I’m looking for the way to name a tvm.tensor from topi without generating from tvm.compute.
Is there any way to achieve something like the following example?
n = tvm.var('n')
x0 = tvm.placeholder((n,), name='x0')
x1 = tvm.placeholder((n,), name='x1')
# this way works
# y = tvm.compute((n,), lambda i: x0[i] + x1[i], name='y')
# however this doesn't work
# there is no name keyward provided
y = topi.add(x0, x1, name='y')
# neither, op.name is read only
y.op.name = 'y'