Blame view
src/nnet2bin/nnet-am-switch-preconditioning.cc
3.54 KB
8dcb6dfcb
first commit
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 |
// nnet2bin/nnet-am-switch-preconditioning.cc
// Copyright 2012 Johns Hopkins University (author: Daniel Povey)
// See ../../COPYING for clarification regarding multiple authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
// WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
// MERCHANTABLITY OR NON-INFRINGEMENT.
// See the Apache 2 License for the specific language governing permissions and
// limitations under the License.
#include "base/kaldi-common.h"
#include "util/common-utils.h"
#include "nnet2/am-nnet.h"
#include "hmm/transition-model.h"
#include "tree/context-dep.h"
int main(int argc, char *argv[]) {
try {
using namespace kaldi;
using namespace kaldi::nnet2;
typedef kaldi::int32 int32;
const char *usage =
"Copy a (cpu-based) neural net and its associated transition model,
"
"and switch it to online preconditioning, i.e. change any components
"
"derived from AffineComponent to components of type
"
"AffineComponentPreconditionedOnline.
"
"
"
"Usage: nnet-am-switch-preconditioning [options] <nnet-in> <nnet-out>
"
"e.g.:
"
" nnet-am-switch-preconditioning --binary=false 1.mdl text.mdl
";
int32 rank_in = 20, rank_out = 80, update_period = 4;
BaseFloat num_samples_history = 2000.0;
BaseFloat alpha = 4.0;
bool binary_write = true;
ParseOptions po(usage);
po.Register("binary", &binary_write, "Write output in binary mode");
po.Register("rank-in", &rank_in,
"Rank used in online-preconditioning on input side of each layer");
po.Register("rank-out", &rank_out,
"Rank used in online-preconditioning on output side of each layer");
po.Register("update-period", &update_period,
"Affects how frequently we update the Fisher-matrix estimate (every "
"this-many minibatches).");
po.Register("num-samples-history", &num_samples_history,
"Number of samples of history to use in online preconditioning "
"(affects speed vs accuracy of update of Fisher matrix)");
po.Register("alpha", &alpha,
"Parameter that affects amount of smoothing with unit matrix "
"in online preconditioning (larger -> more smoothing)");
po.Read(argc, argv);
if (po.NumArgs() != 2) {
po.PrintUsage();
exit(1);
}
std::string nnet_rxfilename = po.GetArg(1),
nnet_wxfilename = po.GetArg(2);
TransitionModel trans_model;
AmNnet am_nnet;
{
bool binary;
Input ki(nnet_rxfilename, &binary);
trans_model.Read(ki.Stream(), binary);
am_nnet.Read(ki.Stream(), binary);
}
am_nnet.GetNnet().SwitchToOnlinePreconditioning(rank_in, rank_out, update_period,
num_samples_history, alpha);
{
Output ko(nnet_wxfilename, binary_write);
trans_model.Write(ko.Stream(), binary_write);
am_nnet.Write(ko.Stream(), binary_write);
}
KALDI_LOG << "Copied neural net from " << nnet_rxfilename
<< " to " << nnet_wxfilename;
return 0;
} catch(const std::exception &e) {
std::cerr << e.what() << '
';
return -1;
}
}
|