Azure Machine Learning Dataset Runtime

1.62.0 · active · verified Tue Apr 14

azureml-dataset-runtime is an internal Python package within the Azure Machine Learning SDK ecosystem, primarily designed to coordinate dependencies among other AzureML packages. It is not intended for direct use by end-users. As of version 1.62.0, released in February 2026, it maintains an active release cadence, reflecting ongoing development in the broader AzureML SDK.

Warnings

breaking Azure Machine Learning SDK v1, which includes `azureml-dataset-runtime`, is deprecated as of March 31, 2025, with support ending on June 30, 2026. Users are strongly advised to migrate to Azure Machine Learning SDK v2 to avoid potential security risks or breaking changes after this date.
Fix: Transition to Azure Machine Learning SDK v2 and its corresponding data handling mechanisms. Refer to official Microsoft documentation for migration guides.
gotcha This package is explicitly internal and not intended for direct public use. Attempting to import or use components directly from `azureml-dataset-runtime` is unsupported and highly discouraged, as its internal APIs are subject to change without notice.
Fix: Do not directly import or rely on `azureml-dataset-runtime`. Instead, use the public-facing APIs of the main Azure Machine Learning SDK (`azureml-core`, `azure.ai.ml` for v2) which depend on this package internally.
gotcha The `azureml-dataset-runtime` package, and by extension other `azureml` SDK components, often have strict compatibility requirements for `numpy` versions (e.g., `!=1.19.3,<1.24`). This can lead to dependency resolution issues with newer Python versions (like 3.12) or other data science libraries that require a different `numpy` version.
Fix: When creating environments for AzureML SDK v1, carefully manage `numpy` versions. It's often necessary to explicitly pin `numpy` to a compatible version (e.g., `numpy==1.23.5`) in your `requirements.txt` or `conda` environment file. Check the specific dependency constraints for your `azureml-core` version.
gotcha Compatibility with newer Python versions (e.g., 3.12+) can be challenging due to upstream dependencies like `numpy` required by `azureml-dataset-runtime`. While some AzureML SDK v1 components support Python 3.8+, the practical upper limit for fully compatible environments often lies around Python 3.11 due to these constraints.
Fix: When setting up your environment for AzureML SDK v1, prioritize Python versions known to be compatible, typically Python 3.8 to 3.11. Always test your environment rigorously for dependency conflicts.

Install

pip install azureml-dataset-runtime Latest stable version

Quickstart

Given that `azureml-dataset-runtime` is an internal package not meant for direct interaction, there is no typical 'quickstart' code for its use. Instead, it is installed as a dependency by other Azure Machine Learning SDK packages. The example demonstrates how it might appear in a `conda` environment configuration, highlighting the common practice of pinning `numpy` versions to avoid dependency conflicts with `azureml-dataset-runtime` and other SDK components.

# This package is internal and not intended for direct use.
# Users typically install it as a dependency for other AzureML SDK components.
# Here's an example of a conda environment file where it might appear as a dependency:

# name: my-azureml-env
# channels:
#   - conda-forge
#   - defaults
# dependencies:
#   - python=3.11  # Note: Python 3.12+ might have numpy compatibility issues
#   - pip
#   - pip:
#     - azureml-core==1.58.0
#     - azureml-dataset-runtime # Pulled in by azureml-core or other components
#     - numpy==1.23.5 # Pinning numpy is often necessary to avoid conflicts
#     - pandas
#     - scikit-learn

view raw JSON →