GPU wizards and clusters/queues....
Is your feature request related to a problem? Please describe. Trying to run gAutomatch wizard in a server where GPUs are only available after the queue.
Describe the solution you'd like GPU wizard to work, using the queue? (hard, I know)
Describe alternatives you've considered Otherwise tell me that is not available if there is no GPU.
In my opinion, wizards are by design interactive, if you don't have a local GPU, submit the job to the queue then. I don't see the reason to implement this.
I agree with @azazellochg. One option is to submit an interactive job into the queue, log into the assigned node, launch the wizard interactively there...and let the job running (or re-submit into more nodes once the parameters are tuned)
then, do not show the wizzard? Because now wizzard is shown, picking windows pop up but nothing happens. The user does not know is parameters are wrong or is not working.
How this can be implemented?.. check if CUDA=True or CUDA_LIB exists in config file?
There are Pyrrhon modules that can check if there are GPUs available
El jue., 8 nov. 2018 22:12, Gregory Sharov [email protected] escribió:
How this can be implemented?.. check if CUDA=True in config file?
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/I2PC/scipion/issues/1820#issuecomment-437157650, or mute the thread https://github.com/notifications/unsubscribe-auth/AAv84XidG9iG79Z2Wv2U_eaetputq6hCks5utJ4zgaJpZM4YP7i9 .
@delarosatrevin I assigned you since I dont know about such python modules...
I think, scipion should provide that information to the plugin first, and then wizards, maybe, have a flag? needsGPU?