torchaudio 2.0.2__cp39-cp39-win_amd64.whl → 2.1.1__cp39-cp39-win_amd64.whl

torchaudio/_backend/utils.py

@@ -1,268 +1,25 @@
 import os
-import re
-from abc import ABC, abstractmethod
 from functools import lru_cache
-from typing import BinaryIO, Dict, Optional, Tuple, Union
+from typing import BinaryIO, Dict, Optional, Tuple, Type, Union
 
 import torch
-import torchaudio.backend.soundfile_backend as soundfile_backend
-from torchaudio._extension import _FFMPEG_INITIALIZED, _SOX_INITIALIZED
-from torchaudio.backend.common import AudioMetaData
 
-if _FFMPEG_INITIALIZED:
-    from torchaudio.io._compat import info_audio, info_audio_fileobj, load_audio, load_audio_fileobj, save_audio
+from torchaudio._extension import _FFMPEG_EXT, _SOX_INITIALIZED
+from torchaudio.io import CodecConfig
 
+from . import soundfile_backend
 
-class Backend(ABC):
-    @staticmethod
-    @abstractmethod
-    def info(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str], buffer_size: int = 4096) -> AudioMetaData:
-        raise NotImplementedError
-
-    @staticmethod
-    @abstractmethod
-    def load(
-        uri: Union[BinaryIO, str, os.PathLike],
-        frame_offset: int = 0,
-        num_frames: int = -1,
-        normalize: bool = True,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        buffer_size: int = 4096,
-    ) -> Tuple[torch.Tensor, int]:
-        raise NotImplementedError
-
-    @staticmethod
-    @abstractmethod
-    def save(
-        uri: Union[BinaryIO, str, os.PathLike],
-        src: torch.Tensor,
-        sample_rate: int,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        encoding: Optional[str] = None,
-        bits_per_sample: Optional[int] = None,
-        buffer_size: int = 4096,
-    ) -> None:
-        raise NotImplementedError
-
-    @staticmethod
-    @abstractmethod
-    def can_decode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        raise NotImplementedError
-
-    @staticmethod
-    @abstractmethod
-    def can_encode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        raise NotImplementedError
-
-
-def _map_encoding(encoding: str) -> str:
-    for dst in ["PCM_S", "PCM_U", "PCM_F"]:
-        if dst in encoding:
-            return dst
-    if encoding == "PCM_MULAW":
-        return "ULAW"
-    elif encoding == "PCM_ALAW":
-        return "ALAW"
-    return encoding
-
-
-def _get_bits_per_sample(encoding: str, bits_per_sample: int) -> str:
-    if m := re.search(r"PCM_\w(\d+)\w*", encoding):
-        return int(m.group(1))
-    elif encoding in ["PCM_ALAW", "PCM_MULAW"]:
-        return 8
-    return bits_per_sample
-
-
-class FFmpegBackend(Backend):
-    @staticmethod
-    def info(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str], buffer_size: int = 4096) -> AudioMetaData:
-        if hasattr(uri, "read"):
-            metadata = info_audio_fileobj(uri, format, buffer_size=buffer_size)
-        else:
-            metadata = info_audio(os.path.normpath(uri), format)
-        metadata.bits_per_sample = _get_bits_per_sample(metadata.encoding, metadata.bits_per_sample)
-        metadata.encoding = _map_encoding(metadata.encoding)
-        return metadata
-
-    @staticmethod
-    def load(
-        uri: Union[BinaryIO, str, os.PathLike],
-        frame_offset: int = 0,
-        num_frames: int = -1,
-        normalize: bool = True,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        buffer_size: int = 4096,
-    ) -> Tuple[torch.Tensor, int]:
-        if hasattr(uri, "read"):
-            return load_audio_fileobj(
-                uri,
-                frame_offset,
-                num_frames,
-                normalize,
-                channels_first,
-                format,
-                buffer_size,
-            )
-        else:
-            return load_audio(os.path.normpath(uri), frame_offset, num_frames, normalize, channels_first, format)
-
-    @staticmethod
-    def save(
-        uri: Union[BinaryIO, str, os.PathLike],
-        src: torch.Tensor,
-        sample_rate: int,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        encoding: Optional[str] = None,
-        bits_per_sample: Optional[int] = None,
-        buffer_size: int = 4096,
-    ) -> None:
-        save_audio(
-            uri,
-            src,
-            sample_rate,
-            channels_first,
-            format,
-            encoding,
-            bits_per_sample,
-            buffer_size,
-        )
-
-    @staticmethod
-    def can_decode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        return True
