Struct AudioEncoderConfig 

pub struct AudioEncoderConfig {

    pub d_model: usize,
    pub num_mel_bins: usize,
    pub num_hidden_layers: usize,
    pub encoder_attention_heads: usize,
    pub encoder_ffn_dim: usize,
    pub downsample_hidden_size: usize,
    pub output_dim: usize,
    pub max_source_positions: usize,
    pub n_window: usize,
    pub n_window_infer: usize,
    pub conv_chunksize: usize,
    pub activation_function: String,
    pub scale_embedding: bool,
}

Qwen3-Omni audio encoder (“qwen3_asr_audio_encoder”) parameters.

Fields

d_model: usize

Transformer width (d_model).

num_mel_bins: usize

Input log-mel channels.

num_hidden_layers: usize

Number of transformer encoder layers (num_hidden_layers; the HF config also carries a redundant encoder_layers that serde ignores).

encoder_attention_heads: usize

Attention heads per encoder layer.

encoder_ffn_dim: usize

Feed-forward inner width.

downsample_hidden_size: usize

Conv2d channel count (downsample_hidden_size).

output_dim: usize

Adapter output width (== text hidden_size).

max_source_positions: usize

Length of the sinusoidal positional table.

n_window: usize

Half-window size in raw mel frames (chunk = 2 * n_window).

n_window_infer: usize

Inference attention-window size in raw mel frames.

conv_chunksize: usize

Convolution batch chunk size (irrelevant for correctness, kept for parity).

activation_function: String

scale_embedding: bool

Implementations

impl AudioEncoderConfig

pub fn head_dim(&self) -> usize

Per-head attention dimension (d_model / heads).

pub fn chunk_frames(&self) -> usize

Raw mel frames per pre-CNN chunk (2 * n_window).

Trait Implementations

impl Clone for AudioEncoderConfig

fn clone(&self) -> AudioEncoderConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AudioEncoderConfig

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for AudioEncoderConfig

fn deserialize<__D>( __deserializer: __D, ) -> Result<AudioEncoderConfig, <__D as Deserializer<'de>>::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.