Spaces:
Running
Running
File size: 5,217 Bytes
864affd |
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 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 |
# To use this file, the dependency (https://github.com/vesis84/kaldi-io-for-python)
# needs to be installed. This is a light wrapper around kaldi_io that returns
# torch.Tensors.
from typing import Any, Callable, Iterable, Tuple
import torch
from torch import Tensor
from torchaudio._internal import module_utils as _mod_utils
if _mod_utils.is_module_available("numpy"):
import numpy as np
__all__ = [
"read_vec_int_ark",
"read_vec_flt_scp",
"read_vec_flt_ark",
"read_mat_scp",
"read_mat_ark",
]
def _convert_method_output_to_tensor(
file_or_fd: Any, fn: Callable, convert_contiguous: bool = False
) -> Iterable[Tuple[str, Tensor]]:
r"""Takes a method invokes it. The output is converted to a tensor.
Args:
file_or_fd (str/FileDescriptor): File name or file descriptor
fn (Callable): Function that has the signature (file name/descriptor) and converts it to
Iterable[Tuple[str, Tensor]].
convert_contiguous (bool, optional): Determines whether the array should be converted into a
contiguous layout. (Default: ``False``)
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is vec/mat
"""
for key, np_arr in fn(file_or_fd):
if convert_contiguous:
np_arr = np.ascontiguousarray(np_arr)
yield key, torch.from_numpy(np_arr)
@_mod_utils.requires_module("kaldi_io", "numpy")
def read_vec_int_ark(file_or_fd: Any) -> Iterable[Tuple[str, Tensor]]:
r"""Create generator of (key,vector<int>) tuples, which reads from the ark file/stream.
Args:
file_or_fd (str/FileDescriptor): ark, gzipped ark, pipe or opened file descriptor
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is the vector read from file
Example
>>> # read ark to a 'dictionary'
>>> d = { u:d for u,d in torchaudio.kaldi_io.read_vec_int_ark(file) }
"""
import kaldi_io
# Requires convert_contiguous to be True because elements from int32 vector are
# sorted in tuples: (sizeof(int32), value) so strides are (5,) instead of (4,) which will throw an error
# in from_numpy as it expects strides to be a multiple of 4 (int32).
return _convert_method_output_to_tensor(file_or_fd, kaldi_io.read_vec_int_ark, convert_contiguous=True)
@_mod_utils.requires_module("kaldi_io", "numpy")
def read_vec_flt_scp(file_or_fd: Any) -> Iterable[Tuple[str, Tensor]]:
r"""Create generator of (key,vector<float32/float64>) tuples, read according to Kaldi scp.
Args:
file_or_fd (str/FileDescriptor): scp, gzipped scp, pipe or opened file descriptor
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is the vector read from file
Example
>>> # read scp to a 'dictionary'
>>> # d = { u:d for u,d in torchaudio.kaldi_io.read_vec_flt_scp(file) }
"""
import kaldi_io
return _convert_method_output_to_tensor(file_or_fd, kaldi_io.read_vec_flt_scp)
@_mod_utils.requires_module("kaldi_io", "numpy")
def read_vec_flt_ark(file_or_fd: Any) -> Iterable[Tuple[str, Tensor]]:
r"""Create generator of (key,vector<float32/float64>) tuples, which reads from the ark file/stream.
Args:
file_or_fd (str/FileDescriptor): ark, gzipped ark, pipe or opened file descriptor
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is the vector read from file
Example
>>> # read ark to a 'dictionary'
>>> d = { u:d for u,d in torchaudio.kaldi_io.read_vec_flt_ark(file) }
"""
import kaldi_io
return _convert_method_output_to_tensor(file_or_fd, kaldi_io.read_vec_flt_ark)
@_mod_utils.requires_module("kaldi_io", "numpy")
def read_mat_scp(file_or_fd: Any) -> Iterable[Tuple[str, Tensor]]:
r"""Create generator of (key,matrix<float32/float64>) tuples, read according to Kaldi scp.
Args:
file_or_fd (str/FileDescriptor): scp, gzipped scp, pipe or opened file descriptor
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is the matrix read from file
Example
>>> # read scp to a 'dictionary'
>>> d = { u:d for u,d in torchaudio.kaldi_io.read_mat_scp(file) }
"""
import kaldi_io
return _convert_method_output_to_tensor(file_or_fd, kaldi_io.read_mat_scp)
@_mod_utils.requires_module("kaldi_io", "numpy")
def read_mat_ark(file_or_fd: Any) -> Iterable[Tuple[str, Tensor]]:
r"""Create generator of (key,matrix<float32/float64>) tuples, which reads from the ark file/stream.
Args:
file_or_fd (str/FileDescriptor): ark, gzipped ark, pipe or opened file descriptor
Returns:
Iterable[Tuple[str, Tensor]]: The string is the key and the tensor is the matrix read from file
Example
>>> # read ark to a 'dictionary'
>>> d = { u:d for u,d in torchaudio.kaldi_io.read_mat_ark(file) }
"""
import kaldi_io
return _convert_method_output_to_tensor(file_or_fd, kaldi_io.read_mat_ark)
|