If you can update to Theano dev version and revert your work around, it
would allow to know if this problem is fixed or not in Theano. It would be
great to fix it if it isn't. It it don't, can you give the new error
message? The lines number would change and maybe other small other thing
that can
I was able to work around it, I think. I had a TensorVariable that I was
adding a dimension to with np.newaxis and I replaced it with a reshape -
that removed the NoneConst from the tree, though I'm not sure why.
I would like to use the existing convolution code, but I'm interpolating a
On Sun, Oct 16, 2016, Pascal Lamblin wrote:
> That optimization probably makes some assumptions that are not true in
> your case, hence the error message, but the only consequence should be
> that this particular optimization gets skipped. This should not make
> your program crash or return
>From what I can see, that message comes from an optimization trying to
move some operation out of a scan loop (to avoid it being repeated at
each time step, given that it operates on a non-sequence).
That optimization probably makes some assumptions that are not true in
your case, hence the
I'm attempting to write a differentiable, parameterized image rotation
layer in Theano with decent performance - as result I'm doing some slightly
unusual things in terms of indexing in order to get decent performance.
When I try to optimize the resulting theano graph using fast_run I get a