search

dhlab-epfl / dhSegment

Generic framework for historical document processing
https://dhlab-epfl.github.com/dhSegment
GNU General Public License v3.0
370 stars 116 forks source link

Which CUDA, cudnn and tensorflow versions are meant to be used together #54

Closed kamilspiewak closed 4 years ago

kamilspiewak commented 4 years ago

I am doing some small project for multilabel classification, but during training I'm keep getting errors about image

image Tests were made with TensorflowGPU 1.13.1, CUDA 10.1 and cudnn 7.6.5.32 under Windows 10

Is this related to my bad configuration of project, incorrect installation of CUDA and cudnn or versions missmatch between those three?

solivr commented 4 years ago

Yes it is most probably due to versions missmatch between tensorflow, cuda and cudnn. On Ubuntu 18.04, I have tensorflow 1.13, cuda 9.0 and cudnn 7.6.3. Maybe try to have a look at tensorfow issues related to this error message.

Gawajn commented 4 years ago

you can check https://www.tensorflow.org/install/source_windows#gpu (Linux: https://www.tensorflow.org/install/source#gpu) for compatible cuda/cudnn/tensorflow versions