PaddleSpeech docker develop-gpu-cuda10.2-cudnn7-latest 缺失 libsndfile1-dev 和 環境參數CUDA_VISIBLE_DEVICES

Paddle可以說是各種坑，但支持國產，含淚試用了百度飛漿的Speech。

https://github.com/PaddlePaddle/PaddleSpeech/issues/3158

1. 坑點

Docker develop-gpu-cuda10.2-cudnn7-latest 缺失:
1. libsndfile1-dev
2. CUDA_VISIBLE_DEVICES

 

2. 安裝教程

也沒什麼安裝教程。下載docker鏡像和項目源碼。

?

1	docker pull paddlecloud/paddlespeech:develop-gpu-cuda10.2-cudnn7-latest

　　docker鏡像

?

1	git clone https://github.com/PaddlePaddle/PaddleSpeech

　　項目源碼

 

3. 測試顯示卡

?

1	nvidia-docker run --runtime=nvidia -it -v ~/PycharmProjects/PaddleSpeech:/home paddlecloud/paddlespeech:develop-gpu-cuda10.2-cudnn7-latest /bin/bash

　　進入Docker鏡像

 

?

1	nvidia-smi

 

?

1

nvcc -V

?

1	./test_cuda

 

4. Bug

?

1	cd demos/style_fs2 && ./run.sh

　　出現error

 

4.1 缺失 libsndfile1-dev

?

1 2 3 4 5 6

Traceback (most recent call last):   File "style_syn.py", line 22, in <module>     import soundfile as sf   File "/usr/local/python3.7.0/lib/python3.7/site-packages/soundfile.py", line 192, in <module>     _snd = _ffi.dlopen(_explicit_libname) OSError: cannot load library 'libsndfile.so': libsndfile.so: cannot open shared object file: No such file or directory

　　我猜測，這是CUDA10.2官方docker的bug，里面缺少了libsndfile1-dev。

?

1	apt install libsndfile1-dev

　　成功解決

 

4.2 缺失 環境參數CUDA_VISIBLE_DEVICES

?

1 2 3 4 5 6 7 8 9 10 11

Traceback (most recent call last):   File "style_syn.py", line 228, in <module>     main()   File "style_syn.py", line 209, in main     paddle.set_device("gpu")   File "/usr/local/python3.7.0/lib/python3.7/site-packages/paddle/device/__init__.py", line 313, in set_device     place = _convert_to_place(device)   File "/usr/local/python3.7.0/lib/python3.7/site-packages/paddle/device/__init__.py", line 204, in _convert_to_place     place = core.CUDAPlace(ParallelEnv().dev_id) OSError: (External) CUDA error(100), no CUDA-capable device is detected.   [Hint: 'cudaErrorNoDevice'. This indicates that no CUDA-capable devices were detected by the installed CUDA driver. ] (at /paddle/paddle/phi/backends/gpu/cuda/cuda_info.cc:66)

　　因缺失 環境參數CUDA_VISIBLE_DEVICES，導致PaddleSpeech失敗獲取GPU。

?

1 2	import os os.environ["CUDA_VISIBLE_DEVICES"] = "0"

　　當設置CUDA_VISIBLE_DEVICES=0後，成功解決。

 

5.0 成功