-
-    @staticmethod
-    def can_encode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        return True
-
-
-class SoXBackend(Backend):
-    @staticmethod
-    def info(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str], buffer_size: int = 4096) -> AudioMetaData:
-        if hasattr(uri, "read"):
-            raise ValueError(
-                "SoX backend does not support reading from file-like objects. ",
-                "Please use an alternative backend that does support reading from file-like objects, e.g. FFmpeg.",
-            )
-        else:
-            sinfo = torch.ops.torchaudio.sox_io_get_info(uri, format)
-            if sinfo:
-                return AudioMetaData(*sinfo)
-            else:
-                raise RuntimeError(f"Failed to fetch metadata for {uri}.")
-
-    @staticmethod
-    def load(
-        uri: Union[BinaryIO, str, os.PathLike],
-        frame_offset: int = 0,
-        num_frames: int = -1,
-        normalize: bool = True,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        buffer_size: int = 4096,
-    ) -> Tuple[torch.Tensor, int]:
-        if hasattr(uri, "read"):
-            raise ValueError(
-                "SoX backend does not support loading from file-like objects. ",
-                "Please use an alternative backend that does support loading from file-like objects, e.g. FFmpeg.",
-            )
-        else:
-            ret = torch.ops.torchaudio.sox_io_load_audio_file(
-                uri, frame_offset, num_frames, normalize, channels_first, format
-            )
-            if not ret:
-                raise RuntimeError(f"Failed to load audio from {uri}.")
-            return ret
-
-    @staticmethod
-    def save(
-        uri: Union[BinaryIO, str, os.PathLike],
-        src: torch.Tensor,
-        sample_rate: int,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        encoding: Optional[str] = None,
-        bits_per_sample: Optional[int] = None,
-        buffer_size: int = 4096,
-    ) -> None:
-        if hasattr(uri, "write"):
-            raise ValueError(
-                "SoX backend does not support writing to file-like objects. ",
-                "Please use an alternative backend that does support writing to file-like objects, e.g. FFmpeg.",
-            )
-        else:
-            torch.ops.torchaudio.sox_io_save_audio_file(
-                uri,
-                src,
-                sample_rate,
-                channels_first,
-                None,
-                format,
-                encoding,
-                bits_per_sample,
-            )
-
-    @staticmethod
-    def can_decode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        # i.e. not a file-like object.
-        return not hasattr(uri, "read")
-
-    @staticmethod
-    def can_encode(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str]) -> bool:
-        # i.e. not a file-like object.
-        return not hasattr(uri, "write")
-
-
-class SoundfileBackend(Backend):
-    @abstractmethod
-    def info(uri: Union[BinaryIO, str, os.PathLike], format: Optional[str], buffer_size: int = 4096) -> AudioMetaData:
-        return soundfile_backend.info(uri, format)
-
-    @abstractmethod
-    def load(
-        uri: Union[BinaryIO, str, os.PathLike],
-        frame_offset: int = 0,
-        num_frames: int = -1,
-        normalize: bool = True,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        buffer_size: int = 4096,
-    ) -> Tuple[torch.Tensor, int]:
-        return soundfile_backend.load(uri, frame_offset, num_frames, normalize, channels_first, format)
-
-    @abstractmethod
-    def save(
-        uri: Union[BinaryIO, str, os.PathLike],
-        src: torch.Tensor,
-        sample_rate: int,
-        channels_first: bool = True,
-        format: Optional[str] = None,
-        encoding: Optional[str] = None,
-        bits_per_sample: Optional[int] = None,
-        buffer_size: int = 4096,
-    ) -> None:
-        soundfile_backend.save(
-            uri, src, sample_rate, channels_first, format=format, encoding=encoding, bits_per_sample=bits_per_sample
-        )
-
-    @abstractmethod
-    def can_decode(uri, format) -> bool:
-        return True
-
-    @abstractmethod
-    def can_encode(uri, format) -> bool:
-        return True
+from .backend import Backend
+from .common import AudioMetaData
+from .ffmpeg import FFmpegBackend
+from .soundfile import SoundfileBackend
+from .sox import SoXBackend
 
