Blame view

Scripts/steps/train_diag_ubm.sh 4.88 KB
ec85f8892   bigot benjamin   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
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
  #!/bin/bash
  
  # Copyright Johns Hopkins University (Author: Daniel Povey),  2012.  
  # Apache 2.0.
  
  # Train a diagonal mixture of Gaussians.  This is trained without
  # reference to class labels-- except that, optionally, you can down-weight
  # silence phones, and alignments are needed for that.
  #
  # The current use for this is in fMMI training.
  
  # Begin configuration section.
  nj=4
  cmd=run.pl
  num_iters=3
  silence_weight=
  stage=-2
  # The value "intermediate" is a number of Gaussians we first obtain by clustering
  # the Gaussians within each state of the model, before clustering down to
  # $num_Gauss.  This is for efficiency.  It's not a very important parameter,
  # as far as I know.
  intermediate=2000
  num_gselect=50 # Number of Gaussian-selection indices to use while training
                 # the model.
  # End configuration section.
  
  echo "$0 $@"  # Print the command line for logging
  
  [ -f ./path.sh ] && . ./path.sh; # source the path.
  . parse_options.sh || exit 1;
  
  
  if [ $# != 5 ]; then
    echo "Usage: steps/train_diag_ubm.sh <num-gauss> <data> <lang> <alignment-dir|src-dir> <dir>"
    echo " e.g.: steps/train_diag_ubm.sh 400 data/train_si84 data/lang exp/tri2b_ali_si84 exp/ubm3c"
    echo "Options: "
    echo "  --silence-weight <sil-weight>                  # default 1.0.  Use to down-weight silence."
    echo "  --cmd (utils/run.pl|utils/queue.pl <queue opts>) # how to run jobs."
    echo "  --nj <num-job>                                 # number of parallel jobs to run."
    echo "  --num-iters <niter>                            # number of iterations of training (default: $num_iters)"
    echo "  --stage <stage>                                # stage to do partial re-run from."
    exit 1;
  fi
  
  num_gauss=$1
  data=$2
  lang=$3
  alidir=$4
  dir=$5
  
  silphonelist=`cat $lang/phones/silence.csl` || exit 1;
  
  sdata=$data/split$nj
  splice_opts=`cat $alidir/splice_opts 2>/dev/null`
  mkdir -p $dir/log
  [[ -d $sdata && $data/feats.scp -ot $sdata ]] || split_data.sh $data $nj || exit 1;
  echo $nj > $dir/num_jobs
  
  if [ -f $alidir/final.mat ]; then feat_type=lda; else feat_type=delta; fi
  echo "$0: feature type is $feat_type"
  
  case $feat_type in
    delta) feats="ark,s,cs:apply-cmvn --norm-vars=false --utt2spk=ark:$sdata/JOB/utt2spk scp:$sdata/JOB/cmvn.scp scp:$sdata/JOB/feats.scp ark:- | add-deltas ark:- ark:- |";;
    lda) feats="ark,s,cs:apply-cmvn --norm-vars=false --utt2spk=ark:$sdata/JOB/utt2spk scp:$sdata/JOB/cmvn.scp scp:$sdata/JOB/feats.scp ark:- | splice-feats $splice_opts ark:- ark:- | transform-feats $alidir/final.mat ark:- ark:- |"
      cp $alidir/final.mat $dir    
      ;;
    *) echo "Invalid feature type $feat_type" && exit 1;
  esac
  
  if [ -f $alidir/trans.1 ]; then
    echo Using transforms from $alidir;
    [ "$nj" -ne "`cat $alidir/num_jobs`" ] && \
      echo "The number of jobs differs from alignment directory $alidir." && exit 1;
    feats="$feats transform-feats --utt2spk=ark:$sdata/JOB/utt2spk ark:$alidir/trans.JOB ark:- ark:- |"
  fi
  
  if [ ! -z "$silence_weight" ]; then
    [ ! -f $alidir/ali.1.gz ] && \
      echo "You specified weighting for silence but $alidir/ali.1.gz does not exist." && exit 1;
    [ "$nj" -ne "`cat $alidir/num_jobs`" ] && \
      echo "You specified silence weight but $alidir has different #jobs." && exit 1;
    weights="--weights='ark,s,cs:gunzip -c $alidir/ali.JOB.gz | ali-to-post ark:- ark:- | weight-silence-post $silence_weight $silphonelist $alidir/final.mdl ark:- ark:- | post-to-weights ark:- ark:- |'"
  else
    weights=
  fi
  
  # $intermediate should be more than $num_gauss..
  [ $[$num_gauss*2] -gt $intermediate ] && intermediate=$[$num_gauss*2] \
    && echo "Setting intermediate=$intermediate (it was too small)";
  
  if [ $stage -le -2 ]; then
   echo "Clustering Gaussians in $alidir/final.mdl"
   $cmd $dir/log/cluster.log \
    init-ubm --fullcov-ubm=false --intermediate-num-gauss=$intermediate \
      --ubm-num-gauss=$num_gauss $alidir/final.mdl $alidir/final.occs $dir/0.dubm   || exit 1;
  fi
  
  # Store Gaussian selection indices on disk-- this speeds up the training passes.
  if [ $stage -le -1 ]; then
    echo Getting Gaussian-selection info
    $cmd JOB=1:$nj $dir/log/gselect.JOB.log \
      gmm-gselect --n=$num_gselect $dir/0.dubm "$feats" \
        "ark:|gzip -c >$dir/gselect.JOB.gz" || exit 1;
  fi
  
  for x in `seq 0 $[$num_iters-1]`; do
    echo "Training pass $x"
    if [ $stage -le $x ]; then
    # Accumulate stats.
      $cmd JOB=1:$nj $dir/log/acc.$x.JOB.log \
        gmm-global-acc-stats $weights "--gselect=ark,s,cs:gunzip -c $dir/gselect.JOB.gz|" \
        $dir/$x.dubm "$feats" $dir/$x.JOB.acc || exit 1;
      if [ $x -lt $[$num_iters-1] ]; then # Don't remove low-count Gaussians till last iter,
        opt="--remove-low-count-gaussians=false" # or gselect info won't be valid any more.
      fi
      $cmd $dir/log/update.$x.log \
        gmm-global-est $opt $dir/$x.dubm "gmm-global-sum-accs - $dir/$x.*.acc|" \
        $dir/$[$x+1].dubm || exit 1;
      rm $dir/$x.*.acc $dir/$x.dubm
    fi
  done
  
  rm $dir/gselect.*.gz
  mv $dir/$num_iters.dubm $dir/final.dubm || exit 1;
  exit 0;