Struct ModelConfig 

pub struct ModelConfig {

    pub dim: usize,
    pub n_layers: usize,
    pub head_dim: usize,
    pub input_dim: usize,
    pub encoder_output_dim: usize,
    pub encoder_latent_downsample_factor: usize,
    pub t_dim: usize,
    pub max_seqlen: usize,
    pub rope_dim: usize,
    pub rope_theta: f64,
    pub norm_eps: f64,
    pub ffn_dim_multiplier: Option<f64>,
    pub multiple_of: usize,
    pub stft_global_sigma: f64,
}

Model and runtime configuration for ZUNA inference.

ModelConfig is deserialised from the HuggingFace config.json (the "model" sub-object). Field names must match exactly.

Fields

dim: usize

n_layers: usize

head_dim: usize

input_dim: usize

encoder_output_dim: usize

encoder_latent_downsample_factor: usize

t_dim: usize

max_seqlen: usize

rope_dim: usize

rope_theta: f64

norm_eps: f64

ffn_dim_multiplier: Option<f64>

multiple_of: usize

stft_global_sigma: f64

Implementations

impl ModelConfig

pub fn n_heads_fallback(&self) -> usize

n_heads is NOT dim/head_dim for this checkpoint. It must be inferred from the wq weight shape at load time. Use WeightMap::infer_n_heads() instead of calling this.

pub fn ffn_hidden_dim(&self) -> usize

Feed-forward hidden dim (matches Python FeedForward.init): hidden = int(2 * 4 * dim / 3) → 2730 hidden = 256 * ceil(2730 / 256) → 2816

Trait Implementations

impl Clone for ModelConfig

fn clone(&self) -> ModelConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ModelConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for ModelConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.