Connecting with ML Pipelines
Last updated September 6, 2024
Connecting Evidently AI with Machine Learning Pipelines
Evidently AI is a valuable tool for monitoring your machine learning models and data, but to fully leverage its benefits, you need to seamlessly integrate it into your existing machine learning pipelines. This article outlines how to connect Evidently AI with popular ML pipelines and frameworks.
Connecting with ML Pipelines
- Identify Your Pipeline: Understand the structure and components of your ML pipeline. Identify the stages where you can integrate Evidently AI for data drift monitoring, model performance checks, or data quality evaluation.
- Use the `Report` Object: Use the `Report` object provided by Evidently AI. You can create a report object in your pipeline code. This will enable you to generate monitoring reports as part of your pipeline's flow.
- Configure the Report: Define the metrics, presets, and custom checks you want to include in your report. Adjust the report to fit your specific requirements and monitoring needs.
- Run the Report: Execute the report within your pipeline workflow. This will typically happen after your model training or data processing steps, allowing you to analyze the data as it flows through your pipeline.
- Save the Report: Save the report in a format compatible with your pipeline and monitoring infrastructure. The HTML format is a commonly used option for interactive reports.
- Use Libraries and Frameworks: Many popular machine learning libraries and frameworks can be integrated with Evidently AI:
- Scikit-learn (sklearn): Integrate Evidently AI into your scikit-learn pipelines by using the `Report` class after model training or predictions.
- TensorFlow: Use Evidently AI within your TensorFlow pipelines to monitor models, data quality, and the performance of TensorFlow models.
- PyTorch: Connect Evidently AI with your PyTorch pipelines to analyze models, data, and predictions as they are generated using PyTorch.
- Keras: Use Evidently AI for monitoring Keras models (a part of TensorFlow).
- Example using scikit-learn:
Was this article helpful?