init v0.10.0

CONTRIBUTING.md

+# Contributing to Torchaudio
+We want to make contributing to this project as easy and transparent as possible.
+
+## TL;DR
+
+Please let us know if you encounter a bug by filing an [issue](https://github.com/pytorch/audio/issues).
+
+We appreciate all contributions. If you are planning to contribute back
+bug-fixes, please do so without any further discussion.
+
+If you plan to contribute new features, utility functions or extensions to the
+core, please first open an issue and discuss the feature with us. Sending a PR
+without discussion might end up resulting in a rejected PR, because we might be
+taking the core in a different direction than you might be aware of.
+
+Facebook has a [bounty program](https://www.facebook.com/whitehat/) for the
+safe disclosure of security bugs. In those cases, please go through the
+process outlined on that page and do not file a public issue.
+
+Fixing bugs and implementing new features are not the only way you can
+contribute. It also helps the project when you report problems you're facing,
+and when you give a :+1: on issues that others reported and that are relevant
+to you.
+
+You can also help by improving the documentation. This is no less important
+than improving the library itself! If you find a typo in the documentation,
+do not hesitate to submit a pull request.
+
+If you're not sure what you want to work on, you can pick an issue from the
+[list of open issues labelled as "help
+wanted"](https://github.com/pytorch/audio/issues?q=is%3Aopen+is%3Aissue+label%3A%22help+wanted%22).
+Comment on the issue that you want to work on it and send a PR with your fix
+(see below).
+
+## Contributor License Agreement ("CLA")
+In order to accept your pull request, we need you to submit a CLA. You only need
+to do this once to work on any of Facebook's open source projects.
+
+Complete your CLA here: <https://code.facebook.com/cla>
+
+## Development installation
+
+We recommend using a `conda` environment to contribute efficiently to
+torchaudio.
+
+### Install PyTorch Nightly
+
+```bash
+conda install pytorch -c pytorch-nightly
+```
+
+### Install Torchaudio
+
+```bash
+# Install build-time dependencies
+pip install cmake ninja pkgconfig
+```
+
+```bash
+# Build torchaudio
+git clone https://github.com/pytorch/audio.git
+cd audio
+git submodule update --init --recursive
+python setup.py develop
+# or, for OSX
+# MACOSX_DEPLOYMENT_TARGET=10.9 CC=clang CXX=clang++ python setup.py develop
+```
+
+Some environmnet variables that change the build behavior
+- `BUILD_SOX`: Deteremines whether build and bind libsox in non-Windows environments. (no effect in Windows as libsox integration is not available) Default value is 1 (build and bind). Use 0 for disabling it.
+- `USE_CUDA`: Determines whether build the custom CUDA kernel. Default to the availability of CUDA-compatible GPUs.
+
+If you built sox, set the `PATH` variable so that the tests properly use the newly built `sox` binary:
+
+```bash
+export PATH="<path_to_torchaudio>/third_party/install/bin:${PATH}"
+```
+
+The following dependencies are also needed for testing:
+
+```bash
+pip install typing pytest scipy numpy parameterized
+```
+
+Optional packages to install if you want to run related tests:
+
+- `librosa`
+- `requests`
+- `soundfile`
+- `kaldi_io`
+- `transformers`
+- `fairseq` (it has to be newer than `0.10.2`, so you will need to install from
+  source. Commit `e6eddd80` is known to work.)
+- `unidecode` (dependency for testing text preprocessing functions for examples/pipeline_tacotron2)
+- `inflect` (dependency for testing text preprocessing functions for examples/pipeline_tacotron2)
+
+## Development Process
+
+If you plan to modify the code or documentation, please follow the steps below:
+
+1. Fork the repository and create your branch from `main`: `$ git checkout main && git checkout -b my_cool_feature`
+2. If you have modified the code (new feature or bug-fix), [please add tests](test/torchaudio_unittest/).
+3. If you have changed APIs, [update the documentation](#Documentation).
+
+For more details about pull requests,
+please read [GitHub's guides](https://docs.github.com/en/github/collaborating-with-issues-and-pull-requests/creating-a-pull-request).
+
+If you would like to contribute a new model, please see [here](#New-model).
+
+If you would like to contribute a new dataset, please see [here](#New-dataset).
+
+## Testing
+
+Please refer to our [testing guidelines](test/torchaudio_unittest/) for more
+details.
+
+## Documentation
+
+Torchaudio uses [Google style](http://sphinxcontrib-napoleon.readthedocs.io/en/latest/example_google.html)
+for formatting docstrings. Length of line inside docstrings block must be limited to 120 characters.
+
+To build the docs, first install the requirements:
+
+```bash
+cd docs
+pip install -r requirements.txt
+```
+
+Then:
+
+```bash
+cd docs
+make html
+```
+
+The built docs should now be available in `docs/build/html`
+
+## Conventions
+
+As a good software development practice, we try to stick to existing variable
+names and shape (for tensors).
+The following are some of the conventions that we follow.
+
+- We use an ellipsis "..." as a placeholder for the rest of the dimensions of a
+  tensor, e.g. optional batching and channel dimensions. If batching, the
+  "batch" dimension should come in the first diemension.
+- Tensors are assumed to have "channel" dimension coming before the "time"
+  dimension. The bins in frequency domain (freq and mel) are assumed to come
+  before the "time" dimension but after the "channel" dimension. These
+  ordering makes the tensors consistent with PyTorch's dimensions.
+- For size names, the prefix `n_` is used (e.g. "a tensor of size (`n_freq`,
+  `n_mels`)") whereas dimension names do not have this prefix (e.g. "a tensor of
+  dimension (channel, time)")
+
+Here are some of the examples of commonly used variables with thier names,
+meanings, and shapes (or units):
+
+* `waveform`: a tensor of audio samples with dimensions (..., channel, time)
+* `sample_rate`: the rate of audio dimensions (samples per second)
+* `specgram`: a tensor of spectrogram with dimensions (..., channel, freq, time)
+* `mel_specgram`: a mel spectrogram with dimensions (..., channel, mel, time)
+* `hop_length`: the number of samples between the starts of consecutive frames
+* `n_fft`: the number of Fourier bins
+* `n_mels`, `n_mfcc`: the number of mel and MFCC bins
+* `n_freq`: the number of bins in a linear spectrogram
+* `f_min`: the lowest frequency of the lowest band in a spectrogram
+* `f_max`: the highest frequency of the highest band in a spectrogram
+* `win_length`: the length of the STFT window
+* `window_fn`: for functions that creates windows e.g. `torch.hann_window`
+
+## License
+
+By contributing to Torchaudio, you agree that your contributions will be licensed
+under the LICENSE file in the root directory of this source tree.

