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

optimize with num_jobs on save_audios

Browse Source
This commit is contained in:
Kinan Martin 2025-05-02 07:22:38 +09:00
parent 88249f0eb4
commit 547f5c5cfb
1 changed files with 5 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
egs/mls_english/ASR/local/compute_fbank_mls_english.py
View File

@ -120,6 +120,11 @@ def main():
cut_sets = make_cutset_blueprints(mls_eng_hf_dataset_path)
for part, cut_set in cut_sets:
logging.info(f"Processing {part}")
cut_set = cut_set.save_audios(
num_jobs=num_jobs,
storage_path=(args.audio_dir / part).as_posix(),
) # makes new cutset that loads audio from paths to actual audio files
cut_set = cut_set.compute_and_store_features(
extractor=extractor,
num_jobs=num_jobs,
@ -127,7 +132,6 @@ def main():
storage_type=LilcomChunkyWriter,
)
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.")