Struct EmbeddingModelWeights 

pub struct EmbeddingModelWeights { /* private fields */ }

Embedding model weights. Same as quantized_qwen3::ModelWeights but without lm_head. Returns last-token hidden states for embedding.

Implementations

impl EmbeddingModelWeights

pub fn from_gguf<R: Read + Seek>( ct: Content, reader: &mut R, device: &CandleDevice, ) -> Result<Self>

Load from a GGUF file. Same format as generative Qwen3 models.

pub fn forward(&mut self, input: &Tensor, offset: usize) -> Result<Tensor>

Forward pass returning last-token hidden state (1024-dim for 0.6B).

Uses causal attention mask (same as generative model) since Qwen3-Embedding is architecturally Qwen3ForCausalLM fine-tuned for embedding. Returns shape: (batch, hidden_size). Forward pass returning last-token hidden state. Only supports batch_size=1 (causal mask not broadcast for larger batches).

pub fn clear_kv_cache(&mut self)

pub fn hidden_size(&self) -> usize

Trait Implementations

impl Clone for EmbeddingModelWeights

fn clone(&self) -> EmbeddingModelWeights

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for EmbeddingModelWeights

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

Formats the value using the given formatter.