Torchaudio Compilation Error #674
Description
🐛 Bug
When trying to compile torchaudio from source after python setup.py install within anaconda, it fails with an error that configure: error: cannot find flac (see below for full information).
To Reproduce
Steps to reproduce the behavior:
- Compile PyTorch from source in conda
- Compile Torchaudio from source in conda
See code as below:
module load cuda/10.1
conda create -n audio-master python=3.7
conda activate audio-master
# install cuda with needed version
conda install -c pytorch magma-cuda101
conda install numpy ninja pyyaml mkl mkl-include setuptools cmake cffi typing
git clone --recursive https://github.com/pytorch/pytorch
cd pytorch
git submodule sync
git submodule update --init --recursive
git submodule sync
git submodule update --init --recursive
python setup.py clean --all
export CMAKE_PREFIX_PATH=${CONDA_PREFIX:-"$(dirname $(which conda))/../"}
python setup.py install
cd ..
python -c "import torch; print(torch.__version__); print(torch.cuda.is_available());"
conda install -c conda-forge sox
git clone https://github.com/pytorch/audio
cd audio
python setup.py clean --all
python setup.py install
cd ..
python -c "import torchaudio;"
Error message:
After python setup.py install for torchaudio, it shows:
Building sox
configure: error: in `/private/home/jimchen90/audio/third_party/tmp/sox-14.4.2':
configure: error: cannot find flac
See `config.log' for more details
subprocess.CalledProcessError: Command '['/private/home/jimchen90/audio/build_tools/setup_helpers/build_third_party.sh']' returned non-zero exit status 1.
Expected behavior
Torchaudio can be imported outside the source folder.
Environment
- PyTorch Version: 1.6.0a0+78acc9d
- OS: Linux
- How you installed PyTorch: Conda
- Build command you used : As described above
- Python version: 3.7
- CUDA/cuDNN version: CUDA 10.1
Additional context
With the help of Moto @mthrok , torchaudio can be imported after adding an external sox file in the audio directory by:
conda uninstall sox
rm -rf third_party
touch .use_external_sox
python setup.py clean develop
The result of python -m torch.utils.collect_env
Collecting environment information...
PyTorch version: 1.6.0a0+78acc9d
Is debug build: No
CUDA used to build PyTorch: 10.1
OS: Ubuntu 18.04.3 LTS
GCC version: (Ubuntu 7.4.0-1ubuntu1~18.04.1) 7.4.0
CMake version: version 3.14.0
Python version: 3.7
Is CUDA available: Yes
CUDA runtime version: 10.1.105
GPU models and configuration:
GPU 0: Quadro GP100
GPU 1: Quadro GP100
Nvidia driver version: 418.116.00
cuDNN version: Could not collect
Versions of relevant libraries:
[pip] numpy==1.18.1
[pip] torch==1.6.0a0+78acc9d
[pip] torchaudio==0.6.0a0+fd73d8f
[conda] blas 1.0 mkl
[conda] magma-cuda101 2.5.2 1 pytorch
[conda] mkl 2020.1 217
[conda] mkl-include 2020.1 217
[conda] mkl-service 2.3.0 py37he904b0f_0
[conda] mkl_fft 1.0.15 py37ha843d7b_0
[conda] mkl_random 1.1.1 py37h0573a6f_0
[conda] numpy 1.18.1 py37h4f9e942_0
[conda] numpy-base 1.18.1 py37hde5b4d6_1
[conda] torch 1.6.0a0+78acc9d pypi_0 pypi
[conda] torchaudio 0.6.0a0+fd73d8f dev_0 <develop>
Related issues can be seen in #666 and #669. Thank @vincentqb and @mthrok for helping me solve this issue.