"""
Create pseudo data
Authors
* Leo 2022
"""
import random
import shutil
import tempfile
from pathlib import Path
from typing import Any, List
import torch
import torchaudio
from torch.nn.utils.rnn import pad_sequence
SAMPLE_RATE = 16000
__all__ = [
"pseudo_audio",
"get_pseudo_wavs",
]
[docs]class pseudo_audio:
"""
This context manager returns filepaths (List[str]) and num_samples (List[int]) on entering
"""
def __init__(self, secs: List[float], sample_rate: int = SAMPLE_RATE):
self.tempdir = Path(tempfile.TemporaryDirectory().name)
self.tempdir.mkdir(parents=True, exist_ok=True)
self.num_samples = []
for n, sec in enumerate(secs):
wav = torch.randn(1, round(sample_rate * sec))
torchaudio.save(
str(self.tempdir / f"audio_{n}.wav"), wav, sample_rate=sample_rate
)
self.num_samples.append(wav.size(-1))
self.filepaths = [
str(self.tempdir / f"audio_{i}.wav") for i in range(len(secs))
]
def __enter__(self):
return self.filepaths, self.num_samples
def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> None:
shutil.rmtree(self.tempdir)
[docs]def get_pseudo_wavs(
seed: int = 0,
n: int = 2,
min_secs: int = 1,
max_secs: int = 3,
sample_rate: int = SAMPLE_RATE,
device: str = "cpu",
padded: bool = False,
):
random.seed(seed)
torch.manual_seed(seed)
wavs = []
wavs_len = []
for _ in range(n):
wav_length = random.randint(min_secs * sample_rate, max_secs * sample_rate)
wav = torch.randn(wav_length, requires_grad=True).to(device)
wavs_len.append(wav_length)
wavs.append(wav)
if not padded:
return wavs
else:
return pad_sequence(wavs, batch_first=True), torch.LongTensor(wavs_len)