Skip to content

module-02

Directory actions

More options

Directory actions

More options

Latest commit

 

History

History

module-02

README.md

Module 2: PyTorch Workflow Fundamentals

Learn the complete deep learning workflow from data to deployment

Quick Start

  1. Read the conceptual guides in Part 1 (start with Introduction & Setup)
  2. Practice with hands-on exercises in ../../module-02/pytorch-workflow/
Learn:  docs/module-02/                  →  Theory and concepts
Do:     module-02/pytorch-workflow/      →  Hands-on exercises

Module Overview

This module teaches the end-to-end PyTorch workflow for building, training, and deploying deep learning models. You'll work through a complete linear regression example, learning each step of the process: from data preparation through model saving.

What we'll build: A linear regression model that learns the relationship y = 0.7*X + 0.3 from synthetic data.

Learning Objectives

By the end of this module, you will be able to:

Workflow Fundamentals

  • Understand the 6-step PyTorch workflow (data → build → train → evaluate → save → deploy)
  • Create and split datasets into train/validation/test sets
  • Build models by subclassing nn.Module

Training & Evaluation

  • Implement the 5-step training loop (forward, loss, zero grad, backward, optimizer step)
  • Use loss functions and optimizers effectively
  • Evaluate models on test data

Model Persistence

  • Save trained models using state_dict
  • Load models for inference
  • Write device-agnostic code that works on CPU and GPU

Study Path

Part 1: The PyTorch Workflow

Master the foundational steps of the deep learning workflow.

# Topic Description Practice
1 Introduction & Setup The big picture, learning mottos, environment setup -
2 Data Preparation Creating synthetic data, train/val/test splits, visualization 01_data_preparation.py
3 Building Models nn.Module, nn.Parameter, forward method 02_building_models.py

Part 2: Training & Deployment

Complete the workflow with training, evaluation, and model persistence.

# Topic Description Practice
4 Training Loop Loss functions, optimizers, 5-step training loop, inference 03_training_models.py
5 Saving & Loading Model persistence, checkpoints, device-agnostic code 04_inference_and_saving.py
05_complete_workflow.py

Exercises

The Learning Mottos

Throughout this module, apply these three core principles:

1. If in doubt, run the code!

Don't just read about concepts—execute them. Seeing the output builds intuition faster than studying theory.

2. Experiment, experiment, experiment!

Modify parameters, break things intentionally, try different approaches. Active learning creates deeper understanding than passive reading.

3. Visualize, visualize, visualize!

Plot your data, your training progress, your predictions. Visual patterns reveal insights that numbers alone cannot.

Prerequisites

  • Module 1 completed: Deep Learning Foundations with PyTorch
  • PyTorch installed: Install Guide
  • matplotlib installed: pip install matplotlib (for visualizations)

Running the Exercises

cd module-02/pytorch-workflow
python 01_data_preparation.py
python 02_building_models.py
python 03_training_models.py
python 04_inference_and_saving.py
python 05_complete_workflow.py

Key Concepts

The 6-Step PyTorch Workflow

Step Action Description
1 Data Preparation Create and split data into train/val/test sets
2 Build Model Define architecture by subclassing nn.Module
3 Train Implement training loop with loss and optimizer
4 Evaluate Test model on unseen data
5 Save Persist trained parameters with state_dict
6 Load Reload model for inference or deployment

The Three Data Splits

Split Purpose Typical Usage
Training Fit model parameters 70% of data
Validation Tune hyperparameters 15% of data
Test Final evaluation 15% of data

The 5-Step Training Loop

# 1. Forward pass
y_pred = model(X_train)

# 2. Calculate loss
loss = criterion(y_pred, y_train)

# 3. Zero gradients
optimizer.zero_grad()

# 4. Backward pass
loss.backward()

# 5. Update parameters
optimizer.step()

Module Summary

After completing this module, you should:

  • Understand the complete PyTorch workflow from start to finish
  • Be able to build, train, and save your own models
  • Know how to evaluate models and make predictions
  • Be ready for Module 3: Neural Network Classification

Additional Resources

External References

Internal Documentation

Next Steps

  1. Start with Introduction & Setup to understand the big picture
  2. Complete all exercises in the pytorch-workflow/ directory
  3. Review the key concepts summary above
  4. Move to Module 3: Neural Network Classification

Module Overview: ../../module-02/

Last Updated: January 2026