. ./path.sh

set -euo pipefail

output_dir=data/dump/librispeech
wham_output_dir=data/dump/wham_noise
wham_root=/F00120240032/wham_noise/wham_noise/wham_noise/
librispeech_root=/F00120240032/librispeech/corpus_librispeech/corpus
librispeech_alignments_root=/F00120240032/LibriSpeech-Alignments
librispeech_subsets=(
  train_clean_100
  train_clean_360
  train_other_500
  dev_clean
  dev_other
  test_clean
  test_other
)
wham_subsets=(
  tr
  cv
  tt
)

mkdir -p ${output_dir}

python data/nemo/scripts/get_librispeech_data.py \
  --data_root ${librispeech_root} \
  --data_sets ALL

for dataset in "${librispeech_subsets[@]}"; do
  _subset_output_dir=${output_dir}/${dataset}
  mkdir -p ${_subset_output_dir}

  python data/nemo/scripts/create_alignment_manifest.py \
    --input_manifest_filepath ${librispeech_root}/${dataset}.json \
    --base_alignment_path ${librispeech_alignments_root} \
    --output_manifest_filepath ${_subset_output_dir}/align.json \
    --ctm_output_directory ${_subset_output_dir}/ctm_out \
    --libri_dataset_split ${dataset}
done

for dataset in "${wham_subsets[@]}"; do
  _subset_output_dir=${wham_output_dir}/${dataset}
  mkdir -p ${_subset_output_dir}

  _paths2audio_files=${wham_output_dir}/${dataset}/wav.list
  realpath ${wham_root}/${dataset}/* > ${_paths2audio_files}


  python data/shared/scripts/pathfiles_to_diarize_manifest.py \
    --paths2audio_files ${_paths2audio_files} \
    --manifest_filepath ${_subset_output_dir}/bg_noise.json
done

# mkdir -p ${output_dir}/train_960
# cat data/dump/librispeech/train_{clean_100,clean_360,other_500}/align.json > data/dump/librispeech/train_960/align.json