vagrant@vagrant-kaldi-african:~/FONGBE$ ./run.sh
~/FONGBE/data/local/dict ~/FONGBE
~/FONGBE
utils/prepare_lang.sh data/local/dict <UNK> data/local/lang data/lang
Checking data/local/dict/silence_phones.txt ...
--> reading data/local/dict/silence_phones.txt
--> data/local/dict/silence_phones.txt is OK
Checking data/local/dict/optional_silence.txt ...
--> reading data/local/dict/optional_silence.txt
--> data/local/dict/optional_silence.txt is OK
Checking data/local/dict/nonsilence_phones.txt ...
--> reading data/local/dict/nonsilence_phones.txt
--> data/local/dict/nonsilence_phones.txt is OK
Checking disjoint: silence_phones.txt, nonsilence_phones.txt
--> disjoint property is OK.
Checking data/local/dict/lexicon.txt
--> reading data/local/dict/lexicon.txt
--> data/local/dict/lexicon.txt is OK
Checking data/local/dict/extra_questions.txt ...
--> data/local/dict/extra_questions.txt is empty (this is OK)
--> SUCCESS [validating dictionary directory data/local/dict]
**Creating data/local/dict/lexiconp.txt from data/local/dict/lexicon.txt
fstaddselfloops data/lang/phones/wdisambig_phones.int data/lang/phones/wdisambig_words.int
prepare_lang.sh: validating output directory
utils/validate_lang.pl data/lang
Checking data/lang/phones.txt ...
--> data/lang/phones.txt is OK
Checking words.txt: #0 ...
--> data/lang/words.txt is OK
Checking disjoint: silence.txt, nonsilence.txt, disambig.txt ...
--> silence.txt and nonsilence.txt are disjoint
--> silence.txt and disambig.txt are disjoint
--> disambig.txt and nonsilence.txt are disjoint
--> disjoint property is OK
Checking sumation: silence.txt, nonsilence.txt, disambig.txt ...
--> summation property is OK
Checking data/lang/phones/context_indep.{txt, int, csl} ...
--> 10 entry/entries in data/lang/phones/context_indep.txt
--> data/lang/phones/context_indep.int corresponds to data/lang/phones/context_indep.txt
--> data/lang/phones/context_indep.csl corresponds to data/lang/phones/context_indep.txt
--> data/lang/phones/context_indep.{txt, int, csl} are OK
Checking data/lang/phones/nonsilence.{txt, int, csl} ...
--> 120 entry/entries in data/lang/phones/nonsilence.txt
--> data/lang/phones/nonsilence.int corresponds to data/lang/phones/nonsilence.txt
--> data/lang/phones/nonsilence.csl corresponds to data/lang/phones/nonsilence.txt
--> data/lang/phones/nonsilence.{txt, int, csl} are OK
Checking data/lang/phones/silence.{txt, int, csl} ...
--> 10 entry/entries in data/lang/phones/silence.txt
--> data/lang/phones/silence.int corresponds to data/lang/phones/silence.txt
--> data/lang/phones/silence.csl corresponds to data/lang/phones/silence.txt
--> data/lang/phones/silence.{txt, int, csl} are OK
Checking data/lang/phones/optional_silence.{txt, int, csl} ...
--> 1 entry/entries in data/lang/phones/optional_silence.txt
--> data/lang/phones/optional_silence.int corresponds to data/lang/phones/optional_silence.txt
--> data/lang/phones/optional_silence.csl corresponds to data/lang/phones/optional_silence.txt
--> data/lang/phones/optional_silence.{txt, int, csl} are OK
Checking data/lang/phones/disambig.{txt, int, csl} ...
--> 2 entry/entries in data/lang/phones/disambig.txt
--> data/lang/phones/disambig.int corresponds to data/lang/phones/disambig.txt
--> data/lang/phones/disambig.csl corresponds to data/lang/phones/disambig.txt
--> data/lang/phones/disambig.{txt, int, csl} are OK
Checking data/lang/phones/roots.{txt, int} ...
--> 32 entry/entries in data/lang/phones/roots.txt
--> data/lang/phones/roots.int corresponds to data/lang/phones/roots.txt
--> data/lang/phones/roots.{txt, int} are OK
Checking data/lang/phones/sets.{txt, int} ...
--> 32 entry/entries in data/lang/phones/sets.txt
--> data/lang/phones/sets.int corresponds to data/lang/phones/sets.txt
--> data/lang/phones/sets.{txt, int} are OK
Checking data/lang/phones/extra_questions.{txt, int} ...
--> 9 entry/entries in data/lang/phones/extra_questions.txt
--> data/lang/phones/extra_questions.int corresponds to data/lang/phones/extra_questions.txt
--> data/lang/phones/extra_questions.{txt, int} are OK
Checking data/lang/phones/word_boundary.{txt, int} ...
--> 130 entry/entries in data/lang/phones/word_boundary.txt
--> data/lang/phones/word_boundary.int corresponds to data/lang/phones/word_boundary.txt
--> data/lang/phones/word_boundary.{txt, int} are OK
Checking optional_silence.txt ...
--> reading data/lang/phones/optional_silence.txt
--> data/lang/phones/optional_silence.txt is OK
Checking disambiguation symbols: #0 and #1
--> data/lang/phones/disambig.txt has "#0" and "#1"
--> data/lang/phones/disambig.txt is OK
Checking topo ...
Checking word_boundary.txt: silence.txt, nonsilence.txt, disambig.txt ...
--> data/lang/phones/word_boundary.txt doesn't include disambiguation symbols
--> data/lang/phones/word_boundary.txt is the union of nonsilence.txt and silence.txt
--> data/lang/phones/word_boundary.txt is OK
Checking word-level disambiguation symbols...
--> data/lang/phones/wdisambig.txt exists (newer prepare_lang.sh)
Checking word_boundary.int and disambig.int
--> generating a 9 word sequence
--> resulting phone sequence from L.fst corresponds to the word sequence
--> L.fst is OK
--> generating a 57 word sequence
--> resulting phone sequence from L_disambig.fst corresponds to the word sequence
--> L_disambig.fst is OK
Checking data/lang/oov.{txt, int} ...
--> 1 entry/entries in data/lang/oov.txt
--> data/lang/oov.int corresponds to data/lang/oov.txt
--> data/lang/oov.{txt, int} are OK
--> data/lang/L.fst is olabel sorted
--> data/lang/L_disambig.fst is olabel sorted
--> SUCCESS [validating lang directory data/lang]
arpa2fst -
LOG (arpa2fst:Read():arpa-file-parser.cc:90) Reading \data\ section.
LOG (arpa2fst:Read():arpa-file-parser.cc:145) Reading \1-grams: section.
LOG (arpa2fst:Read():arpa-file-parser.cc:145) Reading \2-grams: section.
LOG (arpa2fst:Read():arpa-file-parser.cc:145) Reading \3-grams: section.
