Reproducible training pipelines in MLOps - Mini Project: Build & Apply

Reproducible training pipelines

📖 Scenario: You are working as a machine learning engineer. Your team wants to create a training pipeline that always produces the same results when given the same data and code. This helps avoid surprises and makes debugging easier.To do this, you will build a simple reproducible training pipeline step-by-step.

🎯 Goal: Build a reproducible training pipeline that loads data, sets a fixed random seed, trains a simple model, and prints the model accuracy. This pipeline should produce the same accuracy every time it runs.

📋 What You'll Learn

Create a dataset variable with fixed data

Set a fixed random seed for reproducibility

Train a simple model using the fixed data and seed

Print the model accuracy as the final output

💡 Why This Matters

🌍 Real World

Reproducible training pipelines help teams avoid bugs and inconsistencies when training machine learning models repeatedly.

💼 Career

Understanding reproducibility is essential for ML engineers and data scientists to build reliable and trustworthy models.

1

Create the dataset

Create a variable called data that is a list of tuples with these exact entries: ([0, 0], 0), ([1, 1], 1), ([1, 0], 1), ([0, 1], 0).

MLOps

# Create the dataset variable called data
# Your code here

Need a hint?

Use a list of tuples where each tuple has a list of features and a label.

2

Set a fixed random seed

Import the random module and set the random seed to 42 using random.seed(42).

MLOps

data = [([0, 0], 0), ([1, 1], 1), ([1, 0], 1), ([0, 1], 0)]
# Import random and set seed to 42
# Your code here

Need a hint?

Use import random at the top and then random.seed(42) to fix randomness.

3

Train a simple model

Create a function called train_model that takes data as input and returns a dictionary model with keys 'threshold' set to 0.5. Then call train_model(data) and save the result in a variable called model.

MLOps

import random
random.seed(42)
data = [([0, 0], 0), ([1, 1], 1), ([1, 0], 1), ([0, 1], 0)]
# Define train_model function and call it with data
# Your code here

Need a hint?

Define a function that returns a fixed model dictionary and call it.

4

Print the model accuracy

Calculate the accuracy by comparing the model's prediction with the true label for each data point. Use the rule: predict 1 if sum of features >= model['threshold'], else 0. Print the accuracy as a float with two decimals using print(f"Accuracy: {accuracy:.2f}").

MLOps

import random
random.seed(42)
data = [([0, 0], 0), ([1, 1], 1), ([1, 0], 1), ([0, 1], 0)]

def train_model(data):
    model = {'threshold': 0.5}
    return model

model = train_model(data)
# Calculate accuracy and print it
# Your code here

Need a hint?

Loop over data, predict using threshold, count correct predictions, then print accuracy.