将帖子翻译为中文 Hi, I am doing a research project. We are trying to ship a package that can do model selection on Neural Network without time-consuming cross validation. Our alogirhtm requires accurate gradient and hessian and thats why we choose Theano. We find out that the compilation process takes a long time. I am wondering if we can precomile the model and pickle a theano function associate with an output of the model. This is more user-friendly than only pickling the model and tensor variable. We successfully pickle the model and tensor variables but when we try to pickle the theano function which is the step that takes the most of time, it yields error "Cannot pickle fortran objects".
We appreciate any help from you guys. Best, DreamSoul -- --- 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.
