nnet3-discriminative-subset-egs.cc File Reference

#include "base/kaldi-common.h"
#include "util/common-utils.h"
#include "nnet3/nnet-discriminative-example.h"

Include dependency graph for nnet3-discriminative-subset-egs.cc:

Go to the source code of this file.

Functions
int	main (int argc, char *argv[])

Function Documentation

main()

int main	(	int	argc,
		char *	argv[]
	)

Definition at line 25 of file nnet3-discriminative-subset-egs.cc.

References SequentialTableReader< Holder >::Done(), ParseOptions::GetArg(), rnnlm::i, KALDI_LOG, SequentialTableReader< Holder >::Key(), rnnlm::n, SequentialTableReader< Holder >::Next(), ParseOptions::NumArgs(), ParseOptions::PrintUsage(), kaldi::RandInt(), ParseOptions::Read(), ParseOptions::Register(), SequentialTableReader< Holder >::Value(), and kaldi::WithProb().

                                 {
  try {
    using namespace kaldi;
    using namespace kaldi::nnet3;
    typedef kaldi::int32 int32;
    typedef kaldi::int64 int64;

    const char *usage =
        "Creates a random subset of the input examples, of a specified size.\n"
        "Uses no more memory than the size of the subset.\n"
        "\n"
        "Usage:  nnet3-discriminative-subset-egs [options] <degs-rspecifier> [<degs-wspecifier2> ...]\n"
        "\n"
        "e.g.\n"
        "nnet3-discriminative-copy-egs [args] ark:degs.1.ark ark:- | nnet-discriminative-subset-egs --n=1000 ark:- ark:subset.egs\n";
    
    int32 srand_seed = 0;
    int32 n = 1000;
    bool randomize_order = true;
    ParseOptions po(usage);
    po.Register("srand", &srand_seed, "Seed for random number generator ");
    po.Register("n", &n, "Number of examples to output");
    po.Register("randomize-order", &randomize_order, "If true, randomize the order "
                "of the output");
    
    po.Read(argc, argv);
    
    srand(srand_seed);
    
    if (po.NumArgs() != 2) {
      po.PrintUsage();
      exit(1);
    }

    std::string examples_rspecifier = po.GetArg(1),
        examples_wspecifier = po.GetArg(2);

    std::vector<std::pair<std::string, NnetDiscriminativeExample> > egs;
    egs.reserve(n);
    
    SequentialNnetDiscriminativeExampleReader example_reader(examples_rspecifier);

    int64 num_read = 0;
    for (; !example_reader.Done(); example_reader.Next()) {
      num_read++;
      if (num_read <= n) {
        egs.resize(egs.size() + 1);
        egs.back().first = example_reader.Key();
        egs.back().second = example_reader.Value();
      } else {
        BaseFloat keep_prob = n / static_cast<BaseFloat>(num_read);
        if (WithProb(keep_prob)) { // With probability "keep_prob"
          int32 index = RandInt(0, n-1);
          egs[index].first = example_reader.Key();
          egs[index].second = example_reader.Value();
        }
      }
    }
    if (randomize_order)
      std::random_shuffle(egs.begin(), egs.end());

    NnetDiscriminativeExampleWriter writer(examples_wspecifier);
    for (size_t i = 0; i < egs.size(); i++) {
      writer.Write(egs[i].first, egs[i].second);
    }
    
    KALDI_LOG << "Selected a subset of " << egs.size() << " out of " << num_read
              << " neural-network discriminative training examples ";
    
    return (num_read != 0 ? 0 : 1);
  } catch(const std::exception &e) {
    std::cerr << e.what() << '\n';
    return -1;
  }
}