Merge pull request #131 from dabraude/master

adds script for custom voices

idlak-data/en/ga/lexicon-cmu0.7a.xml

@@ -479,6 +479,7 @@
 <lex pron="ae0 k s eh1 n ch ah0 w ey2 t ih0 d" entry="full" default="true">accentuated</lex>
 <lex pron="ae0 k s eh1 n ch uw0 ey0 t s" entry="full" default="true">accentuates</lex>
 <lex pron="ae0 k s eh1 n ch ah0 w ey2 t ih0 ng" entry="full" default="true">accentuating</lex>
+<lex pron="ae1 k s eh0 n ch er0" entry="full" default="true">accenture</lex>
 <lex pron="ae0 k s eh1 p t" entry="full" default="true">accept</lex>
 <lex pron="ah0 k s eh1 p t" entry="1" default="false">accept</lex>
 <lex pron="ah0 k s eh2 p t ah0 b ih1 l ah0 t iy0" entry="full" default="true">acceptability</lex>
@@ -11428,7 +11429,8 @@
 <lex pron="b ay0 m ah1 n th l iy0" entry="full" default="true">bimonthly</lex>
 <lex pron="b ih1 n" entry="full" default="true">bin</lex>
 <lex pron="b iy1 n ah0" entry="full" default="true">bina</lex>
-<lex pron="b ay1 n er0 iy0" entry="full" default="true">binary</lex>
+<lex pron="b ay1 n ax0 r iy0" entry="full" default="true">binary</lex>
+<lex pron="b ay1 n er0 iy0" entry="full" default="false">binary</lex>
 <lex pron="b ay0 n ae1 sh ah0 n ah0 l" entry="full" default="true">binational</lex>
 <lex pron="b ih1 n ch iy0" entry="full" default="true">binchy</lex>
 <lex pron="b ay1 n d" entry="full" default="true">bind</lex>
@@ -19537,6 +19539,7 @@
 <lex pron="s eh1 r ah0 m ow2 n iy0 z" entry="full" default="true">ceremonies</lex>
 <lex pron="s eh1 r ah0 m ow2 n iy0" entry="full" default="true">ceremony</lex>
 <lex pron="s er0 ey1 n ow0" entry="full" default="true">cereno</lex>
+<lex pron="s eh1 r ax0 p r ao2 k" entry="full" default="true">cereproc</lex>
 <lex pron="s ih1 r iy0 z" entry="full" default="true">ceres</lex>
 <lex pron="s eh2 r ey1 z ow0" entry="full" default="true">cerezo</lex>
 <lex pron="s er0 ey1 z ow0" entry="1" default="false">cerezo</lex>

idlak-egs/tts_tangle_custom/s2/cmd.sh

@@ -0,0 +1,23 @@
+# you can change cmd.sh depending on what type of queue you are using.
+# If you have no queueing system and want to run on a local machine, you
+# can change all instances 'queue.pl' to run.pl (but be careful and run
+# commands one by one: most recipes will exhaust the memory on your
+# machine).  queue.pl works with GridEngine (qsub).  slurm.pl works
+# with slurm.  Different queues are configured differently, with different
+# queue names and different ways of specifying things like memory;
+# to account for these differences you can create and edit the file
+# conf/queue.conf to match your queue's configuration.  Search for
+# conf/queue.conf in http://kaldi-asr.org/doc/queue.html for more information,
+# or search for the string 'default_config' in utils/queue.pl or utils/slurm.pl.
+
+export train_cmd=run.pl
+export decode_cmd="run.pl --mem 2G"
+# the use of cuda_cmd is deprecated, used only in 'nnet1',
+export cuda_cmd="run.pl --gpu 1"
+
+if [[ "$(hostname -f)" == "*.fit.vutbr.cz" ]]; then
+  queue_conf=$HOME/queue_conf/default.conf # see example /homes/kazi/iveselyk/queue_conf/default.conf,
+  export train_cmd="queue.pl --config $queue_conf --mem 2G --matylda 0.2"
+  export decode_cmd="queue.pl --config $queue_conf --mem 3G --matylda 0.1"
+  export cuda_cmd="queue.pl --config $queue_conf --gpu 1 --mem 10G --tmp 40G"
+fi

idlak-egs/tts_tangle_custom/s2/conf/2-layer-nn.conf

@@ -0,0 +1,5 @@
+hid_layers=2
+hid-dim=1000
+splice_lr=0
+apply_cmvn=true
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/3-layer-nn-2splice10.conf

@@ -0,0 +1,4 @@
+hid_layers=3
+insplice_lr=10
+splice_lr=10
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/3-layer-nn-splice10.conf

@@ -0,0 +1,4 @@
+hid_layers=3
+insplice_lr=10
+splice_lr=0
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/3-layer-nn-splice5.conf

@@ -0,0 +1,5 @@
+hid_layers=3
+insplice_lr=5
+insplice_step=1
+splice_lr=0
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/3-layer-nn.conf

@@ -0,0 +1,3 @@
+hid_layers=3
+splice_lr=0
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/4-layer-nn.conf

