Struct rust_bert::distilbert::DistilBertModelMaskedLM

pub struct DistilBertModelMaskedLM { /* fields omitted */ }

DistilBERT for masked language model

Base DistilBERT model with a masked language model head to predict missing tokens, for example "Looks like one [MASK] is missing" -> "person" It is made of the following blocks:

• distil_bert_model: Base DistilBertModel
• vocab_transform:linear layer for classification of size (hidden_dim, hidden_dim)
• vocab_layer_norm: layer normalization
• vocab_projector: linear layer for classification of size (hidden_dim, vocab_size) with weights tied to the token embeddings

Methods

impl DistilBertModelMaskedLM

pub fn new(p: &Path, config: &DistilBertConfig) -> DistilBertModelMaskedLM

Build a new DistilBertModelMaskedLM for sequence classification

Arguments

• p - Variable store path for the root of the DistilBertModelMaskedLM model
• config - DistilBertConfig object defining the model architecture

Example

use tch::{nn, Device};
use rust_bert::Config;
use std::path::Path;
use rust_bert::distilbert::{DistilBertConfig, DistilBertModelMaskedLM};

let config_path = Path::new("path/to/config.json");
let device = Device::Cpu;
let p = nn::VarStore::new(device);
let config = DistilBertConfig::from_file(config_path);
let distil_bert = DistilBertModelMaskedLM::new(&(&p.root() / "distilbert"), &config);

pub fn forward_t(
    &self,
    input: Option<Tensor>,
    mask: Option<Tensor>,
    input_embeds: Option<Tensor>,
    train: bool
) -> Result<(Tensor, Option<Vec<Tensor>>, Option<Vec<Tensor>>), &'static str>

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
• 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) representing the logits for position and vocabulary index
• 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

use rust_bert::distilbert::{DistilBertConfig, DistilBertModelMaskedLM};
 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 (output, all_hidden_states, all_attentions) = no_grad(|| {
   distilbert_model
        .forward_t(Some(input_tensor),
                   Some(mask),
                   None,
                   false).unwrap()
   });