Added support for CPU only tensorflow

[tomhog]

Hi

I'm running on macOS and cuda support is pretty flaky so i've made some minor modifications to allow the code to run on cpu only if cuda is not available.

I'm an absolute beginner when it comes to python so perhaps it's possible to store the results of device = torch.device("cuda" if torch.cuda.is_available() else "cpu") Then res use it when needed.

Thanks Tom

[bertjiazheng]

Hi @tomhog ,

Thanks for your PR, but I think the best way to run and inference the network is on the GPU devices. If anyone needs to run the network on the CPU, they can modify these things like what you did.

Best, Jia

[tomhog]

Hi Jia

No problem, the code should prioritize GPU usage if it's available but I understand you not wanting to merge it. I just wanted to play fair as you've shared such awesome code. I'll leave my changes on a branch of my fork so others can use it if need be, Mac users don't have great CUDA support.

Thanks Tom

On Tue, 13 Aug 2019 at 12:11, Jia Zheng notifications@github.com wrote:

Hi @tomhog https://github.com/tomhog ,

Thanks for your PR, but I think the best way to run and inference the network is on the GPU devices. If anyone needs to run the network on the CPU, they can modify these things like what you did.

Best, Jia

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/svip-lab/PlanarReconstruction/pull/10?email_source=notifications&email_token=AAC3KAIFOIU26DAANUQBJW3QEKJG3A5CNFSM4IBZ4AI2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD4FKRRA#issuecomment-520792260, or mute the thread https://github.com/notifications/unsubscribe-auth/AAC3KAONKG7BU7KVKVMXEMDQEKJG3ANCNFSM4IBZ4AIQ .

[bertjiazheng]

@tomhog

I am really busy these days. I will rectify the code to support the CPU device once I have time. For now, I will refer the readers to your repo for CPU support in README.

Best, Jia