torchaudio 2.1.0__cp39-cp39-macosx_11_0_arm64.whl → 2.1.1__cp39-cp39-macosx_11_0_arm64.whl

torchaudio/_backend/backend.py

@@ -3,6 +3,7 @@ from abc import ABC, abstractmethod
 from typing import BinaryIO, Optional, Tuple, Union
 
 from torch import Tensor
+from torchaudio.io import CodecConfig
 
 from .common import AudioMetaData
 
@@ -37,6 +38,7 @@ class Backend(ABC):
         encoding: Optional[str] = None,
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
+        compression: Optional[Union[CodecConfig, float, int]] = None,
     ) -> None:
         raise NotImplementedError
 

torchaudio/_backend/ffmpeg.py

@@ -253,6 +253,7 @@ def save_audio(
     encoding: Optional[str] = None,
     bits_per_sample: Optional[int] = None,
     buffer_size: int = 4096,
+    compression: Optional[torchaudio.io.CodecConfig] = None,
 ) -> None:
     ext = None
     if hasattr(uri, "write"):
@@ -275,6 +276,7 @@ def save_audio(
         format=_get_sample_format(src.dtype),
         encoder=encoder,
         encoder_format=enc_fmt,
+        codec_config=compression,
     )
     with s.open():
         s.write_audio_chunk(0, src)
@@ -343,7 +345,13 @@ class FFmpegBackend(Backend):
         encoding: Optional[str] = None,
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
+        compression: Optional[Union[torchaudio.io.CodecConfig, float, int]] = None,
     ) -> None:
+        if not isinstance(compression, (torchaudio.io.CodecConfig, type(None))):
+            raise ValueError(
+                "FFmpeg backend expects non-`None` value for argument `compression` to be of ",
+                f"type `torchaudio.io.CodecConfig`, but received value of type {type(compression)}",
+            )
         save_audio(
             uri,
             src,
@@ -353,6 +361,7 @@ class FFmpegBackend(Backend):
             encoding,
             bits_per_sample,
             buffer_size,
+            compression,
         )
 
     @staticmethod

torchaudio/_backend/soundfile.py

@@ -2,6 +2,7 @@ import os
 from typing import BinaryIO, Optional, Tuple, Union
 
 import torch
+from torchaudio.io import CodecConfig
 
 from . import soundfile_backend
 from .backend import Backend
@@ -35,7 +36,11 @@ class SoundfileBackend(Backend):
         encoding: Optional[str] = None,
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
+        compression: Optional[Union[CodecConfig, float, int]] = None,
     ) -> None:
+        if compression:
+            raise ValueError("soundfile backend does not support argument `compression`.")
+
         soundfile_backend.save(
             uri, src, sample_rate, channels_first, format=format, encoding=encoding, bits_per_sample=bits_per_sample
         )

torchaudio/_backend/sox.py

@@ -2,6 +2,7 @@ import os
 from typing import BinaryIO, Optional, Tuple, Union
 
 import torch
+from torchaudio.io import CodecConfig
 
 from .backend import Backend
 from .common import AudioMetaData
@@ -55,7 +56,13 @@ class SoXBackend(Backend):
         encoding: Optional[str] = None,
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
+        compression: Optional[Union[CodecConfig, float, int]] = None,
     ) -> None:
+        if not isinstance(compression, (float, int, type(None))):
+            raise ValueError(
+                "SoX backend expects non-`None` value for argument `compression` to be of ",
+                f"type `float` or `int`, but received value of type {type(compression)}",
+            )
         if hasattr(uri, "write"):
             raise ValueError(
                 "SoX backend does not support writing to file-like objects. ",
@@ -67,7 +74,7 @@ class SoXBackend(Backend):
                 src,
                 sample_rate,
                 channels_first,
-                None,
+                compression,
                 format,
                 encoding,
                 bits_per_sample,

torchaudio/_backend/utils.py

@@ -5,6 +5,7 @@ from typing import BinaryIO, Dict, Optional, Tuple, Type, Union
 import torch
 
 from torchaudio._extension import _FFMPEG_EXT, _SOX_INITIALIZED
+from torchaudio.io import CodecConfig
 
 from . import soundfile_backend
 
@@ -229,6 +230,7 @@ def get_save_func():
         bits_per_sample: Optional[int] = None,
         buffer_size: int = 4096,
         backend: Optional[str] = None,
+        compression: Optional[Union[CodecConfig, float, int]] = None,
     ):
         """Save audio data to file.
 
