MLFlow - Problem with aliases and metrics

[claudevs]

Our team is trying to use MLFlow to manage the models created in fabric, however this has been a hard task since many features of MLFlow Python library are apparently unavailable in Fabric. We tried to implement aliases in our models without any success, the same thing happened when we try to access metrics via models (not runs). Every time we instanciate a model only some data are available like model_name, version and tags, but we can´t access to params and metrcis asociated to that model.

I would appreciate any help, as I don't know if these issues are due to a limited implementation of MLFlow's capabilities in Fabric or incorrect use of the MLFlow API. If anyone could share some code that allows acess to model metrics and parameters using any type of model object (LoggedModel, ModelVersion, ModelInfo, etc) would be very appreciated.

[v-karpurapud]

Hi @claudevs 

We have not received a response from you regarding the query and were following up to check if you had the opportunity to review the information provided. Please feel free to contact us if you have any further questions.

Thank You.

[v-karpurapud]

Hi @claudevs 

Thank you for posting your query in the Microsoft Fabric Community Forum, and thanks to @bariscihan for sharing valuable insights.

Could you please confirm if your query has been resolved by the provided solution?If you have any more questions, please let us know and we’ll be happy to help.

Regards,

Microsoft Fabric Community Support Team

[bariscihan · Accepted solution]

What you’re seeing is expected MLflow behavior, and it can be confusing at first in Fabric.

Key point: in MLflow, metrics/params live on the Run, not on the Model Registry objects. A ModelVersion (or ModelInfo, LoggedModel, etc.) typically exposes registry metadata (name, version, tags, source, run_id), but not the run’s params/metrics directly. The right pattern is:

ModelVersion → run_id → client.get_run(run_id)

If aliases are failing in Fabric, it’s usually because the hosted registry endpoint doesn’t support the alias APIs (or the underlying MLflow server feature set differs). In that case, a practical workaround is to use model version tags as a “pseudo-alias” (e.g., champion=true, env=prod) and resolve your “production” version via tags.

References:

• MLflow Client / Registry API: https://mlflow.org/docs/latest/python_api/mlflow.client.html

• MLflow Model Registry (aliases & versions): https://mlflow.org/docs/latest/ml/model-registry/ and https://mlflow.org/docs/latest/ml/model-registry/tutorial/

• Fabric ML model experience (run tracking + registering): https://learn.microsoft.com/en-us/fabric/data-science/machine-learning-model
  
  
  

import mlflow
from mlflow.tracking import MlflowClient

client = MlflowClient()

model_name = "<YOUR_MODEL_NAME>"
model_version = "<YOUR_VERSION_NUMBER>"  # e.g. "3"

# 1) Registry -> ModelVersion
mv = client.get_model_version(name=model_name, version=model_version)
print("Model:", mv.name, "Version:", mv.version, "RunId:", mv.run_id)

# 2) Run -> params/metrics
run = client.get_run(mv.run_id)

print("\nParams:")
print(run.data.params)

print("\nMetrics:")
print(run.data.metrics)

# Optional: metric history (if you logged multiple values over time)
# hist = client.get_metric_history(mv.run_id, "accuracy")
# print(hist)

[claudevs]

Thank you for your answer @bariscihan

At the end, your solution is part of how we use MLFlow, nevertheless I´m still finding out more missing features of MLFlow in fabric, although these features also can be replaced with other approaches, It would be awesome if they could be implemented in the future, as MLFLow is a powerful tool that helps a lot in the ML model lifecycle proccess

[claudevs]

Thank you for your answer @bariscihan

At the end, your solution is part of how we use MLFlow, nevertheless I´m still finding out more missing features of MLFlow in fabric, although these features also can be replaced with other approaches, It would be awesome if they could be implemented in the future, as MLFLow is a powerful tool that helps a lot in the ML model lifecycle process