Implementation:Mlflow Mlflow Version Module

Knowledge Sources	Mlflow_Mlflow
Domains	Version Management, Core Infrastructure
Last Updated	2026-02-13 20:00 GMT

Overview

Defines the canonical MLflow version string and package variant detection flags that control import behavior across the MLflow ecosystem.

Description

mlflow/version.py is the central version definition for the MLflow project. It serves two primary purposes:

Version String: The VERSION constant (currently "3.10.1.dev0") is the single source of truth for the MLflow package version. The is_release_version() function checks whether the current VERSION matches the release format (X.Y.Z with no suffix), returning False for development (.dev0) and release candidate (rcN) versions. This version string is read and updated by the dev/update_mlflow_versions.py script during the release process.

Package Variant Detection: The module uses importlib.metadata to detect which MLflow package variant is installed in the current environment:

• IS_TRACING_SDK_ONLY: Set to True when neither mlflow nor mlflow-skinny is installed, indicating only the tracing SDK (mlflow-tracing) is available. When True, modules that depend on dependencies not included in the tracing SDK are not imported.
• IS_MLFLOW_SKINNY: Set to True when mlflow-skinny is installed but the full mlflow package is not. The skinny package excludes heavy dependencies like SQL libraries and ML framework integrations.
• IS_FULL_MLFLOW: Set to True when the full mlflow package is installed.

The helper function _is_package_installed() uses importlib.metadata.version() to check package availability, catching PackageNotFoundError for packages that are not installed.

Usage

This module is imported by mlflow/__init__.py to control conditional imports based on the installed package variant. Other modules reference VERSION for version reporting, API headers, and compatibility checks.

Code Reference

Source Location

• Repository: Mlflow_Mlflow
• File: mlflow/version.py
• Lines: 1-29

Signature

VERSION = "3.10.1.dev0"

def is_release_version() -> bool: ...

def _is_package_installed(package_name: str) -> bool: ...

IS_TRACING_SDK_ONLY: bool
IS_MLFLOW_SKINNY: bool
IS_FULL_MLFLOW: bool

Import

from mlflow.version import VERSION
from mlflow.version import IS_TRACING_SDK_ONLY, IS_MLFLOW_SKINNY, IS_FULL_MLFLOW
from mlflow.version import is_release_version

I/O Contract

Inputs

Name	Type	Required	Description
(none)	-	-	This module requires no inputs; it reads version from its own source and checks installed packages via importlib.metadata

Outputs

Name	Type	Description
VERSION	str	The current MLflow version string (e.g., "3.10.1.dev0")
is_release_version()	bool	True if VERSION matches the release format (X.Y.Z with no suffix)
IS_TRACING_SDK_ONLY	bool	True if only mlflow-tracing is installed (no mlflow or mlflow-skinny)
IS_MLFLOW_SKINNY	bool	True if mlflow-skinny is installed but not the full mlflow package
IS_FULL_MLFLOW	bool	True if the full mlflow package is installed

Package Variant Detection Logic

Installed Package	IS_FULL_MLFLOW	IS_MLFLOW_SKINNY	IS_TRACING_SDK_ONLY
mlflow	True	False	False
mlflow-skinny	False	True	False
mlflow-tracing only	False	False	True

Usage Examples

Check Version

from mlflow.version import VERSION, is_release_version

print(f"MLflow version: {VERSION}")
if is_release_version():
    print("This is a release version")
else:
    print("This is a development or RC version")

Conditional Import Based on Package Variant

from mlflow.version import IS_TRACING_SDK_ONLY

if not IS_TRACING_SDK_ONLY:
    # Import modules that require full MLflow or skinny dependencies
    from mlflow.tracking import MlflowClient

Related Pages

• Environment:Mlflow_Mlflow_Python_Runtime_Environment