LICENSE

+BSD 2-Clause License
+
+Copyright (c) 2017 Facebook Inc. (Soumith Chintala), 
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are met:
+
+* Redistributions of source code must retain the above copyright notice, this
+  list of conditions and the following disclaimer.
+
+* Redistributions in binary form must reproduce the above copyright notice,
+  this list of conditions and the following disclaimer in the documentation
+  and/or other materials provided with the distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

README.md

-# Torchaudio
+torchaudio: an audio library for PyTorch
+========================================
 
+[![Build Status](https://circleci.com/gh/pytorch/audio.svg?style=svg)](https://app.circleci.com/pipelines/github/pytorch/audio)
+[![Documentation](https://img.shields.io/badge/dynamic/json.svg?label=docs&url=https%3A%2F%2Fpypi.org%2Fpypi%2Ftorchaudio%2Fjson&query=%24.info.version&colorB=brightgreen&prefix=v)](https://pytorch.org/audio/)
+
+The aim of torchaudio is to apply [PyTorch](https://github.com/pytorch/pytorch) to
+the audio domain. By supporting PyTorch, torchaudio follows the same philosophy
+of providing strong GPU acceleration, having a focus on trainable features through
+the autograd system, and having consistent style (tensor names and dimension names).
+Therefore, it is primarily a machine learning library and not a general signal
+processing library. The benefits of PyTorch can be seen in torchaudio through
+having all the computations be through PyTorch operations which makes it easy
+to use and feel like a natural extension.
+
+- [Support audio I/O (Load files, Save files)](http://pytorch.org/audio/stable/)
+  - Load a variety of audio formats, such as `wav`, `mp3`, `ogg`, `flac`, `opus`, `sphere`, into a torch Tensor using SoX
+  - [Kaldi (ark/scp)](http://pytorch.org/audio/stable/kaldi_io.html)
+- [Dataloaders for common audio datasets](http://pytorch.org/audio/stable/datasets.html)
+- Common audio transforms
+    - [Spectrogram, AmplitudeToDB, MelScale, MelSpectrogram, MFCC, MuLawEncoding, MuLawDecoding, Resample](http://pytorch.org/audio/stable/transforms.html)
+- Compliance interfaces: Run code using PyTorch that align with other libraries
+    - [Kaldi: spectrogram, fbank, mfcc](https://pytorch.org/audio/stable/compliance.kaldi.html)
+
+Dependencies
+------------
+* PyTorch (See below for the compatible versions)
+* [optional] vesis84/kaldi-io-for-python commit cb46cb1f44318a5d04d4941cf39084c5b021241e or above
+
+The following are the corresponding ``torchaudio`` versions and supported Python versions.
+
+| ``torch``                | ``torchaudio``           | ``python``                      |
+| ------------------------ | ------------------------ | ------------------------------- |
+| ``master`` / ``nightly`` | ``main`` / ``nightly``   | ``>=3.6``, ``<=3.9``            |
+| ``1.9.0``                | ``0.9.0``                | ``>=3.6``, ``<=3.9``            |
+| ``1.8.0``                | ``0.8.0``                | ``>=3.6``, ``<=3.9``            |
+| ``1.7.1``                | ``0.7.2``                | ``>=3.6``, ``<=3.9``            |
+| ``1.7.0``                | ``0.7.0``                | ``>=3.6``, ``<=3.8``            |
+| ``1.6.0``                | ``0.6.0``                | ``>=3.6``, ``<=3.8``            |
+| ``1.5.0``                | ``0.5.0``                | ``>=3.5``, ``<=3.8``            |
+| ``1.4.0``                | ``0.4.0``                | ``==2.7``, ``>=3.5``, ``<=3.8`` |
+
+
+Installation
+------------
+
+### Binary Distributions
+
+To install the latest version using anaconda, run:
+
+```
+conda install -c pytorch torchaudio
+```
+
+To install the latest pip wheels, run:
+
+```
+pip install torchaudio -f https://download.pytorch.org/whl/torch_stable.html
+```
+
+(If you do not have torch already installed, this will default to installing
+torch from PyPI. If you need a different torch configuration, preinstall torch
+before running this command.)
+
+### Nightly build
+
+Note that nightly build is built on PyTorch's nightly build. Therefore, you need to install the latest PyTorch when you use nightly build of torchaudio.
+
+**pip**
+
+```
+pip install --pre torchaudio -f https://download.pytorch.org/whl/nightly/torch_nightly.html
+```
+
+**conda**
+
+```
+conda install -y -c pytorch-nightly torchaudio
+```
+
+### From Source
+
+On non-Windows platforms, the build process builds libsox and codecs that torchaudio need to link to. It will fetch and build libmad, lame, flac, vorbis, opus, and libsox before building extension. This process requires `cmake` and `pkg-config`. libsox-based features can be disabled with `BUILD_SOX=0`.
+The build process also builds the RNN transducer loss. This functionality can be disabled by setting the environment variable `BUILD_RNNT=0`.
+
+```bash
+# Linux
+python setup.py install
+
+# OSX
+MACOSX_DEPLOYMENT_TARGET=10.9 CC=clang CXX=clang++ python setup.py install
+
+# Windows
+# We need to use the MSVC x64 toolset for compilation, with Visual Studio's vcvarsall.bat or directly with vcvars64.bat.
+# These batch files are under Visual Studio's installation folder, under 'VC\Auxiliary\Build\'.
+# More information available at:
+#   https://docs.microsoft.com/en-us/cpp/build/how-to-enable-a-64-bit-visual-cpp-toolset-on-the-command-line?view=msvc-160#use-vcvarsallbat-to-set-a-64-bit-hosted-build-architecture
+call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Auxiliary\Build\vcvarsall.bat" x64 && set BUILD_SOX=0 && python setup.py install
+# or
+call "C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\VC\Auxiliary\Build\vcvars64.bat" && set BUILD_SOX=0 && python setup.py install
+```
+
+This is known to work on linux and unix distributions such as Ubuntu and CentOS 7 and macOS.
+If you try this on a new system and find a solution to make it work, feel free to share it by opening an issue.
+
+Quick Usage
+-----------
+
+```python
+import torchaudio
+
+waveform, sample_rate = torchaudio.load('foo.wav')  # load tensor from file
+torchaudio.save('foo_save.wav', waveform, sample_rate)  # save tensor to file
+```
+
+Backend Dispatch
+----------------
+
+By default in OSX and Linux, torchaudio uses SoX as a backend to load and save files.
+The backend can be changed to [SoundFile](https://pysoundfile.readthedocs.io/en/latest/)
+using the following. See [SoundFile](https://pysoundfile.readthedocs.io/en/latest/)
+for installation instructions.
+
+```python
+import torchaudio
+torchaudio.set_audio_backend("soundfile")  # switch backend
+
+waveform, sample_rate = torchaudio.load('foo.wav')  # load tensor from file, as usual
+torchaudio.save('foo_save.wav', waveform, sample_rate)  # save tensor to file, as usual
+```
+
+**Note**
+- SoundFile currently does not support mp3.
+- "soundfile" backend is not supported by TorchScript.
+
+API Reference
+-------------
+
+API Reference is located here: http://pytorch.org/audio/
+
+Contributing Guidelines
+-----------------------
+
+Please refer to [CONTRIBUTING.md](./CONTRIBUTING.md)
+
+Disclaimer on Datasets
+----------------------
+
+This is a utility library that downloads and prepares public datasets. We do not host or distribute these datasets, vouch for their quality or fairness, or claim that you have license to use the dataset. It is your responsibility to determine whether you have permission to use the dataset under the dataset's license.
+
+If you're a dataset owner and wish to update any part of it (description, citation, etc.), or do not want your dataset to be included in this library, please get in touch through a GitHub issue. Thanks for your contribution to the ML community!
+>>>>>>> init v0.10.0

