How to Deploy MLFlow on the Vultr Kubernetes Engine

[mahesh]

How to Deploy MLFlow on the Vultr Kubernetes Engine

Introduction
MLflow is a versatile open-source Machine Learning (ML) platform for model lifecycle management. It supports experimentation, duplicability, deployment, and a central model registry. MLflow is used to track machine learning workload experiments, packaging and deployment of code, and collaborative management. Frequently used by data scientists and by MLOps professionals.

This article explains how to deploy MLflow on a Vultr Kubernetes Engine (VKE) cluster with a working environment and demo application.

Prerequisites
Before you begin:

Deploy a Vultr Kubernetes Engine(VKE) cluster

Deploy a Vultr Ubuntu instance to use as your management server

Using SSH, access the server

Install and configure Kubectl to access the cluster

Install the Helm package manager

Code Select Expand

  $ sudo snap install helm --classic
Deploy Persistent Volume and Persistent Volume Claims (PVCs)
MLflow requires persistent storage to store artifacts and experiment data. In this section, deploy a PV and PVC to the cluster as described in the steps below.

1.Create a new file mlflow-pv-pvc.yaml

Code Select Expand

$ touch mlflow-pv-pvc.yaml
2.Using a text editor such as nano, edit the file

Code Select Expand

$ nano mlflow-pv-pvc.yaml
3.Add the following contents to the file

 

Code Select Expand

apiVersion: v1
 kind: PersistentVolume
 metadata:
   name: mlflow-pv
   labels:
     type: local
 spec:
   storageClassName: manual
   capacity:
     storage: 10Gi
   accessModes:
     - ReadWriteOnce
   hostPath:
     path: "/mnt/data"
 ---
 apiVersion: v1
 kind: PersistentVolumeClaim
 metadata:
   name: mlflow-pvc
 spec:
   storageClassName: manual
   accessModes:
     - ReadWriteOnce
   resources:
     requests:
       storage: 10Gi
[/code Save the file

4.Apply the configuration to your cluster

[code] $ kubectl apply -f mlflow-pv-pvc.yaml5.Verify the deployed Cluster PVs

 

Code Select Expand

$ kubectl get pv
Output:

 

Code Select Expand

NAME        CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM                STORAGECLASS   REASON   AGE
 mlflow-pv   10Gi       RWO            Retain           Bound    default/mlflow-pvc   manual                
6.Verify the deployed Cluster PVCs

Code Select Expand

$ kubectl get pvc
Output:

Code Select Expand

NAME         STATUS   VOLUME      CAPACITY   ACCESS MODES   STORAGECLASS   AGE
 mlflow-pvc   Bound    mlflow-pv   10Gi       RWO            manual         24s
Install MLflow
1.Using Helm, add community-charts to your repositories

Code Select Expand

$ helm repo add community-charts https://community-charts.github.io/helm-charts
2.Update the Helm repository index

Code Select Expand

$ helm repo update
3.Install MLflow. Replace vultr with your desired MLflow label

Code Select Expand

$ helm install vultr community-charts/mlflow
4.View the cluster deployment to verify that MLflow is ready and available

Code Select Expand

$ kubectl get deployments
Your output should look like the one below:

Code Select Expand

NAME           READY   UP-TO-DATE   AVAILABLE   AGE
 vultr-mlflow    1/1     1            1           19h
Expose the MLflow Forwarding Service For External Access
To access your MLflow deployment over the Internet, set up a forwarding service to expose the application for external access as described in the steps below.

