Qrack cannot return catalyst mid-circuit measurements

[yorgossot]

The following code

import pennylane as qml
from catalyst import qjit, measure


dev = qml.device("qrack.simulator", wires=3, shots=5)

@qjit()
@qml.qnode(dev, mcm_method="one-shot")
def circuit():
    qml.X(0)
    m_res = [measure(0), measure(1)]

    return [qml.sample(m) for m in m_res]

print(circuit())

gives:

[Array([0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1], dtype=int64), Array([0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1], dtype=int64)]

It created:

• 2 IDENTICAL arrays (as many as the elements sampled)
• each array containing the results of ALL 3 of the qubits reversed and then stacked for every shot. i.e.
  1st shot (0, 0, 1,) 2nd shot(0, 0, 1) 3rd shot(0, 0, 1) 4th shot(0, 0, 1) 5th shot(0, 0, 1) gives
  Array([0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 1], dtype=int64)

These values though are the result of some final measurement happening on all qubits at the end of the execution as resetting the qubit in state "1" gives:

import pennylane as qml
from catalyst import qjit, measure


dev = qml.device("qrack.simulator", wires=3, shots=5)

@qjit()
@qml.qnode(dev, mcm_method="one-shot")
def circuit():
    qml.X(0)
    m_res = [measure(0,True), measure(1)]

    return [qml.sample(m) for m in m_res]

print(circuit())

gives:

[Array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=int64), Array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0], dtype=int64)]

tagging @josh146
related issue #1626

[josh146]

Thanks or reporting this! Just so I understand, what's the relation between this issue and #1626, #1611?

[yorgossot]

I am not really sure. It's likely that it is the same fix will resolve both but I thought I should post them as separate ones just in case this is not the case.

[dime10]

The example now gives the following results for me:

[Array([1, 1, 1, 1, 1], dtype=int64), Array([0, 0, 0, 0, 0], dtype=int64)]

which looks exactly right. This is with the latest Qrack release and the Catalyst main branch, could you confirm whether this is fixed for you as well @yorgossot ? :)