Module deberta

DeBERTa :Decoding-enhanced BERT with Disentangled Attention (He et al.)

Implementation of the DeBERTa language model (DeBERTa :Decoding-enhanced BERT with Disentangled Attention He, Liu ,Gao, Chen, 2021). The base model is implemented in the deberta_model::DebertaModel struct. Several language model heads have also been implemented, including:

• Question answering: deberta_model::DebertaForQuestionAnswering
• Sequence classification: deberta_model::DebertaForSequenceClassification
• Token classification (e.g. NER, POS tagging): deberta_model::DebertaForTokenClassification.

Model set-up and pre-trained weights loading

All models expect the following resources:

• Configuration file expected to have a structure following the Transformers library
• Model weights are expected to have a structure and parameter names following the Transformers library. A conversion using the Python utility scripts is required to convert the .bin weights to the .ot format.
• DebertaTokenizer using a vocab.json vocabulary and merges.txt merges file

Pretrained models for a number of language pairs are available and can be downloaded using RemoteResources.

use tch::{nn, Device};
use rust_bert::deberta::{
    DebertaConfig, DebertaConfigResources, DebertaForSequenceClassification,
    DebertaMergesResources, DebertaModelResources, DebertaVocabResources,
};
use rust_bert::resources::{RemoteResource, ResourceProvider};
use rust_bert::Config;
use rust_tokenizers::tokenizer::DeBERTaTokenizer;

let config_resource =
    RemoteResource::from_pretrained(DebertaConfigResources::DEBERTA_BASE_MNLI);
let vocab_resource = RemoteResource::from_pretrained(DebertaVocabResources::DEBERTA_BASE_MNLI);
let merges_resource =
    RemoteResource::from_pretrained(DebertaMergesResources::DEBERTA_BASE_MNLI);
let weights_resource =
    RemoteResource::from_pretrained(DebertaModelResources::DEBERTA_BASE_MNLI);
let config_path = config_resource.get_local_path()?;
let vocab_path = vocab_resource.get_local_path()?;
let merges_path = merges_resource.get_local_path()?;
let weights_path = weights_resource.get_local_path()?;
let device = Device::cuda_if_available();
let mut vs = nn::VarStore::new(device);
let tokenizer = DeBERTaTokenizer::from_file(
    vocab_path.to_str().unwrap(),
    merges_path.to_str().unwrap(),
    true,
)?;
let config = DebertaConfig::from_file(config_path);
let deberta_model = DebertaForSequenceClassification::new(&vs.root(), &config);
vs.load(weights_path)?;

Structs

DebertaConfig

DeBERTa model configuration

DebertaConfigResources

DeBERTa Pretrained model config files

DebertaForMaskedLM

DeBERTa for masked language model

DebertaForQuestionAnswering

DeBERTa for question answering

DebertaForSequenceClassification

DeBERTa for sequence classification

DebertaForTokenClassification

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

DebertaMaskedLMOutput

Container for the DeBERTa masked LM model output.

DebertaMergesResources

DeBERTa Pretrained model merges files

DebertaModel

DeBERTa Base model

DebertaModelResources

DeBERTa Pretrained model weight files

DebertaVocabResources

DeBERTa Pretrained model vocab files

Type Aliases

DebertaQuestionAnsweringOutput

Container for the DeBERTa question answering model output.

DebertaSequenceClassificationOutput

Container for the DeBERTa sequence classification model output.

DebertaTokenClassificationOutput

Container for the DeBERTa token classification model output.