MLflow 실습

Scikit-learn의 Iris 데이터셋에 대해 실습을 해보았다.
https://github.com/yehoon17/mlflow-tutorial

1. MLflow 설정 및 실행

MLflow 트래킹 URI 설정

.env 파일 생성

MLFLOW_TRACKING_URI=file:///path/to/your/root/mlruns

예시: file:///C:/Users/.../mlflow-tutorial/mlruns

MLflow 실행

mlflow ui

http://127.0.0.1:5000에 접속

2. Scikit-learn 모델

환경설정

environment variable 불러오기

import os
from dotenv import load_dotenv

# Load environment variables from .env file
load_dotenv()

# Get MLflow tracking URI from environment variable
mlflow_tracking_uri = os.getenv("MLFLOW_TRACKING_URI")
print(mlflow_tracking_uri)

MLflow 트래킹 URI 설정

import mlflow

# Set MLflow tracking URI
mlflow.set_tracking_uri(mlflow_tracking_uri)

MLflow Experiment 설정

# Create or set MLflow Experiment
experiment_name = "iris_classification"
mlflow.set_experiment(experiment_name)

데이터셋 불러오기

from sklearn.datasets import load_iris

# Load dataset
iris = load_iris()
X = iris.data
y = iris.target

훈련, 테스트 데이터 분할

from sklearn.model_selection import train_test_split

# Split dataset into train and test sets
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

훈련 및 평가

from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score

# Start MLflow run
with mlflow.start_run(run_name="RandomForestClassifier"):
    # Log tags
    mlflow.set_tag("dataset", "Iris")
    mlflow.set_tag("experiment_type", "Classification")
    
    # Initialize and train model
    model = RandomForestClassifier(n_estimators=100, random_state=42)
    model.fit(X_train, y_train)

    # Make predictions
    y_pred = model.predict(X_test)

    # Calculate accuracy
    accuracy = accuracy_score(y_test, y_pred)

    # Log parameters and metrics
    mlflow.log_param("n_estimators", 100)
    mlflow.log_metric("accuracy", accuracy)

    # Save model
    mlflow.sklearn.log_model(model, "random_forest_model")

MLflow Run 이름 설정

with mlflow.start_run(run_name="RandomForestClassifier"):

MLflow Tag 설정

    # Log tags
    mlflow.set_tag("dataset", "Iris")
    mlflow.set_tag("experiment_type", "Classification")

Parameter 및 Metric 기록

    # Log parameters and metrics
    mlflow.log_param("n_estimators", 100)
    mlflow.log_metric("accuracy", accuracy)

모델 저장

    # Save model
    mlflow.sklearn.log_model(model, "random_forest_model")