LOG (arpa2fst:RemoveRedundantStates():arpa-lm-compiler.cc:341) Reduced num-states from 103788 to 42224
remove_oovs.pl: removed 42 lines.
steps/make_mfcc.sh --nj 4 data/train data/exp/make_mfcc/train data/train/mfcc
utils/validate_data_dir.sh: Successfully validated data-directory data/train
steps/make_mfcc.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.
Succeeded creating MFCC features for train
steps/make_mfcc.sh --nj 4 data/test data/exp/make_mfcc/test data/test/mfcc
utils/validate_data_dir.sh: Successfully validated data-directory data/test
steps/make_mfcc.sh: [info]: no segments file exists: assuming wav.scp indexed by utterance.
Succeeded creating MFCC features for test
steps/compute_cmvn_stats.sh data/train data/exp/make_mfcc/train data/train/mfcc/
Succeeded creating CMVN stats for train
steps/compute_cmvn_stats.sh data/test data/exp/make_mfcc/test data/test/mfcc/
Succeeded creating CMVN stats for test
steps/train_mono.sh --nj 20 --cmd utils/run.pl data/train data/lang exp/system1/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 3
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 4
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 5
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 6
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 7
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 8
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 9
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 10
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 11
steps/train_mono.sh: Pass 12
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 13
steps/train_mono.sh: Pass 14
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 15
steps/train_mono.sh: Pass 16
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 17
steps/train_mono.sh: Pass 18
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 19
steps/train_mono.sh: Pass 20
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 21
steps/train_mono.sh: Pass 22
steps/train_mono.sh: Pass 23
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 24
steps/train_mono.sh: Pass 25
steps/train_mono.sh: Pass 26
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 27
steps/train_mono.sh: Pass 28
steps/train_mono.sh: Pass 29
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 30
steps/train_mono.sh: Pass 31
steps/train_mono.sh: Pass 32
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 33
steps/train_mono.sh: Pass 34
steps/train_mono.sh: Pass 35
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 36
steps/train_mono.sh: Pass 37
steps/train_mono.sh: Pass 38
steps/train_mono.sh: Aligning data
steps/train_mono.sh: Pass 39
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/mono
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/mono/log/analyze_alignments.log
5355 warnings in exp/system1/mono/log/align.*.*.log
924 warnings in exp/system1/mono/log/acc.*.*.log
exp/system1/mono: nj=20 align prob=-93.76 over 5.67h [retry=1.4%, fail=0.3%] states=100 gauss=1004
steps/train_mono.sh: Done training monophone system in exp/system1/mono
tree-info exp/system1/mono/tree
tree-info exp/system1/mono/tree
fstdeterminizestar --use-log=true
fstminimizeencoded
fsttablecompose data/lang/L_disambig.fst data/lang/G.fst
fstpushspecial
fstisstochastic data/lang/tmp/LG.fst
-0.0682587 -0.0691356
[info]: LG not stochastic.
fstcomposecontext --context-size=1 --central-position=0 --read-disambig-syms=data/lang/phones/disambig.int --write-disambig-syms=data/lang/tmp/disambig_ilabels_1_0.int data/lang/tmp/ilabels_1_0
fstisstochastic data/lang/tmp/CLG_1_0.fst
-0.0682587 -0.0691356
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/system1/mono/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_1_0 exp/system1/mono/tree exp/system1/mono/final.mdl
fsttablecompose exp/system1/mono/graph/Ha.fst data/lang/tmp/CLG_1_0.fst
fstrmepslocal
fstdeterminizestar --use-log=true
fstrmsymbols exp/system1/mono/graph/disambig_tid.int
fstminimizeencoded
fstisstochastic exp/system1/mono/graph/HCLGa.fst
0.000309922 -0.137695
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/mono/final.mdl
steps/decode.sh --nj 4 --cmd utils/run.pl exp/system1/mono/graph data/test exp/system1/mono/decode_test
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/mono/graph exp/system1/mono/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/mono/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,12,101) and mean=36.8
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/mono/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/mono/graph exp/system1/mono/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
Mono training done.
steps/align_si.sh --boost-silence 1.25 --nj 20 --cmd utils/run.pl data/train data/lang exp/system1/mono exp/system1/mono_ali
steps/align_si.sh: feature type is delta
steps/align_si.sh: aligning data in data/train using model from exp/system1/mono, putting alignments in exp/system1/mono_ali
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/mono_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/mono_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.
steps/train_deltas.sh --boost-silence 1.25 --cmd utils/run.pl 3200 30000 data/train data/lang exp/system1/mono_ali exp/system1/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/system1/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 7
steps/train_deltas.sh: training pass 8
steps/train_deltas.sh: training pass 9
steps/train_deltas.sh: training pass 10
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 11
steps/train_deltas.sh: training pass 12
steps/train_deltas.sh: training pass 13
steps/train_deltas.sh: training pass 14
steps/train_deltas.sh: training pass 15
steps/train_deltas.sh: training pass 16
steps/train_deltas.sh: training pass 17
steps/train_deltas.sh: training pass 18
steps/train_deltas.sh: training pass 19
steps/train_deltas.sh: training pass 20
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 21
steps/train_deltas.sh: training pass 22
steps/train_deltas.sh: training pass 23
steps/train_deltas.sh: training pass 24
steps/train_deltas.sh: training pass 25
steps/train_deltas.sh: training pass 26
steps/train_deltas.sh: training pass 27
steps/train_deltas.sh: training pass 28
steps/train_deltas.sh: training pass 29
steps/train_deltas.sh: training pass 30
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 31
steps/train_deltas.sh: training pass 32
steps/train_deltas.sh: training pass 33
steps/train_deltas.sh: training pass 34
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri1
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri1/log/analyze_alignments.log
34 warnings in exp/system1/tri1/log/init_model.log
452 warnings in exp/system1/tri1/log/align.*.*.log
120 warnings in exp/system1/tri1/log/update.*.log
1 warnings in exp/system1/tri1/log/build_tree.log
1004 warnings in exp/system1/tri1/log/acc.*.*.log
exp/system1/tri1: nj=20 align prob=-89.75 over 5.67h [retry=1.0%, fail=0.4%] states=1665 gauss=30082 tree-impr=4.54
steps/train_deltas.sh: Done training system with delta+delta-delta features in exp/system1/tri1
tree-info exp/system1/tri1/tree
tree-info exp/system1/tri1/tree
fstcomposecontext --context-size=3 --central-position=1 --read-disambig-syms=data/lang/phones/disambig.int --write-disambig-syms=data/lang/tmp/disambig_ilabels_3_1.int data/lang/tmp/ilabels_3_1
fstisstochastic data/lang/tmp/CLG_3_1.fst
0 -0.0691356
[info]: CLG not stochastic.
