MLflow

• MLflow is an open-source platform designed to manage the end-to-end machine learning lifecycle.
• It is integrated with various ML & DevOps tools and frameworks. [because we are using agile methodology]
• It provides tools for tracking experiments, packaging code into reproducible runs, and sharing and deploying models.
• MLflow aims to simplify the process of developing, deploying, and maintaining machine learning models by providing a unified interface for various stages of the machine learning workflow.

Steps to Install MLflow:

• Open your terminal or command prompt.
• conda create -n deploy python=3.9.23
• conda activate deploy
  • to check version of python python --version
• to install requirements.txt file run below command
  • pip install -r requirements.txt
  • to check mlflow version mlflow --version
• to start mlflow ui run below command
  • mlflow ui
  • Open your web browser and go to http://localhost:5000 to access the MLflow UI.
  • VIP: Make sure to keep the terminal running while using the UI. ⚠️
• open another terminal and activate the environment conda activate mlflow_env to run your python scripts.

ML Lifecycle System

MLflow Components

MLflow Tracking

MLflow Structure

experiments  (0)   'it's unique for each project'
    ├── runs (0)
    ├── runs (1)
    ├── runs (2)
    ├── ...

Notes on MLflow Tracking

• mlflow.set_tracking_uri() - Sets the tracking URI for MLflow, which can be a [remote server, database connection string, local directory]. Defaults to local directory 'mlruns' directory.
• mlflow.get_tracking_uri() - Retrieves the current tracking URI.

---

• mlflow.create_experiment(name) - Creates a new experiment and returns its ID. If an experiment with the same name exists, will return error.
• mlflow.set_experiment() - Sets an experiment as active, creating it if it doesn't exist.

---

• mlflow.start_run() - Starts a new run or returns the active one. Automatically called by logging functions if no active run exists.
• mlflow.end_run(status) - Ends the current run with an optional status.

---

• mlflow.log_param() - Logs a single key-value parameter in the active run.
• mlflow.log_params() - Logs a dictionary key-value parameters in the active run.
• mlflow.log_metric() - Logs a single key-value metric in the active run.
• mlflow.log_metrics() - Logs a dictionary key-value metrics in the active run.
• mlflow.log_artifact() - Logs a local file or directory as an artifact, optionally specifying the path within the run's artifact URI.
• mlflow.log_artifacts() - Logs all files in a directory as artifacts, optionally specifying the artifact path.
• mlflow.log_model() - Logs a machine learning model, specifying the model object, artifact path, and optional serialization format.
• mlflow.log_figure() - Logs a matplotlib figure as an artifact in the active run.

---

• mlflow.set_tag(key, value) - Sets a single key-value tag in the active run. 'metadata' information.

🚀 Serving MLflow Models

1. Start the REST Server

mlflow models serve -m "path_to_model" --port 8000 --env-manager=local

Example:

mlflow models serve -m "file:///D:/Data_Science/Deployment/mlruns/266720109967700531/870c9a74719f438fb4cc192815e066ba/artifacts/RandomForestClassifier_SMOTE" --port 8000 --env-manager=local

2. Access the API

• The server runs at:

  http://localhost:8000
  

• The root (/) endpoint is empty → that’s normal.

• Use /invocations for predictions.

3. Important Notes

• ✅ Keep the terminal running while serving. ⚠️
• ✅ Test using Postman, curl, or Python requests. in Step 4.

Prediction Endpoint:

POST http://localhost:8000/invocations

4. Open Postman:

• New HTTP Request
• Method: POST
• URL: http://localhost:8000/invocations
• Body: raw 'JSON'
• paste the below JSON data in the body section.

