Blame view

Scripts/steps/pretrain_dbn.sh 9.84 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
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
  #!/bin/bash
  # Copyright 2013 Karel Vesely
  
  # 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.
  
  # To be run from ..
  #
  # Deep Belief Network pre-training by Contrastive Divergence (CD-1) algorithm.
  # The script can pre-train on plain features (ie. saved fMLLR features), 
  # or modified features (CMN, delta).
  # The script creates feature-transform in nnet format, which contains splice 
  # and shift+scale (zero mean and unit variance on DBN input).
  #
  # For special cases it is possible to use external feature-transform.
  # 
  
  # Begin configuration.
  #
  # nnet config
  nn_depth=6     #number of hidden layers
  hid_dim=2048   #number of units per layer
  # number of iterations
  rbm_iter=1            #number of pre-training epochs (Gaussian-Bernoulli RBM has 2x more)
  rbm_drop_data=0.0     #sample the training set, 1.0 drops all the data, 0.0 keeps all
  # pre-training opts
  rbm_lrate=0.4         #RBM learning rate
  rbm_lrate_low=0.01    #lower RBM learning rate (for Gaussian units)
  rbm_l2penalty=0.0002  #L2 penalty (increases RBM-mixing rate)
  rbm_extra_opts=
  # data processing config
  copy_feats=true    # resave the features randomized consecutively to tmpdir
  # feature config
  feature_transform= # Optionally reuse feature processing front-end (override splice,etc.)
  delta_order=       # Optionally use deltas on the input features
  apply_cmvn=false   # Optionally do CMVN of the input features
  norm_vars=false    # When apply_cmvn=true, this enables CVN
  splice=5           # Temporal splicing
  splice_step=1      # Stepsize of the splicing (1 is consecutive splice, 
                     # value 2 would do [ -10 -8 -6 -4 -2 0 2 4 6 8 10 ] splicing)
  # misc.
  verbose=1 # enable per-cache reports
  # gpu config
  use_gpu_id= # manually select GPU id to run on, (-1 disables GPU) 
  # End configuration.
  
  echo "$0 $@"  # Print the command line for logging
  
  [ -f path.sh ] && . ./path.sh;
  . parse_options.sh || exit 1;
  
  
  if [ $# != 2 ]; then
     echo "Usage: $0 <data> <exp-dir>"
     echo " e.g.: $0 data/train exp/rbm_pretrain"
     echo "main options (for others, see top of script file)"
     echo "  --config <config-file>           # config containing options"
     echo ""
     echo "  --nn-depth <N>                   # number of RBM layers"
     echo "  --hid-dim <N>                    # number of hidden units per layer"
     echo "  --rbm-iter <N>                   # number of CD-1 iterations per layer"
     echo "  --dbm-drop-data <float>          # probability of frame-dropping,"
     echo "                                   # can be used to subsample large datasets"
     echo "  --rbm-lrate <float>              # learning-rate for Bernoulli-Bernoulli RBMs"
     echo "  --rbm-lrate-low <float>          # learning-rate for Gaussian-Bernoulli RBM"
     echo ""
     echo "  --copy-feats <bool>              # copy features to /tmp, to accelerate training"
     echo "  --apply-cmvn <bool>              # normalize input features (opt.)"
     echo "    --norm-vars <bool>               # use variance normalization (opt.)"
     echo "  --splice <N>                     # splice +/-N frames of input features"
     exit 1;
  fi
  
  data=$1
  dir=$2
  
  
  for f in $data/feats.scp; do
    [ ! -f $f ] && echo "$0: no such file $f" && exit 1;
  done
  
  echo "# INFO"
  echo "$0 : Pre-training Deep Belief Network as a stack of RBMs"
  printf "\t dir       : $dir 
  "
  printf "\t Train-set : $data 
  "
  
