InstructorEmbedding<T>

Models & Types · AiDotNet.NeuralNetworks

Instructor/E5 (Instruction-Tuned) embedding model implementation. Uses task-specific instructions to adapt embeddings for different use cases.

Inheritance: NeuralNetworkBase<T> → DeclaredModelLayoutBase<T> → TextEmbeddingModelLayoutBase<T> → TransformerEmbeddingNetwork<T> → InstructorEmbedding<T>

For Beginners

Most AI models read every sentence the same way. “Instructor” models are like specialized scholars. If you tell them “read this like a doctor looking for a diagnosis,” they will focus on medical terms. If you tell them “read this like a poet,” they will focus on the mood. It makes the “coordinates” (embeddings) much more useful for your specific goal.

How It Works

Instructor models are transformer-based encoders trained with instructions. By prepending a task description (e.g., “Represent the Wikipedia sentence for retrieval:”), the model learns to produce embeddings that are optimized for that specific task.

Example

var architecture = new NeuralNetworkArchitecture<float>(inputFeatures: 8, outputSize: 4);
var input = Tensor<float>.CreateRandom(new[] { 1, 512 });
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 InstructorEmbedding<float>(architecture))
    .Build(trainX, trainY);
var embedding = result.Predict(input);

Constructors

Constructor Summary
InstructorEmbedding() Initializes a new instance with default architecture settings.
InstructorEmbedding(NeuralNetworkArchitecture<T> architecture, InstructorEmbeddingOptions options, ITokenizer tokenizer, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer, ILossFunction<T> lossFunction) Initializes a new instance of the InstructorEmbedding model.

Methods

Method Summary
ModelMetadata<T> GetModelMetadata() Retrieves metadata about the Instructor model, including its default instruction.
Task<Matrix<T>> EmbedBatchAsync(IEnumerable<string> texts)
Task<Vector<T>> EmbedAsync(string text)
Vector<T> Embed(string text)
Vector<T> EmbedWithInstruction(string text, string instruction) Encodes text into a normalized embedding vector using a task-specific instruction.
void SetDefaultInstruction(string instruction) Sets the default instruction used for general embedding generation.