# example of data to be sent [postman]

{
    "dataframe_split": {
        "columns": [
            "Age", 
            "CreditScore", 
            "Balance", 
            "EstimatedSalary", 
            "Gender_Male", 
            "Geography_Germany", 
            "Geography_Spain", 
            "HasCrCard", 
            "Tenure", 
            "IsActiveMember", 
            "NumOfProducts"
        ],
        "data": [
            [-0.7541830079917924, 
            0.5780143566720919, 
            0.11375998165198585, 
            -0.14673040749854463, 
            0.0, 
            0.0, 
            0.0, 
            0.0, 
            2.0, 
            0.0, 
            2.0]
        ]
    }
}
# ----------------------------------------------------------------------------
# multiple samples [postman]

{
    "dataframe_split": {
        "columns": [
            "Age",
            "CreditScore",
            "Balance",
            "EstimatedSalary",
            "Gender_Male",
            "Geography_Germany",
            "Geography_Spain",
            "HasCrCard",
            "Tenure",
            "IsActiveMember",
            "NumOfProducts"
        ],
        "data": [
            [-0.7541830079917924, 0.5780143566720919, 0.11375998165198585, -0.14673040749854463, 0.0, 0.0, 0.0, 0.0, 2.0, 0.0, 2.0],
            [-0.5605884106597949, 0.753908347743766, 0.7003528882054108, 1.6923927520037099, 0.0, 1.0, 0.0, 1.0, 9.0, 1.0, 1.0],
            [0.11699268000219652, -0.3221490094005933, 0.5222180917013974, -0.8721429873346316, 1.0, 1.0, 0.0, 1.0, 5.0, 0.0, 2.0],
            [0.6977764719981892, -0.7256705183297281, -1.2170740485175422, 0.07677206232885857, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 2.0]
        ]
    }
}
# ----------------------------------------------------------------------------
# if you want to use curl [git bash]

curl -X POST \
  http://localhost:8000/invocations \
  -H 'Content-Type: application/json' \
  -d '{
    "dataframe_split": {
        "columns": [
            "Age",
            "CreditScore",
            "Balance",
            "EstimatedSalary",
            "Gender_Male",
            "Geography_Germany",
            "Geography_Spain",
            "HasCrCard",
            "Tenure",
            "IsActiveMember",
            "NumOfProducts"
        ],
        "data": [
            [-0.7541830079917924, 0.5780143566720919, 0.11375998165198585, -0.14673040749854463, 0.0, 0.0, 0.0, 0.0, 2.0, 0.0, 2.0],
            [-0.5605884106597949, 0.753908347743766, 0.7003528882054108, 1.6923927520037099, 0.0, 1.0, 0.0, 1.0, 9.0, 1.0, 1.0],
            [0.11699268000219652, -0.3221490094005933, 0.5222180917013974, -0.8721429873346316, 1.0, 1.0, 0.0, 1.0, 5.0, 0.0, 2.0],
            [0.6977764719981892, -0.7256705183297281, -1.2170740485175422, 0.07677206232885857, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 2.0]
        ]
    }
}'
# ----------------------------------------------------------------------------
# if you want to use [Powershell]

Invoke-RestMethod -Uri "http://localhost:8000/invocations" -Method Post -Headers @{"Content-Type" = "application/json"} -Body '{
    "dataframe_split": {
        "columns": [
            "Age",
            "CreditScore",
            "Balance",
            "EstimatedSalary",
            "Gender_Male",
            "Geography_Germany",
            "Geography_Spain",
            "HasCrCard",
            "Tenure",
            "IsActiveMember",
            "NumOfProducts"
        ],
        "data": [
            [-0.7541830079917924, 0.5780143566720919, 0.11375998165198585, -0.14673040749854463, 0.0, 0.0, 0.0, 0.0, 2.0, 0.0, 2.0],
            [-0.5605884106597949, 0.753908347743766, 0.7003528882054108, 1.6923927520037099, 0.0, 1.0, 0.0, 1.0, 9.0, 1.0, 1.0],
            [0.11699268000219652, -0.3221490094005933, 0.5222180917013974, -0.8721429873346316, 1.0, 1.0, 0.0, 1.0, 5.0, 0.0, 2.0],
            [0.6977764719981892, -0.7256705183297281, -1.2170740485175422, 0.07677206232885857, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 2.0]
        ]
    }
}'

MLflow Remote Tracking

1-go to github repo and create repo 2- copy the repo link 3- go to your local machine and run the following command

git clone "repo_link"

4- go to the repo folder and reaad the README.md file i will complete notes on that file.