Testing the code in the paper

[rsln-s]

Adding the tests for snippets in the upcoming paper "Fast Simulation of High-Depth QAOA Circuits" describing QOKit.

[rsln-s]

test_listing_3 is perhaps OK to skip, but test_listing_2 must pass. Currently, I get the following error on a GPU instance:

======================================================== FAILURES ========================================================
_____________________________________________________ test_listing_2 _____________________________________________________

    def test_listing_2():
        import qokit
    
        simclass = qokit.fur.choose_simulator_xycomplete()
        n = 8
        terms = qokit.labs.get_terms(n)
>       sim = simclass(n, terms=terms)

tests/test_examples_in_the_paper.py:30: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
qokit/fur/nbcuda/qaoa_simulator.py:24: in __init__
    super().__init__(n_qubits, costs, terms)
qokit/fur/qaoa_simulator_base.py:63: in __init__
    self._hc_diag = self._diag_from_terms(terms)
qokit/fur/nbcuda/qaoa_simulator.py:32: in _diag_from_terms
    precompute_gpu(0, self.n_qubits, terms, out)
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 

rank = 0, n_local_qubits = 8, terms = [(0, 1, 3, 4), (0, 1, 4, 5), (0, 2, 3, 5), (3, 7), (4, 6), (5, 7), ...]
out = <numba.cuda.cudadrv.devicearray.DeviceNDArray object at 0x7f138519b7c0>

    def compute_costs(rank: int, n_local_qubits: int, terms, out):
        offset = rank << n_local_qubits
        n = len(out)
        zero_init_kernel.forall(n)(out)
    
>       for coef, pos in terms:
E       ValueError: too many values to unpack (expected 2)

qokit/fur/diagonal_precomputation/gpu_numba.py:39: ValueError

[rsln-s]

The same issue happens in CPU-only as evidenced by the GitHub workflow logs:

...
>       for coeff, pos in weighted_terms:
E       ValueError: too many values to unpack (expected 2)
...

[rsln-s]

LGTM modulo the removal of assertion in qokit/labs.py

[rsln-s]

Any reason to drop this check? Now that get_terms returns weights well, should this become assert set(circuit) == set(get_term_indices(N)) or something like that?

[danlkv]

This is already checked with tests, so runinng get_terms seemed redundant. We can keep the assert but then the tests are a bit silly:)

[rsln-s]

Ah, right! Thanks for catching