Can you try using different Theano cache? This will make them independent. Just give each process a different compiledir or base_compiledir Theano flag:
http://deeplearning.net/software/theano/library/config.html#config.base_compiledir http://deeplearning.net/software/theano/library/config.html#config.compiledir That should do the trick. On Fri, Aug 5, 2016 at 4:43 PM, <[email protected]> wrote: > Hi, everyone, > I encounter a strange problem. > When I run two similar programs in the same GPU at the same time, one of > them will finally encounter the error as following when it comes to the > theano function compiling. > > IOError:[Error2]No such file or Directory: compiledir_Liunx-4.1--MANJARO- > x86_64-with-glibc2.2.5--2.7.12-64/tempFDXRY9/key.pkl > > In my program, there is a loop for average in which I redefine the net and > the functions in it. In the first loop of both programs it performs well, > but when it comes to the second loop some gonna fail. > > I really appreciate it for any idea about it. > > Dylan > > -- > > --- > 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.
