`qml.shadow_expval` should respect the device RNG seed for sampling

[andrijapau]

Expected behavior

Setting the seed kwarg at the device should result in a deterministic sampling execution.

Actual behavior

Doesn't get properly seeded.

Additional information

If you seed both at the MP and device level it works as expected,

import pennylane as qp

@qp.set_shots(10)
@qp.qnode(qp.device("default.qubit", seed=123, wires=1))
def circuit():
    qp.H(0)
    return qp.shadow_expval(qp.Z(0), seed=42) # SET HERE

(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))
(array(-1.2), np.float64(0.0))

Source code

import pennylane as qp

for i in range(10):
    @qp.set_shots(10)
    @qp.qnode(qp.device("default.qubit", wires=1, seed=42))
    def circuit():
        qp.H(0)
        return qp.shadow_expval(qp.Z(0)), qp.expval(qp.Z(0))
    print(circuit())

Tracebacks

(array(-0.6), np.float64(-0.2))
(array(0.6), np.float64(-0.2))
(array(0.6), np.float64(-0.2))
(array(-0.3), np.float64(-0.2))
(array(-0.3), np.float64(-0.2))
(array(-0.6), np.float64(-0.2))
(array(-0.3), np.float64(-0.2))
(array(-0.3), np.float64(-0.2))
(array(-0.3), np.float64(-0.2))
(array(0.3), np.float64(-0.2))

System information

PL v0.44

Existing GitHub issues

• I have searched existing GitHub issues to make sure the issue does not already exist.

[albi3ro]

This is a feature, rather than a bug. The seed on the measurement process configures the what the measurement process invovles, while the seed on the device configures the samples that result from those measurements.

[andrijapau]

This is a feature, rather than a bug. The seed on the measurement process configures the what the measurement process invovles, while the seed on the device configures the samples that result from those measurements.

Thanks for the clarification! I'll update the issue.