[][src]Struct rust_bert::distilbert::DistilBertForQuestionAnswering

pub struct DistilBertForQuestionAnswering { /* fields omitted */ }

DistilBERT for question answering

Extractive question-answering model based on a DistilBERT language model. Identifies the segment of a context that answers a provided question. Please note that a significant amount of pre- and post-processing is required to perform end-to-end question answering. See the question answering pipeline (also provided in this crate) for more details. It is made of the following blocks:

  • distil_bert_model: Base DistilBertModel
  • qa_outputs: Linear layer for question answering

Methods

impl DistilBertForQuestionAnswering[src]

pub fn new(
    p: &Path,
    config: &DistilBertConfig
) -> DistilBertForQuestionAnswering
[src]

Build a new DistilBertForQuestionAnswering for sequence classification

Arguments

  • p - Variable store path for the root of the DistilBertForQuestionAnswering 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, DistilBertForQuestionAnswering};

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 = DistilBertForQuestionAnswering::new(&(&p.root() / "distilbert"), &config);

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

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

  • start_scores - Tensor of shape (batch size, sequence_length) containing the logits for start of the answer
  • end_scores - Tensor of shape (batch size, sequence_length) containing the logits for end of the answer
  • 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, DistilBertForQuestionAnswering};
 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 (start_scores, end_score, all_hidden_states, all_attentions) = no_grad(|| {
   distilbert_model
        .forward_t(Some(input_tensor),
                   Some(mask),
                   None,
                   false).unwrap()
   });

Auto Trait Implementations

Blanket Implementations

impl<T> Any for T where
    T: 'static + ?Sized
[src]

impl<T> Borrow<T> for T where
    T: ?Sized
[src]

impl<T> BorrowMut<T> for T where
    T: ?Sized
[src]

impl<T> From<T> for T[src]

impl<T, U> Into<U> for T where
    U: From<T>, 
[src]

impl<T, U> TryFrom<U> for T where
    U: Into<T>, 
[src]

type Error = Infallible

The type returned in the event of a conversion error.

impl<T, U> TryInto<U> for T where
    U: TryFrom<T>, 
[src]

type Error = <U as TryFrom<T>>::Error

The type returned in the event of a conversion error.