On Tue, Aug 23, 2016, [email protected] wrote:
> for instance when I would issue
> from theano.tensor.nnet. -> (right arrow is just the tab for autocompletion)
> I would not have been able to see the conv2d method there
"from ..." expects a module name, but here "theano.tensor.nnet" is the
module name and "conv2d" is the function.
> but I could import it normally as you indicated with
> from theano.tensor.nnet import conv2d
You could also get it with "import theano.tensor.nnet. -> ... as conv2d".
> rval = __import__(module_name, {}, {}, [module_name])
> ImportError: ('The following error happened while compiling the node',
> CorrMM{(5, 5), (4, 4), (1, 1)}(X, Subtensor{::, ::, ::int64, ::int64}.0),
> '\n',
> '/home/user/.theano/compiledir_Linux-2.6-el6.x86_64-x86_64-with-centos-6.6-Final-x86_64-2.7.12-64/tmpipUDuA/ebdb8712d9028a2d5a99048e808bba16.so:
>
> undefined symbol: *openblas_set_num_threads*', '[CorrMM{(5, 5), (4, 4), (1,
> 1)}(X, <TensorType(float32, 4D)>)]')
>
>
> Does anyone have any idea where this error is coming from?
> I have openblas installed locally in my home directory. Some other scripts
> that I'm running with the old conv2d are running fine without any problems
> and their utilizing openblas properly.
The old conv2d is not linking with blas directly, which is why it would work.
Is the path to the openblas library in your LD_LIBRARY_PATH as well as
your LIBRARY_PATH?
>
>
> Many thanks!
>
>
> Regards,
> M.D.
>
>
>
> On Tuesday, August 23, 2016 at 6:07:57 PM UTC+1, Pascal Lamblin wrote:
> >
> > Use theano.tensor.nnet.conv2d instead.
> >
> > On Tue, Aug 23, 2016, [email protected] <javascript:> wrote:
> > > Hi, sorry for the disturbance but when I use the border_mode=(int, int)
> > on
> > > theano.tensor.nnet.conv.conv2d it does't work it always gives me the
> > > following error:
> > >
> > > Traceback (most recent call last):
> > > File "bigger.py", line 285, in <module>
> > > noise_layers = model(X, W, 0.2, 0.5)
> > > File "bigger.py", line 196, in model
> > > h1 = rectify(conv2d(X, W[0], border_mode=(5,5),
> > subsample=(4,4)))
> > > File
> > >
> > "/Users/user/.local/lib/python2.7/site-packages/theano/tensor/nnet/conv.py",
> >
> >
> > > line 151, in conv2d
> > > imshp=imshp, kshp=kshp, nkern=nkern, bsize=bsize, **kargs)
> > > File
> > >
> > "/Users/user/.local/lib/python2.7/site-packages/theano/tensor/nnet/conv.py",
> >
> >
> > > line 533, in __init__
> > > raise Exception("Mode %s not implemented" % str(self.out_mode))
> > > Exception: Mode (5, 5) not implemented
> > >
> > > I'm using theano 0.9 from developer repo.
> > >
> > > When the conv.py implementation in theano it says that is an old
> > deprecated
> > > interface plus in the border_mode I can only see two options which valid
> > > and full. Besides that I don't see any option for using your own border
> > > dimension like a tuple list (int, int)
> > >
> > > Am I missing something? Or maybe I'm doing something wrong?
> > >
> > > Thanks!
> > >
> > >
> > >
> > > On Tuesday, August 23, 2016 at 2:19:38 PM UTC+1, nouiz wrote:
> > > >
> > > > Border mode for convolution is implemented. What do you mean
> > otherwise? Do
> > > > you have links to those precious request?
> > > >
> > > > Le 23 août 2016 06:31, <[email protected] <javascript:>> a écrit
> > :
> > > >
> > > >> Dear all,
> > > >>
> > > >> may I ask why aren't border_modes still not implemented in theano.
> > > >> There's been discussion since 2014, 215 were people have expressed
> > their
> > > >> interest but still yet not implemented. Is there any plan to include
> > this
> > > >> sometime in the near future in Theano. I'm running latest dev
> > version.
> > > >>
> > > >> 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] <javascript:>.
> > > >> 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] <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.
--
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.
