#!/bin/bash # Copyright 2012 Johns Hopkins University (Author: Daniel Povey) # Decoding script that does fMLLR. This can be on top of delta+delta-delta, or # LDA+MLLT features. # This script does an extra pass of lattice generation over and above what the original # script did-- it's for robustness in the case where your original cepstral mean # normalization was way off. # We also added a new option --distribute=true (by default) to # weight-silence-post. This weights the silence frames in a different way, # weighting all posteriors on the frame rather than just the silence ones, which # removes a particular kind of bias that the old approach suffered from. # There are 3 models involved potentially in this script, # and for a standard, speaker-independent system they will all be the same. # The "alignment model" is for the 1st-pass decoding and to get the # Gaussian-level alignments for the "adaptation model" the first time we # do fMLLR. The "adaptation model" is used to estimate fMLLR transforms # and to generate state-level lattices. The lattices are then rescored # with the "final model". # # The following table explains where we get these 3 models from. # Note: $srcdir is one level up from the decoding directory. # # Model Default source: # # "alignment model" $srcdir/final.alimdl --alignment-model # (or $srcdir/final.mdl if alimdl absent) # "adaptation model" $srcdir/final.mdl --adapt-model # "final model" $srcdir/final.mdl --final-model # Begin configuration section first_beam=10.0 # Beam used in initial, speaker-indep. pass first_max_active=2000 # max-active used in first two passes. first_latbeam=4.0 # lattice pruning beam for si decode and first-pass fMLLR decode. # the different spelling from lattice_beam is unfortunate; these scripts # have a history. alignment_model= adapt_model= final_model= cleanup=true stage=0 acwt=0.083333 # Acoustic weight used in getting fMLLR transforms, and also in # lattice generation. max_active=7000 beam=13.0 lattice_beam=6.0 nj=4 silence_weight=0.01 distribute=true # option to weight-silence-post. cmd=run.pl si_dir= fmllr_update_type=full skip_scoring=false num_threads=1 # if >1, will use gmm-latgen-faster-parallel parallel_opts= # If you supply num-threads, you should supply this too. scoring_opts= # 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 [ $# != 3 ]; then echo "Usage: steps/decode_fmllr.sh [options] " echo " e.g.: steps/decode_fmllr.sh exp/tri2b/graph_tgpr data/test_dev93 exp/tri2b/decode_dev93_tgpr" echo "main options (for others, see top of script file)" echo " --config # config containing options" echo " --nj # number of parallel jobs" echo " --cmd # Command to run in parallel with" echo " --adapt-model # Model to compute transforms with" echo " --alignment-model # Model to get Gaussian-level alignments for" echo " # 1st pass of transform computation." echo " --final-model # Model to finally decode with" echo " --si-dir # use this to skip 1st pass of decoding" echo " # Caution-- must be with same tree" echo " --acwt # default 0.08333 ... used to get posteriors" echo " --num-threads # number of threads to use, default 1." echo " --parallel-opts # e.g. '-pe smp 4' if you supply --num-threads 4" echo " --scoring-opts # options to local/score.sh" exit 1; fi graphdir=$1 data=$2 dir=`echo $3 | sed 