#!/bin/bash

set -o pipefail
set -e
# this is run_discriminative.sh

# This script does discriminative training on top of CE nnet3 system.
# note: this relies on having a cluster that has plenty of CPUs as well as GPUs,
# since the lattice generation runs in about real-time, so takes of the order of
# 1000 hours of CPU time.
#
. ./cmd.sh


stage=0
train_stage=-10 # can be used to start training in the middle.
get_egs_stage=-10
use_gpu=true  # for training
cleanup=false  # run with --cleanup true --stage 6 to clean up (remove large things like denlats,
               # alignments and degs).

. ./cmd.sh
. ./path.sh
. ./utils/parse_options.sh

srcdir=exp/nnet3/nnet_ms_a
train_data_dir=data/train_si284_hires
online_ivector_dir=exp/nnet3/ivectors_train_si284
degs_dir=                     # If provided, will skip the degs directory creation
lats_dir=                     # If provided, will skip denlats creation

## Objective options
criterion=smbr
one_silence_class=true

dir=${srcdir}_${criterion}

## Egs options
frames_per_eg=150
frames_overlap_per_eg=30

## Nnet training options
effective_learning_rate=0.0000125
max_param_change=1
num_jobs_nnet=4
num_epochs=4
regularization_opts=          # Applicable for providing --xent-regularize and --l2-regularize options
minibatch_size=64

## Decode options
decode_start_epoch=1 # can be used to avoid decoding all epochs, e.g. if we decided to run more.

if $use_gpu; then
  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.  Otherwise, call this script with --use-gpu false
EOF
  fi
  num_threads=1
else
  # Use 4 nnet jobs just like run_4d_gpu.sh so the results should be
  # almost the same, but this may be a little bit slow.
  num_threads=16
fi

if [ ! -f ${srcdir}/final.mdl ]; then
  echo "$0: expected ${srcdir}/final.mdl to exist; first run run_tdnn.sh or run_lstm.sh"
  exit 1;
fi

if [ $stage -le 1 ]; then
  # hardcode no-GPU for alignment, although you could use GPU [you wouldn't
  # get excellent GPU utilization though.]
  nj=100 # have a high number of jobs because this could take a while, and we might
         # have some stragglers.
  steps/nnet3/align.sh  --cmd "$decode_cmd" --use-gpu false \
    --online-ivector-dir $online_ivector_dir \
     --nj $nj $train_data_dir data/lang $srcdir ${srcdir}_ali ;

fi

if [ -z "$lats_dir" ]; then
  lats_dir=${srcdir}_denlats
  if [ $stage -le 2 ]; then
    nj=50
    # this doesn't really affect anything strongly, except the num-jobs for one of
    # the phases of get_egs_discriminative.sh below.
    num_threads_denlats=6
    subsplit=40 # number of jobs that run per job (but 2 run at a time, so total jobs is 80, giving
    # total slots = 80 * 6 = 480.
    steps/nnet3/make_denlats.sh --cmd "$decode_cmd" --determinize true \
      --online-ivector-dir $online_ivector_dir \
      --nj $nj --sub-split $subsplit --num-threads "$num_threads_denlats" --config conf/decode_dnn.config \
      $train_data_dir data/lang $srcdir ${lats_dir} ;
  fi
fi

model_left_context=`nnet3-am-info $srcdir/final.mdl | grep "left-context:" | awk '{print $2}'`
model_right_context=`nnet3-am-info $srcdir/final.mdl | grep "right-context:" | awk '{print $2}'`

left_context=$[model_left_context + extra_left_context]
right_context=$[model_right_context + extra_right_context]

frame_subsampling_opt=
if [ -f $srcdir/frame_subsampling_factor ]; then
  frame_subsampling_opt="--frame-subsampling-factor $(cat $srcdir/frame_subsampling_factor)"
fi

cmvn_opts=`cat $srcdir/cmvn_opts`

if [ -z "$degs_dir" ]; then
  degs_dir=${srcdir}_degs

  if [ $stage -le 3 ]; then
    if [[ $(hostname -f) == *.clsp.jhu.edu ]] && [ ! -d ${srcdir}_degs/storage ]; then
      utils/create_split_dir.pl \
        /export/b0{1,2,12,13}/$USER/kaldi-data/egs/swbd-$(date +'%m_%d_%H_%M')/s5/${srcdir}_degs/storage ${srcdir}_degs/storage
    fi
    # have a higher maximum num-jobs if
    if [ -d ${srcdir}_degs/storage ]; then max_jobs=10; else max_jobs=5; fi

    steps/nnet3/get_egs_discriminative.sh \
      --cmd "$decode_cmd --max-jobs-run $max_jobs --mem 20G" --stage $get_egs_stage --cmvn-opts "$cmvn_opts" \
      --online-ivector-dir $online_ivector_dir \
      --left-context $left_context --right-context $right_context \
      $frame_subsampling_opt \
      --frames-per-eg $frames_per_eg --frames-overlap-per-eg $frames_overlap_per_eg \
      $train_data_dir data/lang ${srcdir}_ali $lats_dir $srcdir/final.mdl $degs_dir ;
  fi
fi

if [ $stage -le 4 ]; then
  steps/nnet3/train_discriminative.sh --cmd "$decode_cmd" \
    --stage $train_stage \
    --effective-lrate $effective_learning_rate --max-param-change $max_param_change \
    --criterion $criterion --drop-frames true \
    --num-epochs $num_epochs --one-silence-class $one_silence_class --minibatch-size $minibatch_size \
    --num-jobs-nnet $num_jobs_nnet --num-threads $num_threads \
    --regularization-opts "$regularization_opts" \
    ${degs_dir} $dir
fi

if [ $stage -le 5 ]; then
  for x in `seq $decode_start_epoch $num_epochs`; do
    iter=epoch${x}_adj
    for lm_suffix in tgpr bd_tgpr; do
      graph_dir=exp/tri4b/graph_${lm_suffix}
      # use already-built graphs.
      for year in eval92 dev93; do
        (
          steps/nnet3/decode.sh --nj 8 --cmd "$decode_cmd" --iter $iter \
            --online-ivector-dir exp/nnet3/ivectors_test_$year \
            $graph_dir data/test_${year}_hires $dir/decode_${lm_suffix}_${year}_$iter ;
        ) &
      done
    done
  done
fi
wait;

if [ $stage -le 6 ] && $cleanup; then
  # if you run with "--cleanup true --stage 6" you can clean up.
  rm ${lats_dir}/lat.*.gz || true
  rm ${srcdir}_ali/ali.*.gz || true
  steps/nnet2/remove_egs.sh ${srcdir}_degs || true
fi


exit 0;