mrm1001

void Model::computeOutputSoftmax(Vector& hidden, Vector& output) const {
  if (quant_ && args_->qout) {
    output.mul(*qwo_, hidden);
  } else {
    output.mul(*wo_, hidden);
  }
  real max = output[0], z = 0.0;
  for (int32_t i = 0; i < osz_; i++) {
    max = std::max(output[i], max);
  }

mrm1001

void Model::computeHidden(const std::vector<int32_t>& input, Vector& hidden) const {
  assert(hidden.size() == hsz_);
  hidden.zero();
  for (auto it = input.cbegin(); it != input.cend(); ++it) {
    if(quant_) {
      hidden.addRow(*qwi_, *it);
    } else {
      hidden.addRow(*wi_, *it);
    }
  }

mrm1001

void FastText::supervised(
    Model& model,
    real lr,
    const std::vector<int32_t>& line,
    const std::vector<int32_t>& labels) {
  if (labels.size() == 0 || line.size() == 0) return;
  std::uniform_int_distribution<> uniform(0, labels.size() - 1);
  int32_t i = uniform(model.rng);
  model.update(line, labels[i], lr);
}

mrm1001

void Dictionary::addWordNgrams(std::vector<int32_t>& line,
                               const std::vector<int32_t>& hashes,
                               int32_t n) const {
  for (int32_t i = 0; i < hashes.size(); i++) {
    uint64_t h = hashes[i];
    for (int32_t j = i + 1; j < hashes.size() && j < i + n; j++) {
      h = h * 116049371 + hashes[j];
      pushHash(line, h % args_->bucket);
    }
  }

mrm1001

int32_t Dictionary::getLine(std::istream& in,
                            std::vector<int32_t>& words,
                            std::vector<int32_t>& labels) const {
  std::vector<int32_t> word_hashes;
  std::string token;
  int32_t ntokens = 0;

  reset(in);
  words.clear();
  labels.clear();

mrm1001

void FastText::trainThread(int32_t threadId) {
  std::ifstream ifs(args_->input);
  utils::seek(ifs, threadId * utils::size(ifs) / args_->thread);

  Model model(input_, output_, args_, threadId);
  if (args_->model == model_name::sup) {
    model.setTargetCounts(dict_->getCounts(entry_type::label));
  } else {
    model.setTargetCounts(dict_->getCounts(entry_type::word));
  }

mrm1001

  if (args_->pretrainedVectors.size() != 0) {
    loadVectors(args_->pretrainedVectors);
  } else {
    input_ = std::make_shared<Matrix>(dict_->nwords()+args_->bucket, args_->dim);
    input_->uniform(1.0 / args_->dim);
  }

  if (args_->model == model_name::sup) {
    output_ = std::make_shared<Matrix>(dict_->nlabels(), args_->dim);
  } else {

mrm1001

void Dictionary::initNgrams() {
  for (size_t i = 0; i < size_; i++) {
    std::string word = BOW + words_[i].word + EOW;
    words_[i].subwords.clear();
    words_[i].subwords.push_back(i);
    if (words_[i].word != EOS) {
      computeSubwords(word, words_[i].subwords);
    }
  }
}

mrm1001

struct entry {
  std::string word;
  int64_t count;
  entry_type type;
  std::vector<int32_t> subwords;
};

mrm1001

void Dictionary::readFromFile(std::istream& in) {
  std::string word;
  int64_t minThreshold = 1;
  while (readWord(in, word)) {
    add(word);
    if (ntokens_ % 1000000 == 0 && args_->verbose > 1) {
      std::cerr << "\rRead " << ntokens_  / 1000000 << "M words" << std::flush;
    }
    if (size_ > 0.75 * MAX_VOCAB_SIZE) {
      minThreshold++;