Google Cloud (GCPC) SDK 提供了一组预构建的 Kubeflow Pipelines 组件,它们达到生产质量、具备高性能且易于使用。您可以使用 Google Cloud 流水线组件在 Vertex AI Pipelines 和其他符合 Kubeflow Pipelines 的机器学习流水线执行后端中定义和运行机器学习流水线。
例如,您可以使用这些组件来完成以下操作:
创建新数据集,并将不同的数据类型加载到数据集中(图片、表格、文本或视频)。
将数据从数据集导出到 Cloud Storage。
通过 AutoML 使用图片、表格或视频数据训练模型。
使用自定义容器或 Python 软件包运行自定义训练作业。
将现有模型上传到 Vertex AI 以进行批量预测。
创建新端点并向其部署模型以进行在线预测。
此外, Google Cloud 流水线组件在 Vertex AI Pipelines 中支持这些预建的组件,并具有以下优势:
简化调试:显示从组件启动的底层资源以简化调试。
标准化工件类型:提供一致的接口以使用标准工件类型进行输入和输出。Vertex ML Metadata 会跟踪这些标准制品,便于您更轻松地分析流水线制品的沿袭。如需详细了解工件沿袭,请参阅跟踪流水线工件的沿袭。
使用结算标签了解流水线费用:资源标签会自动传播到 Google Cloud 流水线组件在您的流水线运行作业中生成的 Google Cloud 服务。您可以将结算标签与将 Cloud Billing 数据导出到 BigQuery 这一功能结合使用来查看流水线运行作业的费用。如需详细了解如何使用标签了解流水线运行作业的费用,请参阅了解流水线运行作业的费用。如需详细了解标签如何从流水线运行作业传播到由 Google Cloud 流水线组件生成的资源,请参阅 Vertex AI Pipelines 中的资源加标签功能。
成本效益*:Vertex AI Pipelines 通过启动 Google Cloud 资源来优化这些组件的执行,而无需启动容器。这可缩短启动延迟时间并降低频繁等待容器的费用。
[[["易于理解","easyToUnderstand","thumb-up"],["解决了我的问题","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["很难理解","hardToUnderstand","thumb-down"],["信息或示例代码不正确","incorrectInformationOrSampleCode","thumb-down"],["没有我需要的信息/示例","missingTheInformationSamplesINeed","thumb-down"],["翻译问题","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["最后更新时间 (UTC):2025-09-04。"],[],[],null,["# Introduction to Google Cloud Pipeline Components\n\nThe Google Cloud (GCPC) SDK provides a set of prebuilt\nKubeflow Pipelines components that are production quality,\nperformant, and easy to use. You can use Google Cloud Pipeline Components to define and run ML\npipelines in Vertex AI Pipelines and other\nML pipeline execution backends conformant with Kubeflow Pipelines.\n\nFor example, you can use these components to complete the following:\n\n- Create a new dataset and load different data types into the dataset (image, tabular, text, or video).\n- Export data from a dataset to Cloud Storage.\n- Use AutoML to train a model using image, tabular, or video data.\n- Run a custom training job using a custom container or a Python package.\n- Upload an existing model to Vertex AI for batch prediction.\n- Create a new endpoint and deploy a model to it for online predictions.\n\nAdditionally, Google Cloud Pipeline Components supports these prebuilt components\nin Vertex AI Pipelines and offers the following benefits:\n\n- **Easier debugging**: Show the underlying resources launched from the component for simplified debugging.\n- **Standardized artifact types** : Provide consistent interfaces to use [standard artifact types](/vertex-ai/docs/pipelines/artifact-types) for input and output. Vertex ML Metadata tracks these standard artifacts, making it easier for you to analyze the lineage of your pipeline's artifacts. For more details on artifact lineage, see [Tracking the lineage of pipeline\n artifacts](/vertex-ai/docs/pipelines/lineage).\n- **Understand pipeline costs with billing labels** : Resource labels automatically propagate to Google Cloud services generated by the Google Cloud Pipeline Components in your pipeline run. Use billing labels along with Cloud Billing export to BigQuery to review the cost of your pipeline run. For more information about using labels to understand the cost of a pipeline run, see [Understand pipeline run costs](/vertex-ai/docs/pipelines/understand-pipeline-cost-labels). For more information about how labels propagate from a pipeline run to resources spawned by Google Cloud Pipeline Components, see [Resource labeling by Vertex AI Pipelines](/vertex-ai/docs/pipelines/gcpc-label-propagation).\n- **Cost efficiencies** ^\\*^: Vertex AI Pipelines optimizes the execution of these components by launching the Google Cloud resources, without having to launch the container. This reduces the startup latency and reduces the costs of the busy-waiting container.\n\nWhat's next\n-----------\n\n- See all [tutorials that use the Google Cloud SDK](/vertex-ai/docs/pipelines/notebooks).\n- Learn more about specific [Google Cloud Pipeline Components in the reference section](/vertex-ai/docs/pipelines/gcpc-list).\n- Read the official [Google Cloud SDK reference](https://google-cloud-pipeline-components.readthedocs.io/en/google-cloud-pipeline-components-2.19.0/api/v1/index.html).\n- See the Google Cloud Pipeline Components section in the [Kubeflow Pipelines SDK repository](https://github.com/kubeflow/pipelines/tree/master/components/google-cloud)."]]