Hello @Berthold, Jan
Thank you for Reaching out Microsoft Q&A.
At present, Azure Machine Learning’s native MLflow integration is primarily aligned with the MLflow 2.x API surface, and full support for MLflow 3.x functionality is not yet generally available across all Azure ML-integrated MLflow endpoints and features.
The behavior you referenced (including the 404/API compatibility issues discussed in the linked threads) is currently a known limitation. Azure ML today continues to rely on the existing MLflow-compatible tracking endpoints (for example /api/2.0/mlflow/...), which means some newer MLflow 3.x APIs and workflows are not yet fully supported.
Current support status
The following MLflow capabilities continue to work in Azure ML:
- Experiment tracking
- Metrics and parameter logging
- Artifact logging
- Model registration and deployment
- Standard MLflow 2.x tracking workflows
However, some MLflow 3.x-specific functionality may not yet work correctly, including:
-
logged_models
- newer registry APIs/views
- certain GenAI tracing/evaluation APIs
- some MLflow 3.x client expectations/endpoints
This is why using an MLflow 3.x client against an Azure ML workspace can currently result in issues such as:
-
404 responses
- unsupported endpoint errors
- model logging incompatibilities
Current recommendation / workaround
As a temporary workaround, the current recommendation is to use an MLflow client version below 3.0, for example:
pip install "mlflow<3.0"
and continue using the documented MLflow 2.x-compatible workflows in Azure ML.
At this time, there is no publicly announced ETA for complete MLflow 3.x support or full API parity in Azure Machine Learning Studio.
Please refer this
Track local and remote runs with MLflow: https://docs.microsoft.com/azure/machine-learning/how-to-use-mlflow
MLflow compatibility in Azure ML: https://docs.microsoft.com/azure/machine-learning/concept-mlflow?view=azureml-api-2
I Hope this helps. Do let me know if you have any further queries.
If this answers your query, please do click Accept Answer and Yes for was this answer helpful.
Thank you!