Enable GPU support without installing drivers

I installed Deep learning studio after installing Nvidia Cuda toolkit so my version is latest than what DL Studio tries to install

But when I tried to run a pre-installed sample ‘MNIST Handwritten Digits Classifier’ I noticed that entire training is done on CPU and there was 0 load on GPU

I feel that I dont have GPU support enabled, so how would I go about doing that?


Hi Zubair,

For GPU training to work, a compatible NVIDIA driver is required to be present in the system. You can choose to install the compatible driver during Nvidia Cuda toolkit installation.

In case you have skipped the driver installation, then GPU support may fail to get enabled.


Hi Rajendra,

I already have CUDA 9 and cuDNN installed and I am able to train on GPU using python, its just that Deepcognition wanted to installed v8 so I skipped it

How can I enable GPU support if I already have those installed?

Hi Zubair,

In that case you can wait for Deep Learning Studio 2.0 version for Windows which will support CUDA 9. It should be available in few days.


Oh so the studio doesn’t support latest version of CUDA is it?

Yes, current 1.5.1 version only supports CUDA 8. CUDA 8 and CUDA 9 can be installed in parallel but latest driver may not work well with CUDA 8.

Are there any updates on leveraging exisitng cuda9 ? i also skipped installing cuda from DLS as I already have cuda9 installed. I feel GPU is not being levraged. Is there any way to check this?

In logs tab, DLS does a GPU supported check at the boot. So you can logs tab to confirm whether GPU is getting enabled or not.

If it is not getting enabled, then I would recommend you to let CUDA installer complete during DLS installation.