#!/usr/bin/env bash

: <<'END'

sh run.sh --stage 3 --stop_stage 3 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification8 \
--filename_patterns "E:/programmer/asr_datasets/voicemail/wav_finished/en-US/wav_finished/*/*.wav \
E:/programmer/asr_datasets/voicemail/wav_finished/id-ID/wav_finished/*/*.wav" \


sh run.sh --stage 3 --stop_stage 3 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification8 \
--filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav"

sh run.sh --stage 4 --stop_stage 4 --system_version windows --file_folder_name file_dir --final_model_name vm_sound_classification8 \
--filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav"

sh run.sh --stage 4 --stop_stage 4 --system_version centos --file_folder_name file_dir --final_model_name vm_sound_classification8 \
--filename_patterns "/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav"


"

END


# sh run.sh --stage -1 --stop_stage 9
# sh run.sh --stage -1 --stop_stage 5 --system_version centos --file_folder_name task_cnn_voicemail_id_id --final_model_name cnn_voicemail_id_id
# sh run.sh --stage 3 --stop_stage 4
# sh run.sh --stage 4 --stop_stage 4
# sh run.sh --stage 3 --stop_stage 3 --system_version centos --file_folder_name task_cnn_voicemail_id_id

# params
system_version="windows";
verbose=true;
stage=0 # start from 0 if you need to start from data preparation
stop_stage=9

work_dir="$(pwd)"
file_folder_name=file_folder_name
final_model_name=final_model_name
filename_patterns="/data/tianxing/PycharmProjects/datasets/voicemail/*/wav_finished/*/*.wav"
nohup_name=nohup.out

country=en-US

# model params
batch_size=64
max_epochs=200
save_top_k=10
patience=5


# parse options
while true; do
  [ -z "${1:-}" ] && break;  # break if there are no arguments
  case "$1" in
    --*) name=$(echo "$1" | sed s/^--// | sed s/-/_/g);
      eval '[ -z "${'"$name"'+xxx}" ]' && echo "$0: invalid option $1" 1>&2 && exit 1;
      old_value="(eval echo \\$$name)";
      if [ "${old_value}" == "true" ] || [ "${old_value}" == "false" ]; then
        was_bool=true;
      else
        was_bool=false;
      fi

      # Set the variable to the right value-- the escaped quotes make it work if
      # the option had spaces, like --cmd "queue.pl -sync y"
      eval "${name}=\"$2\"";

      # Check that Boolean-valued arguments are really Boolean.
      if $was_bool && [[ "$2" != "true" && "$2" != "false" ]]; then
        echo "$0: expected \"true\" or \"false\": $1 $2" 1>&2
        exit 1;
      fi
      shift 2;
      ;;

    *) break;
  esac
done

file_dir="${work_dir}/${file_folder_name}"
final_model_dir="${work_dir}/../../trained_models/${final_model_name}";

train_dataset="${file_dir}/train.xlsx"
valid_dataset="${file_dir}/valid.xlsx"
vocabulary_dir="${file_dir}/vocabulary"


$verbose && echo "system_version: ${system_version}"
$verbose && echo "file_folder_name: ${file_folder_name}"

if [ $system_version == "windows" ]; then
  alias python3='D:/Users/tianx/PycharmProjects/virtualenv/vm_sound_classification/Scripts/python.exe'
elif [ $system_version == "centos" ] || [ $system_version == "ubuntu" ]; then
  #source /data/local/bin/vm_sound_classification/bin/activate
  alias python3='/data/local/bin/vm_sound_classification/bin/python3'
fi


if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
  $verbose && echo "stage 0: prepare data"
  cd "${work_dir}" || exit 1
  python3 step_1_prepare_data.py \
  --file_dir "${file_dir}" \
  --filename_patterns "${filename_patterns}" \
  --train_dataset "${train_dataset}" \
  --valid_dataset "${valid_dataset}" \

fi


if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
  $verbose && echo "stage 1: make vocabulary"
  cd "${work_dir}" || exit 1
  python3 step_2_make_vocabulary.py \
  --vocabulary_dir "${vocabulary_dir}" \
  --train_dataset "${train_dataset}" \
  --valid_dataset "${valid_dataset}" \

fi


if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
  $verbose && echo "stage 2: train global model"
  cd "${work_dir}" || exit 1
  python3 step_3_train_global_model.py \
  --vocabulary_dir "${vocabulary_dir}" \
  --train_dataset "${train_dataset}" \
  --valid_dataset "${valid_dataset}" \
  --serialization_dir "${file_dir}/global_model" \

fi


if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
  $verbose && echo "stage 3: train country model"
  cd "${work_dir}" || exit 1
  python3 step_4_train_country_model.py \
  --vocabulary_dir "${vocabulary_dir}" \
  --train_dataset "${train_dataset}" \
  --valid_dataset "${valid_dataset}" \
  --country "${country}" \
  --serialization_dir "${file_dir}/country_model" \

fi


if [ ${stage} -le 4 ] && [ ${stop_stage} -ge 4 ]; then
  $verbose && echo "stage 4: train union model"
  cd "${work_dir}" || exit 1
  python3 step_5_train_union.py \
  --vocabulary_dir "${vocabulary_dir}" \
  --train_dataset "${train_dataset}" \
  --valid_dataset "${valid_dataset}" \
  --serialization_dir "${file_dir}/union" \

fi