Expand description
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
.binweights to the.otformat. DebertaTokenizerusing avocab.jsonvocabulary andmerges.txtmerges 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
- DeBERTa model configuration
- DeBERTa Pretrained model config files
- DeBERTa for masked language model
- DeBERTa for question answering
- DeBERTa for sequence classification
- DeBERTa for token classification (e.g. NER, POS)
- Container for the DeBERTa masked LM model output.
- DeBERTa Pretrained model merges files
- DeBERTa Base model
- DeBERTa Pretrained model weight files
- DeBERTa Pretrained model vocab files
Type Aliases
- Container for the DeBERTa question answering model output.
- Container for the DeBERTa sequence classification model output.
- Container for the DeBERTa token classification model output.