@@ -0,0 +1,4 @@
+hid_layers=4
+splice_lr=0
+apply_cmvn=true
+apply_glob_cmvn=true

idlak-egs/tts_tangle_custom/s2/conf/auto-nn.conf

@@ -0,0 +1,5 @@
+hid_layers=2
+splice_lr=0
+apply_glob_cmvn=true
+apply_cmvn=false
+apply_minmax=false

idlak-egs/tts_tangle_custom/s2/conf/bndap-48k.conf

@@ -0,0 +1,8 @@
+--snip-edges=false
+--frame-shift=5
+--sample-frequency=48000
+--num-mel-bins=25
+--scale-type=bark
+--use-hts-bands=true
+--peak-quality=0.4
+--peak-width=0.1

idlak-egs/tts_tangle_custom/s2/conf/bndap.conf

@@ -0,0 +1,9 @@
+--snip-edges=false
+--frame-shift=5
+--sample-frequency=16000
+--num-mel-bins=21
+--use-hts-bands=true
+--peak-quality=0.4
+--peak-width=0.1
+
+

idlak-egs/tts_tangle_custom/s2/conf/decode_dnn.config

@@ -0,0 +1,3 @@
+#beam=18.0 # beam for decoding.  Was 13.0 in the scripts.
+#latbeam=10.0 # this has most effect on size of the lattices.
+parallel_opts=

idlak-egs/tts_tangle_custom/s2/conf/dur-lstm-splice5.conf

@@ -0,0 +1,13 @@
+hid_layers=3
+insplice=
+splice=0
+delta_opts="--delta-order=2"
+minmax_opts=true
+learn_rate=0.001
+hid_dim=100
+network_type=lstm
+momentum=0.8
+proto_opts="--no-softmax --cell-dim=100 --proj-dim=50 --num-layers=2 --activation-final --activation-type=<Sigmoid>"
+train_tool=nnet-train-multistream-perutt-tgtmat
+train_tool_opts="--objective-function=mse --num-streams=10"
+scheduler_opts="--min-iters 15"

idlak-egs/tts_tangle_custom/s2/conf/dur-nn-splice5.conf

@@ -0,0 +1,22 @@
+hid_layers=3
+insplice=5
+splice=0
+delta_opts=
+#"--delta-order=2"
+#minmax_opts=
+
+#proto_opts="--no-softmax --activation-type=<Tanh>"
+#learn_rate=0.0001
+
+# --activation-final=<Sigmoid>
+#proto_opts="--no-softmax --activation-type=<Sigmoid>"
+#learn_rate=0.00001
+cmvn_opts="--norm-means=false --norm-vars=true"
+minmax_opts=
+global_cmvn=
+proto_opts="--no-softmax --activation-type=<Sigmoid>"
+learn_rate=0.000005
+momentum=0.2
+hid_dim=100
+halving_factor=0.75
+max_iters=50

idlak-egs/tts_tangle_custom/s2/conf/fbank.conf

@@ -0,0 +1,4 @@
+# No non-default options for now.
+--frame-shift=5
+--num-mel-bins=50
+--use-energy=true

idlak-egs/tts_tangle_custom/s2/conf/full-lstm-splice5.conf

@@ -0,0 +1,13 @@
+hid_layers=3
+insplice=
+splice=0
+delta_opts="--delta-order=2"
+minmax_opts=true
+learn_rate=0.0001
+hid_dim=100
+network_type=lstm
+momentum=0.8
+proto_opts="--no-softmax --cell-dim=1000 --proj-dim=700 --num-layers=2 --activation-final --activation-type=<Sigmoid>"
+train_tool=nnet-train-multistream-perutt-tgtmat
+train_tool_opts="--objective-function=mse --num-streams=10"
+scheduler_opts="--min-iters 15"

idlak-egs/tts_tangle_custom/s2/conf/full-nn-splice5.conf

@@ -0,0 +1,11 @@
+hid_layers=3
+insplice=5
+splice=0
+delta_opts="--delta-order=2"
+proto_opts="--activation-final=<Sigmoid> --no-softmax --activation-type=<Sigmoid>"
+learn_rate=0.02
+minmax_opts=true
+momentum=0.2
+hid_dim=700
+halving_factor=0.75
+max_iters=50

idlak-egs/tts_tangle_custom/s2/conf/mcep-48k.conf

@@ -0,0 +1,3 @@
+srate=48000
+order=60
+alpha=0.55

idlak-egs/tts_tangle_custom/s2/conf/mcep.conf

@@ -0,0 +1,3 @@
+srate=16000
+order=39
+alpha=0.42

idlak-egs/tts_tangle_custom/s2/conf/mfcc-48k.conf

@@ -0,0 +1,4 @@
+--use-energy=false
+--frame-shift=5
+--snip-edges=false
+--sample-frequency=48000

idlak-egs/tts_tangle_custom/s2/conf/mfcc.conf

@@ -0,0 +1,3 @@
+--use-energy=false
+--frame-shift=5
+--snip-edges=false

idlak-egs/tts_tangle_custom/s2/conf/pitch-48k.conf

@@ -0,0 +1,4 @@
+--snip-edges=false
+--frame-shift=5
+--sample-frequency=48000 
+--frame-length=30

