cudaq.from_state() dosn't work correctly

[yukwangmin]

Required prerequisites

• Make sure you've read the documentation. Your issue may be addressed there.
• Search the issue tracker to verify that this hasn't already been reported. +1 or comment there if it has.
• If possible, make a PR with a failing test to give us a starting point to work on!

Describe the bug

Since it seems that cudaq.from_state() function doesn't work correctly, I made the following code to test cudaq.from_state():

Steps to reproduce the bug

import cudaq
import numpy as np

def main():
    qc = cudaq.make_kernel()
    qr = qc.qalloc(4)

    psi0_list = [-8.32045644e-17-8.90888763e-20j, -1.03601322e-17-4.47087881e-17j, -2.15333550e-17+1.30326534e-16j, 8.71688857e-02+1.47625478e-01j, 2.84452598e-16-2.04288994e-16j, -4.84060082e-01-8.19783350e-01j, 9.02894336e-02+1.52910304e-01j, -6.39974562e-17+1.07964618e-16j, 1.33111503e-16+4.86229405e-17j,  9.02894336e-02+1.52910304e-01j, -1.73288055e-02-2.93473201e-02j,  2.20931787e-17-9.36645781e-17j, 2.29425514e-03+3.88545190e-03j, -7.44241453e-17+1.85423718e-16j, -2.11836598e-16+5.41151020e-16j,  6.59700092e-17-9.11582920e-18j]
    psi0 = np.array(psi0_list)

    cudaq.from_state(qc, qr, psi0)
    psi = cudaq.get_state(qc)

    print(f'psi: {psi}')
    print(f'np.inner(psi, np.conjugate(psi0)): {np.inner(psi, np.conjugate(psi0))}')
    print(f'np.linalg.norm(np.inner(psi, np.conjugate(psi0))): {np.linalg.norm(np.inner(psi, np.conjugate(psi0)))}')

if __name__ == '__main__':
    main()

Output

psi: 2.07782e-17-1.26498e-18j 1.31988e-33+4.42978e-33j 3.28998e-18-6.10936e-18j 0.143081+0.0944437j 6.84596e-17+8.74027e-17j -0.0216218+0.950844j -0.00954028-0.182289j 6.13566e-15-5.25258e-15j 0.130984+0.124653j -6.87235e-18+2.68913e-17j 7.30323e-15-2.22874e-14j 5.98876e-29-3.1875e-29j 0.00287869-0.00337475j 3.92844e-18+8.12333e-19j 0+0j -0.000523755+0.000640495j

np.inner(psi, np.conjugate(psi0)): (-0.7713473234796359-0.5058996935573853j)
np.linalg.norm(np.inner(psi, np.conjugate(psi0))): 0.9224484773583045

Expected behavior

In the above experiment, I called cudaq.from_state(psi0).
After that, I just called cudaq.get_state().

So I expected 1 or -1 for the inner product between psi0 and psi. But the inner product was (-0.7713473234796359-0.5058996935573853j).

I also checked the norms of psi0 and psi as follows:

np.linalg.norm(psi0)
0.9999999995652816

np.linalg.norm(psi)
0.9999997175385762

Is this a regression? If it is, put the last known working version (or commit) here.

Not a regression

Environment

• CUDA Quantum version: Docker image "ghcr.io/1tnguyen/cuda-quantum:mpich-231710"
• Python version: 3.10.12
• C++ compiler:
• Operating system:
  PRETTY_NAME="Ubuntu 22.04.3 LTS"
  NAME="Ubuntu"
  VERSION_ID="22.04"
  VERSION="22.04.3 LTS (Jammy Jellyfish)"
  VERSION_CODENAME=jammy
  ID=ubuntu
  ID_LIKE=debian
  HOME_URL="https://www.ubuntu.com/"
  SUPPORT_URL="https://help.ubuntu.com/"
  BUG_REPORT_URL="https://bugs.launchpad.net/ubuntu/"
  PRIVACY_POLICY_URL="https://www.ubuntu.com/legal/terms-and-policies/privacy-policy"
  UBUNTU_CODENAME=jammy

Suggestions

Qiskit Test

For more comparison, I did the above computation with psi0 in Qiskit as follows:

qr = QuantumRegister(4)
cr = ClassicalRegister(4)
qc = QuantumCircuit(qr,cr)
qc.initialize(psi0)
backend_sim = Aer.get_backend('statevector_simulator')
job = execute(qc, backend_sim)
result = job.result()
qiskit_out = result.get_statevector(qc)

qiskit_out
Statevector([-8.32045644e-17-8.90888763e-20j,
-1.03601322e-17-4.47087881e-17j,
-2.15333550e-17+1.30326534e-16j,
8.71688857e-02+1.47625478e-01j,
2.84452598e-16-2.04288994e-16j,
-4.84060082e-01-8.19783350e-01j,
9.02894336e-02+1.52910304e-01j,
-6.39974562e-17+1.07964618e-16j,
1.33111503e-16+4.86229405e-17j,
9.02894336e-02+1.52910304e-01j,
-1.73288055e-02-2.93473201e-02j,
2.20931787e-17-9.36645781e-17j,
2.29425514e-03+3.88545190e-03j,
-7.44241453e-17+1.85423718e-16j,
-2.11836598e-16+5.41151020e-16j,
6.59700092e-17-9.11582920e-18j],
dims=(2, 2, 2, 2))

np.inner(qiskit_out, np.conjugate(psi0))
(0.9999999991305633+0j)

np.linalg.norm(np.inner(qiskit_out, np.conjugate(psi0)))
0.9999999991305633

Qiskit made (0.9999999991305633+0j) for the inner product between psi0 and qiskit_out as expected.

[boschmitt]

Possible duplicate #819

[schweitzpgi]

#1107

[bettinaheim]

We believe this issue should be resolved with the new state preparation feature. Please open a new issue if there are any issues with the new implementation/API.