矩阵和3D张量之间的theano点积

Question

我有一个矩阵和3D张量定义如下：

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)

虽然它最终给出了正确的答案，但在中间它显示了许多错误

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

我的预期输出是

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

我怎样才能正确地将这两者相乘？