torchaudio 2.9.1__cp311-cp311-manylinux_2_28_aarch64.whl

torchaudio/datasets/__init__.py

@@ -0,0 +1,47 @@
+from .cmuarctic import CMUARCTIC
+from .cmudict import CMUDict
+from .commonvoice import COMMONVOICE
+from .dr_vctk import DR_VCTK
+from .fluentcommands import FluentSpeechCommands
+from .gtzan import GTZAN
+from .iemocap import IEMOCAP
+from .librilight_limited import LibriLightLimited
+from .librimix import LibriMix
+from .librispeech import LIBRISPEECH
+from .librispeech_biasing import LibriSpeechBiasing
+from .libritts import LIBRITTS
+from .ljspeech import LJSPEECH
+from .musdb_hq import MUSDB_HQ
+from .quesst14 import QUESST14
+from .snips import Snips
+from .speechcommands import SPEECHCOMMANDS
+from .tedlium import TEDLIUM
+from .vctk import VCTK_092
+from .voxceleb1 import VoxCeleb1Identification, VoxCeleb1Verification
+from .yesno import YESNO
+
+
+__all__ = [
+    "COMMONVOICE",
+    "LIBRISPEECH",
+    "LibriSpeechBiasing",
+    "LibriLightLimited",
+    "SPEECHCOMMANDS",
+    "VCTK_092",
+    "DR_VCTK",
+    "YESNO",
+    "LJSPEECH",
+    "GTZAN",
+    "CMUARCTIC",
+    "CMUDict",
+    "LibriMix",
+    "LIBRITTS",
+    "TEDLIUM",
+    "QUESST14",
+    "MUSDB_HQ",
+    "FluentSpeechCommands",
+    "VoxCeleb1Identification",
+    "VoxCeleb1Verification",
+    "IEMOCAP",
+    "Snips",
+]

torchaudio/datasets/cmuarctic.py

@@ -0,0 +1,157 @@
+import csv
+import os
+from pathlib import Path
+from typing import Tuple, Union
+
+import torchaudio
+from torch import Tensor
+from torch.utils.data import Dataset
+from torchaudio._internal import download_url_to_file
+from torchaudio.datasets.utils import _extract_tar
+
+URL = "aew"
+FOLDER_IN_ARCHIVE = "ARCTIC"
+_CHECKSUMS = {
+    "http://festvox.org/cmu_arctic/packed/cmu_us_aew_arctic.tar.bz2": "645cb33c0f0b2ce41384fdd8d3db2c3f5fc15c1e688baeb74d2e08cab18ab406",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_ahw_arctic.tar.bz2": "024664adeb892809d646a3efd043625b46b5bfa3e6189b3500b2d0d59dfab06c",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_aup_arctic.tar.bz2": "2c55bc3050caa996758869126ad10cf42e1441212111db034b3a45189c18b6fc",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_awb_arctic.tar.bz2": "d74a950c9739a65f7bfc4dfa6187f2730fa03de5b8eb3f2da97a51b74df64d3c",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_axb_arctic.tar.bz2": "dd65c3d2907d1ee52f86e44f578319159e60f4bf722a9142be01161d84e330ff",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_bdl_arctic.tar.bz2": "26b91aaf48b2799b2956792b4632c2f926cd0542f402b5452d5adecb60942904",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_clb_arctic.tar.bz2": "3f16dc3f3b97955ea22623efb33b444341013fc660677b2e170efdcc959fa7c6",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_eey_arctic.tar.bz2": "8a0ee4e5acbd4b2f61a4fb947c1730ab3adcc9dc50b195981d99391d29928e8a",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_fem_arctic.tar.bz2": "3fcff629412b57233589cdb058f730594a62c4f3a75c20de14afe06621ef45e2",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_gka_arctic.tar.bz2": "dc82e7967cbd5eddbed33074b0699128dbd4482b41711916d58103707e38c67f",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_jmk_arctic.tar.bz2": "3a37c0e1dfc91e734fdbc88b562d9e2ebca621772402cdc693bbc9b09b211d73",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_ksp_arctic.tar.bz2": "8029cafce8296f9bed3022c44ef1e7953332b6bf6943c14b929f468122532717",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_ljm_arctic.tar.bz2": "b23993765cbf2b9e7bbc3c85b6c56eaf292ac81ee4bb887b638a24d104f921a0",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_lnh_arctic.tar.bz2": "4faf34d71aa7112813252fb20c5433e2fdd9a9de55a00701ffcbf05f24a5991a",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_rms_arctic.tar.bz2": "c6dc11235629c58441c071a7ba8a2d067903dfefbaabc4056d87da35b72ecda4",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_rxr_arctic.tar.bz2": "1fa4271c393e5998d200e56c102ff46fcfea169aaa2148ad9e9469616fbfdd9b",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_slp_arctic.tar.bz2": "54345ed55e45c23d419e9a823eef427f1cc93c83a710735ec667d068c916abf1",  # noqa: E501
