Struct rust_bert::electra::ElectraForMaskedLM

pub struct ElectraForMaskedLM { /* fields omitted */ }

Electra for Masked Language Modeling

Masked Language modeling Electra model It is made of the following blocks:

• electra: ElectraModel (based on a BertEncoder and custom embeddings)
• generator_head: ElectraGeneratorHead to generate token predictions of dimension embedding_size
• lm_head: linear layer of dimension (embeddings_size, vocab_size) to project the output to the vocab size

Methods

impl ElectraForMaskedLM

Defines the implementation of the ElectraForMaskedLM.

pub fn new(p: &Path, config: &ElectraConfig) -> ElectraForMaskedLM

Build a new ElectraForMaskedLM

Arguments

• p - Variable store path for the root of the Electra model
• config - ElectraConfig object defining the model architecture

Example

use rust_bert::electra::{ElectraForMaskedLM, ElectraConfig};
use tch::{nn, Device};
use rust_bert::Config;
use std::path::Path;

let config_path = Path::new("path/to/config.json");
let device = Device::Cpu;
let p = nn::VarStore::new(device);
let config = ElectraConfig::from_file(config_path);
let electra_model: ElectraForMaskedLM = ElectraForMaskedLM::new(&p.root(), &config);

pub fn forward_t(
    &self,
    input_ids: Option<Tensor>,
    mask: Option<Tensor>,
    token_type_ids: Option<Tensor>,
    position_ids: Option<Tensor>,
    input_embeds: Option<Tensor>,
    train: bool
) -> (Tensor, Option<Vec<Tensor>>, Option<Vec<Tensor>>)

Forward pass through the model

Arguments

• input_ids - Optional input tensor of shape (batch size, sequence_length). If None, pre-computed embeddings must be provided (see input_embeds)
• mask - Optional mask of shape (batch size, sequence_length). Masked position have value 0, non-masked value 1. If None set to 1
• token_type_ids - Optional segment id of shape (batch size, sequence_length). Convention is value of 0 for the first sentence (incl. [SEP]) and 1 for the second sentence. If None set to 0.
• position_ids - Optional position ids of shape (batch size, sequence_length). If None, will be incremented from 0.
• input_embeds - Optional pre-computed input embeddings of shape (batch size, sequence_length, hidden_size). If None, input ids must be provided (see input_ids)
• train - boolean flag to turn on/off the dropout layers in the model. Should be set to false for inference.

Returns

• output - Tensor of shape (batch size, sequence_length, vocab_size)
• hidden_states - Option<Vec<Tensor>> of length num_hidden_layers with shape (batch size, sequence_length, hidden_size)
• attentions - Option<Vec<Tensor>> of length num_hidden_layers with shape (batch size, sequence_length, hidden_size)

Example

 let (batch_size, sequence_length) = (64, 128);
 let input_tensor = Tensor::rand(&[batch_size, sequence_length], (Int64, device));
 let mask = Tensor::zeros(&[batch_size, sequence_length], (Int64, device));
 let token_type_ids = Tensor::zeros(&[batch_size, sequence_length], (Int64, device));
 let position_ids = Tensor::arange(sequence_length, (Int64, device)).expand(&[batch_size, sequence_length], true);

 let (output, all_hidden_states, all_attentions) = no_grad(|| {
   electra_model
        .forward_t(Some(input_tensor),
                   Some(mask),
                   Some(token_type_ids),
                   Some(position_ids),
                   None,
                   false)
   });