mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-09 01:52:41 +00:00
Code Issues Packages Projects Releases Wiki Activity

instead of on-the-fly features, precompute fbank and manifests in prepare.sh

Browse Source
This commit is contained in:
Kinan Martin 2025-04-23 10:13:15 +09:00
parent 4e2a4fdcd8
commit 4f743993ef
2 changed files with 31 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
egs/mls_english/ASR/local/utils/compute_fbank_mls_english.py
View File

@ -94,6 +94,7 @@ def get_args():
)
parser.add_argument("-m", "--manifest-dir", type=Path)
parser.add_argument("-a", "--audio-dir", type=Path)
parser.add_argument("-d", "--dl-dir", type=Path)
return parser.parse_args()
@ -114,7 +115,7 @@ def main():
)
return
else:
mls_eng_hf_dataset_path = "/root/datasets/parler-tts--mls_eng"
mls_eng_hf_dataset_path = args.dl_dir # "/root/datasets/parler-tts--mls_eng"
cut_sets = make_cutset_blueprints(mls_eng_hf_dataset_path)
for part, cut_set in cut_sets:
logging.info(f"Processing {part}")
@ -125,8 +126,8 @@ def main():
storage_type=LilcomChunkyWriter,
)
# cut_set.save_audios(args.audio_dir)
# cut_set.to_file(args.manifest_dir / f"mls_eng_cuts_{part}.jsonl.gz")
cut_set = cut_set.save_audios(args.audio_dir / part) # makes new cutset that uses paths to actual audio files
cut_set.to_file(args.manifest_dir / f"mls_eng_cuts_{part}.jsonl.gz")
logging.info("All fbank computed for MLS English.")
(args.manifest_dir / ".mls-eng-fbank.done").touch()

31
egs/mls_english/ASR/prepare.sh
View File

@ -41,19 +41,42 @@ if [ $stage -le 0 ] && [ $stop_stage -ge 0 ]; then
fi
fi
if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
log "Stage 1: Prepare MLS English manifests and compute fbank"
# We assume that you have downloaded the MLS English corpus
# to $dl_dir/mls_english
mkdir -p data/manifests
if [ ! -e data/mls_english.done ]; then
# lhotse prepare mls_english -j $nj $dl_dir/mls_english data/manifests
python local/compute_fbank_mls_english.py --manifest-dir data/manifests --audio-dir data/audio --dl-dir $dl_dir/mls_english
touch data/manifests/.mls_english.done
fi
fi
if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
log "Stage 2: Validate MLS English manifests"
if [ ! -e data/manifests/.mls_english-validated.done ]; then
python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_train.jsonl.gz
python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_dev.jsonl.gz
python local/validate_manifest.py --manifest data/manifests/mls_english_cuts_test.jsonl.gz
touch data/manifests/.mls_english-validated.done
fi
fi
mkdir -p data/lang
lang_dir=data/lang
if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
log "Stage 1: Prepare transcript for BPE training"
if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
log "Stage 3: Prepare transcript for BPE training"
if [ ! -f $lang_dir/transcript.txt ]; then
log "Generating transcripts for BPE training"
./local/utils/generate_transcript.py --lang-dir $lang_dir
fi
fi
if [ $stage -le 2 ] && [ $stop_stage -ge 2 ]; then
log "Stage 2: Prepare BPE tokenizer"
if [ $stage -le 4 ] && [ $stop_stage -ge 4 ]; then
log "Stage 4: Prepare BPE tokenizer"
for vocab_size in ${vocab_sizes[@]}; do
log "Training BPE model with vocab_size=${vocab_size}"