+    "http://festvox.org/cmu_arctic/packed/cmu_us_slt_arctic.tar.bz2": "7c173297916acf3cc7fcab2713be4c60b27312316765a90934651d367226b4ea",  # noqa: E501
+}
+
+
+def load_cmuarctic_item(line: str, path: str, folder_audio: str, ext_audio: str) -> Tuple[Tensor, int, str, str]:
+
+    utterance_id, transcript = line[0].strip().split(" ", 2)[1:]
+
+    # Remove space, double quote, and single parenthesis from transcript
+    transcript = transcript[1:-3]
+
+    file_audio = os.path.join(path, folder_audio, utterance_id + ext_audio)
+
+    # Load audio
+    waveform, sample_rate = torchaudio.load(file_audio)
+
+    return (waveform, sample_rate, transcript, utterance_id.split("_")[1])
+
+
+class CMUARCTIC(Dataset):
+    """*CMU ARCTIC* :cite:`Kominek03cmuarctic` dataset.
+
+    Args:
+        root (str or Path): Path to the directory where the dataset is found or downloaded.
+        url (str, optional):
+            The URL to download the dataset from or the type of the dataset to download.
+            (default: ``"aew"``)
+            Allowed type values are ``"aew"``, ``"ahw"``, ``"aup"``, ``"awb"``, ``"axb"``, ``"bdl"``,
+            ``"clb"``, ``"eey"``, ``"fem"``, ``"gka"``, ``"jmk"``, ``"ksp"``, ``"ljm"``, ``"lnh"``,
+            ``"rms"``, ``"rxr"``, ``"slp"`` or ``"slt"``.
+        folder_in_archive (str, optional):
+            The top-level directory of the dataset. (default: ``"ARCTIC"``)
+        download (bool, optional):
+            Whether to download the dataset if it is not found at root path. (default: ``False``).
+    """
+
+    _file_text = "txt.done.data"
+    _folder_text = "etc"
+    _ext_audio = ".wav"
+    _folder_audio = "wav"
+
+    def __init__(
+        self, root: Union[str, Path], url: str = URL, folder_in_archive: str = FOLDER_IN_ARCHIVE, download: bool = False
+    ) -> None:
+
+        if url in [
+            "aew",
+            "ahw",
+            "aup",
+            "awb",
+            "axb",
+            "bdl",
+            "clb",
+            "eey",
+            "fem",
+            "gka",
+            "jmk",
+            "ksp",
+            "ljm",
+            "lnh",
+            "rms",
+            "rxr",
+            "slp",
+            "slt",
+        ]:
+
+            url = "cmu_us_" + url + "_arctic"
+            ext_archive = ".tar.bz2"
+            base_url = "http://www.festvox.org/cmu_arctic/packed/"
+
+            url = os.path.join(base_url, url + ext_archive)
+
+        # Get string representation of 'root' in case Path object is passed
+        root = os.fspath(root)
+
+        basename = os.path.basename(url)
+        root = os.path.join(root, folder_in_archive)
+        if not os.path.isdir(root):
+            os.mkdir(root)
+        archive = os.path.join(root, basename)
+
+        basename = basename.split(".")[0]
+
+        self._path = os.path.join(root, basename)
+
+        if download:
+            if not os.path.isdir(self._path):
+                if not os.path.isfile(archive):
+                    checksum = _CHECKSUMS.get(url, None)
+                    download_url_to_file(url, archive, hash_prefix=checksum)
+                _extract_tar(archive)
+        else:
+            if not os.path.exists(self._path):
+                raise RuntimeError(
+                    f"The path {self._path} doesn't exist. "
+                    "Please check the ``root`` path or set `download=True` to download it"
+                )
+        self._text = os.path.join(self._path, self._folder_text, self._file_text)
+
+        with open(self._text, "r") as text:
+            walker = csv.reader(text)
+            self._walker = list(walker)
+
+    def __getitem__(self, n: int) -> Tuple[Tensor, int, str, str]:
+        """Load the n-th sample from the dataset.
