Which Version of Tensorflow is this running on?

[tiwarikaran]

[zsyzzsoft]

tf 1.14 or 1.15, see here for the requirements.

[usherbob]

Hi @zsyzzsoft @tiwarikaran ,

I have difficulty setting the environment.

My virtual environment variants are listed here:

`` python 3.6.13

tensorflow-gpu 1.14.0

CUDA: 10.2 ``

There is an error when I tried to run_generator with the pretrained model.

tensorflow.python.framework.errors_impl.NotFoundError: /home/wangjiahua1/code/co-mod-gan-master/dnnlib/tflib/_cudacache/fused_bias_act_546f398f36920850dca748ac250ee717.so: undefined symbol: _ZN10tensorflow12OpDefBuilder6OutputESs

Is it necessary to switch the CUDA version from 10.2 to 10.0 ?

[zsyzzsoft]

Hi @zsyzzsoft @tiwarikaran ,

I have difficulty setting the environment.

My virtual environment variants are listed here:

`` python 3.6.13

tensorflow-gpu 1.14.0

CUDA: 10.2 ``

There is an error when I tried to run_generator with the pretrained model.

tensorflow.python.framework.errors_impl.NotFoundError: /home/wangjiahua1/code/co-mod-gan-master/dnnlib/tflib/_cudacache/fused_bias_act_546f398f36920850dca748ac250ee717.so: undefined symbol: _ZN10tensorflow12OpDefBuilder6OutputESs

Is it necessary to switch the CUDA version from 10.2 to 10.0 ?

This is an error related to stylegan2's custom ops. You can try this solution.

[usherbob]

Thanks,

It now works when I change --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=0 to --compiler-options \'-fPIC -D_GLIBCXX_USE_CXX11_ABI=1 here, since I use gcc 5.4.

No need to switch CUDA version.

[tiwarikaran]

can you exactly specify which tf version this is using, the best thing would be add a requirements.txt file, that would be sensational

[zsyzzsoft]

tf 1.14 or 1.15, see here for the requirements. You can use tf 1.15.

[tiwarikaran]

@zsyzzsoft I did and it is still giving errors :/

[tiwarikaran]

if possible, can you add the req,txt file for this?