make-h-transducer --disambig-syms-out=exp/system1/tri1/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_3_1 exp/system1/tri1/tree exp/system1/tri1/final.mdl
fsttablecompose exp/system1/tri1/graph/Ha.fst data/lang/tmp/CLG_3_1.fst
fstdeterminizestar --use-log=true
fstminimizeencoded
fstrmepslocal
fstrmsymbols exp/system1/tri1/graph/disambig_tid.int
fstisstochastic exp/system1/tri1/graph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/tri1/final.mdl
steps/decode.sh --nj 4 --cmd utils/run.pl exp/system1/tri1/graph data/test exp/system1/tri1/decode_test
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/tri1/graph exp/system1/tri1/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri1/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,4,24) and mean=9.6
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri1/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri1/graph exp/system1/tri1/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/align_si.sh --nj 20 --cmd utils/run.pl data/train data/lang exp/system1/tri1 exp/system1/tri1_ali
steps/align_si.sh: feature type is delta
steps/align_si.sh: aligning data in data/train using model from exp/system1/tri1, putting alignments in exp/system1/tri1_ali
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri1_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri1_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.
steps/train_deltas.sh --cmd utils/run.pl 3200 30000 data/train data/lang exp/system1/tri1_ali exp/system1/tri2a
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/system1/tri1_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 7
steps/train_deltas.sh: training pass 8
steps/train_deltas.sh: training pass 9
steps/train_deltas.sh: training pass 10
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 11
steps/train_deltas.sh: training pass 12
steps/train_deltas.sh: training pass 13
steps/train_deltas.sh: training pass 14
steps/train_deltas.sh: training pass 15
steps/train_deltas.sh: training pass 16
steps/train_deltas.sh: training pass 17
steps/train_deltas.sh: training pass 18
steps/train_deltas.sh: training pass 19
steps/train_deltas.sh: training pass 20
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 21
steps/train_deltas.sh: training pass 22
steps/train_deltas.sh: training pass 23
steps/train_deltas.sh: training pass 24
steps/train_deltas.sh: training pass 25
steps/train_deltas.sh: training pass 26
steps/train_deltas.sh: training pass 27
steps/train_deltas.sh: training pass 28
steps/train_deltas.sh: training pass 29
steps/train_deltas.sh: training pass 30
steps/train_deltas.sh: aligning data
steps/train_deltas.sh: training pass 31
steps/train_deltas.sh: training pass 32
steps/train_deltas.sh: training pass 33
steps/train_deltas.sh: training pass 34
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri2a
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri2a/log/analyze_alignments.log
1 warnings in exp/system1/tri2a/log/build_tree.log
1138 warnings in exp/system1/tri2a/log/acc.*.*.log
51 warnings in exp/system1/tri2a/log/init_model.log
128 warnings in exp/system1/tri2a/log/update.*.log
462 warnings in exp/system1/tri2a/log/align.*.*.log
exp/system1/tri2a: nj=20 align prob=-89.76 over 5.67h [retry=1.0%, fail=0.4%] states=1785 gauss=30064 tree-impr=4.90
steps/train_deltas.sh: Done training system with delta+delta-delta features in exp/system1/tri2a
tree-info exp/system1/tri2a/tree
tree-info exp/system1/tri2a/tree
make-h-transducer --disambig-syms-out=exp/system1/tri2a/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_3_1 exp/system1/tri2a/tree exp/system1/tri2a/final.mdl
fsttablecompose exp/system1/tri2a/graph/Ha.fst data/lang/tmp/CLG_3_1.fst
fstrmepslocal
fstminimizeencoded
fstrmsymbols exp/system1/tri2a/graph/disambig_tid.int
fstdeterminizestar --use-log=true
fstisstochastic exp/system1/tri2a/graph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/tri2a/final.mdl
steps/decode.sh --nj 4 --cmd utils/run.pl exp/system1/tri2a/graph data/test exp/system1/tri2a/decode_test
decode.sh: feature type is delta
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/tri2a/graph exp/system1/tri2a/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri2a/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,4,23) and mean=9.3
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri2a/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri2a/graph exp/system1/tri2a/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
Triphone training done.
steps/align_si.sh --nj 20 --cmd utils/run.pl data/train data/lang exp/system1/tri2a exp/system1/tri2a_ali
steps/align_si.sh: feature type is delta
steps/align_si.sh: aligning data in data/train using model from exp/system1/tri2a, putting alignments in exp/system1/tri2a_ali
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri2a_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri2a_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.
steps/train_lda_mllt.sh --cmd utils/run.pl --splice-opts --left-context=3 --right-context=3 2000 20000 data/train data/lang exp/system1/tri2a_ali exp/system1/tri2b
Accumulating LDA statistics.
Accumulating tree stats
Getting questions for tree clustering.
Building the tree
steps/train_lda_mllt.sh: Initializing the model
Converting alignments from exp/system1/tri2a_ali to use current tree
Compiling graphs of transcripts
Training pass 1
Training pass 2
Estimating MLLT
Training pass 3
Training pass 4
Estimating MLLT
Training pass 5
Training pass 6
Estimating MLLT
Training pass 7
Training pass 8
Training pass 9
Training pass 10
Aligning data
Training pass 11
Training pass 12
Estimating MLLT
Training pass 13
Training pass 14
Training pass 15
Training pass 16
Training pass 17
Training pass 18
Training pass 19
Training pass 20
Aligning data
Training pass 21
Training pass 22
Training pass 23
Training pass 24
Training pass 25
Training pass 26
Training pass 27
Training pass 28
Training pass 29
Training pass 30
Aligning data
Training pass 31
Training pass 32
Training pass 33
Training pass 34
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri2b
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri2b/log/analyze_alignments.log
1 warnings in exp/system1/tri2b/log/build_tree.log
34 warnings in exp/system1/tri2b/log/lda_acc.*.log
27 warnings in exp/system1/tri2b/log/update.*.log
591 warnings in exp/system1/tri2b/log/align.*.*.log
21 warnings in exp/system1/tri2b/log/init_model.log
1471 warnings in exp/system1/tri2b/log/acc.*.*.log
exp/system1/tri2b: nj=20 align prob=-46.30 over 5.65h [retry=1.4%, fail=0.6%] states=1576 gauss=20031 tree-impr=5.52 lda-sum=17.08 mllt:impr,logdet=1.09,1.68
Done training system with LDA+MLLT features in exp/system1/tri2b
tree-info exp/system1/tri2b/tree
tree-info exp/system1/tri2b/tree
make-h-transducer --disambig-syms-out=exp/system1/tri2b/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_3_1 exp/system1/tri2b/tree exp/system1/tri2b/final.mdl
fstrmsymbols exp/system1/tri2b/graph/disambig_tid.int
fsttablecompose exp/system1/tri2b/graph/Ha.fst data/lang/tmp/CLG_3_1.fst
fstrmepslocal
fstminimizeencoded
fstdeterminizestar --use-log=true
fstisstochastic exp/system1/tri2b/graph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/tri2b/final.mdl
steps/decode.sh --nj 4 --cmd utils/run.pl exp/system1/tri2b/graph data/test exp/system1/tri2b/decode_test
decode.sh: feature type is lda
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/tri2b/graph exp/system1/tri2b/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri2b/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,3,17) and mean=6.9
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri2b/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri2b/graph exp/system1/tri2b/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
LDA+MLLT training done.
