GPU supprot about unit-tensorflow HOT 8 CLOSED

Do you mean not to use gpu?

Have you checked "watch -n 0.1 nvidia-smi"?

from unit-tensorflow.

Hi, I have a computer with cuda8.0 installed, but I only find this code running in cpu version. Could please tell me how to solve it ? Thank you very much.

from unit-tensorflow.

Check this

1. pip install tensorflow-gpu (do not install tensorflow)

2. set gpu_num
   ex) export CUDA_VISIBLE_DEVICES=0,1,2,3

3. Try it
   python main_multi_gpu.py --gpu_num 4 --batch_size 1

from unit-tensorflow.

Sorry, it doesn't work

from unit-tensorflow.

Well, I'm running well ...
What is the GPU version?

from unit-tensorflow.

My GPU is Nvidia Geforce 1070, Cuda version is 8.0, cudnn version is 5.1.

from unit-tensorflow.

Hi, I suddenly found that I run this code in python2.7, which makes me feel very embarrassed. When I switch to python3, my gpu is now working. Additionly, the tensorflow I used in python3 is 1.2, but this version doesn't hava leaky_relu and instance_norm, one needs to add these two functions.

from unit-tensorflow.

Could you please send me a training set? I try to run the code by https://github.com/mingyuliutw/UNIT/tree/master/datasets/gta2city
but it has some problems. My email is [email protected],thank you very much!

from unit-tensorflow.