 
 @lru_cache(None)
-def get_available_backends() -> Dict[str, Backend]:
-    backend_specs = {}
-    if _FFMPEG_INITIALIZED:
+def get_available_backends() -> Dict[str, Type[Backend]]:
+    backend_specs: Dict[str, Type[Backend]] = {}
+    if _FFMPEG_EXT is not None:
         backend_specs["ffmpeg"] = FFmpegBackend
     if _SOX_INITIALIZED:
         backend_specs["sox"] = SoXBackend
@@ -303,19 +60,19 @@ def get_info_func():
     ) -> AudioMetaData:
         """Get signal information of an audio file.
 
+        Note:
+            When the input type is file-like object, this function cannot
+            get the correct length (``num_samples``) for certain formats,
+            such as ``vorbis``.
+            In this case, the value of ``num_samples`` is ``0``.
+
         Args:
             uri (path-like object or file-like object):
                 Source of audio data. The following types are accepted:
 
-                    * ``path-like``: file path
-                    * ``file-like``: Object with ``read(size: int) -> bytes`` method,
-                      which returns byte string of at most ``size`` length.
-
-                Note:
-                    When the input type is file-like object, this function cannot
-                    get the correct length (``num_samples``) for certain formats,
-                    such as ``vorbis``.
-                    In this case, the value of ``num_samples`` is ``0``.
+                * ``path-like``: File path or URL.
+                * ``file-like``: Object with ``read(size: int) -> bytes`` method,
+                  which returns byte string of at most ``size`` length.
 
             format (str or None, optional):
                 If not ``None``, interpreted as hint that may allow backend to override the detected format.
@@ -325,12 +82,17 @@ def get_info_func():
                 Size of buffer to use when processing file-like objects, in bytes. (Default: ``4096``)
 
             backend (str or None, optional):
-                I/O backend to use. If ``None``, function selects backend given input and available backends.
-                Otherwise, must be one of ["ffmpeg", "sox", "soundfile"], with the corresponding backend available.
+                I/O backend to use.
+                If ``None``, function selects backend given input and available backends.
+                Otherwise, must be one of [``"ffmpeg"``, ``"sox"``, ``"soundfile"``],
+                with the corresponding backend available.
                 (Default: ``None``)
 
+                .. seealso::
+                   :ref:`backend`
+
         Returns:
-            AudioMetaData: Metadata of the given audio.
+            AudioMetaData
         """
         backend = dispatcher(uri, format, backend)
         return backend.info(uri, format, buffer_size)
@@ -362,27 +124,19 @@ def get_load_func():
         buffer_size: int = 4096,
         backend: Optional[str] = None,
     ) -> Tuple[torch.Tensor, int]:
-        """Load audio data from file.
-
-        Note:
-            The formats this function can handle depend on backend availability.
-            This function is tested on the following formats:
-
-            * WAV
-
-                * 32-bit floating-point
-                * 32-bit signed integer
-                * 24-bit signed integer
-                * 16-bit signed integer
-                * 8-bit unsigned integer
-
-            * FLAC
-            * OGG/VORBIS
-            * SPHERE
+        """Load audio data from source.
 
         By default (``normalize=True``, ``channels_first=True``), this function returns Tensor with
         ``float32`` dtype, and the shape of `[channel, time]`.
 
