[−][src]Module rust_bert::distilbert
DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter (Sanh et al.)
Implementation of the DistilBERT language model (https://arxiv.org/abs/1910.01108 Sanh, Debut, Chaumond, Wolf, 2019).
The base model is implemented in the distilbert::DistilBertModel
struct. Several language model heads have also been implemented, including:
- Masked language model:
distilbert::DistilBertForMaskedLM
- Question answering:
distilbert::DistilBertForQuestionAnswering
- Sequence classification:
distilbert::DistilBertForSequenceClassification
- Token classification (e.g. NER, POS tagging):
distilbert::DistilBertForTokenClassification
Model set-up and pre-trained weights loading
A full working example is provided in examples/distilbert_masked_lm.rs
, run with cargo run --example distilbert_masked_lm
.
The example below illustrate a DistilBERT Masked language model example, the structure is similar for other models.
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. BertTokenizer
using avocab.txt
vocabulary Pretrained models are available and can be downloaded using RemoteResources.
use rust_tokenizers::BertTokenizer; use tch::{nn, Device}; use rust_bert::distilbert::{ DistilBertConfig, DistilBertConfigResources, DistilBertModelMaskedLM, DistilBertModelResources, DistilBertVocabResources, }; use rust_bert::resources::{download_resource, LocalResource, RemoteResource, Resource}; use rust_bert::Config; let config_resource = Resource::Local(LocalResource { local_path: PathBuf::from("path/to/config.json"), }); let vocab_resource = Resource::Local(LocalResource { local_path: PathBuf::from("path/to/vocab.txt"), }); let weights_resource = Resource::Local(LocalResource { local_path: PathBuf::from("path/to/model.ot"), }); let config_path = download_resource(&config_resource)?; let vocab_path = download_resource(&vocab_resource)?; let weights_path = download_resource(&weights_resource)?; let device = Device::cuda_if_available(); let mut vs = nn::VarStore::new(device); let tokenizer: BertTokenizer = BertTokenizer::from_file(vocab_path.to_str().unwrap(), true); let config = DistilBertConfig::from_file(config_path); let bert_model = DistilBertModelMaskedLM::new(&vs.root(), &config); vs.load(weights_path)?;
Structs
DistilBertConfig | DistilBERT model configuration |
DistilBertConfigResources | DistilBERT Pretrained model config files |
DistilBertForQuestionAnswering | DistilBERT for question answering |
DistilBertForTokenClassification | DistilBERT for token classification (e.g. NER, POS) |
DistilBertModel | DistilBERT Base model |
DistilBertModelClassifier | DistilBERT for sequence classification |
DistilBertModelMaskedLM | DistilBERT for masked language model |
DistilBertModelResources | DistilBERT Pretrained model weight files |
DistilBertVocabResources | DistilBERT Pretrained model vocab files |
Enums
Activation | Activation function used in the feed-forward layer in the transformer blocks |