steps/align_si.sh --nj 20 --cmd utils/run.pl --use-graphs true data/train data/lang exp/system1/tri2b exp/system1/tri2b_ali
steps/align_si.sh: feature type is lda
steps/align_si.sh: aligning data in data/train using model from exp/system1/tri2b, putting alignments in exp/system1/tri2b_ali
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri2b_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri2b_ali/log/analyze_alignments.log
steps/align_si.sh: done aligning data.
steps/train_sat.sh --cmd utils/run.pl 2000 20000 data/train data/lang exp/system1/tri2b_ali exp/system1/tri3b
steps/train_sat.sh: feature type is lda
steps/train_sat.sh: obtaining initial fMLLR transforms since not present in exp/system1/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_sat.sh: Converting alignments from exp/system1/tri2b_ali to use current tree
steps/train_sat.sh: Compiling graphs of transcripts
Pass 1
Pass 2
Estimating fMLLR transforms
Pass 3
Pass 4
Estimating fMLLR transforms
Pass 5
Pass 6
Estimating fMLLR transforms
Pass 7
Pass 8
Pass 9
Pass 10
Aligning data
Pass 11
Pass 12
Estimating fMLLR transforms
Pass 13
Pass 14
Pass 15
Pass 16
Pass 17
Pass 18
Pass 19
Pass 20
Aligning data
Pass 21
Pass 22
Pass 23
Pass 24
Pass 25
Pass 26
Pass 27
Pass 28
Pass 29
Pass 30
Aligning data
Pass 31
Pass 32
Pass 33
Pass 34
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/tri3b
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri3b/log/analyze_alignments.log
44 warnings in exp/system1/tri3b/log/update.*.log
1898 warnings in exp/system1/tri3b/log/acc.*.*.log
605 warnings in exp/system1/tri3b/log/align.*.*.log
1 warnings in exp/system1/tri3b/log/build_tree.log
2 warnings in exp/system1/tri3b/log/est_alimdl.log
247 warnings in exp/system1/tri3b/log/fmllr.*.*.log
25 warnings in exp/system1/tri3b/log/init_model.log
steps/train_sat.sh: Likelihood evolution:
-51.0015 -51.3744 -50.7316 -50.5481 -49.7994 -49.0101 -48.4451 -48.0401 -47.6986 -47.1501 -46.8402 -46.5549 -46.334 -46.1789 -46.0026 -45.8455 -45.7347 -45.6156 -45.4894 -45.2905 -45.1402 -45.0425 -44.9276 -44.8339 -44.7518 -44.6754 -44.5824 -44.5013 -44.4301 -44.3363 -44.2634 -44.2343 -44.2155 -44.2013
exp/system1/tri3b: nj=20 align prob=-47.34 over 5.64h [retry=1.3%, fail=0.7%] states=1596 gauss=20021 fmllr-impr=2.20 over 3.44h tree-impr=6.80
steps/train_sat.sh: done training SAT system in exp/system1/tri3b
tree-info exp/system1/tri3b/tree
tree-info exp/system1/tri3b/tree
make-h-transducer --disambig-syms-out=exp/system1/tri3b/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_3_1 exp/system1/tri3b/tree exp/system1/tri3b/final.mdl
fstrmepslocal
fstdeterminizestar --use-log=true
fstminimizeencoded
fstrmsymbols exp/system1/tri3b/graph/disambig_tid.int
fsttablecompose exp/system1/tri3b/graph/Ha.fst data/lang/tmp/CLG_3_1.fst
fstisstochastic exp/system1/tri3b/graph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/tri3b/final.mdl
steps/decode_fmllr.sh --nj 4 --cmd utils/run.pl exp/system1/tri3b/graph data/test exp/system1/tri3b/decode_test
steps/decode.sh --scoring-opts --num-threads 1 --skip-scoring false --acwt 0.083333 --nj 4 --cmd utils/run.pl --beam 10.0 --model exp/system1/tri3b/final.alimdl --max-active 2000 exp/system1/tri3b/graph data/test exp/system1/tri3b/decode_test.si
decode.sh: feature type is lda
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/tri3b/graph exp/system1/tri3b/decode_test.si
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b/decode_test.si/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,2,13) and mean=5.6
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b/decode_test.si/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b/decode_test.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 utils/run.pl exp/system1/tri3b/graph exp/system1/tri3b/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,2,10) and mean=4.4
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/align_fmllr.sh --nj 20 --cmd utils/run.pl data/train data/lang exp/system1/tri3b exp/system1/tri3b_ali
steps/align_fmllr.sh: feature type is lda
steps/align_fmllr.sh: compiling training graphs
steps/align_fmllr.sh: aligning data in data/train using exp/system1/tri3b/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 utils/run.pl data/lang exp/system1/tri3b_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/tri3b_ali/log/analyze_alignments.log
53 warnings in exp/system1/tri3b_ali/log/fmllr.*.log
184 warnings in exp/system1/tri3b_ali/log/align_pass2.*.log
189 warnings in exp/system1/tri3b_ali/log/align_pass1.*.log
SAT+FMLLR training done.
steps/make_denlats.sh --nj 20 --cmd utils/run.pl --sub-split 10 --transform-dir exp/system1/tri3b_ali data/train data/lang exp/system1/tri3b exp/system1/tri3b_denlats
Making unigram grammar FST in exp/system1/tri3b_denlats/lang
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing măga with 2
sym2int.pl: replacing ɖĕvo with 2
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing măga with 2
sym2int.pl: replacing ɖĕvo with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing kpɔ́wun with 2
sym2int.pl: replacing εnὲ with 2
sym2int.pl: replacing cakì with 2
sym2int.pl: replacing kpɔ́wun with 2
sym2int.pl: replacing εnὲ with 2
sym2int.pl: replacing ɖɔ̆la with 2
sym2int.pl: replacing ɖokpó with 2
sym2int.pl: replacing ɖokpó with 2
sym2int.pl: not warning for OOVs any more times
** Replaced 3234 instances of OOVs with 2
Compiling decoding graph in exp/system1/tri3b_denlats/dengraph
tree-info exp/system1/tri3b/tree
tree-info exp/system1/tri3b/tree
make-h-transducer --disambig-syms-out=exp/system1/tri3b_denlats/dengraph/disambig_tid.int --transition-scale=1.0 exp/system1/tri3b_denlats/lang/tmp/ilabels_3_1 exp/system1/tri3b/tree exp/system1/tri3b/final.mdl
fstrmepslocal
fsttablecompose exp/system1/tri3b_denlats/dengraph/Ha.fst exp/system1/tri3b_denlats/lang/tmp/CLG_3_1.fst
fstrmsymbols exp/system1/tri3b_denlats/dengraph/disambig_tid.int
fstdeterminizestar --use-log=true
fstminimizeencoded
fstisstochastic exp/system1/tri3b_denlats/dengraph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/tri3b/final.mdl
steps/make_denlats.sh: feature type is lda
steps/make_denlats.sh: using fMLLR transforms from exp/system1/tri3b_ali
Merging archives for data subset 1
Merging archives for data subset 2
Merging archives for data subset 3
Merging archives for data subset 4
Merging archives for data subset 5
Merging archives for data subset 6
Merging archives for data subset 7
Merging archives for data subset 8
Merging archives for data subset 9
Merging archives for data subset 10
Merging archives for data subset 11
Merging archives for data subset 12
Merging archives for data subset 13
Merging archives for data subset 14
Merging archives for data subset 15
Merging archives for data subset 16
Merging archives for data subset 17
Merging archives for data subset 18
Merging archives for data subset 19
Merging archives for data subset 20
steps/make_denlats.sh: done generating denominator lattices.
