Struct VisionTransformer 

pub struct VisionTransformer { /* private fields */ }

Vision Transformer (ViT) model

Implementations

impl VisionTransformer

pub fn new(config: VisionTransformerConfig) -> Result<Self>

Create new Vision Transformer

pub fn build_vit_graph(&self, graph: &mut EinsumGraph) -> Result<Vec<usize>>

Build complete Vision Transformer einsum graph

Graph Inputs

• Tensor 0: Input patches [batch, num_patches, patch_dim]
• Tensor 1: Patch embedding weights [patch_dim, d_model]
• Additional encoder tensors (weights for each layer)

Graph Outputs

• Classification logits [batch, num_classes]

Note: This is a simplified representation. In practice, you would need to provide all encoder layer weights and handle class token prepending/extraction properly.

pub fn config(&self) -> &VisionTransformerConfig

Get configuration

pub fn count_parameters(&self) -> usize

Count total parameters