What is error?
qingtianwu opened this issue · comments
What is error?
Originally posted by @MuhammadAsadJaved in #27 (comment)
- My environment:
my environment
Keras 1.0.2
numpy 1.11.2 or 1.16.2
jupyter 1.0.0
Theano 0.8.2
tensorflow-gpu 1.4.0
h5py 2.8.0
youtube-dl 2016.6.27
Cuda version 10.1.105
CUDNN version 7.5.0
NVIDIA-SMI 418.39
- Errors:
Exception: ("We can't determine the cudnn version as it is not available", 'CUDA not available')
AttributeError: ('The following error happened while compiling the node', <theano.sandbox.cuda.DnnVersion object at 0x7eff3655c510>(), '\n', 'The following error happened while compiling the node', GpuCAReduce{add}{1}(<CudaNdarrayType(float32, vector)>), '\n', "'module' object has no attribute '_get_ndarray_c_version'")
I have listed the two bugs that I have recorded. And sorry for my pressing the "closed" button by mistakes.
sorry dear I can not help you with this problem. I am also facing some problems while running this project. I think we should ask about details system and libraries requirements.
@Nghingtim As I was also trying to run this project with CUDA 10 + cudNN 7.3 on RTX 2080 GPU, there was an error about Theano "this version of Theano is not compatible with cudnn version, degrade cudnn to version 5.0" . Now I am trying to use CUDA 8 + cudNN 5.0 but still not working. now I found the error same as you. Did you find any solutions?
@Nghingtim OK please let me know when you run it successfully . I will also inform you if I get ride of this error.
Hi, I can not recall the exact versions of the CUDA/cudnn software used to run this. I only remember I used a Titan X to do the training. Take into account that this code was developed 3 years ago, so the software used was the latest version at that time. I am aware that with the fast evolution of the Deep Learning frameworks, this code right now is barely runnable. My recommendation is to try to get the weights and architecture and rebuild the same computational graph with a more recent framework like TensorFlow. Sorry I can not help you more with this.
OK. Thank you very much for your response. @albertomontesg