#!/bin/bash
  
  # CER:
  # %WER 16.44 [ 35459 / 215718, 4216 ins, 11278 del, 19965 sub ] exp/chain/tdnn_1b_sp/decode_test/cer_10_0.0
  # %WER 13.72 [ 29605 / 215718, 4678 ins, 8066 del, 16861 sub ] exp/chain/tdnn_1c_sp/decode_test/cer_10_0.0
  
  # steps/info/chain_dir_info.pl exp/chain/tdnn_1c_sp
  # exp/chain/tdnn_1c_sp: num-iters=147 nj=3..16 num-params=17.9M dim=43+100->4528 combine=-0.041->-0.041 (over 2) xent:train/valid[97,146,final]=(-0.845,-0.625,-0.618/-0.901,-0.710,-0.703) logprob:train/valid[97,146,final]=(-0.064,-0.040,-0.039/-0.072,-0.058,-0.057)
  
  set -e
  
  # configs for 'chain'
  affix=1c
  nnet3_affix=_1b
  stage=0
  train_stage=-10
  get_egs_stage=-10
  dir=exp/chain/tdnn  # Note: _sp will get added to this
  decode_iter=
  
  # training options
  num_epochs=6
  initial_effective_lrate=0.00025
  final_effective_lrate=0.000025
  max_param_change=2.0
  final_layer_normalize_target=0.5
  num_jobs_initial=3
  num_jobs_final=16
  minibatch_size=64
  frames_per_eg=150,110,90
  remove_egs=false
  common_egs_dir=
  xent_regularize=0.1
  dropout_schedule='0,0@0.20,0.5@0.50,0'
  
  # 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.
  
  dir=${dir}${affix:+_$affix}_sp
  train_set=train_cleaned_sp
  ali_dir=exp/tri5a_cleaned_sp_ali
  treedir=exp/chain/tri6a_cleaned_tree_sp
  lang=data/lang_chain
  
  
  # if we are using the speed-perturbed data we need to generate
  # alignments for it.
  local/nnet3/run_ivector_common.sh --stage $stage --train-set train_cleaned --gmm tri5a_cleaned --nnet3-affix $nnet3_affix
  
  if [ $stage -le 7 ]; then
    # Get the alignments as lattices (gives the LF-MMI training more freedom).
    # use the same num-jobs as the alignments
    nj=$(cat $ali_dir/num_jobs) || exit 1;
    steps/align_fmllr_lats.sh --nj $nj --cmd "$train_cmd" data/$train_set \
      data/lang exp/tri5a exp/tri5a_sp_lats
    rm exp/tri5a_sp_lats/fsts.*.gz # save space
  fi
  
  if [ $stage -le 8 ]; 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 [ $stage -le 9 ]; then
    # Build a tree using our new topology. This is the critically different
    # step compared with other recipes.
    steps/nnet3/chain/build_tree.sh --frame-subsampling-factor 3 \
        --context-opts "--context-width=2 --central-position=1" \
        --cmd "$train_cmd" 5000 data/$train_set $lang $ali_dir $treedir
  fi
  
  if [ $stage -le 10 ]; then
    echo "$0: creating neural net configs using the xconfig parser";
  
    num_targets=$(tree-info $treedir/tree |grep num-pdfs|awk '{print $2}')
    learning_rate_factor=$(echo "print (0.5/$xent_regularize)" | python)
    affine_opts="l2-regularize=0.01 dropout-proportion=0.0 dropout-per-dim=true dropout-per-dim-continuous=true"
    tdnnf_opts="l2-regularize=0.01 dropout-proportion=0.0 bypass-scale=0.66"
    linear_opts="l2-regularize=0.01 orthonormal-constraint=-1.0"
    prefinal_opts="l2-regularize=0.01"
    output_opts="l2-regularize=0.002"
  
    mkdir -p $dir/configs
    cat <<EOF > $dir/configs/network.xconfig
    input dim=100 name=ivector
    input dim=43 name=input
  
    # please note that it is important to have input layer with the name=input
    # as the layer immediately preceding the fixed-affine-layer to enable
    # the use of short notation for the descriptor
    fixed-affine-layer name=lda input=Append(-1,0,1,ReplaceIndex(ivector, t, 0)) affine-transform-file=$dir/configs/lda.mat
    # the first splicing is moved before the lda layer, so no splicing here
    relu-batchnorm-dropout-layer name=tdnn1 $affine_opts dim=1536
    tdnnf-layer name=tdnnf2 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=1
    tdnnf-layer name=tdnnf3 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=1
    tdnnf-layer name=tdnnf4 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=1
    tdnnf-layer name=tdnnf5 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=0
    tdnnf-layer name=tdnnf6 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf7 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf8 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf9 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf10 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf11 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf12 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf13 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf14 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    tdnnf-layer name=tdnnf15 $tdnnf_opts dim=1536 bottleneck-dim=160 time-stride=3
    linear-component name=prefinal-l dim=256 $linear_opts
    prefinal-layer name=prefinal-chain input=prefinal-l $prefinal_opts big-dim=1536 small-dim=256
    output-layer name=output include-log-softmax=false dim=$num_targets $output_opts
    prefinal-layer name=prefinal-xent input=prefinal-l $prefinal_opts big-dim=1536 small-dim=256
    output-layer name=output-xent dim=$num_targets learning-rate-factor=$learning_rate_factor $output_opts
  EOF
    steps/nnet3/xconfig_to_configs.py --xconfig-file $dir/configs/network.xconfig --config-dir $dir/configs/
  fi
  
  if [ $stage -le 11 ]; then
    steps/nnet3/chain/train.py --stage $train_stage \
      --cmd "$decode_cmd" \
      --feat.online-ivector-dir exp/nnet3$nnet3_affix/ivectors_${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 0.0 \
      --chain.apply-deriv-weights false \
      --chain.lm-opts="--num-extra-lm-states=2000" \
      --trainer.dropout-schedule $dropout_schedule \
      --trainer.add-option="--optimization.memory-compression-level=2" \
      --egs.dir "$common_egs_dir" \
      --egs.stage $get_egs_stage \
      --egs.opts "--frames-overlap-per-eg 0 --constrained false" \
      --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 exp/tri5a_sp_lats \
      --use-gpu wait \
      --dir $dir  || exit 1;
  fi
  
  if [ $stage -le 12 ]; 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 --self-loop-scale 1.0 data/lang_test $dir $dir/graph
  fi
  
  graph_dir=$dir/graph
  if [ $stage -le 13 ]; then
    for test_set in test eval; do
      steps/nnet3/decode.sh --acwt 1.0 --post-decode-acwt 10.0 \
        --nj 10 --cmd "$decode_cmd" \
        --online-ivector-dir exp/nnet3${nnet3_affix:+_$nnet3_affix}/ivectors_$test_set \
        $graph_dir data/${test_set}_hires $dir/decode_${test_set} || exit 1;
    done
    wait;
  fi
  
  exit 0;