Struct T5ForSentenceEmbeddings

pub struct T5ForSentenceEmbeddings { /* private fields */ }

T5 for sentence embeddings

Transformer usable in SentenceEmbeddingsModel.

Implementations

impl T5ForSentenceEmbeddings

pub fn new<'p, P>(p: P, config: &T5Config) -> Self

where P: Borrow<Path<'p>>,

Build a new T5ForSentenceEmbeddings

Arguments

• p - Variable store path for the root of the BART model
• config - T5Config object defining the model architecture

It consists of only an encoder (there is no decoder).

pub fn forward( &self, input_ids: &Tensor, mask: &Tensor, ) -> Result<(Tensor, Option<Vec<Tensor>>), RustBertError>

Forward pass through the model

Arguments

• input_ids - Input of shape (batch size, source_sequence_length).
• mask - Attention mask of shape (batch size, source_sequence_length) for the encoder positions. Positions with a mask with value 0 will be masked.

Returns

• Tuple containing:
  • Tensor of shape (batch size, target_sequence_length, hidden_size) representing the activations of the last encoder hidden state
  • Option<Vec<Tensor>> of length num_encoder_layers of shape (batch size, target_sequence_length, hidden_size) representing attention weights for all layers of the encoder