Modular Co-Design (MoCo) Interpolants

MoCo enables abstracted interpolants for building and sampling from a variety of popular generative model frameworks. Specifically, MoCo supports interpolants for both continuous and discrete data types.

Continuous Data Interpolants

MoCo currently supports the following continuous data interpolants:

• DDPM (Denoising Diffusion Probabilistic Models)
• VDM (Variational Diffusion Models)
• CFM (Conditional Flow Matching)

Discrete Data Interpolants

MoCo also supports the following discrete data interpolants:

• D3PM (Discrete Denoising Diffusion Probabilistic Models)
• MDLM (Masked Diffusion Language Models)
• DFM (Discrete Flow Matching)

Useful Abstractions

MoCo also provides useful wrappers for customizable time distributions and inference time schedules.

Extendible

If the desired interpolant or sampling method is not already supported, MoCo was designed to be easily extended.

Installation

For Conda environment setup, please refer to the environment directory for specific instructions.

Once your environment is set up, you can install this project by running the following command:

pip install -e .

This will install the project in editable mode, allowing you to make changes and see them reflected immediately.

Examples

Please see examples of all interpolants in the examples directory.