diff --git a/egs/ami/s5/run_mdm.sh b/egs/ami/s5/run_mdm.sh
index 577896c23dcebce2df8afe1c173573349b2fcbb8..4389c6b5d81fb539796f2e5f1f2bf323ce0b1fd1 100755 (executable)
--- a/egs/ami/s5/run_mdm.sh
+++ b/egs/ami/s5/run_mdm.sh
. ./cmd.sh
. ./path.sh
-# MDM - Multiple Distant Microphones
-# Assuming text preparation, dict, lang and LM were build as in run_ihm
-
+# MDM - Multiple Distant Microphones,
nmics=8 #we use all 8 channels, possible other options are 2 and 4
-mic=mdm$nmics #subdir name under data/
-AMI_DIR=/disk/data2/amicorpus #root of AMI corpus
-MDM_DIR=/disk/data1/s1136550/ami/mdm #directory for beamformed waves
-
-#1) Download AMI (distant channels)
-
-local/ami_download.sh mdm $AMI_DIR
-
-#2) Beamform
+mic=mdm$nmics
-local/ami_beamform.sh --nj 16 $nmics $AMI_DIR $MDM_DIR
+# Path where AMI gets downloaded (or where locally available):
+AMI_DIR=$PWD/wav_db # Default,
+case $(hostname -d) in
+ fit.vutbr.cz) AMI_DIR=/mnt/matylda5/iveselyk/KALDI_AMI_WAV ;; # BUT,
+ clsp.jhu.edu) AMI_DIR=/export/corpora4/ami/amicorpus ;; # JHU,
+ cstr.ed.ac.uk) AMI_DIR= ;; # Edinburgh,
+esac
-#3) Prepare mdm data directories
+# MDM_DIR is directory for beamformed waves,
+MDM_DIR=$AMI_DIR/beamformed # Default,
+#MDM_DIR=/disk/data1/s1136550/ami/mdm # Edinburgh,
-local/ami_mdm_data_prep.sh $MDM_DIR $mic || exit 1;
-local/ami_mdm_scoring_data_prep.sh $MDM_DIR $mic dev || exit 1;
-local/ami_mdm_scoring_data_prep.sh $MDM_DIR $mic eval || exit 1;
-
-#use the final LM
+[ ! -r data/local/lm/final_lm ] && echo "Please, run 'run_prepare_shared.sh' first!" && exit 1
final_lm=`cat data/local/lm/final_lm`
LM=$final_lm.pr1-7
-DEV_SPK=`cut -d" " -f2 data/$mic/dev/utt2spk | sort | uniq -c | wc -l`
-EVAL_SPK=`cut -d" " -f2 data/$mic/eval/utt2spk | sort | uniq -c | wc -l`
-nj=16
-
-#GENERATE FEATS
-mfccdir=mfcc_$mic
-(
-steps/make_mfcc.sh --nj 5 --cmd "$train_cmd" data/$mic/eval exp/$mic/make_mfcc/eval $mfccdir || exit 1;
-steps/compute_cmvn_stats.sh data/$mic/eval exp/$mic/make_mfcc/eval $mfccdir || exit 1
-)&
-(
-steps/make_mfcc.sh --nj 5 --cmd "$train_cmd" data/$mic/dev exp/$mic/make_mfcc/dev $mfccdir || exit 1;
-steps/compute_cmvn_stats.sh data/$mic/dev exp/$mic/make_mfcc/dev $mfccdir || exit 1
-)&
-(
-steps/make_mfcc.sh --nj 16 --cmd "$train_cmd" data/$mic/train exp/$mic/make_mfcc/train $mfccdir || exit 1;
- steps/compute_cmvn_stats.sh data/$mic/train exp/$mic/make_mfcc/train $mfccdir || exit 1
-)&
-
-wait;
-for dset in train eval dev; do utils/fix_data_dir.sh data/$mic/$dset; done
-
-# Build the systems
-
-# TRAIN THE MODELS
- mkdir -p exp/$mic/mono
- steps/train_mono.sh --nj $nj --cmd "$train_cmd" --feat-dim 39 \
- data/$mic/train data/lang exp/$mic/mono >& exp/$mic/mono/train_mono.log || exit 1;
-
- mkdir -p exp/$mic/mono_ali