build_tools/convert_fairseq_models.py

+#!/usr/bin/env python3
+"""Convert a Wav2Vec2/HuBERT model published by fairseq into torchaudio format
+
+Examples
+
+```
+python convert_fairseq_models.py \
+  --input-file hubert_base_ls960.pt \
+  --output-file hubert_fairseq_base_ls960.pth
+
+python convert_fairseq_models.py \
+  --input-file hubert_large_ll60k.pt \
+  --output-file hubert_fairseq_large_ll60k.pth
+
+python convert_fairseq_models.py \
+  --input-file hubert_large_ll60k_finetune_ls960.pt \
+  --output-file hubert_fairseq_large_ll60k_asr_ls960.pth
+
+python convert_fairseq_models.py \
+  --input-file hubert_xtralarge_ll60k.pt \
+  --output-file hubert_fairseq_xlarge_ll60k.pth
+
+python convert_fairseq_models.py \
+  --input-file hubert_xtralarge_ll60k_finetune_ls960.pt \
+  --output-file hubert_fairseq_xlarge_ll60k_asr_ls960.pth
+"""
+
+import argparse
+
+# Note: Avoiding the import of torch and fairseq on global scope as they are slow
+
+
+def _parse_args():
+    parser = argparse.ArgumentParser(
+        description=__doc__,
+        formatter_class=argparse.RawTextHelpFormatter,
+    )
+    parser.add_argument(
+        '--input-file', required=True,
+        help='Input model file.'
+    )
+    parser.add_argument(
+        '--output-file', required=False,
+        help='Output model file.'
+    )
+    parser.add_argument(
+        '--dict-dir',
+        help=(
+            'Directory where letter vocabulary file, `dict.ltr.txt`, is found. '
+            'Required when loading wav2vec2 model. '
+            'https://dl.fbaipublicfiles.com/fairseq/wav2vec/dict.ltr.txt'
+        )
+    )
+    return parser.parse_args()
+
+
+def _load_model(input_file, dict_dir):
+    import fairseq
+
+    overrides = {} if dict_dir is None else {'data': dict_dir}
+    models, _, _ = fairseq.checkpoint_utils.load_model_ensemble_and_task(
+        [input_file], arg_overrides=overrides,
+    )
+    return models[0]
+
+
+def _import_model(model):
+    from torchaudio.models.wav2vec2.utils import import_fairseq_model
+
+    if model.__class__.__name__ in ['HubertCtc', 'Wav2VecCtc']:
+        model = model.w2v_encoder
+    model = import_fairseq_model(model)
+    return model
+
+
+def _main(args):
+    import torch
+    model = _load_model(args.input_file, args.dict_dir)
+    model = _import_model(model)
+    torch.save(model.state_dict(), args.output_file)
+
+
+if __name__ == '__main__':
+    _main(_parse_args())

build_tools/setup_helpers/__init__.py

+from .extension import *  # noqa

build_tools/setup_helpers/extension.py

+import os
+import platform
+import subprocess
+from pathlib import Path
+import distutils.sysconfig
+
+from setuptools import Extension
+from setuptools.command.build_ext import build_ext
+import torch
+
+__all__ = [
+    'get_ext_modules',
+    'CMakeBuild',
+]
+
+_THIS_DIR = Path(__file__).parent.resolve()
+_ROOT_DIR = _THIS_DIR.parent.parent.resolve()
+_TORCHAUDIO_DIR = _ROOT_DIR / 'torchaudio'
+
+
+def _get_build(var, default=False):
+    if var not in os.environ:
+        return default
+
+    val = os.environ.get(var, '0')
+    trues = ['1', 'true', 'TRUE', 'on', 'ON', 'yes', 'YES']
+    falses = ['0', 'false', 'FALSE', 'off', 'OFF', 'no', 'NO']
+    if val in trues:
+        return True
+    if val not in falses:
+        print(
+            f'WARNING: Unexpected environment variable value `{var}={val}`. '
+            f'Expected one of {trues + falses}')
+    return False
+
+
+_BUILD_SOX = False if platform.system() == 'Windows' else _get_build("BUILD_SOX", True)
+_BUILD_KALDI = False if platform.system() == 'Windows' else _get_build("BUILD_KALDI", True)
+_BUILD_RNNT = _get_build("BUILD_RNNT", True)
+_USE_ROCM = _get_build("USE_ROCM", torch.cuda.is_available() and torch.version.hip is not None)
+_USE_CUDA = _get_build("USE_CUDA", torch.cuda.is_available() and torch.version.hip is None)
+_TORCH_CUDA_ARCH_LIST = os.environ.get('TORCH_CUDA_ARCH_LIST', None)
+
+
+def get_ext_modules():
+    return [
+        Extension(name='torchaudio.lib.libtorchaudio', sources=[]),
+        Extension(name='torchaudio._torchaudio', sources=[]),
+    ]
+
+
+# Based off of
+# https://github.com/pybind/cmake_example/blob/580c5fd29d4651db99d8874714b07c0c49a53f8a/setup.py
+class CMakeBuild(build_ext):
+    def run(self):
+        try:
+            subprocess.check_output(['cmake', '--version'])
+        except OSError:
+            raise RuntimeError("CMake is not available.") from None
+        super().run()
+
+    def build_extension(self, ext):
+        # Since two library files (libtorchaudio and _torchaudio) need to be
+        # recognized by setuptools, we instantiate `Extension` twice. (see `get_ext_modules`)
+        # This leads to the situation where this `build_extension` method is called twice.
+        # However, the following `cmake` command will build all of them at the same time,
+        # so, we do not need to perform `cmake` twice.
+        # Therefore we call `cmake` only for `torchaudio._torchaudio`.
+        if ext.name != 'torchaudio._torchaudio':
+            return
+
+        extdir = os.path.abspath(
+            os.path.dirname(self.get_ext_fullpath(ext.name)))
+
+        # required for auto-detection of auxiliary "native" libs
+        if not extdir.endswith(os.path.sep):
+            extdir += os.path.sep
+
+        cfg = "Debug" if self.debug else "Release"
+
+        cmake_args = [
+            f"-DCMAKE_BUILD_TYPE={cfg}",
+            f"-DCMAKE_PREFIX_PATH={torch.utils.cmake_prefix_path}",
+            f"-DCMAKE_INSTALL_PREFIX={extdir}",
+            '-DCMAKE_VERBOSE_MAKEFILE=ON',
+            f"-DPython_INCLUDE_DIR={distutils.sysconfig.get_python_inc()}",
+            f"-DBUILD_SOX:BOOL={'ON' if _BUILD_SOX else 'OFF'}",
+            f"-DBUILD_KALDI:BOOL={'ON' if _BUILD_KALDI else 'OFF'}",
+            f"-DBUILD_RNNT:BOOL={'ON' if _BUILD_RNNT else 'OFF'}",
+            "-DBUILD_TORCHAUDIO_PYTHON_EXTENSION:BOOL=ON",
+            f"-DUSE_ROCM:BOOL={'ON' if _USE_ROCM else 'OFF'}",
+            f"-DUSE_CUDA:BOOL={'ON' if _USE_CUDA else 'OFF'}",
+        ]
+        build_args = [
+            '--target', 'install'
+        ]
+        # Pass CUDA architecture to cmake
+        if _TORCH_CUDA_ARCH_LIST is not None:
+            # Convert MAJOR.MINOR[+PTX] list to new style one
+            # defined at https://cmake.org/cmake/help/latest/prop_tgt/CUDA_ARCHITECTURES.html
+            _arches = _TORCH_CUDA_ARCH_LIST.replace('.', '').split(";")
+            _arches = [arch[:-4] if arch.endswith("+PTX") else f"{arch}-real" for arch in _arches]
+            cmake_args += [f"-DCMAKE_CUDA_ARCHITECTURES={';'.join(_arches)}"]
+
+        # Default to Ninja
+        if 'CMAKE_GENERATOR' not in os.environ or platform.system() == 'Windows':
+            cmake_args += ["-GNinja"]
+        if platform.system() == 'Windows':
+            import sys
+            python_version = sys.version_info
+            cmake_args += [
+                "-DCMAKE_C_COMPILER=cl",
+                "-DCMAKE_CXX_COMPILER=cl",
+                f"-DPYTHON_VERSION={python_version.major}.{python_version.minor}",
+            ]
+
+        # Set CMAKE_BUILD_PARALLEL_LEVEL to control the parallel build level
+        # across all generators.
+        if "CMAKE_BUILD_PARALLEL_LEVEL" not in os.environ:
+            # self.parallel is a Python 3 only way to set parallel jobs by hand
+            # using -j in the build_ext call, not supported by pip or PyPA-build.
+            if hasattr(self, "parallel") and self.parallel:
+                # CMake 3.12+ only.
+                build_args += ["-j{}".format(self.parallel)]
+
+        if not os.path.exists(self.build_temp):
+            os.makedirs(self.build_temp)
+
+        subprocess.check_call(
+            ["cmake", str(_ROOT_DIR)] + cmake_args, cwd=self.build_temp)
+        subprocess.check_call(
+            ["cmake", "--build", "."] + build_args, cwd=self.build_temp)
+
+    def get_ext_filename(self, fullname):
+        ext_filename = super().get_ext_filename(fullname)
+        ext_filename_parts = ext_filename.split('.')
+        without_abi = ext_filename_parts[:-2] + ext_filename_parts[-1:]
+        ext_filename = '.'.join(without_abi)
+        return ext_filename