+        Note:
+            The formats this function can handle depend on the availability of backends.
+            Please use the following functions to fetch the supported formats.
+
+            - FFmpeg: :py:func:`torchaudio.utils.ffmpeg_utils.get_audio_decoders`
+            - Sox: :py:func:`torchaudio.utils.sox_utils.list_read_formats`
+            - SoundFile: Refer to `the official document <https://pysoundfile.readthedocs.io/>`__.
+
         .. warning::
 
             ``normalize`` argument does not perform volume normalization.
@@ -432,9 +186,13 @@ def get_load_func():
                 Size of buffer to use when processing file-like objects, in bytes. (Default: ``4096``)
 
             backend (str or None, optional):
-                I/O backend to use. If ``None``, function selects backend given input and available backends.
-                Otherwise, must be one of ["ffmpeg", "sox", "soundfile"], with the corresponding
-                backend being available. (Default: ``None``)
+                I/O backend to use.
+                If ``None``, function selects backend given input and available backends.
+                Otherwise, must be one of [``"ffmpeg"``, ``"sox"``, ``"soundfile"``],
+                with the corresponding backend being available. (Default: ``None``)
+
+                .. seealso::
+                   :ref:`backend`
 
         Returns:
             (torch.Tensor, int): Resulting Tensor and sample rate.
@@ -472,22 +230,17 @@ def get_save_func():
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
         backend: Optional[str] = None,
+        compression: Optional[Union[CodecConfig, float, int]] = None,
     ):
         """Save audio data to file.
 
         Note:
             The formats this function can handle depend on the availability of backends.
-            This function is tested on the following formats:
+            Please use the following functions to fetch the supported formats.
 
-            * WAV
-
-                * 32-bit floating-point
-                * 32-bit signed integer
-                * 16-bit signed integer
-                * 8-bit unsigned integer
-
-            * FLAC
-            * OGG/VORBIS
+            - FFmpeg: :py:func:`torchaudio.utils.ffmpeg_utils.get_audio_encoders`
+            - Sox: :py:func:`torchaudio.utils.sox_utils.list_write_formats`
+            - SoundFile: Refer to `the official document <https://pysoundfile.readthedocs.io/>`__.
 
         Args:
             uri (str or pathlib.Path): Path to audio file.
@@ -508,11 +261,11 @@ def get_save_func():
                 This argument is effective only for supported formats, i.e.
                 ``"wav"`` and ``""flac"```. Valid values are
 
-                    - ``"PCM_S"`` (signed integer Linear PCM)
-                    - ``"PCM_U"`` (unsigned integer Linear PCM)
-                    - ``"PCM_F"`` (floating point PCM)
-                    - ``"ULAW"`` (mu-law)
-                    - ``"ALAW"`` (a-law)
+                - ``"PCM_S"`` (signed integer Linear PCM)
+                - ``"PCM_U"`` (unsigned integer Linear PCM)
+                - ``"PCM_F"`` (floating point PCM)
+                - ``"ULAW"`` (mu-law)
+                - ``"ALAW"`` (a-law)
 
             bits_per_sample (int or None, optional): Changes the bit depth for the
                 supported formats.
@@ -524,35 +277,40 @@ def get_save_func():
                 Size of buffer to use when processing file-like objects, in bytes. (Default: ``4096``)
 
             backend (str or None, optional):
-                I/O backend to use. If ``None``, function selects backend given input and available backends.
-                Otherwise, must be one of ["ffmpeg", "sox", "soundfile"], with the corresponding
-                backend being available. (Default: ``None``)
+                I/O backend to use.
+                If ``None``, function selects backend given input and available backends.
+                Otherwise, must be one of [``"ffmpeg"``, ``"sox"``, ``"soundfile"``],
+                with the corresponding backend being available.
+                (Default: ``None``)
 
+                .. seealso::
+                   :ref:`backend`
 
+            compression (CodecConfig, float, int, or None, optional):
+                Compression configuration to apply.
 
-        Supported formats/encodings/bit depth/compression are:
+                If the selected backend is FFmpeg, an instance of :py:class:`CodecConfig` must be provided.
 
-        ``"wav"``
-            - 32-bit floating-point PCM
-            - 32-bit signed integer PCM
-            - 24-bit signed integer PCM
-            - 16-bit signed integer PCM
-            - 8-bit unsigned integer PCM
-            - 8-bit mu-law
-            - 8-bit a-law
+                Otherwise, if the selected backend is SoX, a float or int value corresponding to option ``-C`` of the
+                ``sox`` command line interface must be provided. For instance:
 
-            Note:
-                Default encoding/bit depth is determined by the dtype of
-                the input Tensor.
+                ``"mp3"``
+                    Either bitrate (in ``kbps``) with quality factor, such as ``128.2``, or
+                    VBR encoding with quality factor such as ``-4.2``. Default: ``-4.5``.
 