- steps/align_si.sh --nj $nj --cmd "$train_cmd" data/$mic/train data/lang exp/$mic/mono \
- exp/$mic/mono_ali >& exp/$mic/mono_ali/align.log || exit 1;
-
- mkdir -p exp/$mic/tri1
- steps/train_deltas.sh --cmd "$train_cmd" \
- 5000 80000 data/$mic/train data/lang exp/$mic/mono_ali exp/$mic/tri1 \
- >& exp/$mic/tri1/train.log || exit 1;
-
- mkdir -p exp/$mic/tri1_ali
- steps/align_si.sh --nj $nj --cmd "$train_cmd" \
- data/$mic/train data/lang exp/$mic/tri1 exp/$mic/tri1_ali || exit 1;
-
- mkdir -p exp/$mic/tri2a
- steps/train_deltas.sh --cmd "$train_cmd" \
- 5000 80000 data/$mic/train data/lang exp/$mic/tri1_ali exp/$mic/tri2a \
- >& exp/$mic/tri2a/train.log || exit 1;
-
- for lm_suffix in $LM; do
- (
- graph_dir=exp/$mic/tri2a/graph_${lm_suffix}
- $highmem_cmd $graph_dir/mkgraph.log \
- utils/mkgraph.sh data/lang_${lm_suffix} exp/$mic/tri2a $graph_dir
-
- steps/decode.sh --nj $DEV_SPK --cmd "$decode_cmd" --config conf/decode.conf \
- $graph_dir data/$mic/dev exp/$mic/tri2a/decode_dev_${lm_suffix}
-
- steps/decode.sh --nj $EVAL_SPK --cmd "$decode_cmd" --config conf/decode.conf \
- $graph_dir data/$mic/eval exp/$mic/tri2a/decode_eval_${lm_suffix}
- )
- done
-
-#THE TARGET LDA+MLLT+SAT+BMMI PART GOES HERE:
-mkdir -p exp/$mic/tri2a_ali
-steps/align_si.sh --nj $nj --cmd "$train_cmd" \
- data/$mic/train data/lang exp/$mic/tri2a exp/$mic/tri2_ali || exit 1;
-
-# Train tri3a, which is LDA+MLLT
-mkdir -p exp/$mic/tri3a
-steps/train_lda_mllt.sh --cmd "$train_cmd" \
- --splice-opts "--left-context=3 --right-context=3" \
- 5000 80000 data/$mic/train data/lang exp/$mic/tri2_ali exp/$mic/tri3a \
- >& exp/$mic/tri3a/train.log || exit 1;
-
-for lm_suffix in $LM; do
- (
- graph_dir=exp/$mic/tri3a/graph_${lm_suffix}
- $highmem_cmd $graph_dir/mkgraph.log \
- utils/mkgraph.sh data/lang_${lm_suffix} exp/$mic/tri3a $graph_dir
-
- steps/decode.sh --nj $DEV_SPK --cmd "$decode_cmd" --config conf/decode.conf \
- $graph_dir data/$mic/dev exp/$mic/tri3a/decode_dev_${lm_suffix}
-
- steps/decode.sh --nj $EVAL_SPK --cmd "$decode_cmd" --config conf/decode.conf \
- $graph_dir data/$mic/eval exp/$mic/tri3a/decode_eval_${lm_suffix}
- )
-done
-
+stage=0
+. utils/parse_options.sh
+
+# Set bash to 'debug' mode, it prints the commands (option '-x') and exits on :
+# -e 'error', -u 'undefined variable', -o pipefail 'error in pipeline',
+set -euxo pipefail
+
+# Download AMI corpus (distant channels), You need around 130GB of free space to get whole data ihm+mdm,
+# Avoiding re-download, using 'wget --continue ...',
+if [ $stage -le 0 ]; then
+ [ -e data/local/downloads/wget_mdm.sh ] && \
+ echo "data/local/downloads/wget_mdm.sh already exists, better quit than re-download... (use --stage N)" && \
+ exit 1
+ local/ami_download.sh mdm $AMI_DIR
+fi
+
+# Beamform-it!
