Kaldi简介

$diff cmd.sh
-export train_cmd=queue.pl
-export decode_cmd="queue.pl --mem 4G"
-export mkgraph_cmd="queue.pl --mem 8G"
-export cuda_cmd="queue.pl --gpu 1"
+export train_cmd=run.pl
+export decode_cmd="run.pl --mem 4G"
+export mkgraph_cmd="run.pl --mem 8G"
+export cuda_cmd="run.pl --gpu 1"

thchs=/home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30-openslr

#[ -d $thchs ] || mkdir -p $thchs  || exit 1
#echo "downloading THCHS30 at $thchs ..."
#local/download_and_untar.sh $thchs  http://www.openslr.org/resources/18 data_thchs30  || exit 1
#local/download_and_untar.sh $thchs  http://www.openslr.org/resources/18 resource      || exit 1
#local/download_and_untar.sh $thchs  http://www.openslr.org/resources/18 test-noise    || exit 1

1 他 仅 凭 腰部 的 力量 在 泳道 上下 翻腾 蛹 动 蛇行 状 如 海豚 一直 以 一头 的 优势 领先
2 ta1 jin3 ping2 yao1 bu4 de5 li4 liang4 zai4 yong3 dao4 shang4 xia4 fan1 teng2 yong3 dong4 she2 xing2 zhuang4 ru2 hai3 tun2 yi4 zhi2 yi3 yi4 tou2 de5 you1 shi4 ling3 xian1
3 t a1 j in3 p ing2 ii iao1 b u4 d e5 l i4 l iang4 z ai4 ii iong3 d ao4 sh ang4 x ia4 f an1 t eng2 ii iong3 d ong4 sh e2 x ing2 zh uang4 r u2 h ai3 t un2 ii i4 zh ix2 ii i3 ii i4 t ou2 d e5 ii iu1 sh ix4 l ing3 x ian1

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data$ ls -l train/
total 5744
-rw-rw-r-- 1 lili lili 2175289 Aug 14 12:38 phone.txt
-rw-rw-r-- 1 lili lili   80200 Aug 14 12:39 spk2utt
-rw-rw-r-- 1 lili lili 1257119 Aug 14 12:38 text
-rw-rw-r-- 1 lili lili  120000 Aug 14 12:38 utt2spk
-rw-rw-r-- 1 lili lili  974999 Aug 14 12:38 wav.scp
-rw-rw-r-- 1 lili lili 1257119 Aug 14 12:38 word.txt

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data/train$ head -2 text 
A02_000 绿 是 阳春 烟 景 大块 文章 的 底色 四月 的 林 峦 更是 绿 得 鲜活 秀媚 诗意 盎然
A02_001 他 仅 凭 腰部 的 力量 在 泳道 上下 翻腾 蛹 动 蛇行 状 如 海豚 一直 以 一头 的 优势 领先

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data/train$ head -2 wav.scp 
A02_000 /home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30-openslr/data_thchs30/train/A2_0.wav
A02_001 /home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30-openslr/data_thchs30/train/A2_1.wav

s5# head -3 data/train/segments
sw02001-A_000098-001156 sw02001-A 0.98 11.56
sw02001-A_001980-002131 sw02001-A 19.8 21.31
sw02001-A_002736-002893 sw02001-A 27.36 28.93

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data/train$ head -2 utt2spk 
A02_000 A02
A02_001 A02

utils/utt2spk_to_spk2utt.pl data/train/utt2spk > data/train/spk2utt

A04 A04_000 A04_001 A04_002 A04_003 A04_005 A04_006 A04_007 A04_008 ...

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data/mfcc/train$ head -3 feats.scp 
A02_000 /home/lili/codes/kaldi/egs/thchs30/s5/mfcc/train/raw_mfcc_train.1.ark:8
A02_001 /home/lili/codes/kaldi/egs/thchs30/s5/mfcc/train/raw_mfcc_train.1.ark:12868
A02_002 /home/lili/codes/kaldi/egs/thchs30/s5/mfcc/train/raw_mfcc_train.1.ark:26222

steps/make_mfcc.sh --nj 8 --cmd "$train_cmd" data/train exp/make_mfcc/train $mfccdir

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/data/mfcc/train$ head -2 cmvn.scp 
A02 /home/lili/codes/kaldi/egs/thchs30/s5/mfcc/train/cmvn_train.ark:4
A04 /home/lili/codes/kaldi/egs/thchs30/s5/mfcc/train/cmvn_train.ark:247

steps/compute_cmvn_stats.sh data/train exp/make_mfcc/train $mfccdir

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ ls data/lang
L_disambig.fst  L.fst  oov.int  oov.txt  phones  phones.txt  topo  words.txt

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ ls data/lang/phones
align_lexicon.int  disambig.int         nonsilence.txt        sets.int              wdisambig.txt
align_lexicon.txt  disambig.txt         optional_silence.csl  sets.txt              wdisambig_words.int
context_indep.csl  extra_questions.int  optional_silence.int  silence.csl
context_indep.int  extra_questions.txt  optional_silence.txt  silence.int
context_indep.txt  nonsilence.csl       roots.int             silence.txt
disambig.csl       nonsilence.int       roots.txt             wdisambig_phones.int

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head data/lang/phones.txt 
<eps> 0
sil 1
a1 2
a2 3
a3 4
a4 5
a5 6
aa 7
ai1 8
ai2 9

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head data/lang/words.txt 
<eps> 0
# 1
<SPOKEN_NOISE> 2
SIL 3
一 4
一一 5
一丁点 6
一万 7
一万元 8
一万多 9

