magpie icon indicating copy to clipboard operation
magpie copied to clipboard

question about leveraging multiple GPUs running magpie?

Open JessicaKuo opened this issue 6 years ago • 3 comments

Hello, I would like to ask a question about the use of GPU in Magpie setting.

Now I am only using one GPU: Nvidia 1080 Ti in Magpie.

However, I am curious to use multiple GPUs while running the Magpie. Is it possible to get multiple GPUs. If it's possible to get multiple GPUs, please let me know your view to get it.

Thanks in advance.

JessicaKuo avatar Nov 02 '18 09:11 JessicaKuo

Yes, it's possible. If you set them up correctly and TensorFlow detects them (or any other backend that you're using), Magpie will take advantage of them. There's no custom GPU setup for Magpie itself, if you manage to set it up for TensorFlow - it'll work.

On 2 Nov 2018, at 06:14, Jessica10105009 [email protected] wrote:

Hello, I would like to ask a question about the use of GPU in Magpie setting.

Now I am only using one GPU: Nvidia 1080 Ti in Magpie.

However, I am curious to use multiple GPUs while running the Magpie. Is it possible to get multiple GPUs. If it's possible to get multiple GPUs, please let me know your view to get it.

Thanks in advance.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or mute the thread.

jstypka avatar Nov 02 '18 11:11 jstypka

like @jstypka said you should get the performance "By default" if you've configured correctly, also:

Check if a GPU is underutilized by running nvidia-smi -l 2. If GPU utilization is not approaching 80-100%, then the input pipeline may be the bottleneck.

https://www.tensorflow.org/guide/performance/overview

dorg-ekrolewicz avatar Nov 02 '18 22:11 dorg-ekrolewicz

OK,I will try. Thanks for your opinion!

JessicaKuo avatar Nov 09 '18 07:11 JessicaKuo