build_tools/travis/install.sh

+#!/bin/bash
+# This script is meant to be called by the "install" step defined in
+# .travis.yml. See http://docs.travis-ci.com/ for more details.
+# The behavior of the script is controlled by environment variabled defined
+# in the .travis.yml in the top level folder of the project.
+
+ set -e
+
+ echo 'List files from cached directories'
+if [ -d $HOME/download ]; then
+    echo 'download:'
+    ls $HOME/download
+fi
+if [ -d $HOME/.cache/pip ]; then
+    echo 'pip:'
+    ls $HOME/.cache/pip
+fi
+
+ # Deactivate the travis-provided virtual environment and setup a
+# conda-based environment instead
+deactivate
+
+ # Add the miniconda bin directory to $PATH
+export PATH=/home/travis/miniconda3/bin:$PATH
+echo $PATH
+
+ # Use the miniconda installer for setup of conda itself
+pushd .
+cd
+mkdir -p download
+cd download
+if [[ ! -f /home/travis/miniconda3/bin/activate ]]
+then
+    if [[ ! -f miniconda.sh ]]
+    then
+        wget http://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh \
+             -O miniconda.sh
+    fi
+    chmod +x miniconda.sh && ./miniconda.sh -b -f
+    conda update --yes conda
+    echo "Creating environment to run tests in."
+    conda create -n testenv --yes python="$PYTHON_VERSION"
+fi
+cd ..
+popd
+
+ # Activate the python environment we created.
+source activate testenv
+
+ # Install requirements via pip in our conda environment
+conda install -y pytorch cpuonly -c pytorch-nightly
+pip install -r requirements.txt
+
+ # Install the following only if running tests
+if [[ "$SKIP_INSTALL" != "true" ]]; then
+     # TorchAudio CPP Extensions
+    python setup.py install
+fi
+
+if [[ "$RUN_EXAMPLE_TESTS" == "true" ]]; then
+  # Install dependencies
+  pip install sentencepiece PyAudio
+
+  if [[ ! -d $HOME/download/fairseq ]]; then
+    # Install fairseq from source
+    git clone https://github.com/pytorch/fairseq $HOME/download/fairseq
+  fi
+
+  pushd $HOME/download/fairseq
+  pip install --editable .
+  popd
+
+  mkdir -p $HOME/download/data
+  # Install dictionary, sentence piece model, and model
+  # These are cached so they are not downloaded if they already exist
+  wget -nc -O $HOME/download/data/dict.txt https://download.pytorch.org/models/audio/dict.txt || true
+  wget -nc -O $HOME/download/data/spm.model https://download.pytorch.org/models/audio/spm.model || true
+  wget -nc -O $HOME/download/data/model.pt https://download.pytorch.org/models/audio/checkpoint_avg_60_80.pt || true
+fi
+
+echo "Finished installation"

build_tools/travis/test_script.sh

+#!/bin/bash
+# This script is meant to be called by the "script" step defined in
+# .travis.yml. See http://docs.travis-ci.com/ for more details.
+# The behavior of the script is controlled by environment variabled defined
+# in the .travis.yml in the top level folder of the project.
+set -e
+
+python --version
+python -c 'import torch;print("torch:", torch.__version__)'
+
+run_tests() {
+  # find all the test files that match "test*.py"
+  TEST_FILES="$(find test -type f -name "test*.py" | sort)"
+  echo "Test files are:"
+  echo $TEST_FILES
+
+  echo "Executing tests:"
+  EXIT_STATUS=0
+  for FILE in $TEST_FILES; do
+    # run each file on a separate process. if one fails, just keep going and
+    # return the final exit status.
+    python -m pytest -v $FILE
+    STATUS=$?
+    EXIT_STATUS="$(($EXIT_STATUS+STATUS))"
+  done
+
+  echo "Done, exit status: $EXIT_STATUS"
+  exit $EXIT_STATUS
+}
+
+if [[ "$RUN_FLAKE8" == "true" ]]; then
+  flake8
+fi
+
+if [[ "$SKIP_TESTS" != "true" ]]; then
+  echo "run_tests"
+  run_tests
+fi
+
+if [[ "$RUN_EXAMPLE_TESTS" == "true" ]]; then
+  echo "run_example_tests"
+  pushd examples
+  ASR_MODEL_PATH=$HOME/download/data/model.pt \
+  ASR_INPUT_FILE=interactive_asr/data/sample.wav \
+  ASR_DATA_PATH=$HOME/download/data \
+  ASR_USER_DIR=$HOME/download/fairseq/examples/speech_recognition \
+  python -m unittest test/test_interactive_asr.py
+  popd
+fi