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/oov.txt 
<SPOKEN_NOISE>

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/oov.int 
2

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head -5 data/local/lang/lexiconp.txt 
# 1.0	sil
<SPOKEN_NOISE> 1.0	sil
SIL 1.0	sil
一 1.0	ii i1
一一 1.0	ii i1 ii i1

和 1.0  h e2
和 1.0  h e4
和 1.0  h u2
和 1.0  h uo2
和 1.0  h uo4

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/topo
<Topology>
<TopologyEntry>
<ForPhones>
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
</ForPhones>
<State> 0 <PdfClass> 0 <Transition> 0 0.75 <Transition> 1 0.25 </State>
<State> 1 <PdfClass> 1 <Transition> 1 0.75 <Transition> 2 0.25 </State>
<State> 2 <PdfClass> 2 <Transition> 2 0.75 <Transition> 3 0.25 </State>
<State> 3 </State>
</TopologyEntry>
<TopologyEntry>
<ForPhones>
1
</ForPhones>
<State> 0 <PdfClass> 0 <Transition> 0 0.25 <Transition> 1 0.25 <Transition> 2 0.25 <Transition> 3 0.25 </State>
<State> 1 <PdfClass> 1 <Transition> 1 0.25 <Transition> 2 0.25 <Transition> 3 0.25 <Transition> 4 0.25 </State>
<State> 2 <PdfClass> 2 <Transition> 1 0.25 <Transition> 2 0.25 <Transition> 3 0.25 <Transition> 4 0.25 </State>
<State> 3 <PdfClass> 3 <Transition> 1 0.25 <Transition> 2 0.25 <Transition> 3 0.25 <Transition> 4 0.25 </State>
<State> 4 <PdfClass> 4 <Transition> 4 0.75 <Transition> 5 0.25 </State>
<State> 5 </State>
</TopologyEntry>
</Topology>

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/phones/context_indep.txt
sil

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/phones/silence.txt 
sil

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head -3 data/lang/phones/nonsilence.txt 
a1
a2
a3

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head -3 data/lang/phones/disambig.txt
#0
#1
#2

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/phones/optional_silence.txt
sil

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head -3 data/lang/phones/sets.txt
sil
a1
a2

s5# head -3 data/lang/phones/sets.txt
SIL SIL_B SIL_E SIL_I SIL_S
SPN SPN_B SPN_E SPN_I SPN_S
NSN NSN_B NSN_E NSN_I NSN_S

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ cat data/lang/phones/extra_questions.txt
sil
a1 ai1 an1 ang1 ao1 e1 ei1 en1 eng1 i1 ia1 ian1 iang1 iao1 ie1 in1 ing1 iong1 iu1 ix1 iy1 o1 ong1 ou1 u1 ua1 uai1 uan1 uang1 ueng1 ui1 un1 uo1 v1 van1 ve1 vn1
a2 ai2 an2 ang2 ao2 e2 ei2 en2 eng2 er2 i2 ia2 ian2 iang2 iao2 ie2 in2 ing2 iong2 iu2 ix2 iy2 o2 ong2 ou2 u2 ua2 uai2 uan2 uang2 ui2 un2 uo2 v2 van2 ve2 vn2
a3 ai3 an3 ang3 ao3 e3 ei3 en3 eng3 er3 i3 ia3 ian3 iang3 iao3 ie3 in3 ing3 iong3 iu3 ix3 iy3 o3 ong3 ou3 u3 ua3 uai3 uan3 uang3 ueng3 ui3 un3 uo3 v3 van3 ve3 vn3
a4 ai4 an4 ang4 ao4 e4 ei4 en4 eng4 er4 i4 ia4 ian4 iang4 iao4 ie4 in4 ing4 iong4 iu4 ix4 iy4 iz4 o4 ong4 ou4 u4 ua4 uai4 uan4 uang4 ueng4 ui4 un4 uo4 v4 van4 ve4 vn4
a5 ai5 an5 ang5 ao5 e5 ei5 en5 eng5 er5 i5 ia5 ian5 iang5 iao5 ie5 in5 ing5 iong5 iu5 ix5 iy5 iz5 o5 ong5 ou5 u5 ua5 uai5 uan5 uang5 ueng5 ui5 un5 uo5 v5 van5 ve5 vn5
aa b c ch d ee f g h ii j k l m n oo p q r s sh t uu vv x z zh

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5$ head data/lang/phones/roots.txt
shared split sil
shared split a1
shared split a2
shared split a3
shared split a4
shared split a5
shared split aa
shared split ai1
shared split ai2
shared split ai3

head data/lang/phones/roots.txt
shared split SIL SIL_B SIL_E SIL_I SIL_S
shared split SPN SPN_B SPN_E SPN_I SPN_S
shared split NSN NSN_B NSN_E NSN_I NSN_S
shared split LAU LAU_B LAU_E LAU_I LAU_S
...
shared split B_B B_E B_I B_S

local/thchs-30_data_prep.sh $H $thchs/data_thchs30 || exit 1;

#!/bin/bash
# Copyright 2016  Tsinghua University (Author: Dong Wang, Xuewei Zhang).  Apache 2.0.
#           2016  LeSpeech (Author: Xingyu Na)

#This script pepares the data directory for thchs30 recipe. 
#It reads the corpus and get wav.scp and transcriptions.