@@ -283,8 +285,32 @@ def get_save_func():
 
                 .. seealso::
                    :ref:`backend`
+
+            compression (CodecConfig, float, int, or None, optional):
+                Compression configuration to apply.
+
+                If the selected backend is FFmpeg, an instance of :py:class:`CodecConfig` must be provided.
+
+                Otherwise, if the selected backend is SoX, a float or int value corresponding to option ``-C`` of the
+                ``sox`` command line interface must be provided. For instance:
+
+                ``"mp3"``
+                    Either bitrate (in ``kbps``) with quality factor, such as ``128.2``, or
+                    VBR encoding with quality factor such as ``-4.2``. Default: ``-4.5``.
+
+                ``"flac"``
+                    Whole number from ``0`` to ``8``. ``8`` is default and highest compression.
+
+                ``"ogg"``, ``"vorbis"``
+                    Number from ``-1`` to ``10``; ``-1`` is the highest compression
+                    and lowest quality. Default: ``3``.
+
+                Refer to http://sox.sourceforge.net/soxformat.html for more details.
+
         """
         backend = dispatcher(uri, format, backend)
-        return backend.save(uri, src, sample_rate, channels_first, format, encoding, bits_per_sample, buffer_size)
+        return backend.save(
+            uri, src, sample_rate, channels_first, format, encoding, bits_per_sample, buffer_size, compression
+        )
 
     return save

torchaudio/pipelines/_wav2vec2/aligner.py

@@ -53,7 +53,7 @@ class IAligner(ABC):
             emission (Tensor): Sequence of token probability distributions in log-domain.
                 Shape: `(time, tokens)`.
             tokens (list of integer sequence): Tokenized transcript.
-                Output from :py:class:`Wav2Vec2FABundle.Tokenizer`.
+                Output from :py:class:`torchaudio.pipelines.Wav2Vec2FABundle.Tokenizer`.
 
         Returns:
             (list of TokenSpan sequence): Tokens with time stamps and scores.

torchaudio/pipelines/_wav2vec2/impl.py

@@ -1350,7 +1350,7 @@ WAVLM_LARGE = Wav2Vec2Bundle(
         "encoder_ff_interm_features": 4096,
         "encoder_ff_interm_dropout": 0.0,
         "encoder_dropout": 0.1,
-        "encoder_layer_norm_first": False,
+        "encoder_layer_norm_first": True,
         "encoder_layer_drop": 0.05,
         "aux_num_out": None,
     },

torchaudio/transforms/_transforms.py

@@ -1020,31 +1020,27 @@ class TimeStretch(torch.nn.Module):
     Proposed in *SpecAugment* :cite:`specaugment`.
 
     Args:
-        hop_length (int or None, optional): Length of hop between STFT windows. (Default: ``win_length // 2``)
+        hop_length (int or None, optional): Length of hop between STFT windows.
+            (Default: ``n_fft // 2``, where ``n_fft == (n_freq - 1) * 2``)
         n_freq (int, optional): number of filter banks from stft. (Default: ``201``)
         fixed_rate (float or None, optional): rate to speed up or slow down by.
             If None is provided, rate must be passed to the forward method. (Default: ``None``)
 
+    .. note::
+
+       The expected input is raw, complex-valued spectrogram.
+
     Example
-        >>> spectrogram = torchaudio.transforms.Spectrogram()
+        >>> spectrogram = torchaudio.transforms.Spectrogram(power=None)
         >>> stretch = torchaudio.transforms.TimeStretch()
         >>>
         >>> original = spectrogram(waveform)
-        >>> streched_1_2 = stretch(original, 1.2)
-        >>> streched_0_9 = stretch(original, 0.9)
-
-        .. image:: https://download.pytorch.org/torchaudio/doc-assets/specaugment_time_stretch_1.png
-           :width: 600
-           :alt: Spectrogram streched by 1.2
+        >>> stretched_1_2 = stretch(original, 1.2)
+        >>> stretched_0_9 = stretch(original, 0.9)
 