  [ -e $dir/${nn_depth}.dbn ] && echo "$0 Skipping, already have $dir/${nn_depth}.dbn" && exit 0
  
  mkdir -p $dir/log
  
  ###### PREPARE FEATURES ######
  echo
  echo "# PREPARING FEATURES"
  # shuffle the list
  echo "Preparing train/cv lists"
  cat $data/feats.scp | utils/shuffle_list.pl --srand ${seed:-777} > $dir/train.scp
  # print the list size
  wc -l $dir/train.scp
  
  #re-save the shuffled features, so they are stored sequentially on the disk in /tmp/
  if [ "$copy_feats" == "true" ]; then
    tmpdir=$(mktemp -d); mv $dir/train.scp $dir/train.scp_non_local
    utils/nnet/copy_feats.sh $dir/train.scp_non_local $tmpdir $dir/train.scp
    #remove data on exit...
    trap "echo \"Removing features tmpdir $tmpdir @ $(hostname)\"; rm -r $tmpdir" EXIT
  fi
  
  #create a 10k utt subset for global cmvn estimates
  head -n 10000 $dir/train.scp > $dir/train.scp.10k
  
  
  
  ###### PREPARE FEATURE PIPELINE ######
  
  #read the features
  feats="ark:copy-feats scp:$dir/train.scp ark:- |"
  
  #optionally add per-speaker CMVN
  if [ $apply_cmvn == "true" ]; then
    echo "Will use CMVN statistics : $data/cmvn.scp"
    [ ! -r $data/cmvn.scp ] && echo "Cannot find cmvn stats $data/cmvn.scp" && exit 1;
    cmvn="scp:$data/cmvn.scp"
    feats="$feats apply-cmvn --print-args=false --norm-vars=$norm_vars --utt2spk=ark:$data/utt2spk $cmvn ark:- ark:- |"
    # keep track of norm_vars option
    echo "$norm_vars" >$dir/norm_vars 
  else
    echo "apply_cmvn disabled (per speaker norm. on input features)"
  fi
  
  #optionally add deltas
  if [ "$delta_order" != "" ]; then
    feats="$feats add-deltas --delta-order=$delta_order ark:- ark:- |"
    echo "$delta_order" > $dir/delta_order
  fi
  
  #get feature dim
  echo -n "Getting feature dim : "
  feat_dim=$(feat-to-dim --print-args=false scp:$dir/train.scp -)
  echo $feat_dim
  
  
  # Now we will start building feature_transform which will 
  # be applied in CUDA to gain more speed.
  #
  # We will use 1GPU for both feature_transform and MLP training in one binary tool. 
  # This is against the kaldi spirit, but it is necessary, because on some sites a GPU 
  # cannot be shared accross by two or more processes (compute exclusive mode),
  # and we would like to use single GPU per training instance,
  # so that the grid resources can be used efficiently...
  
  
  if [ ! -z "$feature_transform" ]; then
    echo Using already prepared feature_transform: $feature_transform
    cp $feature_transform $dir/final.feature_transform
  else
    # Generate the splice transform
    echo "Using splice +/- $splice , step $splice_step"
    feature_transform=$dir/tr_splice$splice-$splice_step.nnet
    utils/nnet/gen_splice.py --fea-dim=$feat_dim --splice=$splice --splice-step=$splice_step > $feature_transform
  
    # Renormalize the MLP input to zero mean and unit variance
    feature_transform_old=$feature_transform
    feature_transform=${feature_transform%.nnet}_cmvn-g.nnet
    echo "Renormalizing MLP input features into $feature_transform"
    nnet-forward ${use_gpu_id:+ --use-gpu-id=$use_gpu_id} \
      $feature_transform_old "$(echo $feats | sed 's|train.scp|train.scp.10k|')" \
      ark:- 2>$dir/log/cmvn_glob_fwd.log |\
    compute-cmvn-stats ark:- - | cmvn-to-nnet - - |\
    nnet-concat --binary=false $feature_transform_old - $feature_transform
  
    # MAKE LINK TO THE FINAL feature_transform, so the other scripts will find it ######
    [ -f $dir/final.feature_transform ] && unlink $dir/final.feature_transform
    (cd $dir; ln -s $(basename $feature_transform) final.feature_transform )
  fi
  