+
+        Args:
+            n (int): The index of the sample to be loaded
+
+        Returns:
+            Tuple of the following items;
+
+            Tensor:
+                Waveform
+            int:
+                Sample rate
+            str:
+                Transcript
+            str:
+                Utterance ID
+        """
+        line = self._walker[n]
+        return load_cmuarctic_item(line, self._path, self._folder_audio, self._ext_audio)
+
+    def __len__(self) -> int:
+        return len(self._walker)

torchaudio/datasets/cmudict.py

@@ -0,0 +1,186 @@
+import os
+import re
+from pathlib import Path
+from typing import Iterable, List, Tuple, Union
+
+from torch.utils.data import Dataset
+from torchaudio._internal import download_url_to_file
+
+
+_CHECKSUMS = {
+    "http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b": "209a8b4cd265013e96f4658632a9878103b0c5abf62b50d4ef3ae1be226b29e4",  # noqa: E501
+    "http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b.symbols": "408ccaae803641c6d7b626b6299949320c2dbca96b2220fd3fb17887b023b027",  # noqa: E501
+}
+_PUNCTUATIONS = {
+    "!EXCLAMATION-POINT",
+    '"CLOSE-QUOTE',
+    '"DOUBLE-QUOTE',
+    '"END-OF-QUOTE',
+    '"END-QUOTE',
+    '"IN-QUOTES',
+    '"QUOTE',
+    '"UNQUOTE',
+    "#HASH-MARK",
+    "#POUND-SIGN",
+    "#SHARP-SIGN",
+    "%PERCENT",
+    "&AMPERSAND",
+    "'END-INNER-QUOTE",
+    "'END-QUOTE",
+    "'INNER-QUOTE",
+    "'QUOTE",
+    "'SINGLE-QUOTE",
+    "(BEGIN-PARENS",
+    "(IN-PARENTHESES",
+    "(LEFT-PAREN",
+    "(OPEN-PARENTHESES",
+    "(PAREN",
+    "(PARENS",
+    "(PARENTHESES",
+    ")CLOSE-PAREN",
+    ")CLOSE-PARENTHESES",
+    ")END-PAREN",
+    ")END-PARENS",
+    ")END-PARENTHESES",
+    ")END-THE-PAREN",
+    ")PAREN",
+    ")PARENS",
+    ")RIGHT-PAREN",
+    ")UN-PARENTHESES",
+    "+PLUS",
+    ",COMMA",
+    "--DASH",
+    "-DASH",
+    "-HYPHEN",
+    "...ELLIPSIS",
+    ".DECIMAL",
+    ".DOT",
+    ".FULL-STOP",
+    ".PERIOD",
+    ".POINT",
+    "/SLASH",
+    ":COLON",
+    ";SEMI-COLON",
+    ";SEMI-COLON(1)",
+    "?QUESTION-MARK",
+    "{BRACE",
+    "{LEFT-BRACE",
+    "{OPEN-BRACE",
+    "}CLOSE-BRACE",
+    "}RIGHT-BRACE",
+}
+
+
+def _parse_dictionary(lines: Iterable[str], exclude_punctuations: bool) -> List[str]:
+    _alt_re = re.compile(r"\([0-9]+\)")
+    cmudict: List[Tuple[str, List[str]]] = []
+    for line in lines:
+        if not line or line.startswith(";;;"):  # ignore comments
+            continue
+
+        word, phones = line.strip().split("  ")
+        if word in _PUNCTUATIONS:
+            if exclude_punctuations:
+                continue
+            # !EXCLAMATION-POINT -> !