1.Create a new file mlflow-service.yaml

Code Select Expand

$ touch mlflow-service.yaml
2.Edit the file

Code Select Expand

$ nano mlflow-service.yaml
3.Add the following contents to the file

 

Code Select Expand

apiVersion: v1
 kind: Service
 metadata:
   name: mlflow-service
 spec:
   selector:
     app: mlflow
   ports:
     - protocol: TCP
       port: 80
       targetPort: 5000
   type: LoadBalancer
4.Apply the service to your cluster

Code Select Expand

$ kubectl apply -f mlflow-service.yaml
5.Wait for at least 3 minutes, view the cluster services, and verify the MLflow External IP value

Code Select Expand

  $ kubectl get services
Your output should look like the one below:

Code Select Expand

NAME             TYPE           CLUSTER-IP      EXTERNAL-IP   PORT(S)        AGE
 kubernetes       ClusterIP      10.96.0.1       <none>        443/TCP        40m
 mlflow-service   LoadBalancer   10.110.234.98   192.0.2.100   80:31545/TCP   64s
 vultr-mlflow     ClusterIP      10.102.93.29    <none>        5000/TCP       101s
Create a sample MLflow Experiment
To test the MLflow functionalities, create a sample linear regression experiment as described in the steps below.

1.Create a new directory Models

Code Select Expand

$ mkdir Models
2.Switch to the directory

Code Select Expand

$ cd Models
3.Using pip, install the necessary experiment dependencies

Code Select Expand

$ pip install mlflow scikit-learn shap matplotlib
4.When using Conda, install the dependencies in your environment:

Code Select Expand

$ conda install mlflow scikit-learn shap matplotlib
5.Create a new environment variable with your desired experiment name

Code Select Expand

$ export MLFLOW_EXPERIMENT_NAME='my-sample-experiment'
6.Export a new mlflow_TRACKING_URI with your MLflow service external IP HTTP URL as the value. For example 192.0.2.100

Code Select Expand

$ export MLFLOW_TRACKING_URI='http://192.0.2.100'
7.Create a new file main.py

Code Select Expand

$ touch main.py
8.Edit the file

Code Select Expand

$ nano main.py
Add the following contents to the file

 # Import Libraries

 import os

 import numpy as np
 import shap
 from sklearn.datasets import load_diabetes
 from sklearn.linear_model import LinearRegression

 import mlflow
 from mlflow.artifacts import download_artifacts
 from mlflow.tracking import MlflowClient

 # Prepare the Training Data

 X, y = load_diabetes(return_X_y=True, as_frame=True)
 X = X.iloc[:50, :4]
 y = y.iloc[:50]

 # Train a model

 model = LinearRegression()
 model.fit(X, y)

 # Log an explanation

 with mlflow.start_run() as run:
     mlflow.shap.log_explanation(model.predict, X)

 # List Artifacts

 client = MlflowClient()
 artifact_path = "model_explanations_shap"
 artifacts = [x.path for x in client.list_artifacts(run.info.run_id, artifact_path)]
 print("# artifacts:")
 print(artifacts)

 # Load the logged explanation

 dst_path = download_artifacts(run_id=run.info.run_id, artifact_path=artifact_path)
 base_values = np.load(os.path.join(dst_path, "base_values.npy"))
 shap_values = np.load(os.path.join(dst_path, "shap_values.npy"))

 # Show a Force Plot

 shap.force_plot(float(base_values), shap_values[0, :], X.iloc[0, :], matplotlib=True)
Save and close the file.

The above Python application imports all necessary libraries and uses the diabetes dataset for training data. The dataset contains a total of 442 samples. Then, the model is trained and an explanation is logged to MLflow using SHAP (Shapley Additive Explanations) that plots the output for a visual representation of the data.

Test the MLflow Experiement
1.Run the Python application

Code Select Expand

$ python3 main.py
2.Using a web browser such as Chrome, visit your MLflow external IP address

Code Select Expand

http://192.0.2.100
Verify that the application is included in your MLflow experiments. It uses a linear regression statistical approach to model a relationship between a scalar response and one or more variables.


Conclusion
You have deployed MLflow on a Vultr Kubernetes Engine (VKE) cluster. To test the service operations, implement more examples from the MLflow repository. For more information about MLflow, visit the following documentation resources:

MLflow documentation
MLflow Recipes
MLflow Projects