We have a memory leak in new code. We have a PR that fix it: https://github.com/Theano/Theano/pull/5832
Another fix is to install conda mkl packages. It will use another code path that don't have this leak. If you don't use conda, you can use the system openblas, check out ubuntu installation page. On Tue, Apr 11, 2017 at 10:30 AM Mikkel <[email protected]> wrote: > This may way off, but is it possible your theanorc or theano flags has > allow_gc=False, and that that is causing your problem? > > > On Thursday, April 6, 2017 at 1:14:41 AM UTC+2, Julien wrote: > > I am running this popular tutorial > <http://deeplearning.net/tutorial/mlp.html> that is training a simple MLP > on MNIST. I run on CPU. While monitoring my computer activity in the > training loop, I can see that the memory usage is steadily increasing, > until it reaches 100% and everything crashes (need to reboot). I am on > Ubuntu 16.04 and using the bleeding edge version of Theano. > > > I have made a post <https://github.com/Theano/Theano/issues/5810> on > theano-issues on github and a question > <http://stackoverflow.com/questions/43234170/memory-gradually-increasing-until-crash-while-training-simple-mlp-on-cpu-with-th> > on StackOverflow. > > > Is anyone experiencing the same issue on CPU? > > -- > > --- > You received this message because you are subscribed to the Google Groups > "theano-users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- --- You received this message because you are subscribed to the Google Groups "theano-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