+            # --DASH -> --
+            # ...ELLIPSIS -> ...
+            if word.startswith("..."):
+                word = "..."
+            elif word.startswith("--"):
+                word = "--"
+            else:
+                word = word[0]
+
+        # if a word have multiple pronunciations, there will be (number) appended to it
+        # for example, DATAPOINTS and DATAPOINTS(1),
+        # the regular expression `_alt_re` removes the '(1)' and change the word DATAPOINTS(1) to DATAPOINTS
+        word = re.sub(_alt_re, "", word)
+        phones = phones.split(" ")
+        cmudict.append((word, phones))
+
+    return cmudict
+
+
+class CMUDict(Dataset):
+    """*CMU Pronouncing Dictionary* :cite:`cmudict` (CMUDict) dataset.
+
+    Args:
+        root (str or Path): Path to the directory where the dataset is found or downloaded.
+        exclude_punctuations (bool, optional):
+            When enabled, exclude the pronounciation of punctuations, such as
+            `!EXCLAMATION-POINT` and `#HASH-MARK`.
+        download (bool, optional):
+            Whether to download the dataset if it is not found at root path. (default: ``False``).
+        url (str, optional):
+            The URL to download the dictionary from.
+            (default: ``"http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b"``)
+        url_symbols (str, optional):
+            The URL to download the list of symbols from.
+            (default: ``"http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b.symbols"``)
+    """
+
+    def __init__(
+        self,
+        root: Union[str, Path],
+        exclude_punctuations: bool = True,
+        *,
+        download: bool = False,
+        url: str = "http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b",
+        url_symbols: str = "http://svn.code.sf.net/p/cmusphinx/code/trunk/cmudict/cmudict-0.7b.symbols",
+    ) -> None:
+
+        self.exclude_punctuations = exclude_punctuations
+
+        self._root_path = Path(root)
+        if not os.path.isdir(self._root_path):
+            raise RuntimeError(f"The root directory does not exist; {root}")
+
+        dict_file = self._root_path / os.path.basename(url)
+        symbol_file = self._root_path / os.path.basename(url_symbols)
+        if not os.path.exists(dict_file):
+            if not download:
+                raise RuntimeError(
+                    "The dictionary file is not found in the following location. "
+                    f"Set `download=True` to download it. {dict_file}"
+                )
+            checksum = _CHECKSUMS.get(url, None)
+            download_url_to_file(url, dict_file, checksum)
+        if not os.path.exists(symbol_file):
+            if not download:
+                raise RuntimeError(
+                    "The symbol file is not found in the following location. "
+                    f"Set `download=True` to download it. {symbol_file}"
+                )
+            checksum = _CHECKSUMS.get(url_symbols, None)
+            download_url_to_file(url_symbols, symbol_file, checksum)
+
+        with open(symbol_file, "r") as text:
+            self._symbols = [line.strip() for line in text.readlines()]
+
+        with open(dict_file, "r", encoding="latin-1") as text:
+            self._dictionary = _parse_dictionary(text.readlines(), exclude_punctuations=self.exclude_punctuations)
+
+    def __getitem__(self, n: int) -> Tuple[str, List[str]]:
+        """Load the n-th sample from the dataset.
+
+        Args:
+            n (int): The index of the sample to be loaded.
