chore: import upstream snapshot with attribution

python/paddle/audio/backends/__init__.py

@@ -0,0 +1,27 @@
+# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from . import init_backend
+from .init_backend import (
+    get_current_backend,
+    list_available_backends,
+    set_backend,
+)
+
+init_backend._init_set_audio_backend()
+
+__all__ = [
+    'get_current_backend',
+    'list_available_backends',
+    'set_backend',
+]

python/paddle/audio/backends/backend.py

@@ -0,0 +1,164 @@
+# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING, BinaryIO
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    from paddle import Tensor
+
+
+class AudioInfo:
+    """Audio info, return type of backend info function"""
+
+    sample_rate: int
+    num_samples: int
+    num_channels: int
+    bits_per_sample: int
+    encoding: str
+
+    def __init__(
+        self,
+        sample_rate: int,
+        num_samples: int,
+        num_channels: int,
+        bits_per_sample: int,
+        encoding: str,
+    ) -> None:
+        self.sample_rate = sample_rate
+        self.num_samples = num_samples
+        self.num_channels = num_channels
+        self.bits_per_sample = bits_per_sample
+        self.encoding = encoding
+
+
+def info(filepath: str | BinaryIO) -> AudioInfo:
+    """Get signal information of input audio file.
+
+    Args:
+        filepath: audio path or file object.
+
+    Returns:
+        AudioInfo: info of the given audio.
+
+    Example:
+        .. code-block:: pycon
+
+            >>> import os
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> base_dir = os.getcwd()
+            >>> filepath = os.path.join(base_dir, "test.wav")
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+            >>> wav_info = paddle.audio.info(filepath)
+    """
+    # for API doc
+    raise NotImplementedError("please set audio backend")
+
+
+def load(
+    filepath: str | Path,
+    frame_offset: int = 0,
+    num_frames: int = -1,
+    normalize: bool = True,
+    channels_first: bool = True,
+) -> tuple[Tensor, int]:
+    """Load audio data from file.Load the audio content start form frame_offset, and get num_frames.
+
+    Args:
+        frame_offset: from 0 to total frames,
+        num_frames: from -1 (means total frames) or number frames which want to read,
+        normalize:
+            if True: return audio which norm to (-1, 1), dtype=float32
+            if False: return audio with raw data, dtype=int16
+
+        channels_first:
+            if True: return audio with shape (channels, time)
+
+    Return:
+        Tuple[paddle.Tensor, int]: (audio_content, sample rate)
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import os
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> base_dir = os.getcwd()
+            >>> filepath = os.path.join(base_dir, "test.wav")
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+            >>> wav_data_read, sr = paddle.audio.load(filepath)
+    """
+    # for API doc
+    raise NotImplementedError("please set audio backend")
+
+
+def save(
+    filepath: str,
+    src: Tensor,
+    sample_rate: int,
+    channels_first: bool = True,
+    encoding: str | None = None,
+    bits_per_sample: int | None = 16,
+) -> None:
+    """
+    Save audio tensor to file.
+
+    Args:
+        filepath: saved path
+        src: the audio tensor
+        sample_rate: the number of samples of audio per second.
+        channels_first: src channel information
+            if True, means input tensor is (channels, time)
+            if False, means input tensor is (time, channels)
+        encoding:encoding format, wave_backend only support PCM16 now.
+        bits_per_sample: bits per sample, wave_backend only support 16 bits now.
+
+    Returns:
+        None
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> filepath = "./test.wav"
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+    """
+    # for API doc
+    raise NotImplementedError("please set audio backend")

