Python SDK

Arize AI for Model Monitoring, Troubleshooting, and Explainability

Use the Arize Python package to monitor machine learning predictions to observe your ML models and their features, predicted labels, and actual labels with just a few lines of code.

Installing the package

pip install arize

In addition to the basic functionality installed by the command above, the Arize SDK has additional functionality that can be installed with some extra dependencies:

LLM Tracing

With this extra module, Arize can receive your traces and spans to break down your system's components into discrete inputs and outputs to help debug your LLM application. Learn more here.

pip install arize[Tracing]

Auto Embeddings

With this extra module, Arize properly extracts the embeddings depending on your use case, and we return it to you to include in your Pandas DataFrame. Learn more here. To install the Arize package including this functionality:

pip install arize[AutoEmbeddings]

NLP Metrics

With this extra module, Arize helps you calculate evaluation metrics for your NLP Generative tasks. Learn more here. To install the Arize package including this functionality:

pip install arize[NLP_Metrics]

Mimic Explainer

With this extra module, Arize allows the user to pass a flag with their request to send data that would produce SHAP values using the surrogate explainability approach. Learn more here. To install the Arize package including this functionality:

pip install arize[MimicExplainer]

Logging Options

The Arize Python SDK offers 2 ways of logging data into the platform:

Pandas Batch Logging

Designed for logging a batch of your model inferences using Pandas DataFrames. Go to the following page for more information.

Single Record Logging

Designed for low latency, one-at-a-time, logging of your model inferences. Go to the following page for more information.

End of Support Table

Changelog

Last updated

Copyright © 2023 Arize AI, Inc