Hi, Currenlty we don't have such an op. I'm not sure how to implement the case of the end of the computation. We could have it return an empty tensor, but this don't seem super clean. So such an op should be used with an ifelse op that would skip all the computation... Nothing super fast to implement in all cases.
What prevent you from doing that outside Theano? You could wrap the Theano function in a class/function that would do what you want. Fred On Sat, Jan 7, 2017 at 10:09 AM, Maxim Kochurov <[email protected]> wrote: > Hi I wonder if it is possible to create a custom Op that takes user > defined generator when init and yields form that generator on runtime. I > want to use it for passing minibatches to the graph in such way > > -- > > --- > 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.