# dir可能是"/home/lili/codes/kaldi/egs/thchs30/s5"
dir=$1
# corpus_dir可能是"/home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30-openslr/data_thchs30"
corpus_dir=$2


cd $dir

# 在data目录下创建 train,dev和test子目录
echo "creating data/{train,dev,test}"
mkdir -p data/{train,dev,test}

# 创建 wav.scp, utt2spk.scp, spk2utt.scp, text
(
# x分别取值"train","dev"和"test"
for x in train dev test; do
	echo "cleaning data/$x"
	# 进入到相应目录，比如data/train
	cd $dir/data/$x
	# 删除下面的相关文件
	rm -rf wav.scp utt2spk spk2utt word.txt phone.txt text
	echo "preparing scps and text in data/$x"
	# 在/home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30-openslr/data_thchs30/train
        # 下找所有的wav文件，并且根据名字排序
	for nn in `find  $corpus_dir/$x -name "*.wav" | sort -u | xargs -I {} basename {} .wav`; do
	    # 从文件名中抽取信息，比如从C8_742.wav
	    # spkid 为C8（需要了解简单的awk)
		spkid=`echo $nn | awk -F"_" '{print "" $1}'`
		# spk_char为C
		spk_char=`echo $spkid | sed 's/\([A-Z]\).*/\1/'`
		# spk_num为8
		spk_num=`echo $spkid | sed 's/[A-Z]\([0-9]\)/\1/'`
		# 把spkid变成C08(原来是C8)，因为说话人小于100个，这样就可以保证顺序
		spkid=$(printf '%s%.2d' "$spk_char" "$spk_num")
		# 抽取下划线后面的 742
		utt_num=`echo $nn | awk -F"_" '{print $2}'`
		# 变成C08_742
		uttid=$(printf '%s%.2d_%.3d' "$spk_char" "$spk_num" "$utt_num")
		# 输出到wav.scp的一行，类似
                # "C08_742 /home/lili/codes/kaldi/egs/thchs30/s5/data/thchs30
                #        -openslr/data_thchs30/train/C8_742.wav"
		echo $uttid $corpus_dir/$x/$nn.wav >> wav.scp
		# 输出到utt2spk的一行 "C08_742 C08"
		echo $uttid $spkid >> utt2spk
		# 输出到word.txt "C08_742 这里 说 的 易 牙 蒸 了 他 儿子 给 桀纣 吃
                #    也是 狂人 语 颇 错 杂 无 伦次 的 表现"
		echo $uttid `sed -n 1p $corpus_dir/data/$nn.wav.trn` >> word.txt
		# 输出到phone.txt "C08_742 zh e4 l i3 sh uo1 d e5 ii i4 ii ia2 zh eng1 l e5 t a1 ee 
                # er2 z iy5 g ei3 j ie2 zh ou4 ch ix1 ii ie3 sh ix4 k uang2 r en2 vv v3 p o1 c uo4 
                # z a2 uu u2 l un2 c iy4 d e5 b iao3 x ian4"
		echo $uttid `sed -n 3p $corpus_dir/data/$nn.wav.trn` >> phone.txt
	done 
	# word.txt就是kaldi要的text
	cp word.txt text
	# wav.scp排序
	sort wav.scp -o wav.scp
	# utt2spk排序
	sort utt2spk -o utt2spk
	# text排序
	sort text -o text
	# phone.txt排序
	sort phone.txt -o phone.txt
done
) || exit 1

# 根据utt2spk生成spk2utt
utils/utt2spk_to_spk2utt.pl data/train/utt2spk > data/train/spk2utt
utils/utt2spk_to_spk2utt.pl data/dev/utt2spk > data/dev/spk2utt
utils/utt2spk_to_spk2utt.pl data/test/utt2spk > data/test/spk2utt

echo "creating test_phone for phone decoding"
(
    # 复制data/test目录到data/test_phone
	rm -rf data/test_phone && cp -R data/test data/test_phone  || exit 1
	# 把phone.txt复制到text
	cd data/test_phone && rm text &&  cp phone.txt text || exit 1
)

#produce MFCC features
rm -rf data/mfcc && mkdir -p data/mfcc &&  cp -R data/{train,dev,test,test_phone} data/mfcc || exit 1;
for x in train dev test; do
	#make  mfcc
	steps/make_mfcc.sh --nj $n --cmd "$train_cmd" data/mfcc/$x exp/make_mfcc/$x mfcc/$x || exit 1;
	#compute cmvn
	steps/compute_cmvn_stats.sh data/mfcc/$x exp/mfcc_cmvn/$x mfcc/$x || exit 1;
done
#copy feats and cmvn to test.ph, avoid duplicated mfcc & cmvn
cp data/mfcc/test/feats.scp data/mfcc/test_phone && cp data/mfcc/test/cmvn.scp data/mfcc/test_phone || exit 1;

#prepare language stuff
#build a large lexicon that invovles words in both the training and decoding.
(
  echo "make word graph ..."
  cd $H; mkdir -p data/{dict,lang,graph} && \
  cp $thchs/resource/dict/{extra_questions.txt,nonsilence_phones.txt,optional_silence.txt,silence_phones.txt} data/dict && \
  cat $thchs/resource/dict/lexicon.txt $thchs/data_thchs30/lm_word/lexicon.txt | \
  grep -v '<s>' | grep -v '</s>' | sort -u > data/dict/lexicon.txt || exit 1;
  utils/prepare_lang.sh --position_dependent_phones false data/dict "<SPOKEN_NOISE>" data/local/lang data/lang || exit 1;
  gzip -c $thchs/data_thchs30/lm_word/word.3gram.lm > data/graph/word.3gram.lm.gz || exit 1;
  utils/format_lm.sh data/lang data/graph/word.3gram.lm.gz $thchs/data_thchs30/lm_word/lexicon.txt data/graph/lang || exit 1;
)