steps/train_mmi.sh --cmd utils/run.pl --boost 0.1 data/train data/lang exp/system1/tri3b_ali exp/system1/tri3b_denlats exp/system1/tri3b_mmi_b0.1
steps/train_mmi.sh: feature type is lda
Using transforms from exp/system1/tri3b_ali
Iteration 0 of MMI training
Iteration 0: objf was 0.174887362154921, MMI auxf change was 0.00801516819403321
Iteration 1 of MMI training
Iteration 1: objf was 0.191068070647098, MMI auxf change was 0.00684724087381474
Iteration 2 of MMI training
Iteration 2: objf was 0.206427799834619, MMI auxf change was 0.00674956093235381
Iteration 3 of MMI training
Iteration 3: objf was 0.220842957504162, MMI auxf change was 0.00690826008157874
MMI training finished
steps/decode.sh --nj 4 --cmd utils/run.pl --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_mmi_b0.1/decode_test
decode.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/tri3b/graph exp/system1/tri3b_mmi_b0.1/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b_mmi_b0.1/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,3,17) and mean=7.0
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/tri3b_mmi_b0.1/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_mmi_b0.1/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/train_diag_ubm.sh --silence-weight 0.5 --nj 20 --cmd utils/run.pl 600 data/train data/lang exp/system1/tri3b_ali exp/system1/dubm3b
steps/train_diag_ubm.sh: feature type is lda
Using transforms from exp/system1/tri3b_ali
Clustering Gaussians in exp/system1/tri3b_ali/final.mdl
Getting Gaussian-selection info
Training pass 0
Training pass 1
Training pass 2
steps/train_mmi_fmmi.sh --cmd utils/run.pl --boost 0.1 data/train data/lang exp/system1/tri3b_ali exp/system1/dubm3b exp/system1/tri3b_denlats exp/system1/tri3b_fmmi_a
steps/train_mmi_fmmi.sh: feature type is lda
Using transforms from exp/system1/tri3b_ali
Iteration 0: doing fMMI
On iter 0, objf was -0.142425, auxf improvement from fMMI was 0.278964563095091
Iteration 1: doing fMMI
On iter 1, objf was 0.0986513, auxf improvement from fMMI was 0.256152237018287
Iteration 2: doing fMMI
On iter 2, objf was 0.329543, auxf improvement from fMMI was 0.249287084054451
Iteration 3: doing fMMI
On iter 3, objf was 0.547799, auxf improvement from fMMI was 0.277511420187614
Iteration 4: doing MMI (getting stats)...
On iter 4, objf was 0.788283, auxf improvement was 0.545861101583694
Iteration 5: doing MMI (getting stats)...
On iter 5, objf was 1.56397, auxf improvement was 0.313733209260944
Iteration 6: doing MMI (getting stats)...
On iter 6, objf was 2.08395, auxf improvement was 0.363901331941046
Iteration 7: doing MMI (getting stats)...
On iter 7, objf was 2.62278, auxf improvement was 0.447802742616553
Succeeded with 8 iters iterations of MMI+fMMI training (boosting factor = 0.1)
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 3 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it3
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it3
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 4 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it4
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it4
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 5 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it5
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it5
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 6 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it6
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it6
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 7 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it7
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it7
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 8 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_a/decode_test_it8
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_a/decode_test_it8
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/train_mmi_fmmi_indirect.sh --cmd utils/run.pl --boost 0.1 data/train data/lang exp/system1/tri3b_ali exp/system1/dubm3b exp/system1/tri3b_denlats exp/system1/tri3b_fmmi_indirect
steps/train_mmi_fmmi_indirect.sh: feature type is lda
Using transforms from exp/system1/tri3b_ali
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 0: doing fMMI
On iter 0, objf was -0.142425, auxf improvement from fMMI was 0.288030836351898
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 1: doing MMI update
On iter 1, objf was 0.105493, auxf improvement was 0.0745738374241627
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 2: doing fMMI
On iter 2, objf was 0.319395, auxf improvement from fMMI was 0.34651331818023
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 3: doing MMI update
On iter 3, objf was 0.604193, auxf improvement was 0.108471483608223
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 4: doing fMMI
On iter 4, objf was 0.888063, auxf improvement from fMMI was 0.44293990543007
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 5: doing MMI update
On iter 5, objf was 1.24127, auxf improvement was 0.180229579848308
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 6: doing fMMI
On iter 6, objf was 1.64213, auxf improvement from fMMI was 0.797132294071246
Getting MMI stats (needed for fMMI and MMI iterations).
Iteration 7: doing MMI update
On iter 7, objf was 2.10139, auxf improvement was 0.30820605005445
Succeeded with 8 iters iterations of MMI+fMMI training (boosting factor = 0.1)
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 3 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it3
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it3
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 4 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it4
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it4
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 5 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it5
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it5
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 6 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it6
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it6
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 7 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it7
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it7
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_fmmi.sh --nj 4 --cmd utils/run.pl --iter 8 --transform-dir exp/system1/tri3b/decode_test exp/system1/tri3b/graph data/test exp/system1/tri3b_fmmi_indirect/decode_test_it8
decode_fmmi.sh: feature type is lda
Using fMLLR transforms from exp/system1/tri3b/decode_test
local/score.sh --cmd utils/run.pl data/test exp/system1/tri3b/graph exp/system1/tri3b_fmmi_indirect/decode_test_it8
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
fMMI+MMI training done.
