Struct StaticModel 

pub struct StaticModel { /* private fields */ }

Static embedding model for Model2Vec

Implementations

impl StaticModel

pub fn from_bytes<T, M, C>( tokenizer_bytes: T, model_bytes: M, config_bytes: C, normalize: Option<bool>, ) -> Result<Self>

where T: AsRef<[u8]>, M: AsRef<[u8]>, C: AsRef<[u8]>,

Load a Model2Vec model directly from in-memory bytes.

This path is useful for runtimes that fetch model assets as bytes rather than reading them from a local filesystem.

pub fn from_pretrained<P: AsRef<Path>>( repo_or_path: P, token: Option<&str>, normalize: Option<bool>, subfolder: Option<&str>, ) -> Result<Self>

Load a Model2Vec model from a local folder or the HuggingFace Hub.

Arguments

• repo_or_path - HuggingFace repo ID or local path to the model folder.
• token - Optional HuggingFace token for authenticated downloads.
• normalize - Optional flag to normalize embeddings (default from the resolved config file).
• subfolder - Optional subfolder within the repo or path to look for model files.

pub fn from_owned( tokenizer: Tokenizer, embeddings: Vec<f32>, rows: usize, cols: usize, normalize: bool, weights: Option<Vec<f32>>, token_mapping: Option<Vec<usize>>, ) -> Result<Self>

Construct from owned data.

Arguments

• tokenizer - Pre-deserialized tokenizer
• embeddings - Owned f32 embedding data
• rows - Number of vocabulary entries
• cols - Embedding dimension
• normalize - Whether to L2-normalize output embeddings
• weights - Optional per-token weights for quantized models
• token_mapping - Optional token ID mapping for quantized models

pub fn from_borrowed( tokenizer: Tokenizer, embeddings: &'static [f32], rows: usize, cols: usize, normalize: bool, weights: Option<&'static [f32]>, token_mapping: Option<&'static [usize]>, ) -> Result<Self>

Construct from static slices (zero-copy for embedded binary data).

Arguments

• tokenizer - Pre-deserialized tokenizer
• embeddings - Static f32 embedding data (borrowed, no copy)
• rows - Number of vocabulary entries
• cols - Embedding dimension
• normalize - Whether to L2-normalize output embeddings
• weights - Optional static per-token weights for quantized models
• token_mapping - Optional static token ID mapping for quantized models

pub fn encode_with_args( &self, sentences: &[String], max_length: Option<usize>, batch_size: usize, ) -> Vec<Vec<f32>>

Encode texts into embeddings.

Arguments

• sentences - the list of sentences to encode.
• max_length - max tokens per text.
• batch_size - number of texts per batch.

pub fn encode(&self, sentences: &[String]) -> Vec<Vec<f32>>

Default encode: max_length=512, batch_size=1024

pub fn encode_single(&self, sentence: &str) -> Vec<f32>

Encode a single sentence into a vector.

Trait Implementations

impl Clone for StaticModel

fn clone(&self) -> StaticModel

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for StaticModel

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.