Score-Based Generative Modeling through Stochastic Differential Equations (Score SDE) by Yang Song, Jascha Sohl-Dickstein, Diederik P. Kingma, Abhishek Kumar, Stefano Ermon and Ben Poole.
The abstract of the paper is the following:
Creating noise from data is easy; creating data from noise is generative modeling. We present a stochastic differential equation (SDE) that smoothly transforms a complex data distribution to a known prior distribution by slowly injecting noise, and a corresponding reverse-time SDE that transforms the prior distribution back into the data distribution by slowly removing the noise. Crucially, the reverse-time SDE depends only on the time-dependent gradient field (\aka, score) of the perturbed data distribution. By leveraging advances in score-based generative modeling, we can accurately estimate these scores with neural networks, and use numerical SDE solvers to generate samples. We show that this framework encapsulates previous approaches in score-based generative modeling and diffusion probabilistic modeling, allowing for new sampling procedures and new modeling capabilities. In particular, we introduce a predictor-corrector framework to correct errors in the evolution of the discretized reverse-time SDE. We also derive an equivalent neural ODE that samples from the same distribution as the SDE, but additionally enables exact likelihood computation, and improved sampling efficiency. In addition, we provide a new way to solve inverse problems with score-based models, as demonstrated with experiments on class-conditional generation, image inpainting, and colorization. Combined with multiple architectural improvements, we achieve record-breaking performance for unconditional image generation on CIFAR-10 with an Inception score of 9.89 and FID of 2.20, a competitive likelihood of 2.99 bits/dim, and demonstrate high fidelity generation of 1024 x 1024 images for the first time from a score-based generative model.
The original codebase can be found here.
This pipeline implements the Variance Expanding (VE) variant of the method.
|pipeline_score_sde_ve.py||Unconditional Image Generation||-|
class diffusers.ScoreSdeVePipeline< source >
( unet: UNet2DModel scheduler: DiffusionPipeline )
- This model inherits from DiffusionPipeline. Check the superclass documentation for the generic methods the —
library implements for all the pipelines (such as downloading or saving, running on a particular device, etc.) —
unet (UNet2DModel): U-Net architecture to denoise the encoded image. scheduler (SchedulerMixin):
The ScoreSdeVeScheduler scheduler to be used in combination with
unetto denoise the encoded image.
__call__< source >
batch_size: int = 1
num_inference_steps: int = 2000
generator: typing.Optional[torch._C.Generator] = None
output_type: typing.Optional[str] = 'pil'
return_dict: bool = True
int, optional, defaults to 1) — The number of images to generate.
torch.Generator, optional) — A torch generator to make generation deterministic.
str, optional, defaults to
"pil") — The output format of the generate image. Choose between PIL:
bool, optional, defaults to
True) — Whether or not to return a ImagePipelineOutput instead of a plain tuple.
return_dict is True, otherwise a `tuple. When returning a tuple, the first element is a list with the