Skip to content

Home

Jump to bottom
hlillemark edited this page Jan 12, 2026 · 5 revisions

Welcome to the Flow Equivariant World Models (FloWM) wiki!

This codebase uses Hydra and Pytorch Lightning to run experiments. Configurations are stored under the ./configurations/ directory, and they are configured hierarchically. Specific configuration options can be overrided from the command line.

Our codebase is structured around four primary components:

  • dataset (e.g., blockworld): This defines the dataset you wish to use. It relies on PyTorch's Dataset class and handles data loading and preprocessing.
  • algorithm (e.g., flowm_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.
  • shortcode: This defines experiment/dataset specific configuration options.

All experiments can be launched by:

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

Links to more information

Section Description
Configuration Details on configuring experiments.
Dataset Structure of datasets and guidelines for adding new ones.
Algorithms Structure of algorithms, models, and guidelines for adding new ones.
Training Commands for training all models (FloWM, Baselines).
Inference and Reproducing Results Commands for reproducing the experiment results of the paper.

This repo is forked from Boyuan Chen's research template repo. More specifically it is forked from the DFoT repo.

Clone this wiki locally