#make_phone_graph
(
  echo "make phone graph ..."
  cd $H; mkdir -p data/{dict_phone,graph_phone,lang_phone} && \
  cp $thchs/resource/dict/{extra_questions.txt,nonsilence_phones.txt,optional_silence.txt,silence_phones.txt} data/dict_phone  && \
  cat $thchs/data_thchs30/lm_phone/lexicon.txt | grep -v '<eps>' | sort -u > data/dict_phone/lexicon.txt  && \
  echo "<SPOKEN_NOISE> sil " >> data/dict_phone/lexicon.txt  || exit 1;
  utils/prepare_lang.sh --position_dependent_phones false data/dict_phone "<SPOKEN_NOISE>" data/local/lang_phone data/lang_phone || exit 1;
  gzip -c $thchs/data_thchs30/lm_phone/phone.3gram.lm > data/graph_phone/phone.3gram.lm.gz  || exit 1;
  utils/format_lm.sh data/lang_phone data/graph_phone/phone.3gram.lm.gz $thchs/data_thchs30/lm_phone/lexicon.txt \
  data/graph_phone/lang  || exit 1;
)

utils/prepare_lang.sh --position_dependent_phones false data/dict "<SPOKEN_NOISE>" data/local/lang data/lang

utils/format_lm.sh data/lang data/graph/word.3gram.lm.gz $thchs/data_thchs30/lm_word/lexicon.txt data/graph/lang

#monophone
steps/train_mono.sh --boost-silence 1.25 --nj $n --cmd "$train_cmd" data/mfcc/train data/lang exp/mono || exit 1;
#test monophone model
local/thchs-30_decode.sh --mono true --nj $n "steps/decode.sh" exp/mono data/mfcc &

steps/train_mono.sh --boost-silence 1.25 --nj 4 --cmd run.pl data/mfcc/train data/lang exp/mono
steps/train_mono.sh: Initializing monophone system.
steps/train_mono.sh: Compiling training graphs
steps/train_mono.sh: Aligning data equally (pass 0)
steps/train_mono.sh: Pass 1
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 2
....

steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 39
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/mono
steps/diagnostic/analyze_alignments.sh: see stats in exp/mono/log/analyze_alignments.log
6154 warnings in exp/mono/log/align.*.*.log
153 warnings in exp/mono/log/acc.*.*.log
1051 warnings in exp/mono/log/update.*.log
exp/mono: nj=4 align prob=-100.09 over 25.49h [retry=0.2%, fail=0.0%] states=656 gauss=990
steps/train_mono.sh: Done training monophone system in exp/mono
using monophone to generate graph
WARNING: the --mono, --left-biphone and --quinphone options are now deprecated and ignored.
tree-info exp/mono/tree 
tree-info exp/mono/tree 
fstminimizeencoded 
fstpushspecial 
fsttablecompose data/graph/lang/L_disambig.fst data/graph/lang/G.fst 
fstdeterminizestar --use-log=true 
fstisstochastic data/graph/lang/tmp/LG.fst 
-0.0480882 -0.0488869
[info]: LG not stochastic.
fstcomposecontext --context-size=1 --central-position=0 --read-disambig-syms=data/graph/lang/phones/disambig.int --write-disambig-syms=data/graph/lang/tmp/disambig_ilabels_1_0.int data/graph/lang/tmp/ilabels_1_0.27036 
fstisstochastic data/graph/lang/tmp/CLG_1_0.fst 
-0.0480882 -0.0488869
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/mono/graph_word/disambig_tid.int --transition-scale=1.0 data/graph/lang/tmp/ilabels_1_0 exp/mono/tree exp/mono/final.mdl 
fsttablecompose exp/mono/graph_word/Ha.fst data/graph/lang/tmp/CLG_1_0.fst 
fstrmsymbols exp/mono/graph_word/disambig_tid.int 
fstminimizeencoded 
fstdeterminizestar --use-log=true 
fstrmepslocal 
fstisstochastic exp/mono/graph_word/HCLGa.fst 
0.644531 -0.0974261
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/mono/final.mdl 
steps/decode.sh --cmd run.pl --mem 4G --nj 4 exp/mono/graph_word data/mfcc/test exp/mono/decode_test_word
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/mono/graph_word exp/mono/decode_test_word
steps/diagnostic/analyze_lats.sh: see stats in exp/mono/decode_test_word/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(3,36,176) and mean=69.7
steps/diagnostic/analyze_lats.sh: see stats in exp/mono/decode_test_word/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test exp/mono/graph_word exp/mono/decode_test_word
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
WARNING: the --mono, --left-biphone and --quinphone options are now deprecated and ignored.
tree-info exp/mono/tree 
tree-info exp/mono/tree 
fstminimizeencoded 
fstpushspecial 
fstdeterminizestar --use-log=true 
fsttablecompose data/graph_phone/lang/L_disambig.fst data/graph_phone/lang/G.fst 
fstisstochastic data/graph_phone/lang/tmp/LG.fst 
-0.264706 -0.265648
[info]: LG not stochastic.
fstcomposecontext --context-size=1 --central-position=0 --read-disambig-syms=data/graph_phone/lang/phones/disambig.int --write-disambig-syms=data/graph_phone/lang/tmp/disambig_ilabels_1_0.int data/graph_phone/lang/tmp/ilabels_1_0.31511 
fstisstochastic data/graph_phone/lang/tmp/CLG_1_0.fst 
-0.264706 -0.265648
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/mono/graph_phone/disambig_tid.int --transition-scale=1.0 data/graph_phone/lang/tmp/ilabels_1_0 exp/mono/tree exp/mono/final.mdl 
fstrmepslocal 
fstrmsymbols exp/mono/graph_phone/disambig_tid.int 
fstdeterminizestar --use-log=true 
fstminimizeencoded 
fsttablecompose exp/mono/graph_phone/Ha.fst data/graph_phone/lang/tmp/CLG_1_0.fst 
fstisstochastic exp/mono/graph_phone/HCLGa.fst 
0.00033643 -0.522085
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/mono/final.mdl 
steps/decode.sh --cmd run.pl --mem 4G --nj 4 exp/mono/graph_phone data/mfcc/test_phone exp/mono/decode_test_phone
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/mono/graph_phone exp/mono/decode_test_phone
steps/diagnostic/analyze_lats.sh: see stats in exp/mono/decode_test_phone/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(6,20,62) and mean=28.8
steps/diagnostic/analyze_lats.sh: see stats in exp/mono/decode_test_phone/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test_phone exp/mono/graph_phone exp/mono/decode_test_phone
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/exp/mono/decode_test_word$ cat wer_9_0.0 
compute-wer --text --mode=present ark:exp/mono/decode_test_word/scoring_kaldi/test_filt.txt ark,p:- 
%WER 50.74 [ 41169 / 81139, 511 ins, 2413 del, 38245 sub ]
%SER 100.00 [ 2495 / 2495 ]
Scored 2495 sentences, 0 not present in hyp.

