Struct UniversalSentenceEncoder 

pub struct UniversalSentenceEncoder {
    pub config: UseConfig,
    /* private fields */
}

Universal Sentence Encoder — transformer-based architecture.

Fields

config: UseConfig

Model configuration.

Implementations

impl UniversalSentenceEncoder

pub fn new(config: UseConfig) -> Self

Construct a new USE with the given configuration.

pub fn encode(&self, token_ids: &[usize]) -> Result<Vec<f64>>

Encode token IDs to a fixed-size sentence embedding of length d_model.

pub fn encode_batch(&self, batch: &[Vec<usize>]) -> Result<Vec<Vec<f64>>>

Encode a batch of token ID sequences.

pub fn cosine_similarity(a: &[f64], b: &[f64]) -> f64

Cosine similarity between two sentence embeddings.

pub fn cross_lingual_encode( &self, token_ids: &[usize], lang_id: usize, xl_config: &CrossLingualConfig, ) -> Result<Vec<f64>>

Cross-lingual encoding — adds a language embedding before the stack.

token_ids: vocabulary indices lang_id: 0-indexed language identifier xl_config: cross-lingual configuration