FAQs

Can MLflow 3.x load resources (runs, models, traces, etc.) created with MLflow 2.x?

Yes, MLflow 3.x can load resources such as runs, models, traces, and more that were created with MLflow 2.x. However, the reverse is not true.

warning

When testing MLflow 3.x, we strongly recommend using a separate environment to avoid conflicts with MLflow 2.x.

load_model throws a ResourceNotFound error when loading a model created with MLflow 2.x. What's wrong?

For example, the following code fails to load the model in MLflow 3.x since the model artifacts are NOT stored as run artifacts:

import mlflow

with mlflow.start_run() as run:
    mlflow.sklearn.log_model(my_model, name="model")
    mlflow.sklearn.load_model(mlflow.get_artifact_uri("model"))
    # Throws a `ResourceNotFound` error.

To avoid this error, call mlflow.<flavor>.load_model with the model URI returned by mlflow.<flavor>.log_model:

import mlflow

with mlflow.start_run() as run:
    info = mlflow.sklearn.log_model(my_model, name="model")
    # if the result of `log_model` is available (recommended)
    mlflow.sklearn.load_model(info.model_uri)
    # if only `model_id` is available
    mlflow.sklearn.load_model(f"models:/{info.model_id}")
    # if neither `model_id` nor `model_uri` is available (deprecated and will be removed in future versions)
    mlflow.sklearn.load_model(f"runs:/{run.info.run_id}/model")

Why does this happen? In MLflow 3.x, the model artifacts are stored in a different location than in MLflow 2.x. The following is a comparison of the two versions using the tree format:

# MLflow 2.x
experiments/
  └── <experiment_id>/
    └── <run_id>/
      └── artifacts/
        └── ... # model artifacts are stored here

# MLflow 3.x
experiments/
  └── <experiment_id>/
    └── models/
      └── <model_id>/
        └── artifacts/
          └── ... # model artifacts are stored here

I want to modify requirements.txt of my model. How can I do that?

You can modify requirements.txt of your model using mlflow.models.update_model_requirements() to add or remove requirements. Here's an example of how to add a requirement to the model:

import mlflow


class DummyModel(mlflow.pyfunc.PythonModel):
    def predict(self, context, model_input: list[str]) -> list[str]:
        return model_input


model_info = mlflow.pyfunc.log_model(name="model", python_model=DummyModel())
mlflow.models.update_model_requirements(
    model_uri=model_info.model_uri,
    operation="add",
    requirement_list=["scikit-learn"],
)

I'm still not ready to upgrade to MLflow 3.x. How can I pin my MLflow version to 2.x?

You can pin MLflow to the latest 2.x version by using the following command:

pip install 'mlflow<3'