  
  
  ###### GET THE DIMENSIONS ######
  num_fea=$(feat-to-dim --print-args=false "$feats nnet-forward --use-gpu-id=-1 $feature_transform ark:- ark:- |" - 2>/dev/null)
  num_hid=$hid_dim
  
  
  ###### PERFORM THE PRE-TRAINING ######
  for depth in $(seq 1 $nn_depth); do
    echo
    echo "# PRE-TRAINING RBM LAYER $depth"
    RBM=$dir/$depth.rbm
    [ -f $RBM ] && echo "RBM '$RBM' already trained, skipping." && continue
  
    #The first RBM needs special treatment, because of Gussian input nodes
    if [ "$depth" == "1" ]; then
      #This is Gaussian-Bernoulli RBM
      #initialize
      echo "Initializing '$RBM.init'"
      utils/nnet/gen_rbm_init.py --dim=${num_fea}:${num_hid} --gauss --vistype=gauss --hidtype=bern > $RBM.init || exit 1
      #pre-train
      echo "Pretraining '$RBM' (reduced lrate and 2x more iters)"
      rbm-train-cd1-frmshuff --learn-rate=$rbm_lrate_low --l2-penalty=$rbm_l2penalty \
        --num-iters=$((2*$rbm_iter)) --drop-data=$rbm_drop_data --verbose=$verbose \
        --feature-transform=$feature_transform \
        ${use_gpu_id:+ --use-gpu-id=$use_gpu_id} $rbm_extra_opts \
        $RBM.init "$feats" $RBM 2>$dir/log/rbm.$depth.log || exit 1
    else
      #This is Bernoulli-Bernoulli RBM
      #cmvn stats for init
      echo "Computing cmvn stats '$dir/$depth.cmvn' for RBM initialization"
      if [ ! -f $dir/$depth.cmvn ]; then 
        nnet-forward ${use_gpu_id:+ --use-gpu-id=$use_gpu_id} \
         "nnet-concat $feature_transform $dir/$((depth-1)).dbn - |" \
          "$(echo $feats | sed 's|train.scp|train.scp.10k|')" \
          ark:- 2>$dir/log/cmvn_fwd.$depth.log | \
        compute-cmvn-stats ark:- - 2>$dir/log/cmvn.$depth.log | \
        cmvn-to-nnet - $dir/$depth.cmvn || exit 1
      else
        echo compute-cmvn-stats already done, skipping.
      fi
      #initialize
      echo "Initializing '$RBM.init'"
      utils/nnet/gen_rbm_init.py --dim=${num_hid}:${num_hid} --gauss --vistype=bern --hidtype=bern --cmvn-nnet=$dir/$depth.cmvn > $RBM.init || exit 1
      #pre-train
      echo "Pretraining '$RBM'"
      rbm-train-cd1-frmshuff --learn-rate=$rbm_lrate --l2-penalty=$rbm_l2penalty \
        --num-iters=$rbm_iter --drop-data=$rbm_drop_data --verbose=$verbose \
        --feature-transform="nnet-concat $feature_transform $dir/$((depth-1)).dbn - |" \
        ${use_gpu_id:+ --use-gpu-id=$use_gpu_id} $rbm_extra_opts \
        $RBM.init "$feats" $RBM 2>$dir/log/rbm.$depth.log || exit 1
    fi
  
    #Create DBN stack
    if [ "$depth" == "1" ]; then
      rbm-convert-to-nnet --binary=true $RBM $dir/$depth.dbn
    else 
      rbm-convert-to-nnet --binary=true $RBM - | \
      nnet-concat $dir/$((depth-1)).dbn - $dir/$depth.dbn
    fi
  
  done
  
  echo
  echo "# REPORT"
  echo "# RBM pre-training progress (line per-layer)"
  grep progress $dir/log/rbm.*.log
  echo 
  
  echo "Pre-training finished."
  
  sleep 3
  exit 0