Struct rust_bert::models::gpt_neo::GptNeoForCausalLM

pub struct GptNeoForCausalLM { /* private fields */ }

GPT-Neo Model for causal language modeling

Gpt-Neo model with a vocabulary decoding head. The language model decoding head is tied to the word embedding matrix weights It is made of the following blocks:

• transformer: GptNeoModel Base ProphetNet model

Implementations

impl GptNeoForCausalLM

pub fn new<'p, P>( p: P, config: &GptNeoConfig ) -> Result<GptNeoForCausalLM, RustBertError>

where P: Borrow<Path<'p>>,

Build a new GptNeoForCausalLM

Arguments

• p - Variable store path for the root of the GPT-Neo model
• config - GptNeoConfig object defining the model architecture

Example

use rust_bert::gpt_neo::{GptNeoConfig, GptNeoForCausalLM};
use rust_bert::Config;
use std::path::Path;
use tch::{nn, Device};

let config_path = Path::new("path/to/config.json");
let device = Device::Cpu;
let p = nn::VarStore::new(device);
let config = GptNeoConfig::from_file(config_path);
let gpt_neo_model = GptNeoForCausalLM::new(&p.root(), &config).unwrap();

pub fn forward_t( &self, input_ids: Option<&Tensor>, input_embeds: Option<&Tensor>, token_type_ids: Option<&Tensor>, position_ids: Option<&Tensor>, layer_states: Option<Vec<Option<LayerState>>>, attention_mask: Option<&Tensor>, train: bool ) -> Result<GptNeoModelLMOutput, RustBertError>

Forward pass through the model

Arguments

• input_ids - Optional input tensor of shape (batch size, sequence_length). This or input_embeds must be provided.
• input_embeds - Optional input tensor of shape (batch size, sequence_length, embeddings dimension). This or input_ids must be provided.
• token_type_ids - Optional token type ids used to indicate the portion of the input the token belongs to. If not None, token type embeddings will be added to the token and position embeddings.
• position_ids - Optional position ids of shape (batch size, sequence_length). If None, will be incremented starting from the length of the past input.
• layer_states - Optional Vector Option<Vec<Option<&LayerState>>> of length n_layer containing tuples with the past keys and values for both the self attention of each layer.
• attention_mask - Optional attention mask of shape (batch size, sequence_length) for the encoder positions. Positions with a mask with value 0 will be masked.
• train - boolean flag to turn on/off the dropout layers in the model. Should be set to false for inference.

Returns

• Result<GptNeoModelLMOutput, RustBertError> containing:
  • lm_logits - Tensor of shape (batch size, sequence_length, vocab_size) representing the logits for each vocab item and position
  • next_cache - Option<Vec<Option<LayerState>>> of length n_layer containing the past content for the the attention layers
  • all_hidden_states - Option<Vec<Tensor>> of length n_layer + 1 with shape (batch size, sequence_length, hidden_size)
  • all_attentions - Option<Vec<Tensor>> of length n_layer containing the attention weights for each layer

Example

use rust_bert::gpt_neo::{GptNeoConfig, GptNeoForCausalLM};
let (batch_size, sequence_length) = (64, 128);
let input_tensor = Tensor::rand(&[batch_size, sequence_length], (Int64, device));
let attention_mask = Tensor::ones(&[batch_size, sequence_length], (Int64, device));

let model_output = no_grad(|| {
    gpt_neo_model.forward_t(
        Some(&input_tensor),
        Some(&attention_mask),
        None,
        None,
        None,
        None,
        false,
    )
});