SCottModel<T>

Models & Types · AiDotNet.Diffusion.FastGeneration

SCott (Score Consistency via Optimal Transport) for efficient single/few-step generation.

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

For Beginners

SCott finds the most efficient path from noise to images using optimal transport theory (the math of moving mass efficiently). Like finding the shortest route for delivery trucks, SCott finds the shortest path from random noise to your desired image, enabling generation in just 1-4 steps.

How It Works

SCott combines score-based consistency training with optimal transport theory to learn efficient single-step mappings. Uses a score-matching objective regularized by an optimal transport penalty to produce straight transport paths that are well-suited for few-step generation.

Reference: “SCott: Accelerating Diffusion Models with Stochastic Consistency Distillation”, 2024

Example

var options = new DiffusionModelOptions<float> { LatentChannels = 4, DefaultInferenceSteps = 2 };
var noise = Tensor<float>.CreateRandom(new[] { 1, 4, 64, 64 });
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 SCottModel<float>(options: options))
    .Build(trainX, trainY);
var generated = result.Predict(noise);

Constructors

Constructor Summary
SCottModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, UNetNoisePredictor<T> predictor, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed) Initializes a new SCott model.

Properties

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

Methods

Method Summary
ModelMetadata<T> GetModelMetadata()