Struct LlamaConfig 

pub struct LlamaConfig {
    pub hidden_size: usize,
    pub intermediate_size: usize,
    pub vocab_size: usize,
    pub num_hidden_layers: usize,
    pub num_attention_heads: usize,
    pub num_key_value_heads: Option<usize>,
    pub rms_norm_eps: f64,
    pub rope_theta: f64,
    pub max_position_embeddings: usize,
    pub tie_word_embeddings: bool,
    pub bos_token_id: Option<u32>,
    pub eos_token_id: Option<u32>,
}

Llama model configuration.

Fields

hidden_size: usize

Hidden size (embedding dimension).

intermediate_size: usize

Intermediate size for MLP.

vocab_size: usize

Vocabulary size.

num_hidden_layers: usize

Number of hidden layers.

num_attention_heads: usize

Number of attention heads.

num_key_value_heads: Option<usize>

Number of key-value heads (for GQA).

rms_norm_eps: f64

RMS norm epsilon.

rope_theta: f64

RoPE theta.

max_position_embeddings: usize

Maximum sequence length.

tie_word_embeddings: bool

Tie word embeddings.

bos_token_id: Option<u32>

BOS token ID.

eos_token_id: Option<u32>

EOS token ID.

Implementations

impl LlamaConfig

pub fn num_kv_heads(&self) -> usize

Returns the number of key-value heads.

pub fn head_dim(&self) -> usize

Returns the head dimension.

Trait Implementations

impl Clone for LlamaConfig

fn clone(&self) -> LlamaConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for LlamaConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for LlamaConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.