-        ``"flac"``
-            - 16-bit (default)
-            - 24-bit
+                ``"flac"``
+                    Whole number from ``0`` to ``8``. ``8`` is default and highest compression.
+
+                ``"ogg"``, ``"vorbis"``
+                    Number from ``-1`` to ``10``; ``-1`` is the highest compression
+                    and lowest quality. Default: ``3``.
+
+                Refer to http://sox.sourceforge.net/soxformat.html for more details.
 
-        ``"ogg"``
-            - Doesn't accept changing configuration.
         """
         backend = dispatcher(uri, format, backend)
-        return backend.save(uri, src, sample_rate, channels_first, format, encoding, bits_per_sample, buffer_size)
+        return backend.save(
+            uri, src, sample_rate, channels_first, format, encoding, bits_per_sample, buffer_size, compression
+        )
 
     return save

torchaudio/_extension/__init__.py

@@ -2,9 +2,13 @@ import logging
 import os
 import sys
 
-from torchaudio._internal.module_utils import fail_with_message, is_module_available, no_op
+from torchaudio._internal.module_utils import eval_env, fail_with_message, is_module_available, no_op
 
-from .utils import _check_cuda_version, _fail_since_no_ffmpeg, _init_dll_path, _init_ffmpeg, _init_sox, _load_lib
+try:
+    from .fb import _init_ffmpeg
+except ImportError:
+    from .utils import _init_ffmpeg
+from .utils import _check_cuda_version, _fail_since_no_ffmpeg, _fail_since_no_sox, _init_dll_path, _init_sox, _load_lib
 
 _LG = logging.getLogger(__name__)
 
@@ -14,15 +18,13 @@ _LG = logging.getLogger(__name__)
 # Builder uses it for debugging purpose, so we export it.
 # https://github.com/pytorch/builder/blob/e2e4542b8eb0bdf491214451a1a4128bd606cce2/test/smoke_test/smoke_test.py#L80
 __all__ = [
-    "fail_if_no_kaldi",
     "fail_if_no_sox",
     "fail_if_no_ffmpeg",
     "_check_cuda_version",
     "_IS_TORCHAUDIO_EXT_AVAILABLE",
-    "_IS_KALDI_AVAILABLE",
     "_IS_RIR_AVAILABLE",
     "_SOX_INITIALIZED",
-    "_FFMPEG_INITIALIZED",
+    "_FFMPEG_EXT",
 ]
 
 
@@ -34,11 +36,11 @@ if os.name == "nt" and (3, 8) <= sys.version_info < (3, 9):
 # In case of an error, we do not catch the failure as it suggests there is something
 # wrong with the installation.
 _IS_TORCHAUDIO_EXT_AVAILABLE = is_module_available("torchaudio.lib._torchaudio")
-# Kaldi and RIR features are implemented in _torchaudio extension, but they can be individually
+# RIR features are implemented in _torchaudio extension, but they can be individually
 # turned on/off at build time. Available means that _torchaudio is loaded properly, and
-# Kaldi or RIR features are found there.
+# RIR features are found there.
 _IS_RIR_AVAILABLE = False
-_IS_KALDI_AVAILABLE = False
+_IS_ALIGN_AVAILABLE = False
 if _IS_TORCHAUDIO_EXT_AVAILABLE:
     _load_lib("libtorchaudio")
 
@@ -46,26 +48,45 @@ if _IS_TORCHAUDIO_EXT_AVAILABLE:
 
     _check_cuda_version()
     _IS_RIR_AVAILABLE = torchaudio.lib._torchaudio.is_rir_available()
-    _IS_KALDI_AVAILABLE = torchaudio.lib._torchaudio.is_kaldi_available()
+    _IS_ALIGN_AVAILABLE = torchaudio.lib._torchaudio.is_align_available()
 