+if [ $stage -le 1 ]; then
+ ! hash BeamformIt && echo "Missing BeamformIt, run 'cd ../../../tools/; extras/install_beamformit.sh'" && exit 1
+ local/ami_beamform.sh --cmd "$train_cmd" --nj 20 $nmics $AMI_DIR $MDM_DIR
+fi
+
+# Prepare mdm data directories,
+if [ $stage -le 2 ]; then
+ local/ami_mdm_data_prep.sh $MDM_DIR $mic
+ local/ami_mdm_scoring_data_prep.sh $MDM_DIR $mic dev
+ local/ami_mdm_scoring_data_prep.sh $MDM_DIR $mic eval
+fi
+
+# Here starts the normal recipe, which is mostly shared across mic scenarios,
+# - for ihm we adapt to speaker by fMLLR,
+# - for sdm and mdm we do not adapt for speaker, but for environment only (cmn),
+
+# Feature extraction,
+if [ $stage -le 3 ]; then
+ for dset in train dev eval; do
+ steps/make_mfcc.sh --nj 15 --cmd "$train_cmd" data/$mic/$dset data/$mic/$dset/log data/$mic/$dset/data
+ steps/compute_cmvn_stats.sh data/$mic/$dset data/$mic/$dset/log data/$mic/$dset/data
+ done
+ for dset in train eval dev; do utils/fix_data_dir.sh data/$mic/$dset; done
+fi
+
+if [ $stage -le 4 ]; then
+ # Taking a subset, now unused, can be handy for quick experiments,
+ # Full set 77h, reduced set 10.8h,
+ utils/subset_data_dir.sh data/$mic/train 15000 data/$mic/train_15k
+fi
+
+# Train systems,
+nj=30 # number of parallel jobs,
+nj_dev=$(cat data/$mic/dev/spk2utt | wc -l)
+nj_eval=$(cat data/$mic/eval/spk2utt | wc -l)
+
+if [ $stage -le 5 ]; then
+ # Mono,
+ steps/train_mono.sh --nj $nj --cmd "$train_cmd" --cmvn-opts "--norm-means=true --norm-vars=false" \
+ data/$mic/train data/lang exp/$mic/mono
+ steps/align_si.sh --nj $nj --cmd "$train_cmd" \
+ data/$mic/train data/lang exp/$mic/mono exp/$mic/mono_ali
+
+ # Deltas,
+ steps/train_deltas.sh --cmd "$train_cmd" --cmvn-opts "--norm-means=true --norm-vars=false" \
+ 5000 80000 data/$mic/train data/lang exp/$mic/mono_ali exp/$mic/tri1
+ steps/align_si.sh --nj $nj --cmd "$train_cmd" \
+ data/$mic/train data/lang exp/$mic/tri1 exp/$mic/tri1_ali
+fi
+
+if [ $stage -le 6 ]; then
+ # Deltas again, (full train-set),
+ steps/train_deltas.sh --cmd "$train_cmd" --cmvn-opts "--norm-means=true --norm-vars=false" \
+ 5000 80000 data/$mic/train data/lang exp/$mic/tri1_ali exp/$mic/tri2a
+ steps/align_si.sh --nj $nj --cmd "$train_cmd" \
+ data/$mic/train data/lang exp/$mic/tri2a exp/$mic/tri2_ali
+ # Decode,
+ graph_dir=exp/$mic/tri2a/graph_${LM}
+ $highmem_cmd $graph_dir/mkgraph.log \
+ utils/mkgraph.sh data/lang_${LM} exp/$mic/tri2a $graph_dir
+ steps/decode.sh --nj $nj_dev --cmd "$decode_cmd" --config conf/decode.conf \
+ $graph_dir data/$mic/dev exp/$mic/tri2a/decode_dev_${LM}
+ steps/decode.sh --nj $nj_eval --cmd "$decode_cmd" --config conf/decode.conf \
+ $graph_dir data/$mic/eval exp/$mic/tri2a/decode_eval_${LM}
+fi
+
+# THE TARGET LDA+MLLT+SAT+BMMI PART GOES HERE:
+
+if [ $stage -le 7 ]; then
+ # Train tri3a, which is LDA+MLLT,
+ steps/train_lda_mllt.sh --cmd "$train_cmd" \
+ --splice-opts "--left-context=3 --right-context=3" \
+ 5000 80000 data/$mic/train data/lang exp/$mic/tri2_ali exp/$mic/tri3a
+ # Decode,
+ graph_dir=exp/$mic/tri3a/graph_${LM}
+ $highmem_cmd $graph_dir/mkgraph.log \
+ utils/mkgraph.sh data/lang_${LM} exp/$mic/tri3a $graph_dir
+ steps/decode.sh --nj $nj_dev --cmd "$decode_cmd" --config conf/decode.conf \
+ $graph_dir data/$mic/dev exp/$mic/tri3a/decode_dev_${LM}
+ steps/decode.sh --nj $nj_eval --cmd "$decode_cmd" --config conf/decode.conf \
+ $graph_dir data/$mic/eval exp/$mic/tri3a/decode_eval_${LM}
+fi
# skip SAT, and build MMI models
-steps/make_denlats.sh --nj $nj --cmd "$decode_cmd" --config conf/decode.config \
- data/$mic/train data/lang exp/$mic/tri3a exp/$mic/tri3a_denlats || exit 1;
+nj_mmi=80
+if [ $stage -le 8 ]; then
+ steps/align_si.sh --nj $nj_mmi --cmd "$train_cmd" \
+ data/$mic/train data/lang exp/$mic/tri3a exp/$mic/tri3a_ali
+fi
+# At this point you can already run the DNN script:
+# local/nnet/run_dnn_lda_mllt.sh $mic
+# exit 0
-mkdir -p exp/$mic/tri3a_ali
-steps/align_si.sh --nj $nj --cmd "$train_cmd" \
- data/$mic/train data/lang exp/$mic/tri3a exp/$mic/tri3a_ali || exit 1;
+if [ $stage -le 9 ]; then
+ steps/make_denlats.sh --nj $nj_mmi --cmd "$decode_cmd" --config conf/decode.conf \
