Struct AlbertForTokenClassification

pub struct AlbertForTokenClassification { /* private fields */ }

ALBERT for token classification (e.g. NER, POS)

Token-level classifier predicting a label for each token provided. Note that because of SentencePiece tokenization, the labels predicted are not necessarily aligned with words in the sentence. It is made of the following blocks:

• albert: Base AlbertModel
• dropout: Dropout to apply on the encoder last hidden states
• classifier: Linear layer for token classification

Implementations

impl AlbertForTokenClassification

pub fn new<'p, P>( p: P, config: &AlbertConfig, ) -> Result<AlbertForTokenClassification, RustBertError>

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

Build a new AlbertForTokenClassification

Arguments

• p - Variable store path for the root of the ALBERT model
• config - AlbertConfig object defining the model architecture and decoder status

Example

use rust_bert::albert::{AlbertConfig, AlbertForTokenClassification};
use rust_bert::Config;
use std::path::Path;
use tch::{nn, Device};

let config_path = Path::new("path/to/config.json");
let device = Device::Cpu;
let p = nn::VarStore::new(device);
let config = AlbertConfig::from_file(config_path);
let albert: AlbertForTokenClassification =
    AlbertForTokenClassification::new(&p.root(), &config).unwrap();

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, ) -> AlbertTokenClassificationOutput

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

• AlbertTokenClassificationOutput containing:
  • logits - Tensor of shape (batch size, sequence_length, num_labels) containing the logits for each of the input tokens and classes
  • all_hidden_states - Option<Vec<Tensor>> of length num_hidden_layers with shape (batch size, sequence_length, hidden_size)
  • all_attentions - Option<Vec<Vec<Tensor>>> of length num_hidden_layers of nested length inner_group_num with shape (batch size, sequence_length, hidden_size)

Example

use rust_bert::albert::{AlbertConfig, AlbertForTokenClassification};
 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 model_output = no_grad(|| {
   albert_model
        .forward_t(Some(&input_tensor),
                   Some(&mask),
                   Some(&token_type_ids),
                   Some(&position_ids),
                   None,
                   false)
   });