Augmentation recipe for swbd

egs/swbd/s5c/RESULTS

@@ -191,9 +191,11 @@ exit 0
 %WER 24.3 | 2628 21594 | 78.6 15.0 6.4 2.9 24.3 60.0 | exp/nnet3/tdnn_cnn_sp/decode_eval2000_hires_sw1_tg/score_10_0.0/eval2000_hires.ctm.callhm.filt.sys
 
 
-# current best 'chain' models with TDNNs (see local/chain/run_tdnn_7d.sh)
-%WER 10.4 | 1831 21395 | 90.7 6.1 3.2 1.2 10.4 44.6 | exp/chain/tdnn_7d_sp/decode_eval2000_sw1_fsh_fg/score_11_1.0/eval2000_hires.ctm.swbd.filt.sys
-%WER 11.6 | 1831 21395 | 89.7 7.0 3.3 1.4 11.6 47.0 | exp/chain/tdnn_7d_sp/decode_eval2000_sw1_tg/score_10_1.0/eval2000_hires.ctm.swbd.filt.sys
+# current best 'chain' models with TDNNs (see local/chain/run_tdnn_7g.sh)
+# (2 epoch training on data being speed-perturbed, volume-perturbed and reverberated with room impulse responses)
+%WER 14.6 | 4459 42989 | 87.1 8.7 4.2 1.7 14.6 50.7 | exp/chain/tdnn_7g_sp/decode_eval2000_sw1_fsh_fg/score_10_0.0/eval2000_hires.ctm.filt.sys
+%WER 9.8 | 1831 21395 | 91.2 5.7 3.1 1.1 9.8 43.4 | exp/chain/tdnn_7g_sp/decode_eval2000_sw1_fsh_fg/score_11_0.0/eval2000_hires.ctm.swbd.filt.sys
+%WER 19.3 | 2628 21594 | 83.0 11.5 5.5 2.3 19.3 55.8 | exp/chain/tdnn_7g_sp/decode_eval2000_sw1_fsh_fg/score_10_0.0/eval2000_hires.ctm.callhm.filt.sys
 
 # current best 'chain' models with LSTM (see local/chain/run_lstm_d.sh)
 %WER 15.9 | 4459 42989 | 86.0 9.6 4.3 2.0 15.9 51.7 | exp/chain/lstm_d_ld5_sp/decode_eval2000_sw1_fsh_fg/score_10_0.0/eval2000_hires.ctm.filt.sys

