ankane / youtokentome-ruby

High performance unsupervised text tokenization for Ruby

YouTokenToMe Ruby

YouTokenToMe - high performance unsupervised text tokenization - for Ruby

Learn more about how it works

Installation

Add this line to your application’s Gemfile:

gem "youtokentome"

Getting Started

Dump your text to a file

Blazingly fast tokenization!

Train a model

model = YouTokenToMe::BPE.train(data: "train.txt", model: "model.txt", vocab_size: 30000)

Load a model

model = YouTokenToMe::BPE.new("model.txt")

Get vocab

model.vocab

Encode

model.encode(sentences)

Decode

model.decode(ids)

Convert between ids and subwords

model.subword_to_id(subword)
model.id_to_subword(id)

Options

Train

YouTokenToMe::BPE.train(
  data: "train.txt",   # path to file with training data
  model: "model.txt",  # path to where the trained model will be saved
  vocab_size: 30000,   # number of tokens in the final vocabulary
  coverage: 1.0,       # fraction of characters covered by the model
  n_threads: -1,       # number of parallel threads used to run
  pad_id: 0,           # reserved id for padding
  unk_id: 1,           # reserved id for unknown symbols
  bos_id: 2,           # reserved id for begin of sentence token
  eos_id: 3            # reserved id for end of sentence token
)

Encode

model.encode(
  sentences,
  output_type: :id,    # or :subword
  bos: false,          # add "beginning of sentence" token
  eos: false,          # add "end of sentence" token
  reverse: false,      # reverse output sequence of tokens
  dropout_prob: 0.0    # BPE-dropout probability
)

History

View the changelog

Contributing

Everyone is encouraged to help improve this project. Here are a few ways you can help:

• Report bugs
• Fix bugs and submit pull requests
• Write, clarify, or fix documentation
• Suggest or add new features

To get started with development:

git clone https://github.com/ankane/youtokentome-ruby.git
cd youtokentome-ruby
bundle install
bundle exec rake compile
bundle exec rake test