Diffusion Model

A basic diffusion model based on the seminal paper by Jonathan Ho et al.

Generated Images

These are a few examples of what this model generated after a lot of debugging.

Trained Model

Find a few trained models and other things on HuggingFace.

Dataset

Smiling or Not Face Data from Kaggle.

The smiling or not smiling is not relevant to me, I just want to generate a face. I will combine all the categories of faces into one single dataset.

Usage

Installation

pipenv shell
pipenv install

Download Dataset

git clone https://huggingface.co/datasets/zrthxn/SmilingOrNot data/faces
unzip data/faces/Archive.zip -d data/faces

Dryrun Sanity Check

python main.py train --dryrun

Full Training

python main.py train \
    --batch_size=128 \
    --device=cuda \
    --lr=0.001 \
    --epochs=100

Resources

Papers

• Denoising Diffusion Probabilistic Models
• Deep Unsupervised Learning using Nonequilibrium Thermodynamics
• Improved Denoising Diffusion Probabilistic Models

Videos and Code

• https://www.youtube.com/watch?v=HoKDTa5jHvg
• https://www.youtube.com/watch?v=a4Yfz2FxXiY
• https://amaarora.github.io/2020/09/13/unet.html
• https://lilianweng.github.io/posts/2021-07-11-diffusion-models/
• https://colab.research.google.com/drive/1sjy9odlSSy0RBVgMTgP7s99NXsqglsUL