-        .. image:: https://download.pytorch.org/torchaudio/doc-assets/specaugment_time_stretch_2.png
+        .. image:: https://download.pytorch.org/torchaudio/doc-assets/specaugment_time_stretch.png
            :width: 600
-           :alt: The original spectrogram
-
-        .. image:: https://download.pytorch.org/torchaudio/doc-assets/specaugment_time_stretch_3.png
-           :width: 600
-           :alt: Spectrogram streched by 0.9
-
+           :alt: The visualization of stretched spectrograms.
     """
     __constants__ = ["fixed_rate"]
 
@@ -1067,8 +1063,8 @@ class TimeStretch(torch.nn.Module):
 
         Returns:
             Tensor:
-                Stretched spectrogram. The resulting tensor is of the same dtype as the input
-                spectrogram, but the number of frames is changed to ``ceil(num_frame / rate)``.
+                Stretched spectrogram. The resulting tensor is of the corresponding complex dtype
+                as the input spectrogram, and the number of frames is changed to ``ceil(num_frame / rate)``.
         """
         if overriding_rate is None:
             if self.fixed_rate is None:

torchaudio/version.py

@@ -1,2 +1,2 @@
-__version__ = '2.1.0'
-git_version = '6ea1133706801ec6e81bb29142da2e21a8583a0a'
+__version__ = '2.1.1'
+git_version = 'db624844f5c95bb7618fe5a5f532bf9b68efeb45'

{torchaudio-2.1.0.dist-info → torchaudio-2.1.1.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: torchaudio
-Version: 2.1.0
+Version: 2.1.1
 Summary: An audio package for PyTorch
 Home-page: https://github.com/pytorch/audio
 Author: Soumith Chintala, David Pollack, Sean Naren, Peter Goldsborough, Moto Hira, Caroline Chen, Jeff Hwang, Zhaoheng Ni, Xiaohui Zhang
@@ -24,7 +24,7 @@ Classifier: Topic :: Multimedia :: Sound/Audio
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Description-Content-Type: text/markdown
 License-File: LICENSE
-Requires-Dist: torch (==2.1.0)
+Requires-Dist: torch (==2.1.1)
 