cmake/LoadHIP.cmake

+set(PYTORCH_FOUND_HIP FALSE)
+
+if(NOT DEFINED ENV{ROCM_PATH})
+  set(ROCM_PATH /opt/rocm)
+else()
+  set(ROCM_PATH $ENV{ROCM_PATH})
+endif()
+
+# HIP_PATH
+if(NOT DEFINED ENV{HIP_PATH})
+  set(HIP_PATH ${ROCM_PATH}/hip)
+else()
+  set(HIP_PATH $ENV{HIP_PATH})
+endif()
+
+if(NOT EXISTS ${HIP_PATH})
+  return()
+endif()
+
+# HCC_PATH
+if(NOT DEFINED ENV{HCC_PATH})
+  set(HCC_PATH ${ROCM_PATH}/hcc)
+else()
+  set(HCC_PATH $ENV{HCC_PATH})
+endif()
+
+# HSA_PATH
+if(NOT DEFINED ENV{HSA_PATH})
+  set(HSA_PATH ${ROCM_PATH}/hsa)
+else()
+  set(HSA_PATH $ENV{HSA_PATH})
+endif()
+
+# ROCBLAS_PATH
+if(NOT DEFINED ENV{ROCBLAS_PATH})
+  set(ROCBLAS_PATH ${ROCM_PATH}/rocblas)
+else()
+  set(ROCBLAS_PATH $ENV{ROCBLAS_PATH})
+endif()
+
+# ROCFFT_PATH
+if(NOT DEFINED ENV{ROCFFT_PATH})
+  set(ROCFFT_PATH ${ROCM_PATH}/rocfft)
+else()
+  set(ROCFFT_PATH $ENV{ROCFFT_PATH})
+endif()
+
+# HIPFFT_PATH
+if(NOT DEFINED ENV{HIPFFT_PATH})
+  set(HIPFFT_PATH ${ROCM_PATH}/hipfft)
+else()
+  set(HIPFFT_PATH $ENV{HIPFFT_PATH})
+endif()
+
+# HIPSPARSE_PATH
+if(NOT DEFINED ENV{HIPSPARSE_PATH})
+  set(HIPSPARSE_PATH ${ROCM_PATH}/hipsparse)
+else()
+  set(HIPSPARSE_PATH $ENV{HIPSPARSE_PATH})
+endif()
+
+# THRUST_PATH
+if(DEFINED ENV{THRUST_PATH})
+  set(THRUST_PATH $ENV{THRUST_PATH})
+else()
+  set(THRUST_PATH ${ROCM_PATH}/include)
+endif()
+
+# HIPRAND_PATH
+if(NOT DEFINED ENV{HIPRAND_PATH})
+  set(HIPRAND_PATH ${ROCM_PATH}/hiprand)
+else()
+  set(HIPRAND_PATH $ENV{HIPRAND_PATH})
+endif()
+
+# ROCRAND_PATH
+if(NOT DEFINED ENV{ROCRAND_PATH})
+  set(ROCRAND_PATH ${ROCM_PATH}/rocrand)
+else()
+  set(ROCRAND_PATH $ENV{ROCRAND_PATH})
+endif()
+
+# MIOPEN_PATH
+if(NOT DEFINED ENV{MIOPEN_PATH})
+  set(MIOPEN_PATH ${ROCM_PATH}/miopen)
+else()
+  set(MIOPEN_PATH $ENV{MIOPEN_PATH})
+endif()
+
+# RCCL_PATH
+if(NOT DEFINED ENV{RCCL_PATH})
+  set(RCCL_PATH ${ROCM_PATH}/rccl)
+else()
+  set(RCCL_PATH $ENV{RCCL_PATH})
+endif()
+
+# ROCPRIM_PATH
+if(NOT DEFINED ENV{ROCPRIM_PATH})
+  set(ROCPRIM_PATH ${ROCM_PATH}/rocprim)
+else()
+  set(ROCPRIM_PATH $ENV{ROCPRIM_PATH})
+endif()
+
+# HIPCUB_PATH
+if(NOT DEFINED ENV{HIPCUB_PATH})
+  set(HIPCUB_PATH ${ROCM_PATH}/hipcub)
+else()
+  set(HIPCUB_PATH $ENV{HIPCUB_PATH})
+endif()
+
+# ROCTHRUST_PATH
+if(NOT DEFINED ENV{ROCTHRUST_PATH})
+  set(ROCTHRUST_PATH ${ROCM_PATH}/rocthrust)
+else()
+  set(ROCTHRUST_PATH $ENV{ROCTHRUST_PATH})
+endif()
+
+# ROCTRACER_PATH
+if(NOT DEFINED ENV{ROCTRACER_PATH})
+  set(ROCTRACER_PATH ${ROCM_PATH}/roctracer)
+else()
+  set(ROCTRACER_PATH $ENV{ROCTRACER_PATH})
+endif()
+
+if(NOT DEFINED ENV{PYTORCH_ROCM_ARCH})
+  set(PYTORCH_ROCM_ARCH gfx803;gfx900;gfx906;gfx908)
+else()
+  set(PYTORCH_ROCM_ARCH $ENV{PYTORCH_ROCM_ARCH})
+endif()
+
+# Add HIP to the CMAKE Module Path
+set(CMAKE_MODULE_PATH ${HIP_PATH}/cmake ${CMAKE_MODULE_PATH})
+
+# Disable Asserts In Code (Can't use asserts on HIP stack.)
+add_definitions(-DNDEBUG)
+
+macro(find_package_and_print_version PACKAGE_NAME)
+  find_package("${PACKAGE_NAME}" ${ARGN})
+  message("${PACKAGE_NAME} VERSION: ${${PACKAGE_NAME}_VERSION}")
+endmacro()
+
+# Find the HIP Package
+find_package_and_print_version(HIP 1.0)
+
+if(HIP_FOUND)
+  set(PYTORCH_FOUND_HIP TRUE)
+
+  # Find ROCM version for checks
+  file(READ "${ROCM_PATH}/.info/version-dev" ROCM_VERSION_DEV_RAW)
+  string(REGEX MATCH "^([0-9]+)\.([0-9]+)\.([0-9]+)-.*$" ROCM_VERSION_DEV_MATCH ${ROCM_VERSION_DEV_RAW})
+  if(ROCM_VERSION_DEV_MATCH)
+    set(ROCM_VERSION_DEV_MAJOR ${CMAKE_MATCH_1})
+    set(ROCM_VERSION_DEV_MINOR ${CMAKE_MATCH_2})
+    set(ROCM_VERSION_DEV_PATCH ${CMAKE_MATCH_3})
+    set(ROCM_VERSION_DEV "${ROCM_VERSION_DEV_MAJOR}.${ROCM_VERSION_DEV_MINOR}.${ROCM_VERSION_DEV_PATCH}")
+  endif()
+  message("\n***** ROCm version from ${ROCM_PATH}/.info/version-dev ****\n")
+  message("ROCM_VERSION_DEV: ${ROCM_VERSION_DEV}")
+  message("ROCM_VERSION_DEV_MAJOR: ${ROCM_VERSION_DEV_MAJOR}")
+  message("ROCM_VERSION_DEV_MINOR: ${ROCM_VERSION_DEV_MINOR}")
+  message("ROCM_VERSION_DEV_PATCH: ${ROCM_VERSION_DEV_PATCH}")
+
+  message("\n***** Library versions from dpkg *****\n")
+  execute_process(COMMAND dpkg -l COMMAND grep rocm-dev COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep rocm-libs COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep hsakmt-roct COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep rocr-dev COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep -w hcc COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep hip_base COMMAND awk "{print $2 \" VERSION: \" $3}")
+  execute_process(COMMAND dpkg -l COMMAND grep hip_hcc COMMAND awk "{print $2 \" VERSION: \" $3}")
+
+  message("\n***** Library versions from cmake find_package *****\n")
+
+  set(CMAKE_HCC_FLAGS_DEBUG ${CMAKE_CXX_FLAGS_DEBUG})
+  set(CMAKE_HCC_FLAGS_RELEASE ${CMAKE_CXX_FLAGS_RELEASE})
+  ### Remove setting of Flags when FindHIP.CMake PR #558 is accepted.###
+
+  set(hip_DIR ${HIP_PATH}/lib/cmake/hip)
+  set(hsa-runtime64_DIR ${ROCM_PATH}/lib/cmake/hsa-runtime64)
+  set(AMDDeviceLibs_DIR ${ROCM_PATH}/lib/cmake/AMDDeviceLibs)
+  set(amd_comgr_DIR ${ROCM_PATH}/lib/cmake/amd_comgr)
+  set(rocrand_DIR ${ROCRAND_PATH}/lib/cmake/rocrand)
+  set(hiprand_DIR ${HIPRAND_PATH}/lib/cmake/hiprand)
+  set(rocblas_DIR ${ROCBLAS_PATH}/lib/cmake/rocblas)
+  set(miopen_DIR ${MIOPEN_PATH}/lib/cmake/miopen)
+  set(rocfft_DIR ${ROCFFT_PATH}/lib/cmake/rocfft)
+  set(hipfft_DIR ${HIPFFT_PATH}/lib/cmake/hipfft)
+  set(hipsparse_DIR ${HIPSPARSE_PATH}/lib/cmake/hipsparse)
+  set(rccl_DIR ${RCCL_PATH}/lib/cmake/rccl)
+  set(rocprim_DIR ${ROCPRIM_PATH}/lib/cmake/rocprim)
+  set(hipcub_DIR ${HIPCUB_PATH}/lib/cmake/hipcub)
+  set(rocthrust_DIR ${ROCTHRUST_PATH}/lib/cmake/rocthrust)
+
+  find_package_and_print_version(hip REQUIRED)
+  find_package_and_print_version(hsa-runtime64 REQUIRED)
+  find_package_and_print_version(amd_comgr REQUIRED)
+  find_package_and_print_version(rocrand REQUIRED)
+  find_package_and_print_version(hiprand REQUIRED)
+  find_package_and_print_version(rocblas REQUIRED)
+  find_package_and_print_version(miopen REQUIRED)
+  find_package_and_print_version(rocfft REQUIRED)
+  #if(ROCM_VERSION_DEV VERSION_GREATER_EQUAL "4.1.0")
+    find_package_and_print_version(hipfft REQUIRED)
+  #endif()
+  find_package_and_print_version(hipsparse REQUIRED)
+  find_package_and_print_version(rccl)
+  find_package_and_print_version(rocprim REQUIRED)
+  find_package_and_print_version(hipcub REQUIRED)
+  find_package_and_print_version(rocthrust REQUIRED)
+
+  if(HIP_COMPILER STREQUAL clang)
+    set(hip_library_name amdhip64)
+  else()
+    set(hip_library_name hip_hcc)
+  endif()
+  message("HIP library name: ${hip_library_name}")
+
+  # TODO: hip_hcc has an interface include flag "-hc" which is only
+  # recognizable by hcc, but not gcc and clang. Right now in our
+  # setup, hcc is only used for linking, but it should be used to
+  # compile the *_hip.cc files as well.
+  find_library(PYTORCH_HIP_HCC_LIBRARIES ${hip_library_name} HINTS ${HIP_PATH}/lib)
+  # TODO: miopen_LIBRARIES should return fullpath to the library file,
+  # however currently it's just the lib name
+  find_library(PYTORCH_MIOPEN_LIBRARIES ${miopen_LIBRARIES} HINTS ${MIOPEN_PATH}/lib)
+  # TODO: rccl_LIBRARIES should return fullpath to the library file,
+  # however currently it's just the lib name
+  find_library(PYTORCH_RCCL_LIBRARIES ${rccl_LIBRARIES} HINTS ${RCCL_PATH}/lib)
+  # hiprtc is part of HIP
+  find_library(ROCM_HIPRTC_LIB ${hip_library_name} HINTS ${HIP_PATH}/lib)
+  # roctx is part of roctracer
+  find_library(ROCM_ROCTX_LIB roctx64 HINTS ${ROCTRACER_PATH}/lib)
+  set(roctracer_INCLUDE_DIRS ${ROCTRACER_PATH}/include)
+endif()

