أساسيات MLOps

import mlflow
import pandas as pd
from sklearn.compose import ColumnTransformer
from sklearn.linear_model import Ridge
from sklearn.metrics import mean_absolute_error
from sklearn.model_selection import train_test_split
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import OneHotEncoder, StandardScaler

mlflow.set_tracking_uri("sqlite:///mlflow.db")      # local tracking database
mlflow.set_experiment("apartment-prices")

df = pd.read_csv("apartments.csv")
num, cat = ["area", "rooms", "floor", "age", "metro_km"], ["city"]
X_tr, X_te, y_tr, y_te = train_test_split(df[num + cat], df["price_k"],
                                          test_size=0.2, random_state=42)

for alpha in [0.1, 10, 100, 1000]:                   # try 4 settings
    with mlflow.start_run(run_name=f"ridge-{alpha}"):
        pipe = Pipeline([
            ("prep", ColumnTransformer([("num", StandardScaler(), num),
                                        ("cat", OneHotEncoder(), cat)])),
            ("model", Ridge(alpha=alpha)),
        ]).fit(X_tr, y_tr)
        mae = mean_absolute_error(y_te, pipe.predict(X_te))
        mlflow.log_param("alpha", alpha)             # what you tried
        mlflow.log_metric("mae_k", mae)              # how it did

runs = mlflow.search_runs(order_by=["metrics.mae_k ASC"])
print(runs[["params.alpha", "metrics.mae_k"]].round(1).to_string(index=False))

name: tests
on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: "3.12"
      - run: pip install -r requirements.txt pytest
      - run: python train.py          # train a fresh model
      - run: pytest -q                # then run all the tests