選取您的 Cookie 偏好設定

我們使用提供自身網站和服務所需的基本 Cookie 和類似工具。我們使用效能 Cookie 收集匿名統計資料，以便了解客戶如何使用我們的網站並進行改進。基本 Cookie 無法停用，但可以按一下「自訂」或「拒絕」以拒絕效能 Cookie。

如果您同意，AWS 與經核准的第三方也會使用 Cookie 提供實用的網站功能、記住您的偏好設定，並顯示相關內容，包括相關廣告。若要接受或拒絕所有非必要 Cookie，請按一下「接受」或「拒絕」。若要進行更詳細的選擇，請按一下「自訂」。

偏好設定
聯絡我們
意見回饋
AWS Documentation
建立 AWS 帳戶
MLOE-15: Enable model observability and tracking - Machine Learning Lens
Implementation planDocumentsBlogsVideosExamples
此頁面尚未翻譯為您的語言。 請求翻譯

MLOE-15: Enable model observability and tracking

PDFRSS

Establish model monitoring mechanisms to identify and proactively avoid any inference issues. ML models can degrade in performance over time due to drifts. Monitor metrics that are attributed to your model’s performance. For real time inference endpoints, measure the operational health of the underlying compute resources hosting the endpoint and the health of endpoint responses. Establish lineage to trace hosted models back to versioned inputs and model artifacts for analysis.

Implementation plan

  • Use Amazon SageMaker AI Model Monitor - Continually monitor the quality of Amazon SageMaker AI ML models in production and compare with the results from training using SageMaker AI Model Monitor.

  • Use Amazon CloudWatch - Amazon SageMaker AI Model Monitor automatically sends metrics to Amazon CloudWatch so that you can gather and analyze usage statistics for your ML models.

  • Use SageMaker AI Model Dashboard - View, search, and explore your models in a centralized portal from the SageMaker AI console. Set up monitors with Amazon SageMaker AI Model Monitor and track the performance of your models that are hosted on real-time inference endpoints. Find models that violate thresholds you have set for data quality, model quality, bias, and explainability

  • Use Amazon SageMaker AI Clarify - Identify various types of bias that can emerge during model training or when the model is in production. This helps improve your ML models by detecting potential bias and helping explain the predictions that the models make. SageMaker AI Clarify helps explain how these models make predictions using a feature attribution approach. It also monitors inferences that the models make in production for bias drift or feature attribution drift. SageMaker AI Clarify provides tools to help you generate model governance reports that you can use to inform risk and compliance teams, and external regulators.

  • Track your model pipeline with SageMaker AI ML lineage Tracking – Lineage tracking creates and stores information about the steps of a machine learning workflow from data preparation to model deployment. Keep a running history of model discovery experiments. Establish model governance by tracking model lineage artifacts for auditing and compliance verification.

  • Use SageMaker AI Model Cards to simplify model information gathering – Documentation on model information, such as business requirements, key decisions, and observations during model development and evaluation, is required to support approval workflows, registration, audits, customer inquiries, and monitoring. Amazon SageMaker AI Model Cards provide a single location to store model information (for example, performance goals, and risk rating) and training and evaluation results (for example, bias or accuracy measurements) in the AWS Management Console, streamlining documentation throughout a model’s lifecycle.

  • Use the automated validation capability of Amazon SageMaker AI – Amazon SageMaker AI Inference enables you to compare the performance of new models against production models, using the same real-world inference request data in real time. Amazon SageMaker AI can be used to route a copy of the inference requests received by the production model to the new model and generate a dashboard to display performance differences across key metrics in real time.

Documents

Blogs

Videos

Examples

需要協助？

隱私權網站條款Cookie 偏好設定
© 2025, Amazon Web Services, Inc.或其附屬公司。保留所有權利。