Hi,
I just ran into this problem when using `theano.gradient.jacobian`.
Previous posts show sampling in computation may be the reason, but I think
I have blocked it with consider_constant.
Is there anything wrong?
Simplified code:
import theano
import theano.tensor as T
from theano.sandbox import rng_mrg
srng = rng_mrg.MRG_RandomStreams(seed=7297)
h_samples = srng.normal(avg=0, std=2, size=(10, 20))
bq = T.fmatrix('bq')
q = (h_samples - bq).flatten()
jacob = theano.gradient.jacobian(q, [bq], consider_constant=[h_samples]) #
Exception here
Thanks.
--
---
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.