+ data/$mic/train data/lang exp/$mic/tri3a exp/$mic/tri3a_denlats
+fi
# 4 iterations of MMI seems to work well overall. The number of iterations is
# used as an explicit argument even though train_mmi.sh will use 4 iterations by
# default.
-num_mmi_iters=4
-steps/train_mmi.sh --cmd "$train_cmd" --boost 0.1 --num-iters $num_mmi_iters \
- data/$mic/train data/lang exp/$mic/tri3a_ali exp/$mic/tri3a_denlats \
- exp/$mic/tri3a_mmi_b0.1 || exit 1;
-
-for lm_suffix in $LM; do
- (
- graph_dir=exp/$mic/tri3a/graph_${lm_suffix}
-
- for i in `seq 1 4`; do
- decode_dir=exp/$mic/tri3a_mmi_b0.1/decode_dev_${i}.mdl_${lm_suffix}
- steps/decode.sh --nj $DEV_SPK --cmd "$decode_cmd" --config conf/decode.conf \
- --iter $i $graph_dir data/$mic/dev $decode_dir
- done
-
- i=3 #simply assummed
- decode_dir=exp/$mic/tri3a_mmi_b0.1/decode_eval_${i}.mdl_${lm_suffix}
- steps/decode.sh --nj $EVAL_SPK --cmd "$decode_cmd" --config conf/decode.conf \
+if [ $stage -le 10 ]; then
+ num_mmi_iters=4
+ steps/train_mmi.sh --cmd "$train_cmd" --boost 0.1 --num-iters $num_mmi_iters \
+ data/$mic/train data/lang exp/$mic/tri3a_ali exp/$mic/tri3a_denlats \
+ exp/$mic/tri3a_mmi_b0.1
+fi
+if [ $stage -le 11 ]; then
+ # Decode,
+ graph_dir=exp/$mic/tri3a/graph_${LM}
+ for i in 4 3 2 1; do
+ decode_dir=exp/$mic/tri3a_mmi_b0.1/decode_dev_${i}.mdl_${LM}
+ steps/decode.sh --nj $nj_dev --cmd "$decode_cmd" --config conf/decode.conf \
+ --iter $i $graph_dir data/$mic/dev $decode_dir
+ decode_dir=exp/$mic/tri3a_mmi_b0.1/decode_eval_${i}.mdl_${LM}
+ steps/decode.sh --nj $nj_eval --cmd "$decode_cmd" --config conf/decode.conf \
--iter $i $graph_dir data/$mic/eval $decode_dir
- )
-done
-
-# here goes hybrid stuf
-# in the ASRU paper we used different python nnet code, so someone needs to copy&adjust nnet or nnet2 switchboard commands
-
+ done
+fi
+
+# DNN training. This script is based on egs/swbd/s5b/local/run_dnn.sh
+# Some of them would be out of date.
+if [ $stage -le 12 ]; then
+ local/nnet/run_dnn_lda_mllt.sh $mic
+fi
+
+# nnet3 systems
+if [ $stage -le 13 ]; then
+ # Slightly better WERs can be obtained by using --use-sat-alignments true
+ # however the SAT systems have to be built before that
+
+ #TDNN model + xent training
+ local/nnet3/run_tdnn.sh \
+ --mic $mic \
+ --use-sat-alignments false
+
+ #LSTM model + xent training
+ local/nnet3/run_lstm.sh \
+ --mic $mic \
+ --stage 10 \
+ --train-stage -5 \
+ --use-sat-alignments false
+
+ #BLSTM model + xent training
+ local/nnet3/run_blstm.sh \
+ --mic $mic \
+ --stage 10 \
+ --train-stage -5 \
+ --use-sat-alignments false
+
+ # TDNN model + chain training
+ local/chain/run_tdnn_ami_5.sh --mic sdm1 --affix msl1.5_45wer
+
+fi
+echo "Done."
+
+
+
+
+# Older nnet2 scripts. They are still kept here
+# as we have not yet committed sMBR training scripts for AMI in nnet3
+#if [ $stage -le 13 ]; then
+# local/online/run_nnet2_ms_perturbed.sh \
+# --mic $mic \
+# --hidden-dim 850 \
+# --splice-indexes "layer0/-2:-1:0:1:2 layer1/-1:2 layer2/-3:3 layer3/-7:2 layer4/-3:3" \
+# --use-sat-alignments false
+#
+# local/online/run_nnet2_ms_sp_disc.sh \
+# --mic $mic \
+# --gmm-dir exp/$mic/tri3a \
+# --srcdir exp/$mic/nnet2_online/nnet_ms_sp
+#fi
+#
+echo "Done."