import torchaudio causes error with latest version of PyTorch

[SethHWeidman]

🐛 Bug

import torchaudio with latest version of PyTorch built from source causes error:

>>> import torchaudio
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/sethweidman/anaconda3/lib/python3.7/site-packages/torchaudio/__init__.py", line 5, in <module>
    import _torch_sox
ImportError: dlopen(/Users/sethweidman/anaconda3/lib/python3.7/site-packages/_torch_sox.cpython-37m-darwin.so, 2): Symbol not found: __ZN3c1011CPUTensorIdEv
  Referenced from: /Users/sethweidman/anaconda3/lib/python3.7/site-packages/_torch_sox.cpython-37m-darwin.so
  Expected in: flat namespace
 in /Users/sethweidman/anaconda3/lib/python3.7/site-packages/_torch_sox.cpython-37m-darwin.so

Expected behavior

Nothing.

Environment

Collecting environment information...
PyTorch version: 1.4.0a0+013ca32
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.14.6
GCC version: Could not collect
CMake version: version 3.14.0

Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip] msgpack-numpy==0.4.4.3
[pip] numpy==1.17.2
[pip] numpydoc==0.9.1
[pip] pytorch-sphinx-theme==0.0.24
[pip] torch==1.4.0a0+013ca32
[pip] torchaudio==0.3.0
[pip] torchtext==0.4.0
[pip] torchvision==0.5.0a0+e8b830f
[conda] _tflow_select             2.3.0                       mkl  
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      233  
[conda] mkl-include               2019.4                      233  
[conda] mkl-service               2.3.0            py37hfbe908c_0  
[conda] mkl_fft                   1.0.14           py37h5e564d8_0  
[conda] mkl_random                1.1.0            py37ha771720_0  
[conda] pytorch-sphinx-theme      0.0.24                    dev_0    <develop>
[conda] tensorflow                1.13.1          mkl_py37h70c3834_0  
[conda] tensorflow-base           1.13.1          mkl_py37h66b1bf0_0  
[conda] torch                     1.4.0a0+013ca32          pypi_0    pypi
[conda] torchaudio                0.3.0                    pypi_0    pypi
[conda] torchtext                 0.4.0                    pypi_0    pypi
[conda] torchvision               0.4.0                    pypi_0    pypi

[vincentqb]

I cannot reproduce this on my mac:

conda create -n pytorch python=3.7
conda activate pytorch
conda install -c conda-forge sox
conda install -c pytorch pytorch
conda install -c pytorch torchaudio
python -c "import torchaudio"

but conda didn't install the latest pytorch :)

PyTorch version: 1.2.0
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.14.6
GCC version: Could not collect
CMake version: version 3.15.2

Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip3] numpy==1.16.4
[pip3] torch==1.1.0.post2
[pip3] torchvision==0.3.0
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      233  
[conda] mkl-service               2.3.0            py37hfbe908c_0  
[conda] mkl_fft                   1.0.14           py37h5e564d8_0  
[conda] mkl_random                1.1.0            py37ha771720_0  
[conda] pytorch                   1.2.0                   py3.7_0    pytorch
[conda] torchaudio                0.3.0                      py37    pytorch

[vincentqb]

I can reproduce this when compiling pytorch and torchaudio.

PyTorch version: 1.3.0a0+0a6d33d
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.14.6
GCC version: Could not collect
CMake version: version 3.14.0

Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip3] numpy==1.16.4
[pip3] torch==1.1.0.post2
[pip3] torchvision==0.3.0
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      233  
[conda] mkl-include               2019.4                      233  
[conda] mkl-service               2.3.0            py37hfbe908c_0  
[conda] mkl_fft                   1.0.14           py37h5e564d8_0  
[conda] mkl_random                1.0.2            py37h27c97d8_0  
[conda] torch                     1.3.0a0+26cd4c5          pypi_0    pypi
[conda] torchaudio                0.4.0a0+d64bf8c          pypi_0    pypi

[vincentqb]

I know @ezyang has been regenerating the binaries, so I'm waiting to see the output of pytorch/tutorials#687.

The error may be related to this comment?

[vincentqb]

Successful locally

python setup.py clean --all
conda create -n audio-built python=3.7
conda activate audio-built
conda install -c pytorch pytorch
conda install -c conda-forge sox
MACOSX_DEPLOYMENT_TARGET=10.9 CC=clang CXX=clang++ python setup.py install
python -c "import torchaudio"                                          

on

PyTorch version: 1.3.0
Is debug build: No
CUDA used to build PyTorch: None

OS: Mac OSX 10.14.6
GCC version: Could not collect
CMake version: version 3.14.0

Python version: 3.7
Is CUDA available: No
CUDA runtime version: No CUDA
GPU models and configuration: No CUDA
Nvidia driver version: No CUDA
cuDNN version: No CUDA

Versions of relevant libraries:
[pip3] numpy==1.16.4
[pip3] torch==1.1.0.post2
[pip3] torchvision==0.3.0
[conda] blas                      1.0                         mkl  
[conda] mkl                       2019.4                      233  
[conda] mkl-service               2.3.0            py37hfbe908c_0  
[conda] mkl_fft                   1.0.14           py37h5e564d8_0  
[conda] mkl_random                1.1.0            py37ha771720_0  
[conda] pytorch                   1.3.0                   py3.7_0    pytorch
[conda] torchaudio                0.4.0a0+d64bf8c          pypi_0    pypi

[vincentqb]

conda create -n pytorch python=3.7
conda activate pytorch
conda install -c conda-forge sox
conda install -c pytorch torchaudio
python -c "import torchaudio"

@SethHWeidman -- can you try the above on your laptop?

[vincentqb]

Linking comment in pytorch/tutorials#687

[vincentqb]

As mentioned in #308, the new binaries need to be uploaded to pypi and conda. The former is done, but not the latter.

[vincentqb]

As mentioned here, uploading the new version fixed the issue on my side. Please feel free to re-open if it is not fixed on yours.

conda create -n pytorch-conda python=3.7
conda activate pytorch-conda
conda install -c conda-forge sox
conda install -c pytorch torchaudio
python -c "import torchaudio"