import os
from typing import Dict, List, Optional, Union
import braceexpand
import torch
import webdataset as wd
from nemo.collections.asr.parts.preprocessing.segment import available_formats as valid_sf_formats
from nemo.collections.common.parts.preprocessing import collections
from nemo.core.classes import Dataset, IterableDataset
from nemo.core.neural_types import AudioSignal, LabelsType, LengthsType, NeuralType, RegressionValuesType
from nemo.utils import logging
# List of valid file formats (prioritized by order of importance)
VALID_FILE_FORMATS = ';'.join(['wav', 'mp3', 'flac'] +
[fmt.lower() for fmt in valid_sf_formats.keys()])
def repeat_signal(signal, sig_len, required_length):
"""repeat signal to make short signal to have required_length
Args:
signal (FloatTensor): input signal
sig_len (LongTensor): length of input signal
required_length(float) : length of generated signal
Returns:
signal (FloatTensor): generated signal of required_length by repeating itself.
"""
repeat = int(required_length // sig_len)
rem = int(required_length % sig_len)
sub = signal[-rem:] if rem > 0 else torch.tensor([])
rep_sig = torch.cat(repeat * [signal])
import io
import os
from typing import Dict, List, Optional, Union
import braceexpand
import torch
import webdataset as wd
from nemo.collections.asr.parts.preprocessing.segment import available_formats as valid_sf_formats
from nemo.collections.common.parts.preprocessing import collections
from nemo.core.classes import Dataset, IterableDataset
from nemo.core.neural_types import AudioSignal, LabelsType, LengthsType, NeuralType, RegressionValuesType
from nemo.utils import logging
# List of valid file formats (prioritized by order of importance)
VALID_FILE_FORMATS = ';'.join(['wav', 'mp3', 'flac'] + [fmt.lower() for fmt in valid_sf_formats.keys()])
def repeat_signal(signal, sig_len, required_length):
"""repeat signal to make short signal to have required_length
Args:
signal (FloatTensor): input signal
sig_len (LongTensor): length of input signal
required_length(float) : length of generated signal
Returns:
signal (FloatTensor): generated signal of required_length by repeating itself.
"""
repeat = int(required_length // sig_len)
rem = int(required_length % sig_len)
sub = signal[-rem:] if rem > 0 else torch.tensor([])
rep_sig = torch.cat(repeat * [signal])