MLflow Model Registry doesn’t work on DagsHub hosted MLflow

[vikasupadhyaya]

When attempting to register a model from a DagsHub-hosted MLflow run, I receive a 404 error indicating that the Model Registry API endpoints are not supported.

import mlflow

mlflow.set_tracking_uri("https://dagshub.com/vikasupadhyaya/PCOS_Model_training.mlflow")

RUN_ID = "6c5867499cb04b1c9d5128a7b8a15388"
ARTIFACT_PATH = "model_artifacts"
REGISTERED_MODEL_NAME = "RF_PCOS_Model"

model_uri = f"runs:/{RUN_ID}/{ARTIFACT_PATH}"
mlflow.register_model(model_uri, REGISTERED_MODEL_NAME)

Error I get:

❌ Error registering model: API request to endpoint /api/2.0/mlflow/registered-models/create failed with error code 404 != 200.
mlflow.exceptions.MlflowException: Model Registry features are not supported by the store with URI: 'https://dagshub.com/vikasupadhyaya/PCOS_Model_training.mlflow'

I understand that DagsHub MLflow currently only supports tracking experiments, but having Model Registry support would be amazing for production workflows. Right now, I can log models as artifacts, but I can’t version them, register them, or move them between stages like “Staging” or “Production”.

Some context:

My MLflow client version is 2.10

DagsHub hosted MLflow is v2.7

Python 3.12, macOS ARM64

It would be super helpful if the hosted MLflow could support model registry, or if there’s a recommended workaround for production-ready pipelines.

Thanks! 🙏

[dagshub]

Join the discussion on DagsHub!

[kbolashev]

Hi! We do have the MLflow model registry working, but it's only supported for MLflow 2.x version for now.
There have been changes in the MLflow 3.x related to the model registry, and when people receive 404 errors, it's usually related to that.

The version of MLflow that's hosted on DagsHub is 2.22.0 right now.
What version of mlflow client are you running? You can check it by doing pip freeze | grep mlflow in the venv