Traceback (most recent call last):
File "/usr/local/bin/deepspeech", line 7, in <module>
from deepspeech.client import main
File "/usr/local/lib/python2.7/dist-packages/deepspeech/client.py", line 10, in <module>
from deepspeech.model import Model
File "/usr/local/lib/python2.7/dist-packages/deepspeech/model.py", line 21, in <module>
_model = swig_import_helper()
File "/usr/local/lib/python2.7/dist-packages/deepspeech/model.py", line 20, in swig_import_helper
return importlib.import_module('_model')
File "/usr/lib/python2.7/importlib/__init__.py", line 37, in import_module
__import__(name)
ImportError: No module named _model
Yes, I went through this all as well, you want CUDA 8.0 and cuDNN 6. Lucky cuDNN is just a few lib files copied to the right dir so its really easy.
Heres a guide: https://yangcha.github.io/Install-CUDA8/
Also be aware that the nvidia-smi tool wont show any memory or GPU usage because your card isn’t a pro level card like Quadro or Tesla etc.
For example:
Then run the deepspeech command and you wont see the process attach. However, if you run with and without the deepspeech-gpu package installed, you should notice a performance difference.
Also when deepspeech runs it will output the Tensorflow GPU device information. Good luck!
@lissyx You are right. Because deepspeech needs tensorflow1.4, I have to reinstall it and use CuDNN v6. So I removed the old deepspeech project and use command
pip install https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.gpu/artifacts/public/deepspeech_gpu-0.1.0-cp27-cp27mu-manylinux1_x86_64.whl
. Finally it’s fine and no error.@lissyx thanks
-i removed the old install by
sudo rm -r deepspeech*
-then installing in my case withsudo -E pip2 install https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.gpu/artifacts/public/deepspeech_gpu-0.1.0-cp27-cp27mu-manylinux1_x86_64.whl
now it run without any PATH modification, thanks