Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
71 commits
Select commit Hold shift + click to select a range
05f7fbd
Fixed OpenFST building on Windows/cygwin64 (#1182)
binghe Nov 11, 2016
49eb733
Change Travis build to use shared libraries to avoid 'no space left o…
danpovey Nov 12, 2016
a2f5464
Adding the reverberation script to AMI (#1178)
tomkocse Nov 12, 2016
d8382ff
Fix to utils/lang/make_unk_lm.sh, thanks to Xingyu Na for finding the…
danpovey Nov 12, 2016
463ef75
WIP: BackpropTruncationComponent (#1126)
freewym Nov 13, 2016
f8df6ff
add results
emc2lie Nov 14, 2016
e38fc92
add results (#1188)
jtrmal Nov 14, 2016
af2386b
adding fix where trailing whitespace will be removed
jtrmal Nov 14, 2016
05104b6
check for swig and produce meaningful info
jtrmal Nov 14, 2016
288890c
check for swig and produce meaningful info (#1192)
danpovey Nov 14, 2016
aca60d8
improving the check
jtrmal Nov 15, 2016
7a42231
style fixes
jtrmal Nov 15, 2016
465b73c
Fix nnet3 endpointing to correctly use frame subsampling factor (#1184)
Nov 15, 2016
91233b8
Merge pull request #1191 from jtrmal/arpa2fst_fix
danpovey Nov 16, 2016
d1aca1a
Change FrameSubsamplingFactor() to non-virtual
Nov 16, 2016
f66e83e
Merge pull request #1196 from alumae/chain-online-endpointing-fix
danpovey Nov 16, 2016
6b30404
Added changes to ensure that the correct left/right contexts are used…
vijayaditya Nov 16, 2016
0482e82
fixed max_deriv_time unset issue for BLSTM
freewym Nov 1, 2016
52fabe5
change {left|right}_deriv_truncate to {min|max}_deriv_time in argumen…
freewym Nov 17, 2016
5874bc4
Merge pull request #1165 from freewym/max_deriv_time
danpovey Nov 17, 2016
c093b7d
Added (B)LSTM scripts for ami/s5b and tedlium/s5_r2 (#1198)
freewym Nov 17, 2016
8a7e3c8
Remove unused config variable in utils/prepare_lang.sh
danpovey Nov 21, 2016
60e2fdd
add self_loop_pdf_class to HmmState
Oct 10, 2016
2bd825c
modify transition-model and context-dep interface
naxingyu Oct 17, 2016
d960f73
add 2nd version of GetPdfInfo
naxingyu Oct 18, 2016
dd48170
resolve hmm-utils fails
naxingyu Oct 19, 2016
ad0d269
add test code for new HmmTopology
naxingyu Oct 24, 2016
46c9a26
Add IsHmm and RM validation
naxingyu Oct 25, 2016
bb1d1c3
modify chain code and scripts for new TransitionModel
naxingyu Oct 26, 2016
5d3d6d7
Enumerate possible pdf pairs
naxingyu Nov 11, 2016
33957ab
modify chain code for new transition model
naxingyu Nov 14, 2016
168957c
update comments for new transition model
naxingyu Nov 14, 2016
86234e5
modify HmmTopology and TransitionModel documents
naxingyu Nov 15, 2016
501cc11
Various fixes; add some testing code and scripts. Note: Makefile cha…
danpovey Nov 16, 2016
3746143
renaming pdf classes and ids, update RM base
naxingyu Nov 18, 2016
7788327
update RM results and HMM document
naxingyu Nov 21, 2016
ac1f932
Fix to nnet3-chain-copy-egs.cc to do frame-shifting before limiting c…
danpovey Nov 21, 2016
620e516
update EnumeratePairs for more complete set of pdfs
naxingyu Nov 21, 2016
b224814
discriminative: Fixed script bug
vimalmanohar Nov 21, 2016
07a5d51
Xconfigs : extension (#1197)
vijayaditya Nov 21, 2016
6200255
chain-train : corrected a minor bug where the chain/train.py was chec…
vijayaditya Nov 21, 2016
6549009
Merge pull request #1202 from vimalmanohar/master
danpovey Nov 21, 2016
45d53f1
Merge pull request #1105 from naxingyu/modify-transition-model
danpovey Nov 21, 2016
1a8a3ad
Some small cosmetic updates regarding last commit [chain transition m…
danpovey Nov 21, 2016
2e1974b
Very small fix to voxforge_prepare_dict.sh to avoid lexicon/lexiconp …
