Me bad.... after some farther investigation the above pickle code seems to
save the trained dbn....
On Wednesday, 15 February 2017 13:47:11 UTC+2, Akataios wrote:
>
> Hi,
>
> I am trying to pickle a trained dbn for later use (from deeplearning 01
> docum. examples) but dont get the results i am expecting.
>
> I use to pickle:
>
> pickle.dump(dbn, f)
>
> and to unpickle, in a predict function:
>
> dbnt=pickle.load(open('trainedDBN.pkl', 'rb'))
>
> DBN object:
>
> dbn = DBN(numpy_rng=numpy_rng,
> n_ins=in_num,
> hidden_layers_sizes=[n_l_1,n_l_2],
> n_outs=out_num)
> I suspect that numpy_rng that is used to initialize sub objects of dbn is
> the cause (the loaded dbn gets new values from numpy_rng and not the ones
> it was trained).
> If numpy_rng is the cause , how do i pickle a trained DBN?
>
> Thanks in advance...
>
--
---
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.
