Ah yes ! That happened to me before... Sorry for the post and thanks for the answer :)
Le mardi 13 septembre 2016 20:27:12 UTC+2, Pascal Lamblin a écrit : > > Especially if you are drawing random numbers, please make sure to > include the "update" dictionary returned by `scan(...)` when you call > theano.function(...). > > On Tue, Sep 13, 2016, Léo Crestel wrote: > > Hi, > > > > I'm trying to write my own version of the Rnn-rbm model > > (http://deeplearning.net/tutorial/code/rnnrbm.py) so that i can make > some > > modifications. > > Since I failed to reproduce the same results with my implementation, i > > wanted to observe some variables, among them v_sample. > > > > To do so, i compile a function : > > > > self.*obs = theano.function(inputs=[v],* > > * outputs=[v_sample]* > > * )* > > > > > > that i inserted line 239, right after this part : > > > > (v, v_sam*ple, cost, monitor, params, updates_train, v_t,* > > * updates_generate, bv_t, bh_t, fe_positive, fe_negative) = > \* > > * build_rnnrbm(* > > * r[1] - r[0],* > > * n_hidden,* > > * n_hidden_recurrent* > > * )* > > > > > > However, i got the following error, which suggests that an input of the > > graph is missing : > > > > theano.gof.fg.MissingInpu*tError: A variable that is an input to the > graph > > was neither provided as an input to the function nor given a value. A > chain > > of variables leading from this input to an output is [v[t-1], dot.0, > > Elemwise{add,no_inplace}.0, sigmoid.0, Shape.0, > RandomFunction{binomial}.1, > > dot.0, Elemwise{add,no_inplace}.0, sigmoid.0, Shape.0, > > RandomFunction{binomial}.0]. This chain may not be unique* > > *Backtrace when the variable is created:* > > * File "test_dd.py", line 344, in <module>* > > * model = test_rnnrbm()* > > * File "test_dd.py", line 336, in test_rnnrbm* > > * model = RnnRbm()* > > * File "test_dd.py", line 236, in __init__* > > * n_hidden_recurrent* > > * File "test_dd.py", line 187, in build_rnnrbm* > > * build_rbm(v, W, bv_t[:], bh_t[:], k=15)* > > * File "test_dd.py", line 82, in build_rbm* > > * n_steps=k)* > > > > > > > > Why ? Using pydotprint, i looked at the graph for v_sample, it is clear > > that v should is the only input node... > > > > -- > > > > --- > > 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] <javascript:>. > > For more options, visit https://groups.google.com/d/optout. > > > -- > Pascal > -- --- 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.