egs/swbd/s5c/local/chain/multi_condition/run_tdnn_7f.sh

@@ -0,0 +1,249 @@
+#!/bin/bash
+
+# This script (multi_condition/run_tdnn_7f.sh) is the reverberated version of
+# tuning/run_tdnn_7f.sh. It reverberates the training data with room impulse responses
+# which leads to better results.
+# (The reverberation of data is done in multi_condition/run_ivector_common.sh)
+# This script assumes a mixing of the original training data with its reverberated copy
+# and results in a 2-fold training set. Thus the number of epochs is halved to
+# keep the same training time. The model converges after 2 epochs of training,
+# The WER doesn't change much with more epochs of training.
+# local/chain/compare_wer.sh tuning/7f multi_condition/7f
+# System                 tuning/7f  multi_condition/7f
+# WER on train_dev(tg)      14.46     14.27
+# WER on train_dev(fg)      13.23     13.16
+# WER on eval2000(tg)        17.0      16.3
+# WER on eval2000(fg)        15.4      14.6
+# Final train prob     -0.0882071 -0.123325
+# Final valid prob      -0.107545 -0.131798
+# Final train prob (xent)      -1.26246   -1.6196
+# Final valid prob (xent)      -1.35525  -1.60244
+
+
+set -e
+
+# configs for 'chain'
+affix=
+stage=1
+train_stage=-10
+get_egs_stage=-10
+speed_perturb=true
+dir=exp/chain/tdnn_7f  # Note: _sp will get added to this if $speed_perturb == true.
+decode_iter=
+ivector_dir=exp/nnet3_rvb
+num_data_reps=1        # number of reverberated copies of data to generate
+input_train_set=train_nodup
+
+
+# TDNN options
+# this script uses the new tdnn config generator so it needs a final 0 to reflect that the final layer input has no splicing
+splice_indexes="-1,0,1 -1,0,1 -1,0,1 -3,0,3 -3,0,3 -6,0,6 0"
+# smoothing options
+self_repair_scale=0.00001
+# training options
+num_epochs=2
+initial_effective_lrate=0.001
+final_effective_lrate=0.0001
+leftmost_questions_truncate=-1
+max_param_change=2.0
+final_layer_normalize_target=0.5
+num_jobs_initial=3
+num_jobs_final=16
+minibatch_size=128
+relu_dim=625
+frames_per_eg=150
+remove_egs=false
+common_egs_dir=
+xent_regularize=0.1
+
+# End configuration section.
+echo "$0 $@"  # Print the command line for logging
+
+. ./cmd.sh
+. ./path.sh
+. ./utils/parse_options.sh
+
+if ! cuda-compiled; then
+  cat <<EOF && exit 1
+This script is intended to be used with GPUs but you have not compiled Kaldi with CUDA
+If you want to use GPUs (and have them), go to src/, and configure and make on a machine
+where "nvcc" is installed.
+EOF
+fi
+
+# The iVector-extraction and feature-dumping parts are the same as the standard
+# nnet3 setup, and you can skip them by setting "--stage 8" if you have already
+# run those things.
+
+suffix=
+if [ "$speed_perturb" == "true" ]; then
+  suffix=_sp
+fi
+
+dir=${dir}${affix:+_$affix}${suffix}_rvb${num_data_reps}
+clean_train_set=${input_train_set}${suffix}
+train_set=${clean_train_set}_rvb${num_data_reps}
+ali_dir=exp/tri4_ali_nodup$suffix
+treedir=exp/chain/tri5_7d_tree$suffix
+lang=data/lang_chain_2y
+clean_lat_dir=exp/tri4_lats_nodup${suffix}
+lat_dir=${clean_lat_dir}_rvb${num_data_reps}
+
+
+# The data reverberation will be done in this script.
+local/nnet3/multi_condition/run_ivector_common.sh --stage $stage \
+  --input-data-dir ${input_train_set} \
+  --ivector-dir $ivector_dir \
+  --speed-perturb $speed_perturb \
+  --num-data-reps $num_data_reps || exit 1;
+
+
+if [ $stage -le 9 ]; then
+  # Get the alignments as lattices (gives the LF-MMI training more freedom).
+  # use the same num-jobs as the alignments
+  nj=$(cat exp/tri4_ali_nodup${suffix}/num_jobs) || exit 1;
+  steps/align_fmllr_lats.sh --nj $nj --cmd "$train_cmd" data/${clean_train_set} \
+    data/lang exp/tri4 $clean_lat_dir
+  rm $clean_lat_dir/fsts.*.gz # save space
+
+
+  # Create the lattices for the reverberated data
+  # We use the lattices/alignments from the clean data for the reverberated data.
+  mkdir -p $lat_dir/temp/
+  lattice-copy "ark:gunzip -c $clean_lat_dir/lat.*.gz |" ark,scp:$lat_dir/temp/lats.ark,$lat_dir/temp/lats.scp
+
+  # copy the lattices for the reverberated data
+  rm -f $lat_dir/temp/combined_lats.scp
+  touch $lat_dir/temp/combined_lats.scp
+  # Here prefix "rev0_" represents the clean set, "rev1_" represents the reverberated set
+  for i in `seq 0 $num_data_reps`; do
+    cat $lat_dir/temp/lats.scp | sed -e "s/^/rev${i}_/" >> $lat_dir/temp/combined_lats.scp
+  done
+  sort -u $lat_dir/temp/combined_lats.scp > $lat_dir/temp/combined_lats_sorted.scp
+
+  lattice-copy scp:$lat_dir/temp/combined_lats_sorted.scp "ark:|gzip -c >$lat_dir/lat.1.gz" || exit 1;
+  echo "1" > $lat_dir/num_jobs
+
+  # copy other files from original lattice dir
+  for f in cmvn_opts final.mdl splice_opts tree; do
+    cp $clean_lat_dir/$f $lat_dir/$f
