Struct VisionConfig 

pub struct VisionConfig {
    pub image_size: usize,
    pub patch_size: usize,
    pub hidden_size: usize,
    pub intermediate_size: usize,
    pub num_hidden_layers: usize,
    pub num_attention_heads: usize,
    pub layer_norm_eps: f64,
    pub scale_factor: usize,
    pub image_token_id: u32,
}

Vision-encoder hyperparameters parsed from config.json::vision_config (plus top-level scale_factor / image_token_id).

Fields

image_size: usize

Square input image size (pixels).

patch_size: usize

Patch size (pixels) of the patch embedding.

hidden_size: usize

Vision hidden size.

intermediate_size: usize

Vision MLP intermediate size.

num_hidden_layers: usize

Vision transformer layers.

num_attention_heads: usize

Vision attention heads (MHA — kv heads == q heads).

layer_norm_eps: f64

LayerNorm epsilon (SigLIP: 1e-6).

scale_factor: usize

Pixel-shuffle scale factor of the connector (scale² patches are merged into one visual token).

image_token_id: u32

Token id whose span in the prompt is replaced by visual embeddings.

Implementations

impl VisionConfig

pub fn image_seq_len(&self) -> usize

Visual tokens per image: (image_size / patch_size)² / scale_factor².

pub fn head_dim(&self) -> usize

Vision per-head dimension.

Trait Implementations

impl Clone for VisionConfig

fn clone(&self) -> VisionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for VisionConfig

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

Formats the value using the given formatter.