3
$\begingroup$

In the Linux command line,type wolfram then use NetTrain[net, mnist, Automatic, "LossEvolutionPlot"] can get the LossEvolutionPlot.But how can we view the LossEvolutionPlot in realtime if don't use frontend?

In the tensorflow there is tensorboard,we can monitor the process of training via browser.Can we do things like this in Mathematica?

enter image description here

$\endgroup$
  • 1
    $\begingroup$ If you want to monitor the training while running a kernel in Linux without a frontend, maybe you should consider using Mathematica's remote kernel functionality. You can use the frontend on your local machine and connect to the kernel on the remote machine. $\endgroup$ – xslittlegrass May 25 '17 at 4:20
  • $\begingroup$ NetTrain produces a dynamic window output you can follow up with. You can use it either on your desktop or in the cloud (I guess in every Wolfram environment). Did you try, and it didn't work? $\endgroup$ – SuTron May 25 '17 at 6:09
  • $\begingroup$ I try.it throws error. ref:mathematica.stackexchange.com/questions/146823/… $\endgroup$ – partida May 25 '17 at 6:24
  • $\begingroup$ Then you have to wait for that question to be answered. $\endgroup$ – SuTron May 25 '17 at 7:26
  • 2
    $\begingroup$ For now, you could use TrainingProgressFunction option to save the losses in a log file every few seconds and then dynamically plot that with some other program (a jupyter notebook connected to the remote machine, for example) or periodically sync that file with your local machine. $\endgroup$ – dan7geo May 25 '17 at 19:18

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.