+
+        Returns:
+            Tuple of a word and its phonemes
+
+            str:
+                Word
+            List[str]:
+                Phonemes
+        """
+        return self._dictionary[n]
+
+    def __len__(self) -> int:
+        return len(self._dictionary)
+
+    @property
+    def symbols(self) -> List[str]:
+        """list[str]: A list of phonemes symbols, such as ``"AA"``, ``"AE"``, ``"AH"``."""
+        return self._symbols.copy()

torchaudio/datasets/commonvoice.py

@@ -0,0 +1,86 @@
+import csv
+import os
+from pathlib import Path
+from typing import Dict, List, Tuple, Union
+
+import torchaudio
+from torch import Tensor
+from torch.utils.data import Dataset
+
+
+def load_commonvoice_item(
+    line: List[str], header: List[str], path: str, folder_audio: str, ext_audio: str
+) -> Tuple[Tensor, int, Dict[str, str]]:
+    # Each line as the following data:
+    # client_id, path, sentence, up_votes, down_votes, age, gender, accent
+
+    if header[1] != "path":
+        raise ValueError(f"expect `header[1]` to be 'path', but got {header[1]}")
+    fileid = line[1]
+    filename = os.path.join(path, folder_audio, fileid)
+    if not filename.endswith(ext_audio):
+        filename += ext_audio
+    waveform, sample_rate = torchaudio.load(filename)
+
+    dic = dict(zip(header, line))
+
+    return waveform, sample_rate, dic
+
+
+class COMMONVOICE(Dataset):
+    """*CommonVoice* :cite:`ardila2020common` dataset.
+
+    Args:
+        root (str or Path): Path to the directory where the dataset is located.
+             (Where the ``tsv`` file is present.)
+        tsv (str, optional):
+            The name of the tsv file used to construct the metadata, such as
+            ``"train.tsv"``, ``"test.tsv"``, ``"dev.tsv"``, ``"invalidated.tsv"``,
+            ``"validated.tsv"`` and ``"other.tsv"``. (default: ``"train.tsv"``)
+    """
+
+    _ext_txt = ".txt"
+    _ext_audio = ".mp3"
+    _folder_audio = "clips"
+
+    def __init__(self, root: Union[str, Path], tsv: str = "train.tsv") -> None:
+
+        # Get string representation of 'root' in case Path object is passed
+        self._path = os.fspath(root)
+        self._tsv = os.path.join(self._path, tsv)
+
+        with open(self._tsv, "r") as tsv_:
+            walker = csv.reader(tsv_, delimiter="\t")
+            self._header = next(walker)
+            self._walker = list(walker)
+
+    def __getitem__(self, n: int) -> Tuple[Tensor, int, Dict[str, str]]:
+        """Load the n-th sample from the dataset.
+
+        Args:
+            n (int): The index of the sample to be loaded
+
+        Returns:
+            Tuple of the following items;
+
+            Tensor:
+                Waveform
+            int:
+                Sample rate
+            Dict[str, str]:
+                Dictionary containing the following items from the corresponding TSV file;
+
+                * ``"client_id"``
+                * ``"path"``
+                * ``"sentence"``
+                * ``"up_votes"``
+                * ``"down_votes"``
+                * ``"age"``
+                * ``"gender"``
+                * ``"accent"``
+        """
+        line = self._walker[n]
+        return load_commonvoice_item(line, self._header, self._path, self._folder_audio, self._ext_audio)
+
+    def __len__(self) -> int:
+        return len(self._walker)

torchaudio/datasets/dr_vctk.py

@@ -0,0 +1,121 @@
+from pathlib import Path
+from typing import Dict, Tuple, Union
+
+import torchaudio
+from torch import Tensor
+from torch.utils.data import Dataset
+from torchaudio._internal import download_url_to_file
+from torchaudio.datasets.utils import _extract_zip
+
+
+_URL = "https://datashare.ed.ac.uk/bitstream/handle/10283/3038/DR-VCTK.zip"
+_CHECKSUM = "781f12f4406ed36ed27ae3bce55da47ba176e2d8bae67319e389e07b2c9bd769"
+_SUPPORTED_SUBSETS = {"train", "test"}
+
+
+class DR_VCTK(Dataset):
+    """*Device Recorded VCTK (Small subset version)* :cite:`Sarfjoo2018DeviceRV` dataset.