's:/$::g'` # remove any trailing slash. srcdir=`dirname $dir`; # Assume model directory one level up from decoding directory. sdata=$data/split$nj; thread_string= [ $num_threads -gt 1 ] && thread_string="-parallel --num-threads=$num_threads" mkdir -p $dir/log [[ -d $sdata && $data/feats.scp -ot $sdata ]] || split_data.sh $data $nj || exit 1; echo $nj > $dir/num_jobs splice_opts=`cat $srcdir/splice_opts 2>/dev/null` # frame-splicing options. silphonelist=`cat $graphdir/phones/silence.csl` || exit 1; # Some checks. Note: we don't need $srcdir/tree but we expect # it should exist, given the current structure of the scripts. for f in $graphdir/HCLG.fst $data/feats.scp $srcdir/tree; do [ ! -f $f ] && echo "$0: no such file $f" && exit 1; done ## Work out name of alignment model. ## if [ -z "$alignment_model" ]; then if [ -f "$srcdir/final.alimdl" ]; then alignment_model=$srcdir/final.alimdl; else alignment_model=$srcdir/final.mdl; fi fi [ ! -f "$alignment_model" ] && echo "$0: no alignment model $alignment_model " && exit 1; ## ## Do the speaker-independent decoding, if --si-dir option not present. ## if [ -z "$si_dir" ]; then # we need to do the speaker-independent decoding pass. si_dir=${dir}.si # Name it as our decoding dir, but with suffix ".si". if [ $stage -le 0 ]; then steps/decode.sh --acwt $acwt --nj $nj --cmd "$cmd" --beam $first_beam --model $alignment_model\ --max-active $first_max_active --parallel-opts "${parallel_opts}" --num-threads $num_threads\ --skip-scoring true $graphdir $data $si_dir || exit 1; fi fi ## ## Some checks, and setting of defaults for variables. [ "$nj" -ne "`cat $si_dir/num_jobs`" ] && echo "Mismatch in #jobs with si-dir" && exit 1; [ ! -f "$si_dir/lat.1.gz" ] && echo "No such file $si_dir/lat.1.gz" && exit 1; [ -z "$adapt_model" ] && adapt_model=$srcdir/final.mdl [ -z "$final_model" ] && final_model=$srcdir/final.mdl for f in $adapt_model $final_model; do [ ! -f $f ] && echo "$0: no such file $f" && exit 1; done ## ## Set up the unadapted features "$sifeats" if [ -f $srcdir/final.mat ]; then feat_type=lda; else feat_type=delta; fi echo "$0: feature type is $feat_type"; case $feat_type in delta) sifeats="ark,s,cs:apply-cmvn --norm-vars=false --utt2spk=ark:$sdata/JOB/utt2spk scp:$sdata/JOB/cmvn.scp scp:$sdata/JOB/feats.scp ark:- | add-deltas ark:- ark:- |";; lda) sifeats="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 $srcdir/final.mat ark:- ark:- |";; *) echo "Invalid feature type $feat_type" && exit 1; esac ## ## Now get the first-pass fMLLR transforms. if [ $stage -le 1 ]; then echo "$0: getting first-pass fMLLR transforms." $cmd JOB=1:$nj $dir/log/fmllr_pass1.JOB.log \ gunzip -c $si_dir/lat.JOB.gz \| \ lattice-to-post --acoustic-scale=$acwt ark:- ark:- \| \ weight-silence-post --distribute=$distribute $silence_weight $silphonelist $alignment_model ark:- ark:- \| \ gmm-post-to-gpost $alignment_model "$sifeats" ark:- ark:- \| \ gmm-est-fmllr-gpost --fmllr-update-type=$fmllr_update_type \ --spk2utt=ark:$sdata/JOB/spk2utt $adapt_model "$sifeats" ark,s,cs:- \ ark:$dir/trans1.JOB || exit 1; fi ## pass1feats="$sifeats transform-feats --utt2spk=ark:$sdata/JOB/utt2spk ark:$dir/trans1.JOB ark:- ark:- |" ## Do the first adapted lattice generation pass. if [ $stage -le 2 ]; then echo "$0: doing first adapted lattice generation phase" $cmd $parallel_opts JOB=1:$nj $dir/log/decode1.JOB.log\ gmm-latgen-faster$thread_string --max-active=$first_max_active --beam=$first_beam --lattice-beam=$first_latbeam \ --acoustic-scale=$acwt --allow-partial=true --word-symbol-table=$graphdir/words.txt \ $adapt_model $graphdir/HCLG.fst "$pass1feats" "ark:|gzip -c > $dir/lat1.JOB.gz" \ || exit 1; fi ## Do a second pass of estimating the transform. Compose the transforms to get ## $dir/trans2.