 原生不支持的语言（如 R）编写，或已为其他用途完成容器化。

控制面板（Control Panel） 包含所有与容器治理相关的功能，包括启用容器使用和管理漏洞。如果未启用容器设置，所有依赖导入容器的 Foundry 作业和部署都将失败。

启用容器工作流后，第一步是创建模型并开始将镜像推送至平台。