Spaces:
Running
Running
File size: 2,615 Bytes
0d735a2 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 |
# Copyright (c) Facebook, Inc. and its affiliates.
#
# This source code is licensed under the MIT license found in the
# LICENSE file in the root directory of this source tree.
import argparse
import logging
from examples.textless_nlp.gslm.speech2unit.pretrained.utils import (
get_and_dump_features,
)
def get_parser():
parser = argparse.ArgumentParser(
description="Compute and dump log mel fbank features."
)
parser.add_argument(
"--feature_type",
type=str,
choices=["logmel", "hubert", "w2v2", "cpc"],
default=None,
help="Acoustic feature type",
)
parser.add_argument(
"--manifest_path",
type=str,
default=None,
help="Manifest file containing the root dir and file names",
)
parser.add_argument(
"--out_features_path",
type=str,
default=None,
help="Features file path to write to",
)
parser.add_argument(
"--checkpoint_path",
type=str,
help="Pretrained acoustic model checkpoint",
)
parser.add_argument(
"--layer",
type=int,
help="The layer of the pretrained model to extract features from",
default=-1,
)
parser.add_argument(
"--sample_pct",
type=float,
help="Percent data to use for K-means training",
default=0.1,
)
parser.add_argument(
"--out_features_path",
type=str,
help="Path to save log mel fbank features",
)
return parser
def get_logger():
log_format = "[%(asctime)s] [%(levelname)s]: %(message)s"
logging.basicConfig(format=log_format, level=logging.INFO)
logger = logging.getLogger(__name__)
return logger
if __name__ == "__main__":
"""
Example command:
python ~/speechbot/clustering/dump_logmelfank_feats.py \
--manifest_path /checkpoint/kushall/data/LJSpeech-1.1/asr_input_wavs_16k/train.tsv
--out_features_path /checkpoint/kushall/experiments/speechbot/logmelfbank/features/ljspeech/train.npy
"""
parser = get_parser()
args = parser.parse_args()
logger = get_logger()
logger.info(args)
logger.info(f"Extracting {args.feature_type} acoustic features...")
get_and_dump_features(
feature_type=args.feature_type,
checkpoint_path=args.checkpoint_path,
layer=args.layer,
manifest_path=args.manifest_path,
sample_pct=args.sample_pct,
flatten=True,
out_features_path=args.out_features_path,
)
logger.info(f"Saved extracted features at {args.out_features_path}")
|