candle-transformers 0.10.2

Minimalist ML framework.
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

/*
 * @Author: SpenserCai
 * @Date: 2026-01-02 11:35:48
 * @version:
 * @LastEditors: SpenserCai
 * @LastEditTime: 2026-01-02 11:48:26
 * @Description: file content
 */
//! Z-Image Model
//!
//! Z-Image is a text-to-image generation model from Alibaba using Flow Matching.
//!
//! - 🤗 [Hugging Face Model](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo)
//! - [Official Website](https://z-image-turbo.org/)
//!
//! # Example
//!
//! ```bash
//! cargo run --features metal --example z_image --release -- \
//!     --prompt "A beautiful landscape" --height 1024 --width 1024
//! ```
//!
//! # Architecture
//!
//! - Transformer: ~24B parameters, 30 main layers + 2 noise_refiner + 2 context_refiner
//! - Text Encoder: Qwen3 (hidden_size=2560, 36 layers)
//! - VAE: AutoencoderKL (diffusers format)
//! - Scheduler: FlowMatchEulerDiscreteScheduler (shift=3.0)

pub mod preprocess;
pub mod sampling;
pub mod scheduler;
pub mod text_encoder;
pub mod transformer;
pub mod vae;

// Re-export main types
pub use preprocess::{prepare_inputs, PreparedInputs};
pub use sampling::{get_noise, get_schedule, postprocess_image};
pub use scheduler::{calculate_shift, FlowMatchEulerDiscreteScheduler, SchedulerConfig};
pub use text_encoder::{TextEncoderConfig, ZImageTextEncoder};
pub use transformer::{Config, ZImageTransformer2DModel};
pub use vae::{AutoEncoderKL, VaeConfig};