Amplitude state array #45
Conversation
| @@ -118,7 +118,8 @@ def run_xir_program(program: XIRProgram) -> List[Union[np.number, np.ndarray]]: | |||
|
|
|||
| num_wires = len(program.wires) | |||
| # TODO: Extract the Fock cutoff dimension from the XIR script. | |||
There was a problem hiding this comment.
This TODO can now be removed, right? Or, perhaps even better, exchanged with a comment saying that a qubit circuit is assumed if no "dimension" is set.
There was a problem hiding this comment.
Not quite. I experimented with using options in this PR but found that the implementation was not easy to get right. Specifically, we'll need to validate that each gate instance is compatible with the dimension of the circuit (which may involve a change to each Gate subclass). It would also be nice to allow implicit cutoff parameters too; however, this is also non-trivial without some kind of reflection hack.
That being said, I am currently working on a separate PR to address these points and remove the TODO 🙂.
There was a problem hiding this comment.
Update: The PR is #47.
There was a problem hiding this comment.
Looks great! 🚀
Thanks @thisac!
Only one thing: I think it might be good to add a test for the "dimension" option as well, using a cutoff value larger than 2.
As I mentioned in the review, I would prefer to do this in a future PR since it is a little tricky to implement correctly.
| @@ -118,7 +118,8 @@ def run_xir_program(program: XIRProgram) -> List[Union[np.number, np.ndarray]]: | |||
|
|
|||
| num_wires = len(program.wires) | |||
| # TODO: Extract the Fock cutoff dimension from the XIR script. | |||
There was a problem hiding this comment.
Not quite. I experimented with using options in this PR but found that the implementation was not easy to get right. Specifically, we'll need to validate that each gate instance is compatible with the dimension of the circuit (which may involve a change to each Gate subclass). It would also be nice to allow implicit cutoff parameters too; however, this is also non-trivial without some kind of reflection hack.
That being said, I am currently working on a separate PR to address these points and remove the TODO 🙂.
Context:
To support the remote execution of XIR programs using Jet, the interpreter used an integral representation of an amplitude state; however, since #28, this is no longer necessary and amplitude states should instead be specified as arrays.
Description of the Change:
stateparameter ofamplitudestatements to be an array.Benefits:
Possible Drawbacks:
None.
Related GitHub Issues:
None.