KimiVL<T>
Models & Types · AiDotNet.VisionLanguage.Reasoning
Kimi-VL: MoE VLM with MoonViT and long-context processing.
Inheritance: NeuralNetworkBase<T> → VisionLanguageModelBase<T> → KimiVL<T>
Implements: IReasoningVLM<T>
For Beginners
Kimi-VL is a vision-language model with MoE architecture for efficient multimodal reasoning. Default values follow the original paper settings.
How It Works
Kimi-VL (Moonshot AI, 2025) is a Mixture of Experts vision-language model featuring the MoonViT visual encoder and long-context processing capabilities. It uses sparse MoE routing for efficient scaling and supports extended context windows for processing high-resolution images and lengthy multi-turn conversations with visual inputs.
References:
- Paper: “Kimi-VL Technical Report” (Moonshot AI, 2025)
Example
// Create a Kimi-VL model for multimodal reasoning
// with MoE architecture, MoonViT encoder, and long-context processing
var architecture = new NeuralNetworkArchitecture<double>(
inputType: InputType.TwoDimensional,
taskType: NeuralNetworkTaskType.ImageClassification,
inputHeight: 224, inputWidth: 224, inputDepth: 3, outputSize: 512);
// ONNX inference mode with pre-trained model
var model = new KimiVL<double>(architecture, "kimivl.onnx");
// Training mode with native layers
var trainModel = new KimiVL<double>(architecture, new KimiVLOptions());
Constructors
| Constructor | Summary |
|---|---|
KimiVL(NeuralNetworkArchitecture<T> architecture, KimiVLOptions options, IGradientBasedOptimizer<T, Tensor<T>, Tensor<T>> optimizer) |
|
KimiVL(NeuralNetworkArchitecture<T> architecture, string modelPath, KimiVLOptions options) |
Properties
| Property | Summary |
|---|---|
int DecoderEmbeddingDim { get; } |
|
int EmbeddingDimension { get; } |
|
int MaxGenerationLength { get; } |
|
string LanguageModelName { get; } |
|
string ReasoningApproach { get; } |
Methods
| Method | Summary |
|---|---|
ModelMetadata<T> GetModelMetadata() |
|
Tensor<T> Chat(Tensor<T> image, IEnumerable<ValueTuple<string, string>> conversationHistory, string userMessage) |
|
Tensor<T> EncodeImage(Tensor<T> image) |
|
Tensor<T> GenerateFromImage(Tensor<T> image, string prompt) |
Generates text from image using Kimi-VL’s MoE architecture with MoonViT encoder. |
Tensor<T> ReasonWithChainOfThought(Tensor<T> image, string question) |
Generates reasoning using Kimi-VL’s MoE-based chain-of-thought with long-context. |