 
-# Similar to libtorchaudio, sox-related features should be importable when present.
-#
-# Note: This will be change in the future when sox is dynamically linked.
-# At that point, this initialization should handle the case where
-# sox integration is built but libsox is not found.
+# Initialize libsox-related features
 _SOX_INITIALIZED = False
-if is_module_available("torchaudio.lib._torchaudio_sox"):
-    _init_sox()
-    _SOX_INITIALIZED = True
+_USE_SOX = False if os.name == "nt" else eval_env("TORCHAUDIO_USE_SOX", True)
+_SOX_MODULE_AVAILABLE = is_module_available("torchaudio.lib._torchaudio_sox")
+if _USE_SOX and _SOX_MODULE_AVAILABLE:
+    try:
+        _init_sox()
+        _SOX_INITIALIZED = True
+    except Exception:
+        # The initialization of sox extension will fail if supported sox
+        # libraries are not found in the system.
+        # Since the rest of the torchaudio works without it, we do not report the
+        # error here.
+        # The error will be raised when user code attempts to use these features.
+        _LG.debug("Failed to initialize sox extension", exc_info=True)
+
+
+if os.name == "nt":
+    fail_if_no_sox = fail_with_message("requires sox extension, which is not supported on Windows.")
+elif not _USE_SOX:
+    fail_if_no_sox = fail_with_message("requires sox extension, but it is disabled. (TORCHAUDIO_USE_SOX=0)")
+elif not _SOX_MODULE_AVAILABLE:
+    fail_if_no_sox = fail_with_message(
+        "requires sox extension, but TorchAudio is not compiled with it. "
+        "Please build TorchAudio with libsox support. (BUILD_SOX=1)"
+    )
+else:
+    fail_if_no_sox = no_op if _SOX_INITIALIZED else _fail_since_no_sox
 
 
 # Initialize FFmpeg-related features
-_FFMPEG_INITIALIZED = False
-if is_module_available("torchaudio.lib._torchaudio_ffmpeg"):
+_FFMPEG_EXT = None
+_USE_FFMPEG = eval_env("TORCHAUDIO_USE_FFMPEG", True)
+if _USE_FFMPEG and _IS_TORCHAUDIO_EXT_AVAILABLE:
     try:
-        _init_ffmpeg()
-        _FFMPEG_INITIALIZED = True
+        _FFMPEG_EXT = _init_ffmpeg()
     except Exception:
         # The initialization of FFmpeg extension will fail if supported FFmpeg
         # libraries are not found in the system.
@@ -75,22 +96,11 @@ if is_module_available("torchaudio.lib._torchaudio_ffmpeg"):
         _LG.debug("Failed to initialize ffmpeg bindings", exc_info=True)
 
 
-fail_if_no_kaldi = (
-    no_op
-    if _IS_KALDI_AVAILABLE
-    else fail_with_message(
-        "requires kaldi extension, but TorchAudio is not compiled with it. Please build TorchAudio with kaldi support."
-    )
-)
-fail_if_no_sox = (
-    no_op
-    if _SOX_INITIALIZED
-    else fail_with_message(
-        "requires sox extension, but TorchAudio is not compiled with it. Please build TorchAudio with libsox support."
-    )
-)
+if _USE_FFMPEG:
+    fail_if_no_ffmpeg = _fail_since_no_ffmpeg if _FFMPEG_EXT is None else no_op
+else:
+    fail_if_no_ffmpeg = fail_with_message("requires ffmpeg extension, but it is disabled. (TORCHAUDIO_USE_FFMPEG=0)")
 
-fail_if_no_ffmpeg = no_op if _FFMPEG_INITIALIZED else _fail_since_no_ffmpeg
 
 fail_if_no_rir = (
     no_op
@@ -99,3 +109,12 @@ fail_if_no_rir = (
         "requires RIR extension, but TorchAudio is not compiled with it. Please build TorchAudio with RIR support."
     )
 )
+
+fail_if_no_align = (
+    no_op
+    if _IS_ALIGN_AVAILABLE
+    else fail_with_message(
+        "Requires alignment extension, but TorchAudio is not compiled with it. \
+        Please build TorchAudio with alignment support."
+    )
+)