Overview

Getting Started

Our codebase is primarily built upon the Hydra and PyTorch Lightning libraries. We use Hydra to configure experiments hierarchically, allowing you to set up static configurations in YAML files within the configurations folder. Moreover, you can override specific parts of these configurations at runtime with command-line arguments. PyTorch Lightning is employed to efficiently manage training and evaluation loops, as well as logging.

Our codebase is structured around three primary components:

• dataset (e.g., realestate10k): This defines the dataset you wish to use. It relies on PyTorch's Dataset class and handles data loading and preprocessing.
• algorithm (e.g., dfot_video): This specifies the algorithm, inheriting from PyTorch Lightning's LightningModule class. It involves initializing a model and defining the training and evaluation processes.
• experiment (e.g., video_generation): This defines the specific task you aim to perform. It leverages PyTorch Lightning's Trainer class and integrates the dataset and algorithm to conduct experiments.

All experiments can be launched by:

python -m main +name=xxx dataset=xxx algorithm=xxx experiment=xxx ...

Template Repository

This repo is forked from Boyuan Chen's research template repo. By its MIT license, you must keep the above sentence in README.md and the LICENSE file to credit the author.