auracaster/piper
3
0
Fork 0
mirror of https://github.com/pstrueb/piper.git synced 2026-04-18 06:15:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add --single-speaker to preprocess

Browse Source
This commit is contained in:
Michael Hansen
2023-02-24 12:54:47 -06:00
parent 5493e5dd4c
commit 657a1fae74
1 changed files with 10 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
src/python/larynx_train/preprocess.py
View File

@@ -42,6 +42,9 @@ def main() -> None:
)
parser.add_argument("--cache-dir", help="Directory to cache processed audio files")
parser.add_argument("--max-workers", type=int)
parser.add_argument(
"--single-speaker", action="store_true", help="Force single speaker dataset"
)
parser.add_argument(
"--debug", action="store_true", help="Print DEBUG messages to the console"
)
@@ -75,7 +78,7 @@ def main() -> None:
_LOGGER.debug("Counting number of speakers/utterances in the dataset")
speaker_counts: Counter[str] = Counter()
num_utterances = 0
for utt in make_dataset(args.input_dir):
for utt in make_dataset(args.input_dir, args.single_speaker):
speaker = utt.speaker or ""
speaker_counts[speaker] += 1
num_utterances += 1
@@ -142,7 +145,9 @@ def main() -> None:
"Processing %s utterance(s) with %s worker(s)", num_utterances, args.max_workers
)
with open(args.output_dir / "dataset.jsonl", "w", encoding="utf-8") as dataset_file:
for utt_batch in batched(make_dataset(args.input_dir), batch_size):
for utt_batch in batched(
make_dataset(args.input_dir, args.single_speaker), batch_size
):
queue_in.put(utt_batch)
_LOGGER.debug("Waiting for jobs to finish")
@@ -226,7 +231,7 @@ class PathEncoder(json.JSONEncoder):
return super().default(o)
def ljspeech_dataset(dataset_dir: Path) -> Iterable[Utterance]:
def ljspeech_dataset(dataset_dir: Path, is_single_speaker: bool) -> Iterable[Utterance]:
# filename|speaker|text
# speaker is optional
metadata_path = dataset_dir / "metadata.csv"
@@ -242,7 +247,7 @@ def ljspeech_dataset(dataset_dir: Path) -> Iterable[Utterance]:
assert len(row) >= 2, "Not enough colums"
speaker: Optional[str] = None
if len(row) == 2:
if is_single_speaker or (len(row) == 2):
filename, text = row[0], row[-1]
else:
filename, speaker, text = row[0], row[1], row[-1]
@@ -269,7 +274,7 @@ def ljspeech_dataset(dataset_dir: Path) -> Iterable[Utterance]:
yield Utterance(text=text, audio_path=wav_path, speaker=speaker)
def mycroft_dataset(dataset_dir: Path) -> Iterable[Utterance]:
def mycroft_dataset(dataset_dir: Path, is_single_speaker: bool) -> Iterable[Utterance]:
for info_path in dataset_dir.glob("*.info"):
wav_path = info_path.with_suffix(".wav")
if wav_path.exists():