Duration
PyTorch from Basics to Cutting-Edge
3 day
What You’ll Learn
- Learn all relevant aspects of PyTorch from simple models to state-of-the-art models
- CNNs (Image-, Audio-Classification; Object Detection), RNNs, Transformers, Style Transfer, Autoencoders, GANs
- Adapt top-notch algorithms like Transformers to custom datasets
- Develop CNN models for image classification, object detection, Style Transfer
- Develop RNN models, Autoencoders, Generative Adversarial Networks
- Learn about new frameworks (e.g. PyTorch Lightning) and new models like OpenAI ChatGPT
- Use transfer learning
Outline
PyTorch Introduction
- System Setup
- How to Get the Course Material
- Setting up the conda environment
- General Environment Setup Error Handling
Machine Learning
- Artificial Intelligence
- Machine Learning
- Machine Learning Models
Deep Learning Introduction
- Deep Learning General Overview
- Deep Learning Modeling
- Performance
- From Perceptron to Neural Network
- Layer Types
- Activation Functions
- Loss Functions
- Optimizers
Model Evaluation
- Underfitting Overfitting
- Train Test Split
- Resampling Techniques
Neural Network from Scratch (opt. but highly recommended)
- Section Overview
- NN from Scratch
- Calculating the dot-product
- NN from Scratch (Data Prep)
- NN from Scratch Modeling __init__ function
- NN from Scratch Modeling Helper Functions
- NN from Scratch Modeling forward function
- NN from Scratch Modeling backward function
- NN from Scratch Modeling optimizer function
- NN from Scratch Modeling train function
- NN from Scratch Model Training
- NN from Scratch Model Evaluation
Tensors
- Section Overview
- From Tensors to Computational Graphs
- Tensor
PyTorch Modeling Introduction
- Section Overview
- Linear Regression from Scratch
- Model Class
- Learning Rate and Number of Epochs
- Batches
- Datasets and Dataloaders
- Saving and Loading Models
- Model Training
- Hyperparameter Tuning
Classification Models
- Section Overview
- Classification Types
- Confusion Matrix
- ROC curve
- Multi-Class 1: Data Prep
- Multi-Class 2: Dataset class
- Multi-Class 3: Network Class
- Multi-Class 5: Loss, Optimizer, and Hyper Parameters
- Multi-Class 6: Training Loop
- Multi-Class 7: Model Evaluation
- Multi-Class 8: Naive Classifier
- Multi-Class 9: Summary
- Multi-Label
CNN: Image Classification
- Section Overview
- CNNs
- Image Preprocessing
- Binary Image Classification
- MultiClass Image Classification
- Layer Calculations
CNN: Audio Classification
- Audio Classification
- Audio Classification (Exploratory Data Analysis)
- Audio Classification (Data Prep-Solution)
- Audio Classification (Model-Solution)
CNN: Object Detection
- Section Overview
- Accuracy Metrics
- Object Detection
- Object Detection with detecto
- Training a Model on GPU for free
- YOLO
- Labeling Formats
- YOLOv7 Project
- YOLOv7 Coding: Setup
- YOLOv7 Coding: Data Prep
- YOLOv7 Coding: Model Training
- YOLOv7 Coding: Model Inference
- YOLOv8 Coding: Model Training and Inference
Style Transfer
- Section Overview
- Style Transfer
Pretrained Networks and Transfer Learning
- Section Overview
- Transfer Learning and Pretrained Networks
- Transfer Learning
Recurrent Neural Networks
- Section Overview
- RNN
- LSTM
Autoencoders
- Section Overview
- utoencoders
Generative Adversarial Networks
- Section Overview
- GANs
Transformers
- Vision Transformers (ViT)
PyTorch Lightning
- PyTorch Lighting
- PyTorch Ligthning
- Early Stopping
Semi-Supervised Learning
- Supvervised Learning (Reference Model, Coding)
- Semi-Supervised Learning (1/2: Dataset and Dataloader)
- Semi-Supervised Learning (2/2 Modeling)
Natural Language Processing (NLP)
- Natural Language Processing
- Word Embeddings Intro
- Sentiment OHE Coding Intro
- Word Embeddings with NN
- GloVe: Get Word Embedding
- Glove: Find closest words
- GloVe: Word Analogy
- Glove Word Cluster
- GloVe Word
- Sentiment with Embedding
- Apply Pre-trained NLP Models
Programme scope, delivery arrangements and suitability are confirmed during enquiry.