idlak-egs/tts_tangle_custom/s2/conf/pitch-lstm-splice5.conf

@@ -0,0 +1,13 @@
+hid_layers=3
+insplice=
+splice=0
+delta_opts="--delta-order=2"
+minmax_opts=true
+learn_rate=0.0005
+hid_dim=100
+network_type=lstm
+momentum=0.8
+proto_opts="--no-softmax --cell-dim=100 --proj-dim=50 --num-layers=2 --activation-final --activation-type=<Sigmoid>"
+train_tool=nnet-train-multistream-perutt-tgtmat
+train_tool_opts="--objective-function=mse --num-streams=10"
+scheduler_opts="--min-iters 15"

idlak-egs/tts_tangle_custom/s2/conf/pitch-nn-splice5.conf

@@ -0,0 +1,22 @@
+hid_layers=3
+insplice=0
+splice=0
+delta_opts="--delta-order=2"
+minmax_opts=true
+#cmvn_opts="--norm-means=true --norm-vars=true"
+proto_opts="--activation-final=<Sigmoid> --no-softmax --activation-type=<Sigmoid>"
+learn_rate=0.02
+momentum=0.2
+hid_dim=128
+halving_factor=0.5
+max_iters=50
+
+#cmvn_opts="--norm-means=false --norm-vars=true"
+#minmax_opts=
+#global_cmvn=true
+#proto_opts="--no-softmax --activation-type=<Sigmoid>"
+#learn_rate=0.02
+#momentum=0.2
+#hid_dim=100
+#halving_factor=0.75
+#max_iters=50

idlak-egs/tts_tangle_custom/s2/conf/pitch.conf

@@ -0,0 +1,4 @@
+--snip-edges=false
+--frame-shift=5
+--sample-frequency=16000 
+--frame-length=30

idlak-egs/tts_tangle_custom/s2/conf/pretrain.conf

@@ -0,0 +1,4 @@
+splice=4
+delta_order=2
+apply_cmvn=true
+copy_feats=true

idlak-egs/tts_tangle_custom/s2/local/apply_map.pl

@@ -0,0 +1,81 @@
+#!/usr/bin/perl -w
+# Copyright 2012  Johns Hopkins University (Author: Daniel Povey)
+# Apache 2.0.
+
+# This program is a bit like ./sym2int.pl in that it applies a map
+# to things in a file, but it's a bit more general in that it doesn't
+# assume the things being mapped to are single tokens, they could
+# be sequences of tokens.
+
+# This program takes two arguments, which may be files or "-" for the
+# standard input.  Both files must have lines with one or more fields,
+# interpreted as a map from the first field (a string) to a list of strings.
+# if the first file has as one of its lines
+# A x y
+# and the second has the lines
+# x P
+# y Q R
+# then the output of this program will be
+# A P Q R
+#
+# Note that if x or y did not appear as the first field of file b, we would
+# print a warning and omit the whole line rather than map it to the empty
+# string.
+
+
+if (@ARGV > 0 && $ARGV[0] eq "-f") {
+  shift @ARGV; 
+  $field_spec = shift @ARGV; 
+  if ($field_spec =~ m/^\d+$/) {
+    $field_begin = $field_spec - 1; $field_end = $field_spec - 1;
+  }
+  if ($field_spec =~ m/^(\d*)[-:](\d*)/) { # accept e.g. 1:10 as a courtesty (properly, 1-10)
+    if ($1 ne "") {
+      $field_begin = $1 - 1;    # Change to zero-based indexing.
+    }
+    if ($2 ne "") {
+      $field_end = $2 - 1;      # Change to zero-based indexing.
+    }
+  }
+  if (!defined $field_begin && !defined $field_end) {
+    die "Bad argument to -f option: $field_spec"; 
+  }
+}
+
+
+if(@ARGV != 1) {
+  print STDERR "Usage: apply_map.pl [options] map <input >output\n" .
+    "options: [-f <field-range> ]\n" .
+    "note: <field-range> can look like 4-5, or 4-, or 5-, or 1.\n" .
+    "e.g.: echo A B | apply_map.pl a.txt\n" .
+    "where a.txt is:\n" .
+    "A a1 a2\n" .
+    "B b\n" .
+    "will produce:\n" .
+    "a1 a2 b\n";
+  exit(1);
+}
+
+($map) = @ARGV;
+open(M, "<$map") || die "Error opening map file $map: $!";
+
+while (<M>) {
+  @A = split(" ", $_);
+  @A >= 1 || die "apply_map.pl: empty line.";
+  $i = shift @A;
+  $o = join(" ", @A);
+  $map{$i} = $o;
+}
+
+while(<STDIN>) {
+  @A = split(" ", $_);
+  for ($x = 0; $x < @A; $x++) {
+    if ( (!defined $field_begin || $x >= $field_begin)
+         && (!defined $field_end || $x <= $field_end)) {
+      $a = $A[$x];
+      if (!defined $map{$a}) { die "compose_maps.pl: undefined key $a\n"; }
+      $A[$x] = $map{$a};
+    }
+  }
+  print join(" ", @A) . "\n";
+}