EasyConsistencyModel<T>
Models & Types · AiDotNet.Diffusion.FastGeneration
Easy Consistency Tuning (ECT) model for simple, stable consistency model training.
Inheritance: DiffusionModelBase<T> → LatentDiffusionModelBase<T> → EasyConsistencyModel<T>
For Beginners
Training consistency models from scratch is tricky — small mistakes in hyperparameters cause training to diverge. ECT makes it easy by starting with small noise differences (easy mappings) and gradually increasing difficulty. Think of it like learning to long-jump: you start with short distances and work up.
How It Works
ECT simplifies consistency model training by using a progressive training curriculum that gradually increases the difficulty of the consistency mapping. This removes the need for careful hyperparameter tuning and produces stable training from any pretrained diffusion model.
Reference: Geng et al., “Consistency Models Made Easy”, NeurIPS 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 EasyConsistencyModel<float>(options: options))
.Build(trainX, trainY);
var generated = result.Predict(noise);
Constructors
| Constructor | Summary |
|---|---|
EasyConsistencyModel(NeuralNetworkArchitecture<T> architecture, DiffusionModelOptions<T> options, INoiseScheduler<T> scheduler, UNetNoisePredictor<T> predictor, StandardVAE<T> vae, IConditioningModule<T> conditioner, int? seed) |
Initializes a new Easy Consistency Tuning 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() |