*. if [ $stage -le 3 ]; then echo "$0: estimating fMLLR transforms a second time." $cmd JOB=1:$nj $dir/log/fmllr_pass2.JOB.log \ lattice-to-post --acoustic-scale=$acwt "ark:gunzip -c $dir/lat1.JOB.gz|" ark:- \| \ weight-silence-post --distribute=$distribute $silence_weight $silphonelist $adapt_model ark:- ark:- \| \ gmm-est-fmllr --fmllr-update-type=$fmllr_update_type \ --spk2utt=ark:$sdata/JOB/spk2utt $adapt_model "$pass1feats" \ ark,s,cs:- ark:$dir/trans1b.JOB '&&' \ compose-transforms --b-is-affine=true ark:$dir/trans1b.JOB ark:$dir/trans1.JOB \ ark:$dir/trans2.JOB || exit 1; if $cleanup; then rm $dir/trans1b.* $dir/trans1.* $dir/lat1.*.gz fi fi ## pass2feats="$sifeats transform-feats --utt2spk=ark:$sdata/JOB/utt2spk ark:$dir/trans2.JOB ark:- ark:- |" # Generate a 3rd set of lattices, with the "adaptation model"; we'll use these # to adapt a 3rd time, and we'll rescore them. Since we should be close to the final # fMLLR, we don't bother dumping un-determinized lattices to disk. ## Do the final lattice generation pass (but we'll rescore these lattices ## after another stage of adaptation.) if [ $stage -le 4 ]; then echo "$0: doing final lattice generation phase" $cmd $parallel_opts JOB=1:$nj $dir/log/decode2.JOB.log\ gmm-latgen-faster$thread_string --max-active=$max_active --beam=$beam --lattice-beam=$lattice_beam \ --acoustic-scale=$acwt --allow-partial=true --word-symbol-table=$graphdir/words.txt \ $adapt_model $graphdir/HCLG.fst "$pass2feats" "ark:|gzip -c > $dir/lat2.JOB.gz" \ || exit 1; fi ## Do a third pass of estimating the transform. Compose the transforms to get ## $dir/trans.*. if [ $stage -le 5 ]; then echo "$0: estimating fMLLR transforms a third time." $cmd JOB=1:$nj $dir/log/fmllr_pass3.JOB.log \ lattice-to-post --acoustic-scale=$acwt "ark:gunzip -c $dir/lat2.JOB.gz|" ark:- \| \ weight-silence-post --distribute=$distribute $silence_weight $silphonelist $adapt_model ark:- ark:- \| \ gmm-est-fmllr --fmllr-update-type=$fmllr_update_type \ --spk2utt=ark:$sdata/JOB/spk2utt $adapt_model "$pass2feats" \ ark,s,cs:- ark:$dir/trans2b.JOB '&&' \ compose-transforms --b-is-affine=true ark:$dir/trans2b.JOB ark:$dir/trans2.JOB \ ark:$dir/trans.JOB || exit 1; if $cleanup; then rm $dir/trans2b.* $dir/trans2.* fi fi ## feats="$sifeats transform-feats --utt2spk=ark:$sdata/JOB/utt2spk ark:$dir/trans.JOB ark:- ark:- |" if [ $stage -le 6 ]; then echo "$0: doing a final pass of acoustic rescoring." $cmd JOB=1:$nj $dir/log/acoustic_rescore.JOB.log \ gmm-rescore-lattice $final_model "ark:gunzip -c $dir/lat2.JOB.gz|" "$feats" \ "ark:|gzip -c > $dir/lat.JOB.gz" || exit 1; if $cleanup; then rm $dir/lat2.*.gz fi fi if ! $skip_scoring ; then [ ! -x local/score.sh ] && \ echo "$0: not scoring because local/score.sh does not exist or not executable." && exit 1; local/score.sh $scoring_opts --cmd "$cmd" $data $graphdir $dir fi exit 0;