 torchaudio: an audio library for PyTorch
 ========================================
@@ -85,6 +85,17 @@ If you find this package useful, please cite as:
 }
 ```
 
+```bibtex
+@misc{hwang2023torchaudio,
+      title={TorchAudio 2.1: Advancing speech recognition, self-supervised learning, and audio processing components for PyTorch}, 
+      author={Jeff Hwang and Moto Hira and Caroline Chen and Xiaohui Zhang and Zhaoheng Ni and Guangzhi Sun and Pingchuan Ma and Ruizhe Huang and Vineel Pratap and Yuekai Zhang and Anurag Kumar and Chin-Yun Yu and Chuang Zhu and Chunxi Liu and Jacob Kahn and Mirco Ravanelli and Peng Sun and Shinji Watanabe and Yangyang Shi and Yumeng Tao and Robin Scheibler and Samuele Cornell and Sean Kim and Stavros Petridis},
+      year={2023},
+      eprint={2310.17864},
+      archivePrefix={arXiv},
+      primaryClass={eess.AS}
+}
+```
+
 Disclaimer on Datasets
 ----------------------
 

{torchaudio-2.1.0.dist-info → torchaudio-2.1.1.dist-info}/RECORD

@@ -1,4 +1,4 @@
-torchaudio/version.py,sha256=qWqmxRpZJgrXDKiBtU2ZfWwMsRHsLYOFf5Ah8P5LnlM,79
+torchaudio/version.py,sha256=eq_ONyJOfoJ46gk7zrY5gZ4iyHKprXTjQLJa9mrAz-I,79
 torchaudio/kaldi_io.py,sha256=TwS2YgSLlJwOXjNNsHBuXyxhKeKKpptVHLBV7QYZCas,5073
 torchaudio/__init__.py,sha256=JNqWeRjt7G9AbkqH-zqgtJ41tBbuKXxzYFwSR6BBwRY,1117
 torchaudio/_internal/__init__.py,sha256=gjU8g9HhVd9hHrHXJM0xOlZL6cT8ktO60MN8RHI6ZbA,241
@@ -35,10 +35,10 @@ torchaudio/pipelines/_source_separation_pipeline.py,sha256=WxngB1d13H5IVqs4RPqSL
 torchaudio/pipelines/__init__.py,sha256=Xy8NmInKwTcNBHwLTTjHjrfczRLuQq8a67ENt1OTVXM,2745
 torchaudio/pipelines/rnnt_pipeline.py,sha256=Qy37z7v6d1jLOHd67zbRu21dgL6Fml1rTd7j4Jl1NsM,13749
 torchaudio/pipelines/_squim_pipeline.py,sha256=JROOpywuRFl6JnWVMhgqM2eu3REKeXj-_Te5QJzOwoA,7170
-torchaudio/pipelines/_wav2vec2/aligner.py,sha256=xRKPcGlIRU_uPlsD4Xy1fo_C4K_oS2eKal53klV6dOA,2688
+torchaudio/pipelines/_wav2vec2/aligner.py,sha256=pIWRgQ-kdYUxtL8bdc0qk9wBjwRrHY1uSWL3L4e2vxs,2709
 torchaudio/pipelines/_wav2vec2/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 torchaudio/pipelines/_wav2vec2/utils.py,sha256=Q8_fWOR2JDnHu0TTRmHzRjI3BOJa0hGIAl0cjtALgsQ,6971
-torchaudio/pipelines/_wav2vec2/impl.py,sha256=bX57aQb8LTYoRwGCDIutWl9tcghEIK-1BL7ydD_0y60,65562
+torchaudio/pipelines/_wav2vec2/impl.py,sha256=zdXFjytJO5MvnB-3aygzUUFKxCTkQGU_OX_rhUh9c0k,65561
 torchaudio/pipelines/_tts/interface.py,sha256=yUaS0UK3PTRruYXRWFil7lAhr-1iYiyBaDBLmEnJPUQ,10224
 torchaudio/pipelines/_tts/__init__.py,sha256=PP7l8XzVURqelwuMJFgfOCv4fvzZunDiy90ZQlRkv7g,426
 torchaudio/pipelines/_tts/utils.py,sha256=0rLyoFWS78n5jn9AC99pgIwAjaXSw-MVbj_pjSaOHiM,4616
