[theano-users] theano dot product between a matrix and 3D tensor

[shyam krishna khadka]

I have a matrix and 3D tensor defined as below :


import numpy as np
import theano
import theano.tensor as T

a = T.matrix('a', dtype='float32')
c = T.tensor3('c',dtype='float32')
d = T.batched_dot(c, a)
g = theano.function([a,c],d)
Y = [[[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, 0]], [[0, 0
,0, 0], [0, 1, 0, 0],[0, 0, 1, 0],[0, 0, 0, 1]]]
X = [[ 0.5052417 , 0.22012063, 0.21787818, 0.41821062, 1, 1, 1, 0], [
0.48668074, 0.26137591, 0.240702 , 0.41308364, 0, 1, 1, 1]]
x = np.array(X, dtype='float32')
y = np.array(Y, dtype='float32')
print g(x[:,0:4], y)


Although it gives correct answer in the end, but in the middle it shows
many error as


ValueError: get_scalar_constant_value detected deterministic IndexError: x.
shape[2] when x.ndim=2. x=Subtensor{int64}.0
ERROR (theano.gof.opt): Optimization failure due to:
local_gpua_gemmbatch
ERROR (theano.gof.opt): node: BatchedDot(c, a)
ERROR (theano.gof.opt): TRACEBACK:
ValueError: get_scalar_constant_value detected deterministic IndexError:
x.shape[2] when x.ndim=2. x=Subtensor{int64}.0



My expected output is

[[ 0.50524169 0.22012062 0.21787818 0. ]
[ 0. 0.2613759 0.240702 0.41308364]]

And I am getting that output as well in the end.

How can I correctly multiply those two, without any errors ?