Struct MultiModalContrastiveLearning 

pub struct MultiModalContrastiveLearning { /* private fields */ }

Contrastive learning for multi-modal graphs

Implementations

impl MultiModalContrastiveLearning

pub fn new( modalities: Vec<Modality>, modality_dims: HashMap<Modality, usize>, projection_dim: usize, temperature: f32, ) -> Result<Self, TorshError>

Create new contrastive learning module

pub fn contrastive_loss( &self, modality1: Modality, features1: &Tensor, modality2: Modality, features2: &Tensor, ) -> Result<f32, TorshError>

Compute contrastive loss between modalities

pub fn generate_contrastive_pairs( &self, mm_graph: &MultiModalGraphData, modality1: Modality, modality2: Modality, ) -> Vec<(Tensor, Tensor, bool)>

Generate positive and negative pairs for contrastive learning

Trait Implementations

impl Debug for MultiModalContrastiveLearning

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

Formats the value using the given formatter.