+  done
+fi
+
+
+if [ $stage -le 10 ]; then
+  # Create a version of the lang/ directory that has one state per phone in the
+  # topo file. [note, it really has two states.. the first one is only repeated
+  # once, the second one has zero or more repeats.]
+  rm -rf $lang
+  cp -r data/lang $lang
+  silphonelist=$(cat $lang/phones/silence.csl) || exit 1;
+  nonsilphonelist=$(cat $lang/phones/nonsilence.csl) || exit 1;
+  # Use our special topology... note that later on may have to tune this
+  # topology.
+  steps/nnet3/chain/gen_topo.py $nonsilphonelist $silphonelist >$lang/topo
+fi
+
+if [ $stage -le 11 ]; then
+  # Build a tree using our new topology. This is the critically different
+  # step compared with other recipes.
+  # we build the tree using the clean alignments as we empirically found that this was better.
+  steps/nnet3/chain/build_tree.sh --frame-subsampling-factor 3 \
+      --leftmost-questions-truncate $leftmost_questions_truncate \
+      --context-opts "--context-width=2 --central-position=1" \
+      --cmd "$train_cmd" 7000 data/${clean_train_set} $lang $ali_dir $treedir
+fi
+
+if [ $stage -le 12 ]; then
+  echo "$0: creating neural net configs";
+  if [ ! -z "$relu_dim" ]; then
+    dim_opts="--relu-dim $relu_dim"
+  else
+    dim_opts="--pnorm-input-dim $pnorm_input_dim --pnorm-output-dim  $pnorm_output_dim"
+  fi
+
+  # create the config files for nnet initialization
+  repair_opts=${self_repair_scale:+" --self-repair-scale-nonlinearity $self_repair_scale "}
+
+  steps/nnet3/tdnn/make_configs.py \
+    $repair_opts \
+    --feat-dir data/${train_set}_hires \
+    --ivector-dir $ivector_dir/ivectors_${train_set} \
+    --tree-dir $treedir \
+    $dim_opts \
+    --splice-indexes "$splice_indexes" \
+    --use-presoftmax-prior-scale false \
+    --xent-regularize $xent_regularize \
+    --xent-separate-forward-affine true \
+    --include-log-softmax false \
+    --final-layer-normalize-target $final_layer_normalize_target \
+    $dir/configs || exit 1;
+fi
+
+
+
+if [ $stage -le 13 ]; then
+  if [[ $(hostname -f) == *.clsp.jhu.edu ]] && [ ! -d $dir/egs/storage ]; then
+    utils/create_split_dir.pl \
+     /export/b0{5,6,7,8}/$USER/kaldi-data/egs/swbd-reverb-$(date +'%m_%d_%H_%M')/s5c/$dir/egs/storage $dir/egs/storage
+  fi
+
+  steps/nnet3/chain/train.py --stage $train_stage \
+    --cmd "$decode_cmd" \
+    --feat.online-ivector-dir $ivector_dir/ivectors_${train_set} \
+    --feat.cmvn-opts "--norm-means=false --norm-vars=false" \
+    --chain.xent-regularize $xent_regularize \
+    --chain.leaky-hmm-coefficient 0.1 \
+    --chain.l2-regularize 0.00005 \
+    --chain.apply-deriv-weights false \
+    --chain.lm-opts="--num-extra-lm-states=2000" \
+    --egs.dir "$common_egs_dir" \
+    --egs.stage $get_egs_stage \
+    --egs.opts "--frames-overlap-per-eg 0" \
+    --egs.chunk-width $frames_per_eg \
+    --trainer.num-chunk-per-minibatch $minibatch_size \
+    --trainer.frames-per-iter 1500000 \
+    --trainer.num-epochs $num_epochs \
+    --trainer.optimization.num-jobs-initial $num_jobs_initial \
+    --trainer.optimization.num-jobs-final $num_jobs_final \
+    --trainer.optimization.initial-effective-lrate $initial_effective_lrate \
+    --trainer.optimization.final-effective-lrate $final_effective_lrate \
+    --trainer.max-param-change $max_param_change \
+    --cleanup.remove-egs $remove_egs \
+    --feat-dir data/${train_set}_hires \
+    --tree-dir $treedir \
+    --lat-dir $lat_dir \
+    --dir $dir  || exit 1;
+
+fi
+
+if [ $stage -le 14 ]; then
+  # Note: it might appear that this $lang directory is mismatched, and it is as
+  # far as the 'topo' is concerned, but this script doesn't read the 'topo' from
+  # the lang directory.
+  utils/mkgraph.sh --left-biphone --self-loop-scale 1.0 data/lang_sw1_tg $dir $dir/graph_sw1_tg
+fi
+
+decode_suff=sw1_tg
+graph_dir=$dir/graph_sw1_tg
+if [ $stage -le 15 ]; then
+  iter_opts=
+  if [ ! -z $decode_iter ]; then
+    iter_opts=" --iter $decode_iter "
+  fi
+  for decode_set in train_dev eval2000; do
+      (
+      steps/nnet3/decode.sh --acwt 1.0 --post-decode-acwt 10.0 \
+          --nj 50 --cmd "$decode_cmd" $iter_opts \
+          --online-ivector-dir $ivector_dir/ivectors_${decode_set} \
+          $graph_dir data/${decode_set}_hires $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_${decode_suff} || exit 1;
+      if $has_fisher; then
+          steps/lmrescore_const_arpa.sh --cmd "$decode_cmd" \
+            data/lang_sw1_{tg,fsh_fg} data/${decode_set}_hires \
+            $dir/decode_${decode_set}${decode_iter:+_$decode_iter}_sw1_{tg,fsh_fg} || exit 1;
+      fi
+      ) &
+  done
+fi
+wait;
+exit 0;

egs/swbd/s5c/local/chain/run_tdnn.sh

@@ -1 +1 @@
-tuning/run_tdnn_7e.sh
+tuning/run_tdnn_7f.sh