@@ -82,27 +82,27 @@ torchaudio/models/wav2vec2/utils/import_fairseq.py,sha256=oCwG6qpG0bCXue2V56fjDc
 torchaudio/compliance/__init__.py,sha256=hhNObUS0c-fS-VMudM7zl3-CvupvCDmESlikntSMn5g,48
 torchaudio/compliance/kaldi.py,sha256=XL6hpYTd6nSPb2imIdeU4TM06I2fqh1AmG968y8ZbSk,36666
 torchaudio/.dylibs/libc++.1.0.dylib,sha256=5vOhaSs9NAFEhrtA2e5_Pp9CADGzxQDYm5I1BejVQiI,1023744
-torchaudio/_backend/backend.py,sha256=g63FIX949ei_u6qRWyMnGYMf1piPD86axuQdCGyzufc,1457
-torchaudio/_backend/ffmpeg.py,sha256=thqemsSahEsuNgHoQHny8z7dYuyUuUQIwvq1B8haM4I,12107
+torchaudio/_backend/backend.py,sha256=hSrfZcj5FMzx5ZpwubN-LLMvBFb7ENyw7HvT_6pVYVU,1565
+torchaudio/_backend/ffmpeg.py,sha256=lgof8oWCJA1IgBETLz3bzgURwGs7QphJQPTWnRkL044,12633
 torchaudio/_backend/__init__.py,sha256=3li8KAJkJN2gkw9j9NSTFbxvXFdLiDAsaefdM72oYBU,1730
-torchaudio/_backend/sox.py,sha256=qSSI3ojVbPXY-ZLrSQCdXd8tgdrK2pehxlTy3ySTR2s,2993
+torchaudio/_backend/sox.py,sha256=2ZQQ62C9bEm_m1no1bPxmiziRSUlkOy6kuSx-nLfDl0,3401
 torchaudio/_backend/common.py,sha256=55Y0r0MsdW6gvTOT_Zy60UGFXc60DfdJ7uvycJKK3is,1783
-torchaudio/_backend/utils.py,sha256=_cL3g0R9FYsHVCGPvp-TXI7rcG39iEyNilY_sfDFJEw,12047
-torchaudio/_backend/soundfile.py,sha256=4PiQ5-aeT7Cwhvcvdz-sBD7G-EPkB6fJURw86nE6faU,1479
+torchaudio/_backend/utils.py,sha256=j1LEBoRk1mbJ5yqB1c19qklEQrRUZgKKYCeNCBjqQUM,13221
+torchaudio/_backend/soundfile.py,sha256=n0Epw0J9rBb89xVJWTXaDfb96YFz0-i2xarXIdDd-Cw,1703
 torchaudio/_backend/soundfile_backend.py,sha256=qJHEEXU1egCkPJ2Y9uJWFvVhW3AqDZ7z7P7mkJjJJWM,17376
 torchaudio/transforms/__init__.py,sha256=Tp1o4haiJAV3MRJenmvGXFbmt-RE4qM_pd6U3Ghohqw,1270
-torchaudio/transforms/_transforms.py,sha256=vKoAZalBVlUmzKqlTP0t0IuUH4nIys_v2fmsTKsXCts,86780
+torchaudio/transforms/_transforms.py,sha256=Tw_xZGT3YAGPX_R_5d8Dr_ARaQA4lc73x6WOyylrNkk,86608
 torchaudio/transforms/_multi_channel.py,sha256=GZ2rrwFt2KtSG7At7kS9Bqh1KmYYw0HwcUnEjc-AWr8,22221
-torchaudio/lib/libtorchaudio.so,sha256=NAx2-wpTGHgmiVBbAmB5i47k_X2tCRGZTmt8SFe7Siw,535920
-torchaudio/lib/_torchaudio_ffmpeg4.so,sha256=S4Z7ZKCAwGW0XE_32hEEuX0AsJdHt2lngduxS6tMswE,504816
-torchaudio/lib/libtorchaudio_ffmpeg4.so,sha256=AtM_lOvldRY7lD6TPhrzK61kH1-rG2tWX_hJ5PnOR7U,571616
-torchaudio/lib/_torchaudio_ffmpeg5.so,sha256=wRswR9oBWsoMx8J1TY_A-naMjhMgy-Fq4tZPG0fPAss,504816
-torchaudio/lib/libtorchaudio_ffmpeg5.so,sha256=3IAmdWkdrTxSc9kulaO8Si8zuzb2aiMR2sXcgizUdE0,571616
-torchaudio/lib/_torchaudio_ffmpeg6.so,sha256=arjxog1Vrpncg2mATXDDje6gBvbwGwVRsFHRHsaM5PU,504816
-torchaudio/lib/_torchaudio.so,sha256=niaqvas_kjd1h43gD7CBAafdx0ZKbEYOfe69cmLY1cE,187584
-torchaudio/lib/libtorchaudio_ffmpeg6.so,sha256=54qZPQ-Kx-Z5e1D5BAEwZ7h5J5gwl5RjVOeGejeNx0M,571616