+
+    Args:
+        root (str or Path): Root directory where the dataset's top level directory is found.
+        subset (str): The subset to use. Can be one of ``"train"`` and ``"test"``. (default: ``"train"``).
+        download (bool):
+            Whether to download the dataset if it is not found at root path. (default: ``False``).
+        url (str): The URL to download the dataset from.
+            (default: ``"https://datashare.ed.ac.uk/bitstream/handle/10283/3038/DR-VCTK.zip"``)
+    """
+
+    def __init__(
+        self,
+        root: Union[str, Path],
+        subset: str = "train",
+        *,
+        download: bool = False,
+        url: str = _URL,
+    ) -> None:
+        if subset not in _SUPPORTED_SUBSETS:
+            raise RuntimeError(
+                f"The subset '{subset}' does not match any of the supported subsets: {_SUPPORTED_SUBSETS}"
+            )
+
+        root = Path(root).expanduser()
+        archive = root / "DR-VCTK.zip"
+
+        self._subset = subset
+        self._path = root / "DR-VCTK" / "DR-VCTK"
+        self._clean_audio_dir = self._path / f"clean_{self._subset}set_wav_16k"
+        self._noisy_audio_dir = self._path / f"device-recorded_{self._subset}set_wav_16k"
+        self._config_filepath = self._path / "configurations" / f"{self._subset}_ch_log.txt"
+
+        if not self._path.is_dir():
+            if not archive.is_file():
+                if not download:
+                    raise RuntimeError("Dataset not found. Please use `download=True` to download it.")
+                download_url_to_file(url, archive, hash_prefix=_CHECKSUM)
+            _extract_zip(archive, root)
+
+        self._config = self._load_config(self._config_filepath)
+        self._filename_list = sorted(self._config)
+
+    def _load_config(self, filepath: str) -> Dict[str, Tuple[str, int]]:
+        # Skip header
+        skip_rows = 2 if self._subset == "train" else 1
+
+        config = {}
+        with open(filepath) as f:
+            for i, line in enumerate(f):
+                if i < skip_rows or not line:
+                    continue
+                filename, source, channel_id = line.strip().split("\t")
+                config[filename] = (source, int(channel_id))
+        return config
+
+    def _load_dr_vctk_item(self, filename: str) -> Tuple[Tensor, int, Tensor, int, str, str, str, int]:
+        speaker_id, utterance_id = filename.split(".")[0].split("_")
+        source, channel_id = self._config[filename]
+        file_clean_audio = self._clean_audio_dir / filename
+        file_noisy_audio = self._noisy_audio_dir / filename
+        waveform_clean, sample_rate_clean = torchaudio.load(file_clean_audio)
+        waveform_noisy, sample_rate_noisy = torchaudio.load(file_noisy_audio)
+        return (
+            waveform_clean,
+            sample_rate_clean,
+            waveform_noisy,
+            sample_rate_noisy,
+            speaker_id,
+            utterance_id,
+            source,
+            channel_id,
+        )
+
+    def __getitem__(self, n: int) -> Tuple[Tensor, int, Tensor, int, str, str, str, int]:
+        """Load the n-th sample from the dataset.
+
+        Args:
+            n (int): The index of the sample to be loaded
+
+        Returns:
+            Tuple of the following items;
+
+            Tensor:
+                Clean waveform
+            int:
+                Sample rate of the clean waveform
+            Tensor:
+                Noisy waveform
+            int:
+                Sample rate of the noisy waveform
+            str:
+                Speaker ID
+            str:
+                Utterance ID
+            str:
+                Source
+            int:
+                Channel ID
+        """
+        filename = self._filename_list[n]
+        return self._load_dr_vctk_item(filename)
+
+    def __len__(self) -> int:
+        return len(self._filename_list)