danpovey Nov 22, 2016
a3d4430
swbd : (#1205)
vijayaditya Nov 23, 2016
ab308f8
swbd: updated TDNN+LSTM recipe. Interleaving these layers provides go…
vijayaditya Nov 23, 2016
bfe3159
Fix to problem encountered by Vincent Nguyen when using multi-fst inp…
danpovey Nov 23, 2016
a5bd0ef
Remove compute capability 1.3 from the list of supported capabilities…
danpovey Nov 24, 2016
7346471
Modifications to align-text.cc and create_segments_from_ctm.pl to all…
danpovey Nov 25, 2016
c364071
Fix to compilation problem on MacOS X Sierra by removing use of signal.h
danpovey Nov 25, 2016
d8493b6
Small change to nnet3 compilation so that for updatable components wi…
danpovey Nov 25, 2016
cc413a6
adding low-rank module in xent-branch (#1213)
GaofengCheng Nov 26, 2016
0cbc041
use native double atomicAdd for CC6.0 and devices; fix bugs;
kangshiyin Nov 26, 2016
0dfaa77
Fix problem encountered by alexandre nanchen, where LG determinizatio…
danpovey Nov 26, 2016
5ead2c0
Merge pull request #1218 from kangshiyin/double-precision
danpovey Nov 26, 2016
f447910
new kernel: norm per row
kangshiyin Nov 26, 2016
c887a51
speed test and unit test for normalize per row
kangshiyin Jun 1, 2016
b878fdb
some scripts fixes in blstm_6j and tdnn_lstm recipes (#1220)
freewym Nov 28, 2016
a5e2a04
Look in right location for new style subdirectories
Nov 28, 2016
2266cf2
Merge pull request #1217 from kangshiyin/normalize-prop
danpovey Nov 28, 2016
7b47ebe
Use correct subdirectories in Fisher English recipes
Nov 28, 2016
2bd1c4a
Merge pull request #1223 from aevernon/master
danpovey Nov 28, 2016
b710d78
WIP : original LM from TedliumRelease2 (#1164)
vince62s Nov 28, 2016
e95aeee
changed the definition of deriv-truncate-margin option. Now the margi…
freewym Nov 29, 2016
87465f5
Fixes to CUDA problems introduced by recent commits: fix normalize co…
danpovey Nov 30, 2016
fa64e62
fix assert; use fmax overloading that has been documented. (#1230)
kangshiyin Nov 30, 2016
4435248
ami : added TDNN+LSTM scripts for AMI. (#1212)
vijayaditya Nov 30, 2016
5f51486
ami/TDNN+LSTM: changed left-deriv-truncate to deriv-truncate-margin (…
vijayaditya Nov 30, 2016
335ae0b
Reduce wget num-tries and timeout for http://openfst.cs.nyu.edu/ to f…
danpovey Dec 1, 2016
03e6b92
Lexicon learning (#976)
xiaohui-zhang Dec 1, 2016
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion egs/ami/s5/local/chain/run_blstm_ami_5.sh
Original file line number Diff line number Diff line change
Expand Up @@ -118,7 +118,6 @@ if [ $stage -le 17 ]; then
--chain.l2-regularize 0.00005 \
--chain.apply-deriv-weights false \
--chain.lm-opts="--num-extra-lm-states=2000" \
--chain.left-deriv-truncate 0 \
--trainer.num-chunk-per-minibatch 128 \
--trainer.frames-per-iter 1200000 \
--trainer.max-param-change 2.0 \
Expand All @@ -129,6 +128,7 @@ if [ $stage -le 17 ]; then
--trainer.optimization.initial-effective-lrate 0.001 \
--trainer.optimization.final-effective-lrate 0.0001 \
--trainer.optimization.momentum 0.0 \
--trainer.deriv-truncate-margin 8 \
--egs.stage $get_egs_stage \
--egs.opts "--frames-overlap-per-eg 0" \
--egs.chunk-width $chunk_width \
Expand Down
19 changes: 19 additions & 0 deletions egs/ami/s5b/RESULTS_ihm
Original file line number Diff line number Diff line change
Expand Up @@ -48,6 +48,13 @@
%WER 24.2 | 13098 94477 | 79.3 12.2 8.6 3.5 24.2 57.1 | -0.178 | exp/ihm/nnet3/tdnn_sp/decode_dev/ascore_11/dev_hires.ctm.filt.sys
%WER 25.4 | 12643 89970 | 77.6 13.7 8.7 3.0 25.4 56.3 | -0.067 | exp/ihm/nnet3/tdnn_sp/decode_eval/ascore_12/eval_hires.ctm.filt.sys