-torchaudio/lib/libtorchaudio_sox.so,sha256=E-S8L_ZCJzGqaVMu_fIeAS-cAUZMDkYAI0TMyW5wv1o,289952
-torchaudio/lib/_torchaudio_sox.so,sha256=0US0cwXqiNOzDPMCv-KYDeEtK6oDYTsg4suQhDquv3g,194688
+torchaudio/lib/libtorchaudio.so,sha256=leiPjavnzKaAUR8jdjRhx3I48cxVOVCf_gZ-Wkb20OA,535920
+torchaudio/lib/_torchaudio_ffmpeg4.so,sha256=2JAxEBiesfFAIISLX1mX1Mf1A_nBJb_Ud5AWtiZ-nEU,504816
+torchaudio/lib/libtorchaudio_ffmpeg4.so,sha256=ZgyyBReUHIybdygA4bYGGwZvbuZuPlBsA-G5NigJnUs,571616
+torchaudio/lib/_torchaudio_ffmpeg5.so,sha256=g7q7ypdNSzGhSmIM8dJao74J2xniCi7LPlU5oCTUugI,504816
+torchaudio/lib/libtorchaudio_ffmpeg5.so,sha256=lCKkfBEbrC0iu7LNiUsyYRF0_grfyNGFbMbaVkpujBo,571616
+torchaudio/lib/_torchaudio_ffmpeg6.so,sha256=GrqB903WpEpaJp_WYdCgjCm5F-oXy0IC4rVOcyxSs9Q,504816
+torchaudio/lib/_torchaudio.so,sha256=Vkr_cctnWlOLH5str_FLYwKRDBobFdRLiB-L9wrZ8KY,187584
+torchaudio/lib/libtorchaudio_ffmpeg6.so,sha256=O2HSdXWB_GKd-5ID_Zq9rOuGo1rMzSDZ_HS0txniknY,571616
+torchaudio/lib/libtorchaudio_sox.so,sha256=LyG09q1et9Xcvu4UCOPNTshSNkUOtURGzfeyg-XNMss,289952
+torchaudio/lib/_torchaudio_sox.so,sha256=Y1uuJnNdwJIuCHcsFhxFcN-ue9SIoPBDrgJKMXsXoBI,194688
 torchaudio/_extension/__init__.py,sha256=gMNUxx7DJmCR5z1tDKlJFihpEfmmoo53JOfKI5FUI_0,4262
 torchaudio/_extension/utils.py,sha256=jhWYhsh-QTh9Y4dxX2P0w9X8HrOIghCFKwnAK0ekz5s,8679
 torchaudio/sox_effects/sox_effects.py,sha256=aVs0GddGxQg1eU9fae4KtYYzQu3Nx6zPeJ7aso2WT-Y,11089
@@ -111,8 +111,8 @@ torchaudio/functional/filtering.py,sha256=or_OVPpqJjQBYIj_oi2-mAnYSUnCMI-07pZW7p
 torchaudio/functional/__init__.py,sha256=l-gC2WyY5COabU0lhkUS8EnwOYdEYR_6234OyoAIgnU,2357
 torchaudio/functional/_alignment.py,sha256=wmDeohWvuoYORYDeIRxnYUhUqv1uCUkaCZYLEK_ryUg,4695
 torchaudio/functional/functional.py,sha256=xLXVwb26048CMPOjPNA7d0H5GK53jziFi6Cw750bWLc,95982
-torchaudio-2.1.0.dist-info/RECORD,,
-torchaudio-2.1.0.dist-info/LICENSE,sha256=k6WIYahYzBCOa2uDPgjnbosqZjOeSoAHyKWowf-cQNY,1338
-torchaudio-2.1.0.dist-info/WHEEL,sha256=oITnL8r4K9lhj-wWSHlE2XUA0J8WLXdegrKtvi_34zA,108
-torchaudio-2.1.0.dist-info/top_level.txt,sha256=mPKWMIRWWW2JwbJN6wRckeN1gpbjhifapAF0Z9t7SMo,11
-torchaudio-2.1.0.dist-info/METADATA,sha256=Htp97OrVr2VWGqDHtidztmWcG6EfSY7qSe-0QORubeQ,5655
+torchaudio-2.1.1.dist-info/RECORD,,
+torchaudio-2.1.1.dist-info/LICENSE,sha256=k6WIYahYzBCOa2uDPgjnbosqZjOeSoAHyKWowf-cQNY,1338
+torchaudio-2.1.1.dist-info/WHEEL,sha256=oITnL8r4K9lhj-wWSHlE2XUA0J8WLXdegrKtvi_34zA,108
+torchaudio-2.1.1.dist-info/top_level.txt,sha256=mPKWMIRWWW2JwbJN6wRckeN1gpbjhifapAF0Z9t7SMo,11
+torchaudio-2.1.1.dist-info/METADATA,sha256=XpDpZl81t3G7h2UKbfF2-bE99asIWyUafMmavF51NL0,6351