Open paul0403 opened 1 month ago
Thanks @paul0403! I just wanted to check with @erick-xanadu, is this expected with our current support for quantum transforms?
If I recall correctly, we integrated PL tape transforms assuming that the QNode being transformed was a straight line program only (e.g., that no for loops or conditionals were present).
Thanks @paul0403! I just wanted to check with @erick-xanadu, is this expected with our current support for quantum transforms?
If I recall correctly, we integrated PL tape transforms assuming that the QNode being transformed was a straight line program only (e.g., that no for loops or conditionals were present).
We already disallow MCMs with transforms that produce multiple tapes, but I think we are not restricted enough. Disallowing all hybrid ops for the moment should prevent the issue at least.
So the fix here would simply be additional validation? Sounds good!
@josh146 yes, the assumption is that QNode being transformed was a straight line program only. I think there was some initial validation, but perhaps it did not cover all possible cases.