Struct Trainer 

pub struct Trainer {
    pub history: Vec<EpochStats>,
    /* private fields */
}

Manages the full training pipeline.

Fields

history: Vec<EpochStats>

Loss history per epoch (populated during train).

Implementations

impl Trainer

pub fn new(config: Config) -> Self

Create a new trainer with the given configuration.

use word2vec::{Config, Trainer};
let trainer = Trainer::new(Config::default());
assert!(trainer.history.is_empty());

pub fn train(&mut self, corpus: &[String]) -> Result<Embeddings>

Train on a corpus of sentences.

Returns Embeddings which wraps the trained model and vocabulary for inference queries.

Errors

Returns Word2VecError if the config is invalid or the corpus is too small to train.

Example

use word2vec::{Config, ModelType, Trainer};

let corpus = vec![
    "paris is the capital of france".to_string(),
    "berlin is the capital of germany".to_string(),
    "tokyo is the capital of japan".to_string(),
];

let mut trainer = Trainer::new(Config {
    epochs: 3,
    embedding_dim: 50,
    ..Config::default()
});

let embeddings = trainer.train(&corpus).unwrap();
println!("Vocab size: {}", embeddings.vocab_size());

pub fn save_history(&self, path: impl AsRef<Path>) -> Result<()>

Save training history as JSON.

use word2vec::{Config, Trainer};
let mut trainer = Trainer::new(Config::default());
// trainer.train(...);
// trainer.save_history("history.json").unwrap();