regression

This commit is contained in:
Yuekai Zhang 2024-01-23 17:53:55 +08:00
parent f66b266aa4
commit 08db3051ad

View File

@ -137,44 +137,44 @@ def compute_fbank_wenetspeech_splits(args):
set_audio_duration_mismatch_tolerance(0.01) # 10ms tolerance
set_caching_enabled(False)
with get_executor() as ex: # Initialize the executor only once.
for i in range(start, stop):
idx = f"{i + 1}".zfill(num_digits)
logging.info(f"Processing {idx}/{num_splits}")
#with get_executor() as ex: # Initialize the executor only once.
for i in range(start, stop):
idx = f"{i + 1}".zfill(num_digits)
logging.info(f"Processing {idx}/{num_splits}")
cuts_path = output_dir / f"cuts_{subset}.{idx}.jsonl.gz"
if cuts_path.is_file():
logging.info(f"{cuts_path} exists - skipping")
continue
cuts_path = output_dir / f"cuts_{subset}.{idx}.jsonl.gz"
if cuts_path.is_file():
logging.info(f"{cuts_path} exists - skipping")
continue
raw_cuts_path = output_dir / f"cuts_{subset}_raw.{idx}.jsonl.gz"
raw_cuts_path = output_dir / f"cuts_{subset}_raw.{idx}.jsonl.gz"
logging.info(f"Loading {raw_cuts_path}")
cut_set = CutSet.from_file(raw_cuts_path)
logging.info(f"Loading {raw_cuts_path}")
cut_set = CutSet.from_file(raw_cuts_path)
logging.info("Splitting cuts into smaller chunks.")
cut_set = cut_set.trim_to_supervisions(
keep_overlapping=False, min_duration=None
)
logging.info("Splitting cuts into smaller chunks.")
cut_set = cut_set.trim_to_supervisions(
keep_overlapping=False, min_duration=None
)
logging.info("Computing features")
# cut_set = cut_set.compute_and_store_features_batch(
# extractor=extractor,
# storage_path=f"{output_dir}/feats_{subset}_{idx}",
# num_workers=args.num_workers,
# batch_duration=args.batch_duration,
# storage_type=LilcomChunkyWriter,
# overwrite=True,
# )
cut_set = cut_set.compute_and_store_features(
extractor=extractor,
storage_path=f"{output_dir}/feats_{subset}_{idx}",
num_jobs=args.num_workers,
executor=ex,
storage_type=LilcomChunkyWriter,
)
logging.info(f"Saving to {cuts_path}")
cut_set.to_file(cuts_path)
logging.info("Computing features")
cut_set = cut_set.compute_and_store_features_batch(
extractor=extractor,
storage_path=f"{output_dir}/feats_{subset}_{idx}",
num_workers=args.num_workers,
batch_duration=args.batch_duration,
storage_type=LilcomChunkyWriter,
overwrite=True,
)
# cut_set = cut_set.compute_and_store_features(
# extractor=extractor,
# storage_path=f"{output_dir}/feats_{subset}_{idx}",
# num_jobs=args.num_workers,
# executor=ex,
# storage_type=LilcomChunkyWriter,
# )
logging.info(f"Saving to {cuts_path}")
cut_set.to_file(cuts_path)
def main():