Crate mold_inference 

Re-exports

pub use device::reclaim_gpu_memory;

pub use engine::InferenceEngine;

pub use engine::LoadStrategy;

pub use flux::FluxEngine;

pub use flux2::Flux2Engine;

pub use ltx2::Ltx2Engine;

pub use ltx_video::LtxVideoEngine;

pub use model_registry::known_models;

pub use progress::ProgressEvent;

pub use qwen_image::QwenImageEngine;

pub use sd15::SD15Engine;

pub use sd3::SD3Engine;

pub use sdxl::SDXLEngine;

pub use upscaler::create_upscale_engine;

pub use upscaler::UpscaleEngine;

pub use wuerstchen::WuerstchenEngine;

pub use zimage::ZImageEngine;

Modules

attention

Pluggable attention backend for FLUX-family transformers.

controlnet

device

engine

error

flux

flux2

img_utils

Image decoding and preprocessing utilities for img2img.

loader

Loaders for non-diffusers checkpoint formats.

ltx2

ltx_video

model_registry

progress

qwen_image

scheduler

sd3

sd15

sdxl

shared_pool

upscaler

vae_tiling

Tiled VAE decode with OOM fallback.

wuerstchen

zimage

Structs

InferenceError

Re-export anyhow for callers that want a uniform error type. The Error type, a wrapper around a dynamic error type.

Functions

create_engine

Create an inference engine for the given model, auto-detecting the family.

create_engine_with_pool

Create an inference engine with an optional shared tokenizer pool.