docs/Makefile

+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line.
+SPHINXOPTS    = -W  # converts warnings into error
+SPHINXBUILD   = sphinx-build
+SPHINXPROJ    = torchaudio
+SOURCEDIR     = source
+BUILDDIR      = build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+docset: html
+	doc2dash --name $(SPHINXPROJ) --icon $(SOURCEDIR)/_static/img/pytorch-logo-flame.png --enable-js --online-redirect-url http://pytorch.org/audio/ --force $(BUILDDIR)/html/
+
+	# Manually fix because Zeal doesn't deal well with `icon.png`-only at 2x resolution.
+	cp $(SPHINXPROJ).docset/icon.png $(SPHINXPROJ).docset/icon@2x.png
+	convert $(SPHINXPROJ).docset/icon@2x.png -resize 16x16 $(SPHINXPROJ).docset/icon.png
+
+.PHONY: help Makefile docset
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

docs/make.bat

+@ECHO OFF
+
+pushd %~dp0
+
+REM Command file for Sphinx documentation
+
+if "%SPHINXBUILD%" == "" (
+	set SPHINXBUILD=sphinx-build
+)
+set SOURCEDIR=source
+set BUILDDIR=build
+set SPHINXPROJ=torchaudio
+
+if "%1" == "" goto help
+
+%SPHINXBUILD% >NUL 2>NUL
+if errorlevel 9009 (
+	echo.
+	echo.The 'sphinx-build' command was not found. Make sure you have Sphinx
+	echo.installed, then set the SPHINXBUILD environment variable to point
+	echo.to the full path of the 'sphinx-build' executable. Alternatively you
+	echo.may add the Sphinx directory to PATH.
+	echo.
+	echo.If you don't have Sphinx installed, grab it from
+	echo.http://sphinx-doc.org/
+	exit /b 1
+)
+
+%SPHINXBUILD% -M %1 %SOURCEDIR% %BUILDDIR% %SPHINXOPTS%
+goto end
+
+:help
+%SPHINXBUILD% -M help %SOURCEDIR% %BUILDDIR% %SPHINXOPTS%
+
+:end
+popd

docs/requirements.txt

+sphinx==3.5.4
+-e git+git://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme
+sphinxcontrib.katex
+sphinxcontrib.bibtex
+matplotlib

