Diffusers documentation
Philosophy
Get started
Using Diffusers
Loading & Hub
Loading Pipelines, Models, and SchedulersUsing different SchedulersConfiguring Pipelines, Models, and SchedulersLoading and Adding Custom Pipelines
Pipelines for Inference
Unconditional Image GenerationText-to-Image GenerationText-Guided Image-to-ImageText-Guided Image-InpaintingText-Guided Depth-to-ImageReusing seeds for deterministic generationReproducibilityCommunity PipelinesHow to contribute a Pipeline
Taking Diffusers Beyond Images
Optimization/Special Hardware
Training
OverviewUnconditional Image GenerationTextual InversionDreamboothText-to-image fine-tuningLoRA Support in Diffusers
Conceptual Guides
API
Main Classes
Pipelines
OverviewAltDiffusionAudio DiffusionCycle DiffusionDance DiffusionDDIMDDPMDiTLatent DiffusionPaintByExamplePNDMRePaintSafe Stable DiffusionScore SDE VE
Stable Diffusion
OverviewText-to-ImageImage-to-ImageInpaintDepth-to-ImageImage-VariationSuper-ResolutionInstructPix2Pix
Stable Diffusion 2Stochastic Karras VEUnCLIPUnconditional Latent DiffusionVersatile DiffusionVQ DiffusionSchedulers
OverviewDDIMDDPMDEISDPM Discrete SchedulerDPM Discrete Scheduler with ancestral samplingEuler Ancestral SchedulerEuler schedulerHeun SchedulerIPNDMLinear MultistepMultistep DPM-SolverPNDMRePaint SchedulerSinglestep DPM-SolverStochastic Kerras VEVE-SDEVP-SDEVQDiffusionScheduler
Experimental Features
You are viewing v0.12.0 version. A newer version v0.38.0 is available.
Philosophy
- Readability and clarity are preferred over highly optimized code. A strong importance is put on providing readable, intuitive and elementary code design. E.g., the provided schedulers are separated from the provided models and use well-commented code that can be read alongside the original paper.
- Diffusers is modality independent and focuses on providing pretrained models and tools to build systems that generate continuous outputs, e.g. vision and audio. This is one of the guiding goals even if the initial pipelines are devoted to vision tasks.
- Diffusion models and schedulers are provided as concise, elementary building blocks. In contrast, diffusion pipelines are a collection of end-to-end diffusion systems that can be used out-of-the-box, should stay as close as possible to their original implementations and can include components of other libraries, such as text encoders. Examples of diffusion pipelines are Glide, Latent Diffusion and Stable Diffusion.