python/paddle/audio/backends/init_backend.py

@@ -0,0 +1,192 @@
+# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import annotations
+
+import sys
+import warnings
+
+import paddle
+
+from . import backend, wave_backend
+
+
+def _check_version(version: str) -> bool:
+    # require paddleaudio >= 1.0.2
+    ver_arr = version.split('.')
+    v0 = int(ver_arr[0])
+    v1 = int(ver_arr[1])
+    v2 = int(ver_arr[2])
+    if v0 < 1:
+        return False
+    if v0 == 1 and v1 == 0 and v2 <= 1:
+        return False
+    return True
+
+
+def list_available_backends() -> list[str]:
+    """List available backends, the backends in paddleaudio and the default backend.
+
+    Returns:
+        list[str]: The list of available backends.
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> wav_path = "./test.wav"
+
+            >>> current_backend = paddle.audio.backends.get_current_backend()
+            >>> print(current_backend)
+            wave_backend
+
+            >>> backends = paddle.audio.backends.list_available_backends()
+            >>> # default backends is ['wave_backend']
+            >>> # backends is ['wave_backend', 'soundfile'], if have installed paddleaudio >= 1.0.2
+            >>> if 'soundfile' in backends:
+            ...     paddle.audio.backends.set_backend('soundfile')
+            >>> paddle.audio.save(wav_path, waveform, sample_rate)
+
+    """
+    backends = []
+    try:
+        import paddleaudio
+    except ImportError:
+        package = "paddleaudio"
+        warn_msg = (
+            f"Failed importing {package}. \n"
+            "only wave_backend(only can deal with PCM16 WAV) supported.\n"
+            "if want soundfile_backend(more audio type supported),\n"
+            f"please manually installed (usually with `pip install {package} >= 1.0.2`). "
+        )
+        warnings.warn(warn_msg)
+
+    if "paddleaudio" in sys.modules:
+        version = paddleaudio.__version__
+        if not _check_version(version):
+            err_msg = (
+                f"the version of paddleaudio installed is {version},\n"
+                "please ensure the paddleaudio >= 1.0.2."
+            )
+            raise ImportError(err_msg)
+        backends = paddleaudio.backends.list_audio_backends()
+    backends.append("wave_backend")
+    return backends
+
+
+def get_current_backend() -> str:
+    """Get the name of the current audio backend
+
+    Returns:
+        str: The name of the current backend,
+        the wave_backend or backend imported from paddleaudio
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> wav_path = "./test.wav"
+
+            >>> current_backend = paddle.audio.backends.get_current_backend()
+            >>> print(current_backend)
+            wave_backend
+
+            >>> backends = paddle.audio.backends.list_available_backends()
+            >>> # default backends is ['wave_backend']
+            >>> # backends is ['wave_backend', 'soundfile'], if have installed paddleaudio >= 1.0.2
+
+            >>> if 'soundfile' in backends:
+            ...     paddle.audio.backends.set_backend('soundfile')
+            >>> paddle.audio.save(wav_path, waveform, sample_rate)
+
+    """
+    current_backend = None
+    if "paddleaudio" in sys.modules:
+        import paddleaudio
+
+        current_backend = paddleaudio.backends.get_audio_backend()
+        if paddle.audio.load == paddleaudio.load:
+            return current_backend
+    return "wave_backend"
+
+
+def set_backend(backend_name: str) -> None:
+    """Set the backend by one of the list_audio_backend return.
+
+    Args:
+        backend (str): one of the list_audio_backend. "wave_backend" is the default. "soundfile" imported from paddleaudio.
+
+    Returns:
+        None
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> wav_path = "./test.wav"
+
+            >>> current_backend = paddle.audio.backends.get_current_backend()
+            >>> print(current_backend)
+            wave_backend
+
+            >>> backends = paddle.audio.backends.list_available_backends()
+            >>> # default backends is ['wave_backend']
+            >>> # backends is ['wave_backend', 'soundfile'], if have installed paddleaudio >= 1.0.2
+
+            >>> if 'soundfile' in backends:
+            ...     paddle.audio.backends.set_backend('soundfile')
+            >>> paddle.audio.save(wav_path, waveform, sample_rate)
+
+    """
+    if backend_name not in list_available_backends():
+        raise NotImplementedError
+
+    if backend_name == "wave_backend":
+        module = wave_backend
+    else:
+        import paddleaudio
+
+        paddleaudio.backends.set_audio_backend(backend_name)
+        module = paddleaudio
+
+    for func in ["save", "load", "info"]:
+        setattr(backend, func, getattr(module, func))
+        setattr(paddle.audio, func, getattr(module, func))
+
+
+def _init_set_audio_backend() -> None:
+    # init the default wave_backend.
+    for func in ["save", "load", "info"]:
+        setattr(backend, func, getattr(wave_backend, func))

