mirror of
https://github.com/k2-fsa/icefall.git
synced 2025-12-10 22:45:27 +00:00
176 lines
5.6 KiB
Bash
176 lines
5.6 KiB
Bash
#!/usr/bin/env bash
|
|
# Copyright 2023 Johns Hopkins University (Amir Hussein)
|
|
# Apache 2.0 (http://www.apache.org/licenses/LICENSE-2.0)
|
|
|
|
set -eou pipefail
|
|
|
|
nj=20
|
|
stage=0
|
|
stop_stage=7
|
|
|
|
# We assume dl_dir (download dir) contains the following
|
|
# directories and files.
|
|
#
|
|
# - $dl_dir/cts
|
|
#
|
|
# You can download the data from
|
|
#
|
|
#
|
|
# - $dl_dir/musan
|
|
# This directory contains the following directories downloaded from
|
|
# http://www.openslr.org/17/
|
|
#
|
|
# - music
|
|
# - noise
|
|
# - speech
|
|
#
|
|
|
|
dl_dir=cts
|
|
. shared/parse_options.sh || exit 1
|
|
|
|
# vocab size for sentence piece models.
|
|
# It will generate data/lang_bpe_xxx,
|
|
# data/lang_bpe_yyy if the array contains xxx, yyy
|
|
vocab_sizes=(
|
|
5000
|
|
)
|
|
st_vocab_sizes=(
|
|
4000
|
|
)
|
|
|
|
# All files generated by this script are saved in "data".
|
|
# You can safely remove "data" and rerun this script to regenerate it.
|
|
mkdir -p data
|
|
|
|
log() {
|
|
# This function is from espnet
|
|
local fname=${BASH_SOURCE[1]##*/}
|
|
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
|
}
|
|
|
|
log "dl_dir: $dl_dir"
|
|
|
|
if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
|
|
log "Stage 0: Download data"
|
|
|
|
# Download callhome_spanish, fisher_spanish iwslt22_ta and HKUST from LDC
|
|
#
|
|
# you can create a symlink
|
|
#
|
|
# ln -sfv /path/to/data $dl_dir/data
|
|
|
|
# If you have pre-downloaded it to /path/to/musan,
|
|
# you can create a symlink
|
|
#
|
|
# ln -sfv /path/to/musan $dl_dir/
|
|
#
|
|
if [ ! -d $dl_dir/musan ]; then
|
|
lhotse download musan $dl_dir
|
|
fi
|
|
fi
|
|
|
|
fbank=data/fbank
|
|
manifests=data/manifests
|
|
mkdir -p $manifests
|
|
sets="hkust iwslt-ta callhome-sp fisher-sp"
|
|
if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
|
|
log "Stage 0: Prepare telephone manifest"
|
|
# We assume that you have downloaded callhome_spanish, fisher_spanish iwslt22_ta and hkust to $dl_dir/
|
|
for set in $sets; do
|
|
log "Prepare $set manifests"
|
|
if [[ "$set" == "iwslt-ta" ]]; then
|
|
if [ ! -d "iwslt22-dialect" ]; then
|
|
echo "Splits directory (iwslt22-dialect) does not exist"
|
|
echo "Run: git clone https://github.com/kevinduh/iwslt22-dialect"
|
|
exit 1
|
|
else
|
|
lhotse prepare "$set" "$dl_dir/$set" iwslt22-dialect "$manifests"
|
|
fi
|
|
else
|
|
lhotse prepare "$set" "$dl_dir/$set" "$manifests"
|
|
# validate recordings and supervisions
|
|
fi
|
|
# python local/cuts_validate.py \
|
|
# --sup "${manifests}/supervisions.jsonl.gz" \
|
|
# --rec "${manifests}/recordings.jsonl.gz" \
|
|
# --savecut "${manifests}/cuts_${set}.jsonl.gz"
|
|
done
|
|
fi
|
|
|
|
if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
|
|
if [ ! -f ${manifests}/cut_train.jsonl.gz ]; then
|
|
log "Combining conversational data to create train, dev sets"
|
|
# combine train
|
|
