how to enable CUDA in the plugins

[cyf203]

Dear ImagePy team,

I came from cellpose repository and I am a cellpose user. When I use cellpose, I can select CPU or GPU to compute the segregation. But, I didn't see this option here. The another question is when I install cellpose-plgs from imagePy. I saw mxnet and mxnet-mkl was installed. But, my CUDA version is 10.1. If I use cellpose alone, I should install mxnet-cu101. Can you give me some suggestions?

[yxdragon]

I can add the selection in the UI, but imagepy's pluginsmanager use pip solve the dependence, So you need install mxnet-cu101 with your pip.
(do you install imagepy on your local python? or download the release one which with a isolate python?)

[cyf203]

I install imagepy on my local python via pip install. Thanks for your suggestion. But, I wonder even if I installed mxnet-cu101. But, will the plugins automatically use GPU to do the segmentation? Because when I use cellpose alone, I believe no matter in GUI or script, the GPU is always an option, but not sure if it is by default using GPU or not. So, if you have a way to force GPU computation or ask user to select will be very helpful!

[yxdragon]

if you are aware of python, maybe you can see the source code directly. ImagePy 's plugin is very easy to understand (all the cellpose plugin is in about 30 lines).

config "para and view", para are the parameter needed, view indicate how to show.
then call the core function in run.

here is the plugin development document.

about if GPU is used by default, I need ask @carsen-stringer.

use_gpu = utils.use_gpu()
if use_gpu:
    device = mx.gpu()
else:
    device = mx.cpu()

I think that means try to use gpu, if failed, use cpu, right?
I can add a choice in the plugin's dialog.