-
Notifications
You must be signed in to change notification settings - Fork 7
/
Copy pathencode.py
65 lines (52 loc) · 1.94 KB
/
encode.py
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
import argparse
import logging
from pathlib import Path
import numpy as np
from tqdm import tqdm
import torch
import torchaudio
import torchaudio.functional as AF
from urhythmic.model import encode
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
def encode_dataset(args):
logging.info("Loading hubert checkpoint")
hubert = torch.hub.load("bshall/hubert:main", "hubert_soft").cuda()
logging.info(f"Encoding dataset at {args.in_dir}")
for in_path in tqdm(list(args.in_dir.rglob(f"*{args.extension}"))):
wav, sr = torchaudio.load(in_path)
if sr != 16000:
raise ValueError(f"Sample rate: {sr} should be 16kHz.")
wav = wav.unsqueeze(0).cuda()
with torch.inference_mode():
units, log_probs = encode(hubert, wav)
units_out_path = args.out_dir / "soft" / in_path.relative_to(args.in_dir)
units_out_path.parent.mkdir(parents=True, exist_ok=True)
np.save(units_out_path.with_suffix(".npy"), units.squeeze().cpu().numpy())
probs_out_path = args.out_dir / "logprobs" / in_path.relative_to(args.in_dir)
probs_out_path.parent.mkdir(parents=True, exist_ok=True)
np.save(probs_out_path.with_suffix(".npy"), log_probs.squeeze().cpu().numpy())
if __name__ == "__main__":
parser = argparse.ArgumentParser(
description="Encode an audio dataset into soft speech units and the log probabilities of the associated discrete units."
)
parser.add_argument(
"in_dir",
metavar="in-dir",
help="path to the dataset directory.",
type=Path,
)
parser.add_argument(
"out_dir",
metavar="out-dir",
help="path to the output directory.",
type=Path,
)
parser.add_argument(
"--extension",
help="extension of the audio files (defaults to .wav).",
default=".wav",
type=str,
)
args = parser.parse_args()
encode_dataset(args)