lili@lili-Precision-7720:~/codes/kaldi/egs/thchs30/s5/exp/mono/decode_test_word/log$ cat decode.1.log
gmm-latgen-faster --max-active=7000 --beam=13.0 --lattice-beam=6.0 --acoustic-scale=0.083333 --allow-partial=true --word-symbol-table=exp/mono/graph_word/words.txt exp/mono/final.mdl exp/mono/graph_word/HCLG.fst 'ark,s,cs:apply-cmvn  --utt2spk=ark:data/mfcc/test/split4/1/utt2spk scp:data/mfcc/test/split4/1/cmvn.scp scp:data/mfcc/test/split4/1/feats.scp ark:- | add-deltas  ark:- ark:- |' 'ark:|gzip -c > exp/mono/decode_test_word/lat.1.gz'
add-deltas ark:- ark:-
apply-cmvn --utt2spk=ark:data/mfcc/test/split4/1/utt2spk scp:data/mfcc/test/split4/1/cmvn.scp scp:data/mfcc/test/split4/1/feats.scp ark:-
D04_750 受 北京 一些 爱国 驾驶 满载 山 起步 长期 无 说明 艾 根 钱 没 敢 言 奋起 抗战
LOG (gmm-latgen-faster[5.4.232~1-532f3]:DecodeUtteranceLatticeFaster():decoder-wrappers.cc:286) Log-like per frame for utterance D04_750 is -8.33257 over 961 frames.
D04_751 王 金山 被 枪毙 后 部分 余孽 宣传 起来 其次 为 捕 鲸 被 抓获

steps/align_si.sh --boost-silence 1.25 --nj 6 --cmd run.pl data/mfcc/train data/lang exp/mono exp/mono_ali
steps/align_si.sh: feature type is delta
steps/align_si.sh: aligning data in data/mfcc/train using model from exp/mono, putting alignments in exp/mono_ali
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/mono_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/mono_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.

steps/train_deltas.sh --boost-silence 1.25 --cmd run.pl 2000 10000 data/mfcc/train data/lang exp/mono_ali exp/tri1
steps/train_deltas.sh: accumulating tree stats
steps/train_deltas.sh: getting questions for tree-building, via clustering
steps/train_deltas.sh: building the tree
 
steps/train_deltas.sh: converting alignments from exp/mono_ali to use current tree
steps/train_deltas.sh: compiling graphs of transcripts
steps/train_deltas.sh: training pass 1
steps/train_deltas.sh: training pass 2
steps/train_deltas.sh: training pass 3
steps/train_deltas.sh: training pass 4
steps/train_deltas.sh: training pass 5
steps/train_deltas.sh: training pass 6
...
steps/train_deltas.sh: training pass 34
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/tri1
steps/diagnostic/analyze_alignments.sh: see stats in exp/tri1/log/analyze_alignments.log
153 warnings in exp/tri1/log/align.*.*.log
1 warnings in exp/tri1/log/compile_questions.log
281 warnings in exp/tri1/log/update.*.log
124 warnings in exp/tri1/log/acc.*.*.log
9 warnings in exp/tri1/log/init_model.log
9 warnings in exp/tri1/log/questions.log
1 warnings in exp/tri1/log/build_tree.log
exp/tri1: nj=6 align prob=-96.74 over 25.48h [retry=0.5%, fail=0.0%] states=1648 gauss=10031 tree-impr=4.77
steps/train_deltas.sh: Done training system with delta+delta-delta features in exp/tri1
tree-info exp/tri1/tree 
tree-info exp/tri1/tree 
fstcomposecontext --context-size=3 --central-position=1 --read-disambig-syms=data/graph/lang/phones/disambig.int --write-disambig-syms=data/graph/lang/tmp/disambig_ilabels_3_1.int data/graph/lang/tmp/ilabels_3_1.17687 
fstisstochastic data/graph/lang/tmp/CLG_3_1.fst 
0 -0.0488869
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/tri1/graph_word/disambig_tid.int --transition-scale=1.0 data/graph/lang/tmp/ilabels_3_1 exp/tri1/tree exp/tri1/final.mdl 
fstrmepslocal 
fsttablecompose exp/tri1/graph_word/Ha.fst data/graph/lang/tmp/CLG_3_1.fst 
fstrmsymbols exp/tri1/graph_word/disambig_tid.int 
fstdeterminizestar --use-log=true 
fstminimizeencoded 
fstisstochastic exp/tri1/graph_word/HCLGa.fst 
0.644531 -0.128776
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/tri1/final.mdl 

