KolorsModel<T>

Models & Types · AiDotNet.Diffusion.TextToImage

Kolors model — ChatGLM3-powered bilingual text-to-image model by Kwai.

Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → KolorsModel<T>

For Beginners

Kolors is like SDXL but with Chinese language understanding:

Key characteristics:

  • ChatGLM3-6B text encoder: strong Chinese + English understanding
  • SDXL-like U-Net backbone: proven architecture for quality
  • 4096-dim cross-attention from ChatGLM3 embeddings
  • Native 1024x1024 resolution
  • Open-source with Apache 2.0 license

How Kolors works:

  1. Text goes through ChatGLM3-6B (6B parameter language model)
  2. 4096-dim embeddings provide rich text understanding
  3. SDXL-like U-Net denoises with cross-attention to embeddings
  4. VAE decodes to final image

Use Kolors when you need:

  • Chinese text-to-image generation
  • Open-source bilingual model
  • SDXL-quality with multilingual support

How It Works

Kolors is Kwai’s text-to-image model that uses ChatGLM3-6B as the text encoder, providing strong bilingual (Chinese-English) text understanding. It builds on the SDXL U-Net architecture with ChatGLM3’s 4096-dim embeddings for cross-attention.

Technical specifications:

  • Architecture: SDXL U-Net with ChatGLM3-6B text encoder
  • Text encoder: ChatGLM3-6B (4096-dim, 65024 vocab)
  • U-Net: ~2.6B parameters
  • Resolution: 1024x1024
  • Latent space: 4 channels, 8x downsampling
  • Guidance scale: 5.0-7.5 recommended

Reference: Kwai, “Kolors: Effective Training of Diffusion Model for Photorealistic Text-to-Image Synthesis”, 2024

Example

var options = new DiffusionModelOptions<float> { LatentChannels = 4, DefaultInferenceSteps = 25 };
var noise = Tensor<float>.CreateRandom(new[] { 1, 4, 128, 128 });
var trainX = Tensor<float>.CreateRandom(4, 8);
var trainY = Tensor<float>.CreateRandom(4, 2);
var result = new AiModelBuilder<float, Tensor<float>, Tensor<float>>()
    .ConfigureModel(new KolorsModel<float>(options: options))
    .Build(trainX, trainY);
var generated = result.Predict(noise);

Constructors

Constructor Summary
KolorsModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, UNetNoisePredictor<T> unet, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed) Initializes a new instance of KolorsModel with full customization support.

Properties

Property Summary
IConditioningModule<T> Conditioner { get; }
INoisePredictor<T> NoisePredictor { get; }
IVAEModel<T> VAE { get; }
int LatentChannels { get; }

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()
Tensor<T> GenerateFromText(string prompt, string negativePrompt, int width, int height, int numInferenceSteps, double? guidanceScale, int? seed)

Fields

Field Summary
int DefaultHeight
int DefaultWidth