python/paddle/audio/backends/wave_backend.py

@@ -0,0 +1,236 @@
+# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import annotations
+
+import wave
+from typing import TYPE_CHECKING, BinaryIO
+
+import numpy as np
+
+import paddle
+
+from .backend import AudioInfo
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+    from paddle import Tensor
+
+
+def _error_message():
+    package = "paddleaudio"
+    warn_msg = (
+        "only PCM16 WAV supported. \n"
+        "if want support more other audio types, please "
+        f"manually installed (usually with `pip install {package}`). \n "
+        "and use paddle.audio.backends.set_backend('soundfile') to set audio backend"
+    )
+    return warn_msg
+
+
+def info(filepath: str | BinaryIO) -> AudioInfo:
+    """Get signal information of input audio file.
+
+    Args:
+        filepath: audio path or file object.
+
+    Returns:
+        AudioInfo: info of the given audio.
+
+    Example:
+        .. code-block:: pycon
+
+            >>> import os
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> base_dir = os.getcwd()
+            >>> filepath = os.path.join(base_dir, "test.wav")
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+            >>> wav_info = paddle.audio.info(filepath)
+    """
+
+    if hasattr(filepath, 'read'):
+        file_obj = filepath
+    else:
+        file_obj = open(filepath, 'rb')
+
+    try:
+        file_ = wave.open(file_obj)
+    except wave.Error:
+        file_obj.seek(0)
+        file_obj.close()
+        err_msg = _error_message()
+        raise NotImplementedError(err_msg)
+
+    channels = file_.getnchannels()
+    sample_rate = file_.getframerate()
+    sample_frames = file_.getnframes()  # audio frame
+    bits_per_sample = file_.getsampwidth() * 8
+    encoding = "PCM_S"  # default WAV encoding, only support
+    file_obj.close()
+    return AudioInfo(
+        sample_rate, sample_frames, channels, bits_per_sample, encoding
+    )
+
+
+def load(
+    filepath: str | Path,
+    frame_offset: int = 0,
+    num_frames: int = -1,
+    normalize: bool = True,
+    channels_first: bool = True,
+) -> tuple[Tensor, int]:
+    """Load audio data from file. load the audio content start form frame_offset, and get num_frames.
+
+    Args:
+        frame_offset: from 0 to total frames,
+        num_frames: from -1 (means total frames) or number frames which want to read,
+        normalize:
+            if True: return audio which norm to (-1, 1), dtype=float32
+            if False: return audio with raw data, dtype=int16
+
+        channels_first:
+            if True: return audio with shape (channels, time)
+
+    Return:
+        Tuple[paddle.Tensor, int]: (audio_content, sample rate)
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import os
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> base_dir = os.getcwd()
+            >>> filepath = os.path.join(base_dir, "test.wav")
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+            >>> wav_data_read, sr = paddle.audio.load(filepath)
+    """
+    if hasattr(filepath, 'read'):
+        file_obj = filepath
+    else:
+        file_obj = open(filepath, 'rb')
+
+    try:
+        file_ = wave.open(file_obj)
+    except wave.Error:
+        file_obj.seek(0)
+        file_obj.close()
+        err_msg = _error_message()
+        raise NotImplementedError(err_msg)
+
+    channels = file_.getnchannels()
+    sample_rate = file_.getframerate()
+    frames = file_.getnframes()  # audio frame
+
+    audio_content = file_.readframes(frames)
+    file_obj.close()
+
+    # default_subtype = "PCM_16", only support PCM16 WAV
+    audio_as_np16 = np.frombuffer(audio_content, dtype=np.int16)
+    audio_as_np32 = audio_as_np16.astype(np.float32)
+    if normalize:
+        # dtype = "float32"
+        audio_norm = audio_as_np32 / (2**15)
+    else:
+        # dtype = "int16"
+        audio_norm = audio_as_np32
+
+    waveform = np.reshape(audio_norm, (frames, channels))
+    if num_frames != -1:
+        waveform = waveform[frame_offset : frame_offset + num_frames, :]
+    waveform = paddle.to_tensor(waveform)
+    if channels_first:
+        waveform = paddle.transpose(waveform, perm=[1, 0])
+    return waveform, sample_rate
+
+
+def save(
+    filepath: str,
+    src: Tensor,
+    sample_rate: int,
+    channels_first: bool = True,
+    encoding: str | None = None,
+    bits_per_sample: int | None = 16,
+) -> None:
+    """
+    Save audio tensor to file.
+
+    Args:
+        filepath: saved path
+        src: the audio tensor
+        sample_rate: the number of samples of audio per second.
+        channels_first: src channel information
+            if True, means input tensor is (channels, time)
+            if False, means input tensor is (time, channels)
+        encoding: audio encoding format, wave_backend only support PCM16 now.
+        bits_per_sample: bits per sample, wave_backend only support 16 bits now.
+
+    Returns:
+        None
+
+    Examples:
+        .. code-block:: pycon
+
+            >>> import paddle
+
+            >>> sample_rate = 16000
+            >>> wav_duration = 0.5
+            >>> num_channels = 1
+            >>> num_frames = sample_rate * wav_duration
+            >>> wav_data = paddle.linspace(-1.0, 1.0, int(num_frames)) * 0.1
+            >>> waveform = wav_data.tile([num_channels, 1])
+            >>> filepath = "./test.wav"
+
+            >>> paddle.audio.save(filepath, waveform, sample_rate)
+    """
+    assert src.ndim == 2, "Expected 2D tensor"
+
+    audio_numpy = src.numpy()
+
+    # change src shape to (time, channels)
+    if channels_first:
+        audio_numpy = np.transpose(audio_numpy)
+
+    channels = audio_numpy.shape[1]
+
+    # only support PCM16
+    if bits_per_sample not in (None, 16):
+        raise ValueError("Invalid bits_per_sample, only support 16 bit")
+
+    sample_width = int(bits_per_sample / 8)  # 2
+
+    if src.dtype == paddle.float32:
+        audio_numpy = (audio_numpy * (2**15)).astype("<h")
+
+    with wave.open(filepath, 'w') as f:
+        f.setnchannels(channels)
+        f.setsampwidth(sample_width)
+        f.setframerate(sample_rate)
+        f.writeframes(audio_numpy.tobytes())