docs/source/_static/img/pytorch-logo-dark.svg

+<?xml version="1.0" encoding="utf-8"?>
+<!-- Generator: Adobe Illustrator 22.1.0, SVG Export Plug-In . SVG Version: 6.00 Build 0)  -->
+<svg version="1.1" id="Layer_1" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" x="0px" y="0px"
+	 viewBox="0 0 199.7 40.2" style="enable-background:new 0 0 199.7 40.2;" xml:space="preserve">
+<style type="text/css">
+	.st0{fill:#EE4C2C;}
+	.st1{fill:#252525;}
+</style>
+<g>
+	<path class="st0" d="M40.8,9.3l-2.1,2.1c3.5,3.5,3.5,9.2,0,12.7c-3.5,3.5-9.2,3.5-12.7,0c-3.5-3.5-3.5-9.2,0-12.7l0,0l5.6-5.6
+		L32.3,5l0,0V0.8l-8.5,8.5c-4.7,4.7-4.7,12.2,0,16.9s12.2,4.7,16.9,0C45.5,21.5,45.5,13.9,40.8,9.3z"/>
+	<circle class="st0" cx="36.6" cy="7.1" r="1.6"/>
+</g>
+<g>
+	<g>
+		<path class="st1" d="M62.6,20l-3.6,0v9.3h-2.7V2.9c0,0,6.3,0,6.6,0c7,0,10.3,3.4,10.3,8.3C73.2,17,69.1,19.9,62.6,20z M62.8,5.4
+			c-0.3,0-3.9,0-3.9,0v12.1l3.8-0.1c5-0.1,7.7-2.1,7.7-6.2C70.4,7.5,67.8,5.4,62.8,5.4z"/>
+		<path class="st1" d="M85.4,29.2l-1.6,4.2c-1.8,4.7-3.6,6.1-6.3,6.1c-1.5,0-2.6-0.4-3.8-0.9l0.8-2.4c0.9,0.5,1.9,0.8,3,0.8
+			c1.5,0,2.6-0.8,4-4.5l1.3-3.4L75.3,10h2.8l6.1,16l6-16h2.7L85.4,29.2z"/>
+		<path class="st1" d="M101.9,5.5v23.9h-2.7V5.5h-9.3V2.9h21.3v2.5H101.9z"/>
+		<path class="st1" d="M118.8,29.9c-5.4,0-9.4-4-9.4-10.2c0-6.2,4.1-10.3,9.6-10.3c5.4,0,9.3,4,9.3,10.2
+			C128.3,25.8,124.2,29.9,118.8,29.9z M118.9,11.8c-4.1,0-6.8,3.3-6.8,7.8c0,4.7,2.8,7.9,6.9,7.9s6.8-3.3,6.8-7.8
+			C125.8,15,123,11.8,118.9,11.8z"/>
+		<path class="st1" d="M135,29.4h-2.6V10l2.6-0.5v4.1c1.3-2.5,3.2-4.1,5.7-4.1c1.3,0,2.5,0.4,3.4,0.9l-0.7,2.5
+			c-0.8-0.5-1.9-0.8-3-0.8c-2,0-3.9,1.5-5.5,5V29.4z"/>
+		<path class="st1" d="M154.4,29.9c-5.8,0-9.5-4.2-9.5-10.2c0-6.1,4-10.3,9.5-10.3c2.4,0,4.4,0.6,6.1,1.7l-0.7,2.4
+			c-1.5-1-3.3-1.6-5.4-1.6c-4.2,0-6.8,3.1-6.8,7.7c0,4.7,2.8,7.8,6.9,7.8c1.9,0,3.9-0.6,5.4-1.6l0.5,2.4
+			C158.7,29.3,156.6,29.9,154.4,29.9z"/>
+		<path class="st1" d="M176.7,29.4V16.9c0-3.4-1.4-4.9-4.1-4.9c-2.2,0-4.4,1.1-6,2.8v14.7h-2.6V0.9l2.6-0.5c0,0,0,12.1,0,12.2
+			c2-2,4.6-3.1,6.7-3.1c3.8,0,6.1,2.4,6.1,6.6v13.3H176.7z"/>
+	</g>
+</g>
+</svg>

docs/source/_templates/layout.html

+{% extends "!layout.html" %}
+
+{% block sidebartitle %}
+    <div class="version">
+      <a href='https://pytorch.org/audio/versions.html'>{{ version }} &#x25BC</a>
+    </div>
+    {% include "searchbox.html" %}
+{% endblock %}

docs/source/backend.rst

+.. _backend:
+
+torchaudio.backend
+==================
+
+Overview
+~~~~~~~~
+
+:mod:`torchaudio.backend` module provides implementations for audio file I/O functionalities, which are ``torchaudio.info``, ``torchaudio.load``, and ``torchaudio.save``.
+
+There are currently four implementations available.
+
+* :ref:`"sox_io" <sox_io_backend>` (default on Linux/macOS)
+* :ref:`"soundfile" <soundfile_backend>` (default on Windows)
+
+.. note::
+   Instead of calling functions in ``torchaudio.backend`` directly, please use ``torchaudio.info``, ``torchaudio.load``, and ``torchaudio.save`` with proper backend set with :func:`torchaudio.set_audio_backend`.
+
+Availability
+------------
+
+``"sox_io"`` backend requires C++ extension module, which is included in Linux/macOS binary distributions. This backend is not available on Windows.
+
+``"soundfile"`` backend requires ``SoundFile``. Please refer to `the SoundFile documentation <https://pysoundfile.readthedocs.io/en/latest/>`_ for the installation.
+
+Common Data Structure
+~~~~~~~~~~~~~~~~~~~~~
+
+Structures used to report the metadata of audio files.
+
+AudioMetaData
+-------------
+
+.. autoclass:: torchaudio.backend.common.AudioMetaData
+
+.. _sox_io_backend:
+
+Sox IO Backend
+~~~~~~~~~~~~~~
+
+The ``"sox_io"`` backend is available and default on Linux/macOS and not available on Windows.
+
+I/O functions of this backend support `TorchScript <https://pytorch.org/docs/stable/jit.html>`_.
+
+You can switch from another backend to the ``sox_io`` backend with the following;
+
+.. code::
+
+   torchaudio.set_audio_backend("sox_io")
+
+info
+----
+
+.. autofunction:: torchaudio.backend.sox_io_backend.info
+
+load
+----
+
+.. autofunction:: torchaudio.backend.sox_io_backend.load
+
+save
+----
+
+.. autofunction:: torchaudio.backend.sox_io_backend.save
+
+.. _soundfile_backend:
+
+Soundfile Backend
+~~~~~~~~~~~~~~~~~
+
+The ``"soundfile"`` backend is available when `SoundFile <https://pysoundfile.readthedocs.io/en/latest/>`_ is installed. This backend is the default on Windows.
+
+You can switch from another backend to the ``"soundfile"`` backend with the following;
+
+.. code::
+
+   torchaudio.set_audio_backend("soundfile")
+
+info
+----
+
+.. autofunction:: torchaudio.backend.soundfile_backend.info
+
+load
+----
+
+.. autofunction:: torchaudio.backend.soundfile_backend.load
+
+save
+----
+
+.. autofunction:: torchaudio.backend.soundfile_backend.save

docs/source/compliance.kaldi.rst

+.. role:: hidden
+    :class: hidden-section
+
+torchaudio.compliance.kaldi
+============================
+
+.. currentmodule:: torchaudio.compliance.kaldi
+
+The useful processing operations of kaldi_ can be performed with torchaudio.
+Various functions with identical parameters are given so that torchaudio can
+produce similar outputs.
+
+.. _kaldi: https://github.com/kaldi-asr/kaldi
+
+Functions
+---------
+
+:hidden:`spectrogram`
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autofunction:: spectrogram
+
+:hidden:`fbank`
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autofunction:: fbank
+
+:hidden:`mfcc`
+~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. autofunction:: mfcc