steps/decode.sh --cmd run.pl --mem 4G --nj 6 exp/tri1/graph_word data/mfcc/test exp/tri1/decode_test_word
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri1/graph_word exp/tri1/decode_test_word
steps/diagnostic/analyze_lats.sh: see stats in exp/tri1/decode_test_word/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,13,72) and mean=28.3
steps/diagnostic/analyze_lats.sh: see stats in exp/tri1/decode_test_word/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test exp/tri1/graph_word exp/tri1/decode_test_word
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
tree-info exp/tri1/tree 
tree-info exp/tri1/tree 
fstcomposecontext --context-size=3 --central-position=1 --read-disambig-syms=data/graph_phone/lang/phones/disambig.int --write-disambig-syms=data/graph_phone/lang/tmp/disambig_ilabels_3_1.int data/graph_phone/lang/tmp/ilabels_3_1.24591 
fstisstochastic data/graph_phone/lang/tmp/CLG_3_1.fst 
0 -0.265648
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/tri1/graph_phone/disambig_tid.int --transition-scale=1.0 data/graph_phone/lang/tmp/ilabels_3_1 exp/tri1/tree exp/tri1/final.mdl 
fstrmepslocal 
fstminimizeencoded 
fstrmsymbols exp/tri1/graph_phone/disambig_tid.int 
fstdeterminizestar --use-log=true 
fsttablecompose exp/tri1/graph_phone/Ha.fst data/graph_phone/lang/tmp/CLG_3_1.fst 
fstisstochastic exp/tri1/graph_phone/HCLGa.fst 
0.000486116 -0.797725
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/tri1/final.mdl 
steps/decode.sh --cmd run.pl --mem 4G --nj 6 exp/tri1/graph_phone data/mfcc/test_phone exp/tri1/decode_test_phone
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri1/graph_phone exp/tri1/decode_test_phone
steps/diagnostic/analyze_lats.sh: see stats in exp/tri1/decode_test_phone/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(3,10,32) and mean=14.7
steps/diagnostic/analyze_lats.sh: see stats in exp/tri1/decode_test_phone/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test_phone exp/tri1/graph_phone exp/tri1/decode_test_phone
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/align_si.sh --nj 6 --cmd run.pl data/mfcc/train data/lang exp/tri1 exp/tri1_ali
steps/align_si.sh: feature type is delta
steps/align_si.sh: aligning data in data/mfcc/train using model from exp/tri1, putting alignments in exp/tri1_ali
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/tri1_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/tri1_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.

steps/train_lda_mllt.sh --cmd run.pl --splice-opts --left-context=3 --right-context=3 2500 15000 data/mfcc/train data/lang exp/tri1_ali exp/tri2b
steps/train_lda_mllt.sh: Accumulating LDA statistics.
steps/train_lda_mllt.sh: Accumulating tree stats
steps/train_lda_mllt.sh: Getting questions for tree clustering.
steps/train_lda_mllt.sh: Building the tree
steps/train_lda_mllt.sh: Initializing the model
...

steps/train_sat.sh --cmd run.pl 2500 15000 data/mfcc/train data/lang exp/tri2b_ali exp/tri3b
steps/train_sat.sh: feature type is lda
steps/train_sat.sh: obtaining initial fMLLR transforms since not present in exp/tri2b_ali
steps/train_sat.sh: Accumulating tree stats
steps/train_sat.sh: Getting questions for tree clustering.
steps/train_sat.sh: Building the tree
steps/train_sat.sh: Initializing the model
...

