Module rust_bert::gpt2

GPT2 (Radford et al.)

Implementation of the GPT2 language model (Language Models are Unsupervised Multitask Learners Radford, Wu, Child, Luan, Amodei, Sutskever 2019). The base model is implemented in the gpt2::Gpt2Model struct. The model also includes a language model head: gpt2::GPT2LMHeadModel implementing the common generation::LMHeadModel trait shared between the models used for generation (see pipelines for more information).

Model set-up and pre-trained weights loading

A full working example is provided in examples/summarization.rs, run with cargo run --example gpt2. 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.
• Gpt2Tokenizer using a vocab.txt vocabulary and merges.txt 2-gram merges

use rust_tokenizers::Gpt2Tokenizer;
use tch::{nn, Device};
use rust_bert::Config;
use rust_bert::gpt2::{Gpt2Config, GPT2LMHeadModel};

let device = Device::cuda_if_available();
let mut vs = nn::VarStore::new(device);
let tokenizer: Gpt2Tokenizer = Gpt2Tokenizer::from_file(vocab_path.to_str().unwrap(), merges_path.to_str().unwrap(), true);
let config = Gpt2Config::from_file(config_path);
let gpt2_model = GPT2LMHeadModel::new(&vs.root(), &config);
vs.load(weights_path)?;

Structs

GPT2LMHeadModel	GPT2 Language Modeling head
Gpt2Config	GPT2 model configuration
Gpt2Model	GPT2 Base model

Enums

GptActivation

Activation function used in the fully connected layers of the transformer block