as_tensor_variable(range(n)) should work

[williambdean]

as_tensor_variable(range(n)) currently raises NotImplementedError because there is no singledispatch handler for range objects.

Workaround: as_tensor_variable(list(range(n))).

Proposed fix: add a range handler in pytensor/tensor/basic.py that converts to a list and delegates to the existing sequence handler (same behavior as list/tuple). Empty ranges are already covered via the empty-sequence path.

[ricardoV94]

if numpy accepts range (as a non object array) we can do the same yes

[williambdean]

np.array?

[ricardoV94]

np.array?

Yeah, as tensor_variable is kinda the same as np.asarray

[williambdean]

np.array(range(10))
# array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])

[ricardoV94]

sounds good, I suggest implementing as pt.arange