# local/nnet3/run_blstm.sh --mic ihm
# nnet3 xent BLSTM with data cleaning
# for d in exp/ihm/nnet3_cleaned/lstm_bidirectional_sp/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
# Note: the results are with ClipGradientComponent, which may be different from with BackpropTruncationComponent
%WER 22.3 | 13098 94494 | 80.9 11.7 7.4 3.2 22.3 55.7 | -0.618 | exp/ihm/nnet3_cleaned/lstm_bidirectional_sp/decode_dev/ascore_10/dev_hires.ctm.filt.sys
%WER 22.5 | 12643 89962 | 80.2 12.7 7.1 2.7 22.5 53.4 | -0.476 | exp/ihm/nnet3_cleaned/lstm_bidirectional_sp/decode_eval/ascore_10/eval_hires.ctm.filt.sys

############################################

# local/chain/run_tdnn.sh --mic ihm --stage 12 &
Expand All @@ -62,3 +69,15 @@
for d in exp/ihm/chain/tdnn_sp_bi/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 22.4 | 13098 94476 | 80.4 10.4 9.2 2.8 22.4 54.6 | 0.069 | exp/ihm/chain/tdnn_sp_bi/decode_dev/ascore_10/dev_hires.ctm.filt.sys
%WER 22.5 | 12643 89974 | 80.0 12.1 7.9 2.6 22.5 52.8 | 0.157 | exp/ihm/chain/tdnn_sp_bi/decode_eval/ascore_10/eval_hires.ctm.filt.sys

# local/chain/multi_condition/run_tdnn.sh --mic ihm &
# cleanup + chain TDNN model + IHM reverberated data
# for d in exp/ihm/chain_cleaned_rvb/tdnn_sp_bi/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 21.5 | 13098 94486 | 81.8 11.0 7.2 3.3 21.5 54.6 | 0.090 | exp/ihm/chain_cleaned_rvb/tdnn_sp_rvb_bi/decode_dev/ascore_10/dev_hires.ctm.filt.sys
%WER 21.9 | 12643 89985 | 80.8 12.3 6.9 2.7 21.9 52.5 | 0.183 | exp/ihm/chain_cleaned_rvb/tdnn_sp_rvb_bi/decode_eval/ascore_10/eval_hires.ctm.filt.sys


# local/chain/tuning/run_tdnn_lstm_1i.sh --mic ihm --train-set train_cleaned --gmm tri3_cleaned
# cleanup + chain TDNN+LSTM model
%WER 20.8 | 13098 94489 | 82.0 10.0 8.0 2.8 20.8 53.2 | -0.096 | exp/ihm/chain_cleaned/tdnn_lstm1i_sp_bi_ld5/decode_dev/ascore_11/dev_hires.ctm.filt.sys
%WER 20.7 | 12643 89980 | 81.7 11.5 6.8 2.5 20.7 51.8 | 0.015 | exp/ihm/chain_cleaned/tdnn_lstm1i_sp_bi_ld5/decode_eval/ascore_11/eval_hires.ctm.filt.sys
12 changes: 11 additions & 1 deletion egs/ami/s5b/RESULTS_mdm
Original file line number Diff line number Diff line change
Expand Up @@ -65,7 +65,6 @@

# cleanup + chain TDNN model, alignments from IHM data (IHM alignments help).
# local/chain/run_tdnn.sh --mic mdm8 --use-ihm-ali true --stage 12 &
# *** best system ***
# for d in exp/mdm8/chain_cleaned/tdnn_sp_bi_ihmali/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 37.4 | 15286 94509 | 66.6 18.0 15.5 3.9 37.4 62.8 | 0.624 | exp/mdm8/chain_cleaned/tdnn_sp_bi_ihmali/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
%WER 40.6 | 13381 89982 | 62.7 18.9 18.3 3.3 40.6 67.6 | 0.594 | exp/mdm8/chain_cleaned/tdnn_sp_bi_ihmali/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys
Expand All @@ -78,4 +77,15 @@
%WER 37.9 | 15635 94514 | 66.5 19.1 14.4 4.4 37.9 61.2 | 0.646 | exp/mdm8/chain/tdnn_sp_bi_ihmali/decode_dev/ascore_8/dev_hires_o4.ctm.filt.sys
%WER 41.5 | 13884 89975 | 62.3 20.3 17.4 3.8 41.5 66.0 | 0.621 | exp/mdm8/chain/tdnn_sp_bi_ihmali/decode_eval/ascore_8/eval_hires_o4.ctm.filt.sys

