megdiffusion 0.0.2

MegEngine implementation of Diffusion Models

MegDiffusion

MegEngine implementation of Diffusion Models (in early development).

Current maintainer: @MegChai

Usage

Infer with pre-trained models

Now users can use megengine.hub to get pre-trained models directly:

import megengine

repo_info = "MegEngine/MegDiffusion:main"
megengine.hub.list(repo_info)

preatrained_model = "ddpm_cifar10_ema_converted"
megengine.hub.help(repo_info, preatrained_model)

model = megengine.hub.load(repo_info, preatrained_model, pretrained=True)
model.eval()

Note that using megengine.hub will download the whole repository from it's host or using cache.

If you have downloaded or installed MegDiffusion, you can get pre-trained models from pretrain module.

from megdiffusion.model import pretrain

model = pretrain.ddpm_cifar10_ema_converted(pretrained=True)
model.eval()

The sample script shows how to generate 64 CIFAR10-like images and make a grid of them:

python3 -m megdiffusion.pipeline.ddpm.sample

Train from scratch

• Take DDPM CIFAR10 for example:

  python3 -m megdiffusion.pipeline.ddpm.train \
      --config ./configs/ddpm/cifar10.yaml
  

• [Optional] Overwrite arguments:

  python3 -m megdiffusion.pipeline.ddpm.train \
     --config ./configs/ddpm/cifar10.yaml \
     --logdir ./path/to/logdir \
     --parallel --resume
  

See python3 -m megdiffusion.pipeline.ddpm.train --help for more information. For other options like batch_size, we recommend modifying and backing up them in the yaml file.

If you want to sample with model trained by yourself (not the pre-trained model):

python3 -m megdiffusion.pipeline.ddpm.sample --nopretrain \
   --logdir ./path/to/logdir \
   --config ./configs/ddpm/cifar10.yaml  # Coule be your customed file

Development

python3 -m pip install -r requirements.txt
python3 -m pip install -v -e .

Develop this project with a new branch locally, remember to add necessary test codes. If finished, submit Pull Request to the main branch then just wait for review.

Acknowledgment

The following open-sourced projects was referenced here:

• hojonathanho/diffusion: The official Tensorflow implementation of DDPM.
• w86763777/pytorch-ddpm: Unofficial PyTorch implementation of Denoising Diffusion Probabilistic Models.
• pesser/pytorch_diffusion: Unofficial PyTorch implementation of DDPM and provides converted torch checkpoints.
• openai/improved-diffusion: The official codebase for Improved Denoising Diffusion Probabilistic Models.

Thanks to people including @gaohuazuo, @xxr3376, @P2Oileen and other contributors for support in this project. The R&D platform and the resources required for the experiment are provided by MEGVII Inc. The deep learning framework used in this project is MegEngine -- a magic weapon.

Citations

@article{ho2020denoising,
    title   = {Denoising Diffusion Probabilistic Models},
    author  = {Jonathan Ho and Ajay Jain and Pieter Abbeel},
    year    = {2020},
    eprint  = {2006.11239},
    archivePrefix = {arXiv},
    primaryClass = {cs.LG}
}

@article{DBLP,
  title     = {Improved Denoising Diffusion Probabilistic Models},
  author    = {Alex Nichol and Prafulla Dhariwal},
  year      = {2021},
  url       = {https://arxiv.org/abs/2102.09672},
  eprinttype = {arXiv},
  eprint    = {2102.09672},
}