Struct Lfm3AudioEncoder 

pub struct Lfm3AudioEncoder {
    pub cfg: Lfm3AudioConfig,
    pub text_hidden: usize,
    /* private fields */
}

The speech encoder and its projector.

Fields

cfg: Lfm3AudioConfig

text_hidden: usize

Implementations

impl Lfm3AudioEncoder

pub fn from_model( model: &Arc<CmfModel>, pool: Option<Arc<Pool>>, ) -> Result<Self, String>

pub fn encode(&self, mel: &[f32], frames: usize) -> Result<Vec<f32>, String>

Encoder rows [⌈frames/8⌉, hidden] of log-mel features [n_mels][frames], before the projector.

pub fn project(&self, enc: &[f32]) -> Vec<f32>

Projected rows [tokens, text_hidden] of encoder rows.

pub fn rows(&self, wave: &[f32]) -> Result<Vec<f32>, String>

One clip, 16 kHz mono, to the language model’s rows.