🔥 PyTorch Examples Repository

Author: Christopher Gaughan
LinkedIn | Portfolio

This repository contains practical, educational examples built with PyTorch 2.5.1, tested using NVIDIA A100 GPUs on Google Colab. All projects are fully anonymized and designed to showcase my machine learning engineering capabilities — from core tensor operations to transformer components like positional encoding.

---

🧠 Projects Overview

📍 Positional Encodings

Built from scratch to explore transformer positional encoding using sine/cosine signals

• Implements sinusoidal encodings from “Attention Is All You Need” Original Paper
• Visualizes encodings across positions & dimensions
• Explains how sequential token order is encoded in models like BERT
• Code: Positional_Encodings.ipynb

---

🧠 PyTorch Tensors Intro

Hands-on intro to tensor creation, indexing, reshaping, and operations

• Basics of tensors, operations, and gradients
• Great for newcomers or interviews
• Code: PyTorch_Tensor_Intro.ipynb

---

🧩 PyTorch Workflow

End-to-end ML project setup in PyTorch

• Data loading, model definition, training loop
• Clean separation of concerns
• Code: PyTorch_Workflow.ipynb

---

🖼️ Computer Vision Classification

Image classification using CNNs in PyTorch

• End-to-end workflow for image classification
• Uses torchvision transforms and loaders
• Code: ComputerVision_PyTorch.ipynb

---

🧪 Custom Dataset Handling

Shows how to create and load custom datasets in PyTorch

• Custom Dataset and DataLoader implementation
• Ideal for adapting real-world data into training pipelines
• Code: PyTorch_Custom_Datasets.ipynb

---

⚙️ Requirements

Tested with:

• Python 3.11
• PyTorch 2.5.1+cu121
• Jupyter Notebook
• NumPy
• Matplotlib

Install dependencies:

pip install torch matplotlib numpy

Usage

Clone the repository:

git clone https://github.com/christophergaughan/PyTorch.git
cd PyTorch

Launch the notebooks:

jupyter notebook

All notebooks are designed to run with or without GPU. For optimal speed, use Google Colab + GPU runtime.

GPU Acceleration

All code has been tested on:

• NVIDIA A100 GPUs (via Google Colab)

• Accelerated training for heavy models (e.g., CNNs, transformer prototypes)

Future Additions

• 🧠 Generative Adversarial Networks (GANs)

• 🕹️ Reinforcement Learning agents

• 🏥 Transfer learning with medical imaging

📚 References

• PyTorch Documentation

• Attention Is All You Need — Transformer architecture introduction

🪪 License

This project is licensed under the MIT License. See LICENSE file for details.