#!/bin/bash
  
  # This is the "multi-splice" version of the online-nnet2 training script.
  # It's currently the best recipe for aspire.
  # You'll notice that we splice over successively larger windows as we go deeper
  # into the network. The temporal context used for training on reverberant data
  # is larger than that used for other LVCSR recipes.
  
  . ./cmd.sh
  
  
  stage=1
  train_stage=-10
  use_gpu=true
  dir=exp/nnet2_multicondition/nnet_ms_a
  
  set -e
  . ./cmd.sh
  . ./path.sh
  . ./utils/parse_options.sh
  
  
  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
    parallel_opts="--gpu 1"
    num_threads=1
    minibatch_size=512
  
    if [[ $(hostname -f) == *.clsp.jhu.edu ]]; then
      parallel_opts="$parallel_opts --config conf/queue_no_k20.conf --allow-k20 false"
      # that config is like the default config in the text of queue.pl, but adding the following lines.
      # default allow_k20=true
      # option allow_k20=true
      # option allow_k20=false -l 'hostname=!g01&!g02&!b06'
      # It's a workaround for an NVidia CUDA library bug for our currently installed version
      # of the CUDA toolkit, that only shows up on k20's
    fi
    # the _a is in case I want to change the parameters.
  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
    minibatch_size=128
    parallel_opts="--num-threads $num_threads"
  fi
  
  # do the common parts of the script.
  local/multi_condition/run_nnet2_common.sh --stage $stage
  
  
  if [ $stage -le 7 ]; then
    if [[ $(hostname -f) == *.clsp.jhu.edu ]] && [ ! -d $dir/egs/storage ]; then
      utils/create_split_dir.pl \
       /export/b0{3,4,5,6}/$USER/kaldi-data/egs/fisher_english_reverb-$(date +'%m_%d_%H_%M')/s5/$dir/egs/storage $dir/egs/storage
    fi
  
    # The size of the system is kept rather smaller than the run_7a_960.sh system:
    # this is because we want it to be small enough that we could plausibly run it
    # in real-time.
    steps/nnet2/train_multisplice_accel2.sh --stage $train_stage \
      --num-epochs 3 --num-jobs-initial 4 --num-jobs-final 22 \
      --num-hidden-layers 6 --splice-indexes "layer0/-2:-1:0:1:2 layer1/-1:2 layer3/-3:3 layer4/-10:-7:2:5" \
      --feat-type raw \
      --online-ivector-dir exp/nnet2_multicondition/ivectors_train \
      --cmvn-opts "--norm-means=false --norm-vars=false" \
      --num-threads "$num_threads" \
      --minibatch-size "$minibatch_size" \
      --parallel-opts "$parallel_opts" \
      --io-opts "--max-jobs-run 12" \
      --initial-effective-lrate 0.0015 --final-effective-lrate 0.00015 \
      --cmd "$decode_cmd" \
      --pnorm-input-dim 4000 \
      --pnorm-output-dim 400 \
      --mix-up 12000 \
      --frames-per-eg 16 \
      --remove-egs false \
      data/train_rvb_hires data/lang exp/tri5a_rvb_ali $dir  || exit 1;
  fi
  
  if [ $stage -le 8 ]; then
    # dump iVectors for the testing data.
    for data_dir in dev_rvb test_rvb dev_aspire dev test; do
      steps/online/nnet2/extract_ivectors_online.sh --cmd "$train_cmd" --nj 20 \
        data/${data_dir}_hires exp/nnet2_multicondition/extractor exp/nnet2_multicondition/ivectors_${data_dir} || exit 1;
    done
  fi
  
  
  if [ $stage -le 9 ]; then
    # this does offline decoding that should give about the same results as the
    # real online decoding (the one with --per-utt true)
    for data_dir in dev_rvb test_rvb dev_aspire dev test; do
     ( steps/nnet2/decode.sh --nj 30 --cmd "$decode_cmd" --config conf/decode.config \
        --online-ivector-dir exp/nnet2_multicondition/ivectors_${data_dir} \
        exp/tri5a/graph data/${data_dir}_hires $dir/decode_${data_dir} || exit 1;
     ) &
    done
    wait;
  fi
  
  
  if [ $stage -le 10 ]; then
    # If this setup used PLP features, we'd have to give the option --feature-type plp
    # to the script below.
    steps/online/nnet2/prepare_online_decoding.sh --mfcc-config conf/mfcc_hires.conf \
      data/lang exp/nnet2_multicondition/extractor "$dir" ${dir}_online || exit 1;
  fi
  
  if [ $stage -le 11 ]; then
    # do the actual online decoding with iVectors, carrying info forward from
    # previous utterances of the same speaker.
    for data_dir in dev_rvb test_rvb dev_aspire dev test; do
     ( steps/online/nnet2/decode.sh --nj 30 --cmd "$decode_cmd" \
        --config conf/decode.config \
        exp/tri5a/graph data/${data_dir}_hires ${dir}_online/decode_${data_dir} || exit 1;
     ) &
    done
    wait;
  fi
  
  if [ $stage -le 12 ]; then
    # this version of the decoding treats each utterance separately
    # without carrying forward speaker information.
    for data_dir in dev_rvb test_rvb dev_aspire dev test; do
     ( steps/online/nnet2/decode.sh --nj 30 --cmd "$decode_cmd" \
        --config conf/decode.config \
        --per-utt true \
        exp/tri5a/graph data/${data_dir}_hires ${dir}_online/decode_${data_dir} || exit 1;
     ) &
    done
    wait;
  fi
  
  if [ $stage -le 13 ]; then
    # this version of the decoding treats each utterance separately
    # without carrying forward speaker information, but looks to the end
    # of the utterance while computing the iVector (--online false)
    for data_dir in dev_rvb test_rvb dev_aspire dev test; do
     ( steps/online/nnet2/decode.sh --nj 30 --cmd "$decode_cmd" \
        --config conf/decode.config \
        --per-utt true --online false \
        exp/tri5a/graph data/${data_dir}_hires ${dir}_online/decode_${data_dir} || exit 1;
     ) &
    done
  fi
  
  exit 0;