Struct ViTConfig 

pub struct ViTConfig {
    pub image_size: usize,
    pub patch_size: usize,
    pub in_channels: usize,
    pub embed_dim: usize,
    pub num_layers: usize,
    pub num_heads: usize,
    pub mlp_dim: usize,
    pub num_classes: usize,
    pub dropout: f32,
}

Vision Transformer configuration

Fields

image_size: usize

Image size (assumed square)

patch_size: usize

Patch size (assumed square)

in_channels: usize

Number of input channels

embed_dim: usize

Embedding dimension

num_layers: usize

Number of transformer layers

num_heads: usize

Number of attention heads

mlp_dim: usize

MLP hidden dimension

num_classes: usize

Number of output classes

dropout: f32

Dropout rate

Implementations

impl ViTConfig

pub fn vit_base() -> Self

ViT-Base configuration

pub fn vit_large() -> Self

ViT-Large configuration

pub fn vit_huge() -> Self

ViT-Huge configuration

pub fn num_patches(&self) -> usize

Get number of patches

Trait Implementations

impl Clone for ViTConfig

fn clone(&self) -> ViTConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for ViTConfig

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

Formats the value using the given formatter.

impl Default for ViTConfig

fn default() -> Self

Returns the “default value” for a type.