steps/train_quick.sh --cmd run.pl 4200 40000 data/mfcc/train data/lang exp/tri3b_ali exp/tri4b
steps/train_quick.sh: feature type is lda
steps/train_quick.sh: using transforms from exp/tri3b_ali
steps/train_quick.sh: accumulating tree stats
steps/train_quick.sh: Getting questions for tree clustering.
steps/train_quick.sh: Building the tree
steps/train_quick.sh: Initializing the model
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 109 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 121 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 163 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 175 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 176 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 177 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 203 of new tree has no stats.
WARNING (gmm-init-model[5.4.232~1-532f3]:InitAmGmmFromOld():gmm-init-model.cc:147) Leaf 208 of new tree has no stats.
steps/train_quick.sh: This is a bad warning.
steps/train_quick.sh: mixing up old model.
steps/train_quick.sh: converting old alignments
steps/train_quick.sh: compiling training graphs
steps/train_quick.sh: pass 1
steps/train_quick.sh: pass 2
steps/train_quick.sh: pass 3
steps/train_quick.sh: pass 4
steps/train_quick.sh: pass 5
steps/train_quick.sh: pass 6
steps/train_quick.sh: pass 7
steps/train_quick.sh: pass 8
steps/train_quick.sh: pass 9
steps/train_quick.sh: pass 10
steps/train_quick.sh: aligning data
steps/train_quick.sh: pass 11
steps/train_quick.sh: pass 12
steps/train_quick.sh: pass 13
steps/train_quick.sh: pass 14
steps/train_quick.sh: pass 15
steps/train_quick.sh: aligning data
steps/train_quick.sh: pass 16
steps/train_quick.sh: pass 17
steps/train_quick.sh: pass 18
steps/train_quick.sh: pass 19
steps/train_quick.sh: estimating alignment model
Done
tree-info exp/tri4b/tree 
tree-info exp/tri4b/tree 
make-h-transducer --disambig-syms-out=exp/tri4b/graph_word/disambig_tid.int --transition-scale=1.0 data/graph/lang/tmp/ilabels_3_1 exp/tri4b/tree exp/tri4b/final.mdl 
fsttablecompose exp/tri4b/graph_word/Ha.fst data/graph/lang/tmp/CLG_3_1.fst 
fstrmsymbols exp/tri4b/graph_word/disambig_tid.int 
fstdeterminizestar --use-log=true 
fstrmepslocal 
fstminimizeencoded 
fstisstochastic exp/tri4b/graph_word/HCLGa.fst 
0.644531 -0.128947
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/tri4b/final.mdl 
steps/decode_fmllr.sh --cmd run.pl --mem 4G --nj 6 exp/tri4b/graph_word data/mfcc/test exp/tri4b/decode_test_word
steps/decode.sh --scoring-opts  --num-threads 1 --skip-scoring false --acwt 0.083333 --nj 6 --cmd run.pl --mem 4G --beam 10.0 --model exp/tri4b/final.alimdl --max-active 2000 exp/tri4b/graph_word data/mfcc/test exp/tri4b/decode_test_word.si
decode.sh: feature type is lda
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri4b/graph_word exp/tri4b/decode_test_word.si
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_word.si/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,6,35) and mean=13.7
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_word.si/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test exp/tri4b/graph_word exp/tri4b/decode_test_word.si
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmllr.sh: feature type is lda
steps/decode_fmllr.sh: getting first-pass fMLLR transforms.
steps/decode_fmllr.sh: doing main lattice generation phase
steps/decode_fmllr.sh: estimating fMLLR transforms a second time.
steps/decode_fmllr.sh: doing a final pass of acoustic rescoring.
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri4b/graph_word exp/tri4b/decode_test_word
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_word/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,6,35) and mean=13.9
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_word/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test exp/tri4b/graph_word exp/tri4b/decode_test_word
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
tree-info exp/tri4b/tree 
tree-info exp/tri4b/tree 
make-h-transducer --disambig-syms-out=exp/tri4b/graph_phone/disambig_tid.int --transition-scale=1.0 data/graph_phone/lang/tmp/ilabels_3_1 exp/tri4b/tree exp/tri4b/final.mdl 
fstrmsymbols exp/tri4b/graph_phone/disambig_tid.int 
fstminimizeencoded 
fstrmepslocal 
fsttablecompose exp/tri4b/graph_phone/Ha.fst data/graph_phone/lang/tmp/CLG_3_1.fst 
fstdeterminizestar --use-log=true 
fstisstochastic exp/tri4b/graph_phone/HCLGa.fst 
0.00048422 -0.797661
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/tri4b/final.mdl 
steps/decode_fmllr.sh --cmd run.pl --mem 4G --nj 6 exp/tri4b/graph_phone data/mfcc/test_phone exp/tri4b/decode_test_phone
steps/decode.sh --scoring-opts  --num-threads 1 --skip-scoring false --acwt 0.083333 --nj 6 --cmd run.pl --mem 4G --beam 10.0 --model exp/tri4b/final.alimdl --max-active 2000 exp/tri4b/graph_phone data/mfcc/test_phone exp/tri4b/decode_test_phone.si
decode.sh: feature type is lda
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri4b/graph_phone exp/tri4b/decode_test_phone.si
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_phone.si/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(2,6,17) and mean=7.9
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_phone.si/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test_phone exp/tri4b/graph_phone exp/tri4b/decode_test_phone.si
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmllr.sh: feature type is lda
steps/decode_fmllr.sh: getting first-pass fMLLR transforms.
steps/decode_fmllr.sh: doing main lattice generation phase
steps/decode_fmllr.sh: estimating fMLLR transforms a second time.
steps/decode_fmllr.sh: doing a final pass of acoustic rescoring.
steps/diagnostic/analyze_lats.sh --cmd run.pl --mem 4G exp/tri4b/graph_phone exp/tri4b/decode_test_phone
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_phone/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(2,5,14) and mean=6.9
steps/diagnostic/analyze_lats.sh: see stats in exp/tri4b/decode_test_phone/log/analyze_lattice_depth_stats.log
local/score.sh --cmd run.pl --mem 4G data/mfcc/test_phone exp/tri4b/graph_phone exp/tri4b/decode_test_phone
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0

steps/align_fmllr.sh --nj 6 --cmd run.pl data/mfcc/train data/lang exp/tri4b exp/tri4b_ali
steps/align_fmllr.sh: feature type is lda
steps/align_fmllr.sh: compiling training graphs
steps/align_fmllr.sh: aligning data in data/mfcc/train using exp/tri4b/final.alimdl and speaker-independent features.
steps/align_fmllr.sh: computing fMLLR transforms
steps/align_fmllr.sh: doing final alignment.
steps/align_fmllr.sh: done aligning data.
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/tri4b_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/tri4b_ali/log/analyze_alignments.log
72 warnings in exp/tri4b_ali/log/align_pass2.*.log
4 warnings in exp/tri4b_ali/log/fmllr.*.log
49 warnings in exp/tri4b_ali/log/align_pass1.*.log
steps/align_fmllr.sh --nj 6 --cmd run.pl data/mfcc/dev data/lang exp/tri4b exp/tri4b_ali_cv
steps/align_fmllr.sh: feature type is lda
steps/align_fmllr.sh: compiling training graphs
steps/align_fmllr.sh: aligning data in data/mfcc/dev using exp/tri4b/final.alimdl and speaker-independent features.
steps/align_fmllr.sh: computing fMLLR transforms
steps/align_fmllr.sh: doing final alignment.
steps/align_fmllr.sh: done aligning data.
steps/diagnostic/analyze_alignments.sh --cmd run.pl data/lang exp/tri4b_ali_cv
steps/diagnostic/analyze_alignments.sh: see stats in exp/tri4b_ali_cv/log/analyze_alignments.log
6 warnings in exp/tri4b_ali_cv/log/align_pass2.*.log
5 warnings in exp/tri4b_ali_cv/log/align_pass1.*.log