docs/source/conf.py

+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+#
+# PyTorch documentation build configuration file, created by
+# sphinx-quickstart on Fri Dec 23 13:31:47 2016.
+#
+# This file is execfile()d with the current directory set to its
+# containing dir.
+#
+# Note that not all possible configuration values are present in this
+# autogenerated file.
+#
+# All configuration values have a default; values that are commented out
+# serve to show the default.
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+#
+# import os
+# import sys
+# sys.path.insert(0, os.path.abspath('.'))
+import pytorch_sphinx_theme
+
+# -- General configuration ------------------------------------------------
+
+# If your documentation needs a minimal Sphinx version, state it here.
+#
+needs_sphinx = '1.6'
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = [
+    'sphinx.ext.autodoc',
+    'sphinx.ext.autosummary',
+    'sphinx.ext.doctest',
+    'sphinx.ext.intersphinx',
+    'sphinx.ext.todo',
+    'sphinx.ext.coverage',
+    'sphinx.ext.napoleon',
+    'sphinx.ext.viewcode',
+    'sphinxcontrib.katex',
+    'sphinxcontrib.bibtex',
+]
+
+# katex options
+#
+#
+
+katex_options = r'''
+delimiters : [
+   {left: "$$", right: "$$", display: true},
+   {left: "\\(", right: "\\)", display: false},
+   {left: "\\[", right: "\\]", display: true}
+]
+'''
+
+bibtex_bibfiles = ['refs.bib']
+
+napoleon_use_ivar = True
+napoleon_numpy_docstring = False
+napoleon_google_docstring = True
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ['_templates']
+
+# The suffix(es) of source filenames.
+# You can specify multiple suffix as a list of string:
+#
+# source_suffix = ['.rst', '.md']
+source_suffix = '.rst'
+
+# The master toctree document.
+master_doc = 'index'
+
+# General information about the project.
+project = 'Torchaudio'
+copyright = '2018, Torchaudio Contributors'
+author = 'Torchaudio Contributors'
+
+# The version info for the project you're documenting, acts as replacement for
+# |version| and |release|, also used in various other places throughout the
+# built documents.
+#
+# The short X.Y version.
+# TODO: change to [:2] at v1.0
+version = '0.10.0 '
+# The full version, including alpha/beta/rc tags.
+# TODO: verify this works as expected
+release = '0.10.0'
+
+# The language for content autogenerated by Sphinx. Refer to documentation
+# for a list of supported languages.
+#
+# This is also used if you do content translation via gettext catalogs.
+# Usually you set "language" from the command line for these cases.
+language = None
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This patterns also effect to html_static_path and html_extra_path
+exclude_patterns = []
+
+# The name of the Pygments (syntax highlighting) style to use.
+pygments_style = 'sphinx'
+
+# If true, `todo` and `todoList` produce output, else they produce nothing.
+todo_include_todos = True
+
+
+# -- Options for HTML output ----------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+html_theme = 'pytorch_sphinx_theme'
+html_theme_path = [pytorch_sphinx_theme.get_html_theme_path()]
+
+# Theme options are theme-specific and customize the look and feel of a theme
+# further.  For a list of options available for each theme, see the
+# documentation.
+#
+html_theme_options = {
+    'pytorch_project': 'audio',
+    'collapse_navigation': False,
+    'display_version': True,
+    'logo_only': True,
+    'navigation_with_keys': True,
+    'analytics_id': 'UA-117752657-2',
+}
+
+html_logo = '_static/img/pytorch-logo-dark.svg'
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+html_static_path = ['_static']
+html_css_files = [
+    'https://cdn.jsdelivr.net/npm/katex@0.10.0-beta/dist/katex.min.css'
+]
+
+# -- Options for HTMLHelp output ------------------------------------------
+
+# Output file base name for HTML help builder.
+htmlhelp_basename = 'TorchAudiodoc'
+
+
+# -- Options for LaTeX output ---------------------------------------------
+
+latex_elements = {
+    # The paper size ('letterpaper' or 'a4paper').
+    #
+    # 'papersize': 'letterpaper',
+
+    # The font size ('10pt', '11pt' or '12pt').
+    #
+    # 'pointsize': '10pt',
+
+    # Additional stuff for the LaTeX preamble.
+    #
+    # 'preamble': '',
+
+    # Latex figure (float) alignment
+    #
+    # 'figure_align': 'htbp',
+}
+
+# Grouping the document tree into LaTeX files. List of tuples
+# (source start file, target name, title,
+#  author, documentclass [howto, manual, or own class]).
+latex_documents = [
+    (master_doc, 'pytorch.tex', 'Torchaudio Documentation',
+     'Torch Contributors', 'manual'),
+]
+
+
+# -- Options for manual page output ---------------------------------------
+
+# One entry per manual page. List of tuples
+# (source start file, name, description, authors, manual section).
+man_pages = [
+    (master_doc, 'Torchaudio', 'Torchaudio Documentation',
+     [author], 1)
+]
+
+
+# -- Options for Texinfo output -------------------------------------------
+
+# Grouping the document tree into Texinfo files. List of tuples
+# (source start file, target name, title, author,
+#  dir menu entry, description, category)
+texinfo_documents = [
+    (master_doc, 'Torchaudio', 'Torchaudio Documentation',
+     author, 'Torchaudio', 'Load audio files into pytorch tensors.',
+     'Miscellaneous'),
+]
+
+
+# Example configuration for intersphinx: refer to the Python standard library.
+intersphinx_mapping = {
+    'python': ('https://docs.python.org/', None),
+    'numpy': ('https://docs.scipy.org/doc/numpy/', None),
+    'torch': ('https://pytorch.org/docs/stable/', None),
+}
+
+# -- A patch that prevents Sphinx from cross-referencing ivar tags -------
+# See http://stackoverflow.com/a/41184353/3343043
+
+from docutils import nodes
+from sphinx.util.docfields import TypedField
+from sphinx import addnodes
+
+
+def patched_make_field(self, types, domain, items, **kw):
+    # `kw` catches `env=None` needed for newer sphinx while maintaining
+    #  backwards compatibility when passed along further down!
+
+    # type: (list, str, tuple) -> nodes.field
+    def handle_item(fieldarg, content):
+        par = nodes.paragraph()
+        par += addnodes.literal_strong('', fieldarg)  # Patch: this line added
+        # par.extend(self.make_xrefs(self.rolename, domain, fieldarg,
+        #                           addnodes.literal_strong))
+        if fieldarg in types:
+            par += nodes.Text(' (')
+            # NOTE: using .pop() here to prevent a single type node to be
+            # inserted twice into the doctree, which leads to
+            # inconsistencies later when references are resolved
+            fieldtype = types.pop(fieldarg)
+            if len(fieldtype) == 1 and isinstance(fieldtype[0], nodes.Text):
+                typename = u''.join(n.astext() for n in fieldtype)
+                typename = typename.replace('int', 'python:int')
+                typename = typename.replace('long', 'python:long')
+                typename = typename.replace('float', 'python:float')
+                typename = typename.replace('type', 'python:type')
+                par.extend(self.make_xrefs(self.typerolename, domain, typename,
+                                           addnodes.literal_emphasis, **kw))
+            else:
+                par += fieldtype
+            par += nodes.Text(')')
+        par += nodes.Text(' -- ')
+        par += content
+        return par
+
+    fieldname = nodes.field_name('', self.label)
+    if len(items) == 1 and self.can_collapse:
+        fieldarg, content = items[0]
+        bodynode = handle_item(fieldarg, content)
+    else:
+        bodynode = self.list_type()
+        for fieldarg, content in items:
+            bodynode += nodes.list_item('', handle_item(fieldarg, content))
+    fieldbody = nodes.field_body('', bodynode)
+    return nodes.field('', fieldname, fieldbody)
+
+TypedField.make_field = patched_make_field