# local/chain/multi_condition/run_tdnn.sh --mic mdm8 --use-ihm-ali true --train-set train_cleaned --gmm tri3_cleaned
# cleanup + chain TDNN model, MDM original + IHM reverberated data, alignments from IHM data
# for d in exp/mdm8/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 35.8 | 14512 94498 | 68.2 17.2 14.6 4.0 35.8 64.9 | 0.632 | exp/mdm8/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
%WER 39.1 | 13651 89967 | 64.3 18.4 17.3 3.3 39.1 65.2 | 0.607 | exp/mdm8/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys


# local/chain/tuning/run_tdnn_lstm_1i.sh --mic mdm8 --use-ihm-ali true --train-set train_cleaned --gmm tri3_cleaned
# cleanup + chain TDNN+LSTM model, MDM audio and alignments from IHM data
# *** best system ***
%WER 34.6 | 15116 94508 | 69.6 17.6 12.9 4.1 34.6 62.3 | 0.687 | exp/mdm8/chain_cleaned/tdnn_lstm1i_sp_bi_ihmali_ld5/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
%WER 37.1 | 14343 90002 | 66.3 18.8 14.9 3.4 37.1 62.3 | 0.659 | exp/mdm8/chain_cleaned/tdnn_lstm1i_sp_bi_ihmali_ld5/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys
19 changes: 18 additions & 1 deletion egs/ami/s5b/RESULTS_sdm
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,12 @@
%WER 41.6 | 14493 94516 | 63.3 23.5 13.2 4.9 41.6 66.8 | 0.639 | exp/sdm1/nnet3/tdnn_sp_ihmali/decode_dev/ascore_13/dev_hires_o4.ctm.filt.sys
%WER 46.0 | 13597 89967 | 57.5 24.9 17.6 3.6 46.0 68.1 | 0.601 | exp/sdm1/nnet3/tdnn_sp_ihmali/decode_eval/ascore_14/eval_hires_o4.ctm.filt.sys

# xent BLSTM system; cleaned data and IHM alignments.
# local/nnet3/run_blstm.sh --mic sdm1 --use-ihm-ali true
# for d in exp/sdm1/nnet3_cleaned/lstm_bidirectional_sp/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
# Note: the results are with ClipGradientComponent, which may be different from with BackpropTruncationComponent
%WER 37.8 | 14633 94518 | 67.1 22.3 10.7 4.9 37.8 64.2 | 0.745 | exp/sdm1/nnet3_cleaned/lstm_bidirectional_sp_ihmali/decode_dev/ascore_11/dev_hires_o4.ctm.filt.sys
%WER 41.4 | 13809 89628 | 62.7 24.1 13.2 4.1 41.4 65.2 | 0.723 | exp/sdm1/nnet3_cleaned/lstm_bidirectional_sp_ihmali/decode_eval/ascore_11/eval_hires_o4.ctm.filt.sys

# =========================

Expand All @@ -62,7 +68,6 @@
# cleanup + chain TDNN model, alignments from IHM data (IHM alignments help).
# local/chain/run_tdnn.sh --mic sdm1 --use-ihm-ali true --stage 12 &
# cleanup + chain TDNN model, cleaned data and alignments from ihm data.
# *** best system ***
# for d in exp/sdm1/chain_cleaned/tdnn_sp_bi_ihmali/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 40.7 | 14321 94501 | 63.0 19.6 17.4 3.7 40.7 67.7 | 0.592 | exp/sdm1/chain_cleaned/tdnn_sp_bi_ihmali/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
%WER 44.8 | 14293 89976 | 58.6 21.3 20.1 3.3 44.8 64.2 | 0.559 | exp/sdm1/chain_cleaned/tdnn_sp_bi_ihmali/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys
Expand All @@ -75,4 +80,16 @@
%WER 40.7 | 14549 94520 | 63.6 21.4 15.0 4.3 40.7 66.2 | 0.617 | exp/sdm1/chain/tdnn_sp_bi_ihmali/decode_dev/ascore_8/dev_hires_o4.ctm.filt.sys
%WER 45.1 | 13296 89971 | 59.1 23.4 17.6 4.2 45.1 69.5 | 0.591 | exp/sdm1/chain/tdnn_sp_bi_ihmali/decode_eval/ascore_8/eval_hires_o4.ctm.filt.sys