DNN training: stage 0: feature generation
producing fbank for train
steps/make_fbank.sh --nj 6 --cmd run.pl data/fbank/train exp/make_fbank/train fbank/train
utils/validate_data_dir.sh: Successfully validated data-directory data/fbank/train
steps/make_fbank.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.
Succeeded creating filterbank features for train
steps/compute_cmvn_stats.sh data/fbank/train exp/fbank_cmvn/train fbank/train
Succeeded creating CMVN stats for train
producing fbank for dev
steps/make_fbank.sh --nj 6 --cmd run.pl data/fbank/dev exp/make_fbank/dev fbank/dev
utils/validate_data_dir.sh: Successfully validated data-directory data/fbank/dev
steps/make_fbank.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.
Succeeded creating filterbank features for dev
steps/compute_cmvn_stats.sh data/fbank/dev exp/fbank_cmvn/dev fbank/dev
Succeeded creating CMVN stats for dev
producing fbank for test
steps/make_fbank.sh --nj 6 --cmd run.pl data/fbank/test exp/make_fbank/test fbank/test
utils/validate_data_dir.sh: Successfully validated data-directory data/fbank/test
steps/make_fbank.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.
Succeeded creating filterbank features for test
steps/compute_cmvn_stats.sh data/fbank/test exp/fbank_cmvn/test fbank/test
Succeeded creating CMVN stats for test
producing test_fbank_phone
# INFO
steps/nnet/train.sh : Training Neural Network
dir       : exp/tri4b_dnn 
Train-set : data/fbank/train 10000, exp/tri4b_ali 
CV-set    : data/fbank/dev 893 exp/tri4b_ali_cv 

SKIPPING TRAINING... (steps/nnet/train.sh)
nnet already trained : exp/tri4b_dnn/final.nnet (nnet/nnet_iter14_learnrate7.8125e-06_tr1.0435_cv1.3938_final_)
# Accounting: time=1 threads=1
# Ended (code 0) at Tue Aug  7 21:42:03 CST 2018, elapsed time 1 seconds
# steps/nnet/train.sh --copy_feats false --cmvn-opts "--norm-means=true --norm-vars=false" --hid-layers 4 --hid-dim 1024 --learn-rate 0.008 data/fbank/train data/fbank/dev data/lang exp/tri4b_ali exp/tri4b_ali_cv exp/tri4b_dnn 
# Started at Tue Aug 21 11:10:36 CST 2018
#
steps/nnet/train.sh --copy_feats false --cmvn-opts --norm-means=true --norm-vars=false --hid-layers 4 --hid-dim 1024 --learn-rate 0.008 data/fbank/train data/fbank/dev data/lang exp/tri4b_ali exp/tri4b_ali_cv exp/tri4b_dnn

# INFO
steps/nnet/train.sh : Training Neural Network
dir       : exp/tri4b_dnn 
Train-set : data/fbank/train 10000, exp/tri4b_ali 
CV-set    : data/fbank/dev 893 exp/tri4b_ali_cv 

SKIPPING TRAINING... (steps/nnet/train.sh)
nnet already trained : exp/tri4b_dnn/final.nnet (nnet/nnet_iter14_learnrate7.8125e-06_tr1.0435_cv1.3938_final_)
# Accounting: time=0 threads=1
# Ended (code 0) at Tue Aug 21 11:10:36 CST 2018, elapsed time 0 seconds
steps/nnet/align.sh --nj 6 --cmd run.pl data/fbank/train data/lang exp/tri4b_dnn exp/tri4b_dnn_ali
steps/nnet/decode.sh --nj 6 --cmd run.pl --mem 4G --srcdir exp/tri4b_dnn --config conf/decode_dnn.config --acwt 0.1 exp/tri4b/graph_word data/fbank/test exp/tri4b_dnn/decode_test_word
steps/nnet/decode.sh --nj 6 --cmd run.pl --mem 4G --srcdir exp/tri4b_dnn --config conf/decode_dnn.config --acwt 0.1 exp/tri4b/graph_phone data/fbank/test_phone exp/tri4b_dnn/decode_test_phone
steps/nnet/align.sh: aligning data 'data/fbank/train' using nnet/model 'exp/tri4b_dnn', putting alignments in 'exp/tri4b_dnn_ali'
run.pl: 6 / 6 failed, log is in exp/tri4b_dnn/decode_test_phone/log/decode.*.log
run.pl: 6 / 6 failed, log is in exp/tri4b_dnn/decode_test_word/log/decode.*.log
steps/nnet/align.sh: done aligning data.
steps/nnet/make_denlats.sh --nj 6 --cmd run.pl --mem 4G --config conf/decode_dnn.config --acwt 0.1 data/fbank/train data/lang exp/tri4b_dnn exp/tri4b_dnn_denlats
Making unigram grammar FST in exp/tri4b_dnn_denlats/lang
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
sym2int.pl: replacing = with 2
sym2int.pl: replacing l with 2
....