CleanDiffuser: An Easy-to-use Modularized Library for Diffusion Models in Decision Making

TL;DR

This work introduces CleanDiffuser, the first DM library specifically designed for decision-making algorithms, and identifies a set of essential sub-modules that constitute the core of CleanDiffuser, allowing for the implementation of various DM algorithms with simple and flexible building blocks.

Abstract

Leveraging the powerful generative capability of diffusion models (DMs) to build decision-making agents has achieved extensive success. However, there is still a demand for an easy-to-use and modularized open-source library that offers customized and efficient development for DM-based decision-making algorithms. In this work, we introduce CleanDiffuser, the first DM library specifically designed for decision-making algorithms. By revisiting the roles of DMs in the decision-making domain, we identify a set of essential sub-modules that constitute the core of CleanDiffuser, allowing for the implementation of various DM algorithms with simple and flexible building blocks. To demonstrate the reliability and flexibility of CleanDiffuser, we conduct comprehensive evaluations of various DM algorithms implemented with CleanDiffuser across an extensive range of tasks. The analytical experiments provide a wealth of valuable design choices and insights, reveal opportunities and challenges, and lay a solid groundwork for future research. CleanDiffuser will provide long-term support to the decision-making community, enhancing reproducibility and fostering the development of more robust solutions. The code and documentation of CleanDiffuser are open-sourced on the https://github.com/CleanDiffuserTeam/CleanDiffuser.

Figures (11)

• Figure 1: The Architecture of CleanDiffuser.CleanDiffuser is specifically tailored for the decision-making domain, supporting a wide range of Diffusion Models, Network Architectures, and Guided Sampling Methods modules and extra useful features. By simply combining the building blocks into a pipeline, CleanDiffuser integrates 9 popular DM algorithms.
• Figure 2: Diffusion Models Mainly Play Three Roles in Decision-Making Scenarios. Planner janner2022planning: Acting as planners to make better decisions from a long-term perspective. Policy pearce2023imitating: Serving as policies to support complex multimodal-distribution modeling. Data Synthesizer lu2024synthetic: Performing data augmentation to assist model training.
• Figure 3: Features of CleanDiffuser Designed for Decision-Making Introduced in \ref{['sec:features']}.
• Figure 4: Visualization of Implemented Network Architectures in CleanDiffuser.
• Figure 5: Diffuser Implementation with CleanDiffuser. The left part is a minimal code example showcasing simplicity and readability, and the right part provides a code explanation where the algorithm implementation can be entirely represented as a combination of building blocks, showing an example of various pipelines.