Skip to content

Commit

Permalink
Merge pull request #469 from Xiaofei-Wang/xiaofei_merge
Browse files Browse the repository at this point in the history 
AMI recipe improvement 2
  • Loading branch information
@sw005320
sw005320 committed Nov 20, 2018
2 parents 12477de + d9d2668 commit 5ad45a3
Show file tree
Hide file tree
Showing 4 changed files with 99 additions and 13 deletions.
22 changes: 11 additions & 11 deletions egs/ami/asr1/RESULTS
View file
@@ -1,12 +1,12 @@
# initial results (WER
$ grep -e Avg -e SPKR -m 2 exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_*bs64*/result.wrd.txt
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.wrd.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.wrd.txt:| Sum/Avg | 13059 94914 | 69.7 23.3 7.0 5.3 35.7 69.3 |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.wrd.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.wrd.txt:| Sum/Avg | 12612 89635 | 66.9 26.7 6.4 5.5 38.5 66.4 |
# initial results (WER)
$ grep -e Avg -e SPKR -m 2 exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_*bs64_word20000/result.wrd.txt
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.wrd.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.wrd.txt:| Sum/Avg | 13059 94914 | 69.7 23.3 7.0 5.3 35.7 69.3 |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.wrd.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.wrd.txt:| Sum/Avg | 12612 89635 | 66.9 26.7 6.4 5.5 38.5 66.4 |
# initial results (CER)
$ grep -e Avg -e SPKR -m 2 exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_*bs64*/result.txt
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.txt:| Sum/Avg | 13059 452218 | 82.8 8.4 8.9 5.4 22.6 69.3 |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word65000/result.txt:| Sum/Avg | 12612 431997 | 81.6 9.9 8.5 5.6 24.0 66.4 |
$ grep -e Avg -e SPKR -m 2 exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_*bs64_word20000/result.txt
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_dev_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.txt:| Sum/Avg | 13059 452218 | 82.8 8.4 8.9 5.4 22.6 69.3 |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.txt:| SPKR | # Snt # Wrd | Corr Sub Del Ins Err S.Err |
exp/ihm_train_pytorch_blstmp_e8_subsample1_2_2_1_1_unit320_proj320_d1_unit300_location_aconvc10_aconvf100_mtlalpha0.5_adadelta_bs30_mli800_mlo150_lsmunigram0.05/decode_ihm_eval_beam20_emodel.acc.best_p0.2_len0.0-0.0_ctcw0.3_rnnlm0.5_1layer_unit1000_sgd_bs64_word20000/result.txt:| Sum/Avg | 12612 431997 | 81.6 9.9 8.5 5.6 24.0 66.4 |
50 changes: 50 additions & 0 deletions egs/ami/asr1/conf/ami_beamformit.cfg
View file
@@ -0,0 +1,50 @@
#BeamformIt sample configuration file for AMI data (http://groups.inf.ed.ac.uk/ami/download/)

# scrolling size to compute the delays
scroll_size = 250

# cross correlation computation window size
window_size = 500

#amount of maximum points for the xcorrelation taken into account
nbest_amount = 4

#flag wether to apply an automatic noise thresholding
do_noise_threshold = 1

#Percentage of frames with lower xcorr taken as noisy
noise_percent = 10

######## acoustic modelling parameters

#transition probabilities weight for multichannel decoding
trans_weight_multi = 25
trans_weight_nbest = 25

###

#flag wether to print the feaures after setting them, or not
print_features = 1

#flag wether to use the bad frames in the sum process
do_avoid_bad_frames = 1

#flag to use the best channel (SNR) as a reference
#defined from command line
do_compute_reference = 1

#flag wether to use a uem file or not(process all the file)
do_use_uem_file = 0

#flag wether to use an adaptative weights scheme or fixed weights
do_adapt_weights = 1

#flag wether to output the sph files or just run the system to create the auxiliary files
do_write_sph_files = 1

####directories where to store/retrieve info####
#channels_file = ./cfg-files/channels

#show needs to be passed as argument normally, here a default one is given just in case
#show_id = Ttmp

36 changes: 36 additions & 0 deletions egs/ami/asr1/local/beamformit.sh
View file
@@ -0,0 +1,36 @@
#!/bin/bash

# Copyright 2014, University of Edinburgh (Author: Pawel Swietojanski)

. ./path.sh

nj=$1
job=$2
numch=$3
meetings=$4
sdir=$5
odir=$6
wdir=data/local/beamforming

set -e
set -u

utils/split_scp.pl -j $nj $((job-1)) $meetings $meetings.$job

while read line; do

mkdir -p $odir/$line
BeamformIt -s $line -c $wdir/channels_$numch \
--config_file `pwd`/conf/ami_beamformit.cfg \
--source_dir $sdir \
--result_dir $odir/$line
mkdir -p $odir/$line
mv $odir/$line/${line}.del $odir/$line/${line}_MDM$numch.del
mv $odir/$line/${line}.del2 $odir/$line/${line}_MDM$numch.del2
mv $odir/$line/${line}.info $odir/$line/${line}_MDM$numch.info
mv $odir/$line/${line}.weat $odir/$line/${line}_MDM$numch.weat
mv $odir/$line/${line}.wav $odir/$line/${line}_MDM$numch.wav
#mv $odir/$line/${line}.ovl $odir/$line/${line}_MDM$numch.ovl # Was not created!

done < $meetings.$job

4 changes: 2 additions & 2 deletions egs/ami/asr1/run.sh
View file
Expand Up @@ -56,7 +56,7 @@ epochs=15

# rnnlm related
use_wordlm=true # false means to train/use a character LM
lm_vocabsize=65000 # effective only for word LMs
lm_vocabsize=20000 # effective only for word LMs
lm_layers=1 # 2 for character LMs
lm_units=1000 # 650 for character LMs
lm_opt=sgd # adam for character LMs
Expand Down Expand Up @@ -295,7 +295,7 @@ else
fi
mkdir -p ${expdir}

if [ ${stage} -le -4 ]; then
if [ ${stage} -le 4 ]; then
echo "stage 4: Network Training"

${cuda_cmd} --gpu ${ngpu} ${expdir}/train.log \
Expand Down

0 comments on commit 5ad45a3

Please sign in to comment.