steps/train_ubm.sh --cmd utils/run.pl 500 data/train data/lang exp/system1/tri3b_ali exp/system1/ubm5b2
steps/train_ubm.sh: feature type is lda
steps/train_ubm.sh: using transforms from exp/system1/tri3b_ali
steps/train_ubm.sh: clustering model exp/system1/tri3b_ali/final.mdl to get initial UBM
steps/train_ubm.sh: doing Gaussian selection
Pass 0
Pass 1
Pass 2
steps/train_sgmm2.sh --cmd utils/run.pl 5000 12000 data/train data/lang exp/system1/tri3b_ali exp/system1/ubm5b2/final.ubm exp/system1/sgmm2_5b2
steps/train_sgmm2.sh: feature type is lda
steps/train_sgmm2.sh: using transforms from exp/system1/tri3b_ali
steps/train_sgmm2.sh: accumulating tree stats
steps/train_sgmm2.sh: Getting questions for tree clustering.
steps/train_sgmm2.sh: Building the tree
steps/train_sgmm2.sh: Initializing the model
steps/train_sgmm2.sh: doing Gaussian selection
steps/train_sgmm2.sh: compiling training graphs
steps/train_sgmm2.sh: converting alignments
steps/train_sgmm2.sh: training pass 0 ...
steps/train_sgmm2.sh: training pass 1 ...
steps/train_sgmm2.sh: training pass 2 ...
steps/train_sgmm2.sh: training pass 3 ...
steps/train_sgmm2.sh: training pass 4 ...
steps/train_sgmm2.sh: training pass 5 ...
steps/train_sgmm2.sh: re-aligning data
steps/train_sgmm2.sh: training pass 6 ...
steps/train_sgmm2.sh: training pass 7 ...
steps/train_sgmm2.sh: training pass 8 ...
steps/train_sgmm2.sh: training pass 9 ...
steps/train_sgmm2.sh: training pass 10 ...
steps/train_sgmm2.sh: re-aligning data
steps/train_sgmm2.sh: training pass 11 ...
steps/train_sgmm2.sh: training pass 12 ...
steps/train_sgmm2.sh: training pass 13 ...
steps/train_sgmm2.sh: training pass 14 ...
steps/train_sgmm2.sh: training pass 15 ...
steps/train_sgmm2.sh: re-aligning data
steps/train_sgmm2.sh: training pass 16 ...
steps/train_sgmm2.sh: training pass 17 ...
steps/train_sgmm2.sh: training pass 18 ...
steps/train_sgmm2.sh: training pass 19 ...
steps/train_sgmm2.sh: training pass 20 ...
steps/train_sgmm2.sh: training pass 21 ...
steps/train_sgmm2.sh: training pass 22 ...
steps/train_sgmm2.sh: training pass 23 ...
steps/train_sgmm2.sh: training pass 24 ...
steps/train_sgmm2.sh: building alignment model (pass 25)
steps/train_sgmm2.sh: building alignment model (pass 26)
steps/train_sgmm2.sh: building alignment model (pass 27)
905 warnings in exp/system1/sgmm2_5b2/log/acc.*.*.log
111 warnings in exp/system1/sgmm2_5b2/log/update_ali.*.log
1333 warnings in exp/system1/sgmm2_5b2/log/update.*.log
121 warnings in exp/system1/sgmm2_5b2/log/spkvecs.*.*.log
399 warnings in exp/system1/sgmm2_5b2/log/align.*.*.log
186 warnings in exp/system1/sgmm2_5b2/log/acc_ali.*.*.log
1 warnings in exp/system1/sgmm2_5b2/log/build_tree.log
Done
tree-info exp/system1/sgmm2_5b2/tree
tree-info exp/system1/sgmm2_5b2/tree
make-h-transducer --disambig-syms-out=exp/system1/sgmm2_5b2/graph/disambig_tid.int --transition-scale=1.0 data/lang/tmp/ilabels_3_1 exp/system1/sgmm2_5b2/tree exp/system1/sgmm2_5b2/final.mdl
fstrmsymbols exp/system1/sgmm2_5b2/graph/disambig_tid.int
fsttablecompose exp/system1/sgmm2_5b2/graph/Ha.fst data/lang/tmp/CLG_3_1.fst
fstrmepslocal
fstminimizeencoded
fstdeterminizestar --use-log=true
fstisstochastic exp/system1/sgmm2_5b2/graph/HCLGa.fst
0.583089 -0.170404
HCLGa is not stochastic
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/sgmm2_5b2/final.mdl
steps/decode_sgmm2.sh --nj 4 --cmd utils/run.pl --transform-dir exp/system1/tri3b/decode_test exp/system1/sgmm2_5b2/graph data/test exp/system1/sgmm2_5b2/decode_test
steps/decode_sgmm2.sh: feature type is lda
steps/decode_sgmm2.sh: using transforms from exp/system1/tri3b/decode_test
steps/diagnostic/analyze_lats.sh --cmd utils/run.pl exp/system1/sgmm2_5b2/graph exp/system1/sgmm2_5b2/decode_test
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/sgmm2_5b2/decode_test/log/analyze_alignments.log
Overall, lattice depth (10,50,90-percentile)=(1,2,8) and mean=3.7
steps/diagnostic/analyze_lats.sh: see stats in exp/system1/sgmm2_5b2/decode_test/log/analyze_lattice_depth_stats.log
local/score.sh --cmd utils/run.pl data/test exp/system1/sgmm2_5b2/graph exp/system1/sgmm2_5b2/decode_test
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/align_sgmm2.sh --nj 20 --cmd utils/run.pl --transform-dir exp/system1/tri3b_ali --use-graphs true --use-gselect true data/train data/lang exp/system1/sgmm2_5b2 exp/system1/sgmm2_5b2_ali
steps/align_sgmm2.sh: feature type is lda
steps/align_sgmm2.sh: using transforms from exp/system1/tri3b_ali
steps/align_sgmm2.sh: aligning data in data/train using model exp/system1/sgmm2_5b2/final.alimdl
steps/align_sgmm2.sh: computing speaker vectors (1st pass)
steps/align_sgmm2.sh: computing speaker vectors (2nd pass)
steps/align_sgmm2.sh: doing final alignment.
steps/align_sgmm2.sh: done aligning data.
