Machine Learning with Hugging Face Bootcamp: Zero to Mastery

Overview

Learn to do Machine Learning using the Hugging Face ecosystem from scratch. This project-based course will teach you everything from training to deployment using Hugging Face. We'll start from the basics, learn real skills used by Machine Learning Engineers, and have fun along the way!

Find and load datasets with Hugging Face Datasets
Clean, label, and batch text the simple way
Fine-tune a small Transformer for classification
Measure accuracy and improve step by step
Save and share your model on the Hugging Face Hub
Build an easy demo with Gradio
Publish your app on Hugging Face Spaces
Explain your projects with a clear model card

Syllabus

Introduction

Machine Learning with Hugging Face Bootcamp: Zero to Mastery
Exercise: Meet Your Classmates and Instructor
ZTM Plugin + Understanding Your Video Player
Course Resources
Set Your Learning Streak Goal

Course Overview

Overview

Project 0 - Introduction to Text Classification

Introduction to Text Classification
What We're Going To Build!

Project 0 - Let's Get Started!

Getting Setup: Adding Hugging Face Tokens to Google Colab
Getting Setup: Importing Necessary Libraries to Google Colab
Downloading a Text Classification Dataset from Hugging Face Datasets

Project 0 - Preparing Text Data & Evaluation Metric

Preparing Text Data for Use with a Model - Part 1: Turning Our Labels into Numbers
Preparing Text Data for Use with a Model - Part 2: Creating Train and Test Sets
Preparing Text Data for Use with a Model - Part 3: Getting a Tokenizer
Preparing Text Data for Use with a Model - Part 4: Exploring Our Tokenizer
Preparing Text Data for Use with a Model - Part 5: Creating a Function to Tokenize Our Data
Setting Up an Evaluation Metric (to measure how well our model performs)
Let's Have Some Fun (+ More Resources)

Project 0 - Model Training

Introduction to Transfer Learning (a powerful technique to get good results quickly)
Model Training - Part 1: Setting Up a Pretrained Model from the Hugging Face Hub
Model Training - Part 2: Counting the Parameters in Our Model
Model Training - Part 3: Creating a Folder to Save Our Model
Model Training - Part 4: Setting Up Our Training Arguments with TrainingArguments
Model Training - Part 5: Setting Up an Instance of Trainer with Hugging Face Transformers
Model Training - Part 6: Training Our Model and Fixing Errors Along the Way
Model Training - Part 7: Inspecting Our Models Loss Curves
Model Training - Part 8: Uploading Our Model to the Hugging Face Hub
Unlimited Updates

Project 0 - Making Predictions

Making Predictions on the Test Data with Our Trained Model
Turning Our Predictions into Prediction Probabilities with PyTorch
Sorting Our Model's Predictions by Their Probability
Course Check-In

Project 0 - Performing Inference

Performing Inference - Part 1: Discussing Our Options
Performing Inference - Part 2: Using a Transformers Pipeline (one sample at a time)
Performing Inference - Part 3: Using a Transformers Pipeline on Multiple Samples at a Time (Batching)
Performing Inference - Part 4: Running Speed Tests to Compare One at a Time vs. Batched Predictions
Performing Inference - Part 5: Performing Inference with PyTorch
OPTIONAL - Putting It All Together: from Data Loading, to Model Training, to making Predictions on Custom Data
Implement a New Life System

Project 0 - Launching Our Model!

Turning Our Model into a Demo - Part 1: Gradio Overview
Turning Our Model into a Demo - Part 2: Building a Function to Map Inputs to Outputs
Turning Our Model into a Demo - Part 3: Getting Our Gradio Demo Running Locally
Making Our Demo Publicly Accessible - Part 1: Introduction to Hugging Face Spaces and Creating a Demos Directory
Making Our Demo Publicly Accessible - Part 2: Creating an App File
Making Our Demo Publicly Accessible - Part 3: Creating a README File
Making Our Demo Publicly Accessible - Part 4: Making a Requirements File
Making Our Demo Publicly Accessible - Part 5: Uploading Our Demo to Hugging Face Spaces and Making it Publicly Available
Summary Exercises and Extensions

Project 1 - Building a Custom Object Detection Model with Hugging Face Transformers

