Welcome!
As part of Full Stack Deep Learning 2022, we will incrementally develop a complete deep learning codebase to create and deploy a model that understands the content of hand-written paragraphs.
For an overview of the Text Recognizer application architecture, click the badge below to open an interactive Jupyter notebook on Google Colab:
We will use the modern stack of PyTorch and PyTorch Lightning.
We will use the main workhorses of DL today: CNNs and Transformers.
We will manage our experiments using what we believe to be the best tool for the job: Weights & Biases.
We will set up a quality assurance and continuous integration system for our codebase using pre-commit and GitHub Actions.
We will package up the prediction system and deploy it as a Docker container on AWS Lambda.
We will wrap that prediction system in a frontend written in Python using Gradio.
We will set up monitoring that alerts us to potential issues in our model using Gantry.
| Lab | Colab | Video |
|---|---|---|
| Lab Overview | |
|
| Lab 01: Deep Neural Networks in PyTorch | |
|
| Lab 02a: PyTorch Lightning | |
|
| Lab 02b: Training a CNN on Synthetic Handwriting Data | |
|
| Lab 03: Transformers and Paragraphs | |
|
| Lab 04: Experiment Tracking | |