Theano张量类型：Python

Question

CODE:

x=T.dscalar('x')
y=T.dscalar('y')
z=T.dscalar('z')
z=x+y
f= function([x,y],z)

UPON RUNNING:

$ T.dscalar

TensorType(float64, scalar)

$ x.type

TensorType(float64, scalar)

$ z.type

TensorType(float64, scalar)

$ x.type is T.dscalar

True <<<<<<<-------------------------------(1)

$ z.type is T.dscalar

False <<<<<<<-------------------------------(2)

Why isn't (1) & (2) same?

Answer 1

Line z = T.dscalar('z') is pointless because it's not a graph input. It's gets discarded when you do z=x+y . The type object for z is constructed in make_node method of an Op, instead of just using T.dscalar .

Finally, Python is operator compares whether two object is identical rather than equal . If you check z.type == T.dscalar it will be True .