Struct LlamaForward 

pub struct LlamaForward { /* private fields */ }

Real Llama-architecture forward engine backed by safetensors weights.

Implementations

impl LlamaForward

pub fn from_files(info: ModelInfo, weight_paths: &[PathBuf]) -> Result<Self>

pub fn from_files_with_backend( info: ModelInfo, weight_paths: &[PathBuf], backend: LlmBackendKind, ) -> Result<Self>

pub fn from_weights( info: ModelInfo, weights: HashMap<String, Tensor>, ) -> Result<Self>

pub fn from_weights_with_backend( info: ModelInfo, weights: HashMap<String, Tensor>, backend: LlmBackendKind, ) -> Result<Self>

pub fn reset_cache(&mut self)

pub fn forward_logits( &mut self, input_ids: &[u32], use_cache: bool, ) -> Result<Vec<f32>>

Run forward on token ids and return logits for the last token.

pub fn embed(&mut self, input_ids: &[u32]) -> Result<Vec<f32>>

Mean-pooled hidden states for embedding models.