steps/diagnostic/analyze_alignments.sh --cmd utils/run.pl data/lang exp/system1/sgmm2_5b2_ali
steps/diagnostic/analyze_alignments.sh: see stats in exp/system1/sgmm2_5b2_ali/log/analyze_alignments.log
101 warnings in exp/system1/sgmm2_5b2_ali/log/align_pass2.*.log
101 warnings in exp/system1/sgmm2_5b2_ali/log/align_pass1.*.log
30 warnings in exp/system1/sgmm2_5b2_ali/log/spk_vecs2.*.log
60 warnings in exp/system1/sgmm2_5b2_ali/log/spk_vecs1.*.log
steps/make_denlats_sgmm2.sh --nj 20 --cmd utils/run.pl --sub-split 10 --transform-dir exp/system1/tri3b_ali data/train data/lang exp/system1/sgmm2_5b2_ali exp/system1/sgmm2_5b2_denlats
steps/make_denlats_sgmm2.sh: Making unigram grammar FST in exp/system1/sgmm2_5b2_denlats/lang
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing măga with 2
sym2int.pl: replacing ɖĕvo with 2
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing kpoŏ with 2
sym2int.pl: replacing măga with 2
sym2int.pl: replacing ɖĕvo with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing ă with 2
sym2int.pl: replacing kpɔ́wun with 2
sym2int.pl: replacing εnὲ with 2
sym2int.pl: replacing cakì with 2
sym2int.pl: replacing kpɔ́wun with 2
sym2int.pl: replacing εnὲ with 2
sym2int.pl: replacing ɖɔ̆la with 2
sym2int.pl: replacing ɖokpó with 2
sym2int.pl: replacing ɖokpó with 2
sym2int.pl: not warning for OOVs any more times
** Replaced 3234 instances of OOVs with 2
steps/make_denlats_sgmm2.sh: Compiling decoding graph in exp/system1/sgmm2_5b2_denlats/dengraph
tree-info exp/system1/sgmm2_5b2_ali/tree
tree-info exp/system1/sgmm2_5b2_ali/tree
fsttablecompose exp/system1/sgmm2_5b2_denlats/lang/L_disambig.fst exp/system1/sgmm2_5b2_denlats/lang/G.fst
fstdeterminizestar --use-log=true
fstminimizeencoded
fstpushspecial
fstisstochastic exp/system1/sgmm2_5b2_denlats/lang/tmp/LG.fst
8.76188e-05 -3.51126e-05
fstcomposecontext --context-size=3 --central-position=1 --read-disambig-syms=exp/system1/sgmm2_5b2_denlats/lang/phones/disambig.int --write-disambig-syms=exp/system1/sgmm2_5b2_denlats/lang/tmp/disambig_ilabels_3_1.int exp/system1/sgmm2_5b2_denlats/lang/tmp/ilabels_3_1
fstisstochastic exp/system1/sgmm2_5b2_denlats/lang/tmp/CLG_3_1.fst
8.76188e-05 -3.51126e-05
make-h-transducer --disambig-syms-out=exp/system1/sgmm2_5b2_denlats/dengraph/disambig_tid.int --transition-scale=1.0 exp/system1/sgmm2_5b2_denlats/lang/tmp/ilabels_3_1 exp/system1/sgmm2_5b2_ali/tree exp/system1/sgmm2_5b2_ali/final.mdl
fsttablecompose exp/system1/sgmm2_5b2_denlats/dengraph/Ha.fst exp/system1/sgmm2_5b2_denlats/lang/tmp/CLG_3_1.fst
fstrmsymbols exp/system1/sgmm2_5b2_denlats/dengraph/disambig_tid.int
fstdeterminizestar --use-log=true
fstrmepslocal
fstminimizeencoded
fstisstochastic exp/system1/sgmm2_5b2_denlats/dengraph/HCLGa.fst
0.000519499 -0.000478178
add-self-loops --self-loop-scale=0.1 --reorder=true exp/system1/sgmm2_5b2_ali/final.mdl
steps/make_denlats_sgmm2.sh: feature type is lda
steps/make_denlats_sgmm2.sh: using fMLLR transforms from exp/system1/tri3b_ali
steps/make_denlats_sgmm2.sh: Merging archives for data subset 1
steps/make_denlats_sgmm2.sh: Merging archives for data subset 2
steps/make_denlats_sgmm2.sh: Merging archives for data subset 3
steps/make_denlats_sgmm2.sh: Merging archives for data subset 4
steps/make_denlats_sgmm2.sh: Merging archives for data subset 5
steps/make_denlats_sgmm2.sh: Merging archives for data subset 6
steps/make_denlats_sgmm2.sh: Merging archives for data subset 7
steps/make_denlats_sgmm2.sh: Merging archives for data subset 8
steps/make_denlats_sgmm2.sh: Merging archives for data subset 9
steps/make_denlats_sgmm2.sh: Merging archives for data subset 10
steps/make_denlats_sgmm2.sh: Merging archives for data subset 11
steps/make_denlats_sgmm2.sh: Merging archives for data subset 12
steps/make_denlats_sgmm2.sh: Merging archives for data subset 13
steps/make_denlats_sgmm2.sh: Merging archives for data subset 14
steps/make_denlats_sgmm2.sh: Merging archives for data subset 15
steps/make_denlats_sgmm2.sh: Merging archives for data subset 16
steps/make_denlats_sgmm2.sh: Merging archives for data subset 17
steps/make_denlats_sgmm2.sh: Merging archives for data subset 18
steps/make_denlats_sgmm2.sh: Merging archives for data subset 19
steps/make_denlats_sgmm2.sh: Merging archives for data subset 20
steps/make_denlats_sgmm2.sh: done generating denominator lattices with SGMMs.
steps/train_mmi_sgmm2.sh --cmd utils/run.pl --transform-dir exp/system1/tri3b_ali --boost 0.1 data/train data/lang exp/system1/sgmm2_5b2_ali exp/system1/sgmm2_5b2_denlats exp/system1/sgmm2_5b2_mmi_b0.1
steps/train_mmi_sgmm2.sh: feature type is lda
steps/train_mmi_sgmm2.sh: using transforms from exp/system1/tri3b_ali
steps/train_mmi_sgmm2.sh: using speaker vectors from exp/system1/sgmm2_5b2_ali
steps/train_mmi_sgmm2.sh: using Gaussian-selection info from exp/system1/sgmm2_5b2_ali
Iteration 0 of MMI training
Iteration 0: objf was 0.174663745182539, MMI auxf change was 0.0112395846962723
Iteration 1 of MMI training
Iteration 1: objf was 0.188849190646458, MMI auxf change was 0.00563243552104329
Iteration 2 of MMI training
Iteration 2: objf was 0.198712244870434, MMI auxf change was 0.00557044477016133
Iteration 3 of MMI training
Iteration 3: objf was 0.209253941667566, MMI auxf change was 0.00659963824844031
MMI training finished
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 1 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it1
steps/decode_sgmm2_rescore.sh: using speaker vectors from exp/system1/sgmm2_5b2/decode_test
steps/decode_sgmm2_rescore.sh: feature type is lda
steps/decode_sgmm2_rescore.sh: using transforms from exp/system1/tri3b/decode_test
steps/decode_sgmm2_rescore.sh: rescoring lattices with SGMM model in exp/system1/sgmm2_5b2_mmi_b0.1/1.mdl
local/score.sh --cmd utils/run.pl data/test data/lang exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it1
