deep_learning_from_scratch

DeepLearning/AI From Stracth

Introduction

This repo is for practicing deep learning and generative AI. It basically covers the popular deep learning models in CV, NLP, genAI etc. The goal of this repo is to provide a comprehensive and practical understanding of deep learning and generative AI. The repo is structured in such a way that it covers the following topics:

• Prediction models: used for regression, classification etc.
  • googlenet
  • resnet
  • densenet
  • recommendation models
    • dlrm
    • deepfm
    • mmoe
    • bst
• Generative models: used to generate new data
  • Autogressive models
    • pixelCNN
  • VAE
  • energy based model
  • normalizing flow
  • diffusion model
• Discovery models: used to learn the hidden state of the data for downstream task
  • AE
  • self-supervised learning:
    • contrastive learning: SimCLR
  • ViT

The above list did not include all revolutionary works, like GAN, diffusion model are not included. I am on my way of hands on practicing and would keep adding them later.

All models are coded using pytorch and lighting, we also tried tuning hyperparameters using ray tune. Users who have interest may refer to the README.md in each model folder for more details.

How to use

The repo is structured in such a way that each model is in a separate folder. Users can directly go to the folder of interest and run the code. The code is well documented and easy to understand. Users can also refer to the README.md in each model folder for more details.

Setup

1. clone the repo into your local machine

cd path/to/your/workspace
git clone https://github.com/Bingzw/DL_genAI_practicing.git

2. create python virtual environment

python3 -m venv venv
source venv/bin/activate

3. install the required packages

pip install -r requirements.txt

4. go to each model's folder and run the code

cd path/to/model/folder
python train.py