model

package

v0.3.3 Latest Latest Go to latest Published: Apr 10, 2026 License: MIT Imports: 0 Imported by: 0

Details

View Source

const CodeRankEmbedFullDim = 768

View Source

const CodeRankEmbedModelName = "CodeRankEmbed"

CodeRankEmbed model metadata (primary/default model)

View Source

const CodeRankEmbedQuantization = "INT8-dynamic"

View Source

const DefaultEmbeddingDim = CodeRankEmbedFullDim

DefaultEmbeddingDim is the recommended ONNX embedding dimension (CodeRankEmbed full).

View Source

const DefaultMatryoshkaDim = 256

DefaultMatryoshkaDim is the recommended Qwen2 Matryoshka dimension.

View Source

const ModelFullDim = 896 // full hidden_size of the Qwen2 model

View Source

const ModelName = "qwen2-code-embedding"

Qwen2/Jina model metadata (legacy/alternative)

View Source

const ModelQuantization = "INT8" // quantization format (QOperator)

View Source

var MatryoshkaDims = []int{64, 128, 256, 512, 896}

MatryoshkaDims lists the supported Matryoshka embedding dimensions for Qwen2.

This section is empty.

This section is empty.