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 2 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it2
steps/decode_sgmm2_rescore.sh: using speaker vectors from exp/system1/sgmm2_5b2/decode_test
steps/decode_sgmm2_rescore.sh: feature type is lda
steps/decode_sgmm2_rescore.sh: using transforms from exp/system1/tri3b/decode_test
steps/decode_sgmm2_rescore.sh: rescoring lattices with SGMM model in exp/system1/sgmm2_5b2_mmi_b0.1/2.mdl
local/score.sh --cmd utils/run.pl data/test data/lang exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it2
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 3 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it3
steps/decode_sgmm2_rescore.sh: using speaker vectors from exp/system1/sgmm2_5b2/decode_test
steps/decode_sgmm2_rescore.sh: feature type is lda
steps/decode_sgmm2_rescore.sh: using transforms from exp/system1/tri3b/decode_test
steps/decode_sgmm2_rescore.sh: rescoring lattices with SGMM model in exp/system1/sgmm2_5b2_mmi_b0.1/3.mdl
local/score.sh --cmd utils/run.pl data/test data/lang exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it3
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 4 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it4
steps/decode_sgmm2_rescore.sh: using speaker vectors from exp/system1/sgmm2_5b2/decode_test
steps/decode_sgmm2_rescore.sh: feature type is lda
steps/decode_sgmm2_rescore.sh: using transforms from exp/system1/tri3b/decode_test
steps/decode_sgmm2_rescore.sh: rescoring lattices with SGMM model in exp/system1/sgmm2_5b2_mmi_b0.1/4.mdl
local/score.sh --cmd utils/run.pl data/test data/lang exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it4
local/score.sh: scoring with word insertion penalty=0.0,0.5,1.0
steps/train_mmi_sgmm2.sh --cmd utils/run.pl --transform-dir exp/tri3b_ali --boost 0.1 --zero-if-disjoint true data/train data/lang exp/system1/sgmm2_5b2_ali exp/system1/sgmm2_5b2_denlats exp/system1/sgmm2_5b2_mmi_b0.1_z
steps/train_mmi_sgmm2.sh: invalid option --zero-if-disjoint
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 1 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1_z/decode_test_it1
steps/decode_sgmm2_rescore.sh: no such file exp/system1/sgmm2_5b2_mmi_b0.1_z/1.mdl
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 2 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1_z/decode_test_it2
steps/decode_sgmm2_rescore.sh: no such file exp/system1/sgmm2_5b2_mmi_b0.1_z/2.mdl
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 3 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1_z/decode_test_it3
steps/decode_sgmm2_rescore.sh: no such file exp/system1/sgmm2_5b2_mmi_b0.1_z/3.mdl
steps/decode_sgmm2_rescore.sh --cmd utils/run.pl --iter 4 --transform-dir exp/system1/tri3b/decode_test data/lang data/test exp/system1/sgmm2_5b2/decode_test exp/system1/sgmm2_5b2_mmi_b0.1_z/decode_test_it4
steps/decode_sgmm2_rescore.sh: no such file exp/system1/sgmm2_5b2_mmi_b0.1_z/4.mdl
SGMM training done.
%WER 18.33 [ 2119 / 11562, 292 ins, 425 del, 1402 sub ] exp/system1/combine_tri3b_fmmi_indirect_sgmm2_5b2_mmi_b0.1/decode_test_it8_3/wer_16
%WER 35.18 [ 4068 / 11562, 372 ins, 883 del, 2813 sub ] exp/system1/mono/decode_test/wer_8_0.0
%WER 16.68 [ 1929 / 11562, 260 ins, 392 del, 1277 sub ] exp/system1/sgmm2_5b2/decode_test/wer_13_0.5
%WER 17.66 [ 2042 / 11562, 265 ins, 421 del, 1356 sub ] exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it1/wer_16_0.5
%WER 18.38 [ 2125 / 11562, 249 ins, 455 del, 1421 sub ] exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it2/wer_16_1.0
%WER 19.49 [ 2253 / 11562, 281 ins, 451 del, 1521 sub ] exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it3/wer_17_1.0
%WER 19.45 [ 2249 / 11562, 373 ins, 366 del, 1510 sub ] exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it3.mbr/wer_17
%WER 20.29 [ 2346 / 11562, 321 ins, 437 del, 1588 sub ] exp/system1/sgmm2_5b2_mmi_b0.1/decode_test_it4/wer_17_1.0
%WER 27.34 [ 3161 / 11562, 440 ins, 515 del, 2206 sub ] exp/system1/tri1/decode_test/wer_15_0.0
%WER 26.74 [ 3092 / 11562, 401 ins, 502 del, 2189 sub ] exp/system1/tri2a/decode_test/wer_16_0.0
%WER 22.35 [ 2584 / 11562, 353 ins, 487 del, 1744 sub ] exp/system1/tri2b/decode_test/wer_17_0.0
%WER 17.85 [ 2064 / 11562, 260 ins, 418 del, 1386 sub ] exp/system1/tri3b/decode_test/wer_15_0.5
%WER 23.46 [ 2713 / 11562, 368 ins, 444 del, 1901 sub ] exp/system1/tri3b/decode_test.si/wer_16_0.0
%WER 28.70 [ 3318 / 11562, 309 ins, 885 del, 2124 sub ] exp/system1/tri3b_fmmi_a/decode_test_it3/wer_15_0.0
%WER 42.27 [ 4887 / 11562, 477 ins, 1423 del, 2987 sub ] exp/system1/tri3b_fmmi_a/decode_test_it4/wer_13_0.0
%WER 61.32 [ 7090 / 11562, 352 ins, 3759 del, 2979 sub ] exp/system1/tri3b_fmmi_a/decode_test_it5/wer_12_0.0
%WER 75.00 [ 8671 / 11562, 419 ins, 4738 del, 3514 sub ] exp/system1/tri3b_fmmi_a/decode_test_it6/wer_10_0.0
%WER 81.44 [ 9416 / 11562, 355 ins, 5256 del, 3805 sub ] exp/system1/tri3b_fmmi_a/decode_test_it7/wer_12_0.0
%WER 88.00 [ 10175 / 11562, 273 ins, 6076 del, 3826 sub ] exp/system1/tri3b_fmmi_a/decode_test_it8/wer_12_0.0
%WER 23.34 [ 2698 / 11562, 322 ins, 586 del, 1790 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it3/wer_16_0.0
%WER 27.92 [ 3228 / 11562, 401 ins, 697 del, 2130 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it4/wer_16_0.0
%WER 31.33 [ 3622 / 11562, 511 ins, 703 del, 2408 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it5/wer_14_0.0
%WER 37.48 [ 4333 / 11562, 595 ins, 948 del, 2790 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it6/wer_15_0.0
%WER 39.40 [ 4556 / 11562, 583 ins, 998 del, 2975 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it7/wer_16_0.0
%WER 46.11 [ 5331 / 11562, 691 ins, 1254 del, 3386 sub ] exp/system1/tri3b_fmmi_indirect/decode_test_it8/wer_17_0.0
%WER 26.94 [ 3115 / 11562, 435 ins, 632 del, 2048 sub ] exp/system1/tri3b_mmi_b0.1/decode_test/wer_17_0.0
vagrant@vagrant-kaldi-african:~/FONGBE$