nnet3-train.cc File Reference

#include "base/kaldi-common.h"
#include "util/common-utils.h"
#include "nnet3/nnet-training.h"
#include "cudamatrix/cu-allocator.h"

Include dependency graph for nnet3-train.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-train.cc.

References SequentialTableReader< Holder >::Done(), ParseOptions::GetArg(), KALDI_LOG, SequentialTableReader< Holder >::Next(), ParseOptions::NumArgs(), NnetTrainer::PrintTotalStats(), ParseOptions::PrintUsage(), ParseOptions::Read(), kaldi::ReadKaldiObject(), NnetTrainerOptions::Register(), ParseOptions::Register(), kaldi::RegisterCuAllocatorOptions(), NnetTrainer::Train(), SequentialTableReader< Holder >::Value(), and kaldi::WriteKaldiObject().

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

    const char *usage =
        "Train nnet3 neural network parameters with backprop and stochastic\n"
        "gradient descent.  Minibatches are to be created by nnet3-merge-egs in\n"
        "the input pipeline.  This training program is single-threaded (best to\n"
        "use it with a GPU); see nnet3-train-parallel for multi-threaded training\n"
        "that is better suited to CPUs.\n"
        "\n"
        "Usage:  nnet3-train [options] <raw-model-in> <training-examples-in> <raw-model-out>\n"
        "\n"
        "e.g.:\n"
        "nnet3-train 1.raw 'ark:nnet3-merge-egs 1.egs ark:-|' 2.raw\n";

    int32 srand_seed = 0;
    bool binary_write = true;
    std::string use_gpu = "yes";
    NnetTrainerOptions train_config;

    ParseOptions po(usage);
    po.Register("srand", &srand_seed, "Seed for random number generator ");
    po.Register("binary", &binary_write, "Write output in binary mode");
    po.Register("use-gpu", &use_gpu,
                "yes|no|optional|wait, only has effect if compiled with CUDA");

    train_config.Register(&po);
    RegisterCuAllocatorOptions(&po);

    po.Read(argc, argv);

    srand(srand_seed);

    if (po.NumArgs() != 3) {
      po.PrintUsage();
      exit(1);
    }

#if HAVE_CUDA==1
    CuDevice::Instantiate().SelectGpuId(use_gpu);
#endif

    std::string nnet_rxfilename = po.GetArg(1),
        examples_rspecifier = po.GetArg(2),
        nnet_wxfilename = po.GetArg(3);

    Nnet nnet;
    ReadKaldiObject(nnet_rxfilename, &nnet);

    NnetTrainer trainer(train_config, &nnet);

    SequentialNnetExampleReader example_reader(examples_rspecifier);

    for (; !example_reader.Done(); example_reader.Next())
      trainer.Train(example_reader.Value());

    bool ok = trainer.PrintTotalStats();

#if HAVE_CUDA==1
    CuDevice::Instantiate().PrintProfile();
#endif
    WriteKaldiObject(nnet, nnet_wxfilename, binary_write);
    KALDI_LOG << "Wrote model to " << nnet_wxfilename;
    return (ok ? 0 : 1);
  } catch(const std::exception &e) {
    std::cerr << e.what() << '\n';
    return -1;
  }
}