# local/chain/multi_condition/run_tdnn.sh --mic sdm1 --use-ihm-ali true --train-set train_cleaned --gmm tri3_cleaned &
# cleanup + chain TDNN model, SDM original + IHM reverberated data, alignments from ihm data.
# *** best system ***
# for d in exp/sdm1/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done
%WER 38.6 | 14760 94502 | 65.3 19.3 15.4 3.9 38.6 64.9 | 0.599 | exp/sdm1/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_dev/ascore_9/dev_hires_o4.ctm.filt.sys
%WER 42.7 | 14070 89982 | 60.9 21.0 18.0 3.6 42.7 64.5 | 0.571 | exp/sdm1/chain_cleaned_rvb/tdnn_sp_rvb_bi_ihmali/decode_eval/ascore_9/eval_hires_o4.ctm.filt.sys


# local/chain/tuning/run_tdnn_lstm_1i.sh --mic sdm1 --use-ihm-ali true --train-set train_cleaned --gmm tri3_cleaned
# cleanup + chain TDNN model, SDM audio + alignments from ihm data.
# *** best system ***
%WER 37.6 | 15122 94495 | 66.1 18.7 15.1 3.7 37.6 63.2 | 0.646 | exp/sdm1/chain_cleaned/tdnn_lstm1i_sp_bi_ihmali_ld5/decode_dev/ascore_10/dev_hires_o4.ctm.filt.sys
%WER 40.9 | 13807 89961 | 62.4 20.0 17.6 3.3 40.9 65.7 | 0.612 | exp/sdm1/chain_cleaned/tdnn_lstm1i_sp_bi_ihmali_ld5/decode_eval/ascore_10/eval_hires_o4.ctm.filt.sys
53 changes: 53 additions & 0 deletions egs/ami/s5b/local/chain/compare_wer_general.sh
Original file line number Diff line number Diff line change
@@ -0,0 +1,53 @@
#!/bin/bash

mic=$1;
shift;

echo -n "System "
for x in $*; do printf "% 10s" $x; done
echo

#for d in exp/sdm1/chain_cleaned/tdnn*/decode_*; do grep Sum $d/*sc*/*ys | utils/best_wer.sh; done|grep eval_hires


echo -n "WER on dev "
for x in $*; do
wer=$(grep Sum exp/$mic/chain_cleaned/${x}/decode_dev*/*sc*/*ys | utils/best_wer.sh | awk '{print $2}')
printf "% 10s" $wer
done
echo

echo -n "WER on eval "
for x in $*; do
wer=$(grep Sum exp/$mic/chain_cleaned/${x}/decode_eval*/*sc*/*ys | utils/best_wer.sh | awk '{print $2}')
printf "% 10s" $wer
done
echo

echo -n "Final train prob "
for x in $*; do
prob=$(grep Overall exp/$mic/chain_cleaned/${x}/log/compute_prob_train.final.log | grep -v xent | awk '{print $8}')
printf "% 10s" $prob
done
echo

echo -n "Final valid prob "
for x in $*; do
prob=$(grep Overall exp/$mic/chain_cleaned/${x}/log/compute_prob_valid.final.log | grep -v xent | awk '{print $8}')
printf "% 10s" $prob
done
echo

echo -n "Final train prob (xent) "
for x in $*; do
prob=$(grep Overall exp/$mic/chain_cleaned/${x}/log/compute_prob_train.final.log | grep -w xent | awk '{print $8}')
printf "% 10s" $prob
done
echo

echo -n "Final valid prob (xent) "
for x in $*; do
prob=$(grep Overall exp/$mic/chain_cleaned/${x}/log/compute_prob_valid.final.log | grep -w xent | awk '{print $8}')
printf "% 10s" $prob
done
echo
Loading