Project 1 Overview and Resources
Introduction
Setting Up Google Colab with Hugging Face Tokens
Installing Necessary Dependencies
Getting an Object Detection Dataset
Inspecting the Features of Our Dataset
Creating a Colour Palette to Visualize Our Classes
Creating a Helper Function to Halve Our Image Sizes
Creating a Helper Function to Halve Our Box Sizes
Testing our Helper Functions
Outlining the Steps to Draw Boxes on an Image
Plotting Bounding Boxes on a Single Image Step by Step
Different Bounding Box Formats

Project 1 - Getting and Object Detection Model and Image Preprocessor

Getting an Object Detection Model
Transfer Learning Overview
Downloading our Model from the Hugging Face Hub and Trying it Out
Inspecting the Layers of Our Model
Counting the Number of Parameters in Our Model
Creating a Function to Build Our Custom Model
Passing a Single Image Sample Through Our Model - Part 1
OPTIONAL: Data Preprocessor Model Workflow
Loading Our Models Image Preprocessor and Customizing it for Our Use Case
Exercise: Imposter Syndrome

Project 1 - Getting Hands-on with Different Bounding Box Formats

Bounding Box Formats 101
Discussing the Format Our Model Expects Our Annotations In (COCO)
Creating Dataclasses to Hold the COCO Format
Creating a Function to Turn Our Annotations into COCO Format
Preprocessing a Single Image Sample and COCO Formatted Annotations
Post Processing a Single Output
Plotting a Single Post Processed Sample onto an Image

Project 1 - OPTIONAL: Reproducing our Model's Post Processed Outputs by Hand

OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 1: Overview
OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 2: Replicating Scores by Hand
OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 3: Replicating Labels by Hand
OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 4: Replicating Boxes by Hand Overview
OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 5: Replicating Boxes by Hand Implementation
OPTIONAL: Reproducing Our Models Post Processed Outputs by Hand - Part 6: Plotting Our Manual Post Processed Outputs on an Image

Project 1 - Preparing our Data for Training our Object Detection Model

Preparing Our Data at Scale - Part 1: Concept Overview
Preparing Our Data at Scale - Part 2: Creating Train Validation and Test Splits
Preparing Our Data at Scale - Part 3: Preprocessing Multiple Samples at a Time Overview
Preparing our Data at Scale - Part 4: Making a Function to Preprocess Multiple Samples at a Time
Preparing our Data at Scale - Part 5: Applying Our Preprocessing Function to Our Datasets
Preparing Our Data at Scale - Part 6: Creating a Data Collation Function

Project 1 - Training a Custom Object Detection Model for Trashify

Training a Custom Model - Part 1: Overview
Training a Custom Model - Part 2: Creating a Model and Folder to Save Our Model to
Training a Custom Model - Part 3: Creating TrainingArguments for Our Model Overview
Training a Custom Model - Part 4: Creating our First TrainingArguments
Training a Custom Model - Part 5: Finishing Off the TrainingArguments
Training a Custom Model - Part 6: OPTIONAL - Creating a Custom Optimizer for Different Learning Rates
Training a Custom Model - Part 7: Creating an Evaluation Function for Our Model Overview
Training a Custom Model - Part 8: Creating an Evaluation Function for Our Model Targets Processing
Training a Custom Model - Part 9: Creating an Evaluation Function for Our Model Predictions Processing
Training a Custom Model - Part 10: Training Our Model with Trainer
Training a Custom Model - Part 11: Plotting Our Models Loss Curves

Project 1 - Evaluating our Trained Object Detection Model

Evaluating Our Model on the Test Dataset
Making Predictions on Test Data and Visualizing Them
Plotting Our Models Predictions vs. the Ground Truth Images
Trying Our Model on Images from the Wild
Uploading Our Trained Model to the Hugging Face Hub

Project 1 - Bringing Trashify to Life: Turning our Custom Model into a Shareable Demo

Turning Our Model into a Demo - Part 1: Gradio and Hugging Face Spaces Overview
Turning Our Model into a Demo - Part 2: Creating an App File Overview
Turning Our Model into a Demo - Part 3: Building the Main Function of Our App File
Turning Our Model into a Demo - Part 4: Finishing Off Our App File and Testing Our Demo
Turning Our Model into a Demo - Part 5: Creating a Readme and Requirements File
Turning Our Model into a Demo - Part 6: Getting Example Images for Our Demo
Turning Our Model into a Demo - Part 7: Uploading Our Demo to the Hugging Face Hub
Turning Our Model into a Demo - Part 8: Embedding Our Demo into Our Notebook
Summary, Extensions and Extra-Curriculum