Lecture notes, readings, code samples and resources for teaching yourself how PyTorch and Tensorflow.
This class was formerly taught as a bootcamp but is now a self-paced online resource.
See Brad's GitHub Stars for a curated collection of helpful resources.
| Session | Topic | TODO |
|---|---|---|
| Preparation | N/A | See preparation.md |
| 📅 Week 1 | 📚 Roles, Machine Learning Basics, Tech Stack | How Deep Learning Works and Intro to Free Software |
| 📅 Week 1 | 🎬 "Foundations" Live Demos | SuperDataScience Code Interpreter Guide |
| 📅 Week 2 | 💡 Final Project Ideas | Pick a project! |
| 📅 Week 2 | 📊 Universal Machine Learning Template | The Universal ML Workflow and The Regression Theory of Everything |
| 📅 Week 3 | 👥 Pair Programming, Project Q&A | work on your project |
| 📅 Week 3 | 📝 Data Engineering, ETL Basics and Data Wrangling Techniques | work on your project |
| 📅 Week 4 | 📊 OpenRouter Sommelier App in AWS, Azure and GCP Cloud Setup Demo | Setup AWS Free Tier |
| 📅 Week 5 | 🧠 Metrics and Loss Functions, Model Architecture and Hyperparameters | 60 Minute Pytorch Blitz |
| 📅 Week 7 | 🎓 After The Bootcamp... | go forth and win! |
| 📅 Weeks 7-8 | 🎉 Project Presentations | brag about yourself online |
Participants will dedicate substantial time to final projects aligned with their ✨career aspirations✨. For an array of past presentations and source code, 📺 visit Brad's Youtube Channel.
| Step | Details |
|---|---|
| Utilize 🐍 Medusa-ML Template to kickstart your project. Generate synthetic text data using Scikit-LLM. Explore datasets on Kaggle or Hugging Face. | |
| Beginner: Follow TensorFlow quickstart for basic image/text classifiers. Recommended: Dive into 60 Minute PyTorch Blitz, then explore 📜 PyTorch Text Classification or PyTorch Image Classification. More Advanced: Fine-tune pretrained models with Hugging Face for 📝 text or 📸 image classification. Apple Nerds Only Use MLX. | |
| ⭐ Challenge and Experiment (Optional) | Advanced: Explore cross-platform frameworks like Keras Core. 🏆 Very Advanced: Attempt to Replicate a Winning Model from Kaggle. |