Stable Diffusion

This project uses diffusion models to generate flowers from initial noise. This project is created using PyTorch only. Everything is created from scratch including the U-Net architecture. The data used for training is the Flowers 102 dataset.

Stable Diffusion Real-Time Game Engine

This project uses diffusion models to generate the next frame in real-time. The Model is trained on the video game Super Mario Kart. It takes in the last four frames and actions as input and generates the next frame. The training data was collected by an AI agent learning to play the game.

Proximal Policy Optimization for Continuous Control

This project uses the Proximal Policy Optimization algorithm to train an agent on a continuous control problem. The environment used is the Humanoid environment from the OpenAI Gym. The agent has the objective to walk as far as possible in the environment.

Smooth Particle Hydrodynamics Simulation

This project uses a Smooth Particle Hydrodynamics simulation to simulate fluid dynamics. The project is created using raylib and C++. The project is still in development and CUDA acceleration is planned.