Overview¶
MMGeneration is a powerful toolkit for generative models, including GANs and diffusion models. It is based on PyTorch and MMCV. The 1.x branch works with PyTorch 1.5+.
supported tasks¶
Now, MMGeneration support 4 tasks of image generation. Lists are as follows.
Unconditional GANs
Conditional GANs
Image2Image Translation
Internal Learning
Diffusion Models
highlight¶
High-quality Training Performance: MMGeneration currently support training on Unconditional GANs, Conditional GANs, Internal GANs, Image Translation Models, and diffusion models.
Powerful Application Toolkit: A toolkit that provides plentiful applications to users. MMGeneration supports GAN interpolation, GAN projection, GAN manipulations and many other popular GAN’s applications. It’s time to play with your GANs! (Tutorial for applications)
Efficient Distributed Training for Generative Models: With support of MMSeparateDistributedDataParallel, distributed training for dynamic architectures can be easily implemented.
New Modular Design for Flexible Combination: A new design for complex loss modules is proposed for customizing the links between modules, which can achieve flexible combination among different modules.(Tutorial for losses)
get started¶
To get started with our repo, please refer to get_started.md.
user guides¶
For elementary guides on basic usage, please refer to user_guides.
advanced guides¶
To learn design and structure of MMGeneration, as well as how to extend the repo, how to use multiple repos and other advanced usages, please refer to advanced_guides.