lhotse combine $manifests/iwslt-ta_supervisions_train.jsonl.gz $manifests/hkust_supervisions_train.jsonl.gz $manifests/fisher-sp_supervisions_train.jsonl.gz ${manifests}/cts_supervisions_train.jsonl.gz
|
|
|
|
lhotse combine $manifests/iwslt-ta_recordings_train.jsonl.gz $manifests/hkust_recordings_train.jsonl.gz $manifests/fisher-sp_recordings_train.jsonl.gz ${manifests}/cts_recordings_train.jsonl.gz
|
|
|
|
# python local/cuts_validate.py --sup $manifests/cts_supervisions_train.jsonl.gz --rec ${manifests}/cts_recordings_train.jsonl.gz
|
|
|
|
# combine dev
|
|
lhotse combine $manifests/iwslt-ta_supervisions_dev1.jsonl.gz $manifests/hkust_supervisions_dev1.jsonl.gz $manifests/fisher-sp_supervisions_dev.jsonl.gz ${manifests}/cts_supervisions_dev.jsonl.gz
|
|
|
|
lhotse combine $manifests/iwslt-ta_recordings_dev1.jsonl.gz $manifests/fisher-spanish_recordings_dev.jsonl.gz $manifests/hkust_recordings_dev1.jsonl.gz $manifests/fisher-sp_recordings_dev.jsonl.gz ${manifests}/cts_recordings_dev.jsonl.gz
|
|
|
|
# python local/cuts_validate.py --sup ${manifests}/cts_supervisions_dev.jsonl.gz --rec ${manifests}/cts_recordings_dev.jsonl.gz
|
|
fi
|
|
fi
|
|
|
|
if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
|
|
log "Stage 2: Prepare musan manifest"
|
|
# We assume that you have downloaded the musan corpus
|
|
# to data/musan
|
|
if [ ! -f ${manifests}/musan_recordings_speech.jsonl.gz ]; then
|
|
mkdir -p $manifests
|
|
lhotse prepare musan $dl_dir/musan $manifests
|
|
fi
|
|
fi
|
|
|
|
if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
|
|
log "Stage 3: Compute fbank features"
|
|
mkdir -p ${fbank}
|
|
./local/compute_fbank_gpu.py
|
|
./local/compute_fbank_gpu.py --test
|
|
fi
|
|
|
|
if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
|
|
log "Stage 4: Compute fbank for musan"
|
|
./local/compute_fbank_musan.py
|
|
fi
|
|
|
|
if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
|
|
log "Stage 6: Prepare BPE based lang"
|
|
for vocab_size in ${vocab_sizes[@]}; do
|
|
lang_dir=data/lang_bpe_${vocab_size}
|
|
mkdir -p ${lang_dir}
|
|
cp data/lang_phone/words.txt $lang_dir
|
|
if [ ! -f $lang_dir/transcript_words.txt ]; then
|
|
log "Generate text for BPE training from data/fbank/cuts_train.jsonl.gz"
|
|
python local/prepare_transcripts.py --cut ${fbank}/cuts_train.jsonl.gz --langdir ${lang_dir}
|
|
fi
|
|
./local/train_bpe_model.py \
|
|
--lang-dir $lang_dir \
|
|
--vocab-size $vocab_size \
|
|
--transcript $lang_dir/transcript_words.txt
|
|
done
|
|
fi
|
|
|
|
|
|
if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
|
|
log "Stage 7: Prepare BPE ST based lang"
|
|
for vocab_size in ${st_vocab_sizes[@]}; do
|
|
lang_dir=data/lang_st_bpe_${vocab_size}
|
|
mkdir -p ${lang_dir}
|
|
|
|
if [ ! -f $lang_dir/st_words.txt ]; then
|
|
log "Generate text for BPE training from data/fbank/cuts_train.jsonl.gz"
|
|
python local/prepare_st_transcripts.py --cut ${fbank}/cuts_train.jsonl.gz --langdir ${lang_dir}
|
|
fi
|
|
./local/train_bpe_model.py \
|
|
--lang-dir $lang_dir \
|
|
--vocab-size $vocab_size \
|
|
--transcript $lang_dir/st_words.txt
|
|
done
|
|
fi
|
|
|