Contradicting error messages on where to set max_qubits/n_qubits when running emulation

[rainij]

Today I wanted to try out qnexus job submission and followed the tutorial here. I made a small modification in replacing repeat_until_success(q: qubit, attempts: int) by

@guppy
def cnot_chain() -> None:
    q0: qubit = qubit()
    q1: qubit = qubit()
    q2: qubit = qubit()

    h(q0)
    cx(q0, q1)
    cx(q1, q2)

    result("c0", measure(q0))
    result("c1", measure(q1))
    result("c2", measure(q2))

to make the compilation to hugr work (otherwise I get an entrypoint error, see also #288). Unfortunately I still get an error in qnx.start_execute_job. The error message is essentially this:

ResourceCreateFailed: Failed to create resource with status code: 400, message:
{"message":"Helios emulation must have max-qubits set in QuantinuumConfig.compiler_options set."}

It seems the error message wants me to change my code like this:

compiler_options = QuantinuumCompilerOptions(max_qubits=4) # <- new

config = qnx.QuantinuumConfig(
    device_name="Helios-1E",
    max_cost=6, #np.ceil(prediction)
    compiler_options=compiler_options, # <- new
)

But note that actually the QuantinuumCompilerOptions has no field for max_qubits. It didn't complain about it though. When I run this get another error message:

ResourceCreateFailed: Failed to create resource with status code: 400, message:
{"message":"max-qubits/n_qubits must currently be set explicitly in the BackendConfig."}

This sounds as if it contradicts the previous error message. Anyway, I tried out this

config = qnx.QuantinuumConfig(
    device_name="Helios-1E",
    max_cost=6, #np.ceil(prediction)
    compiler_options=compiler_options,
    max_qubits=4, # <- new (again typechecker complains though)
    n_qubits=4,   # <- new (again typechecker complains though)
)

But I get the same error (second) message. And if I comment out the compiler_options I get back the first error message.

I believe it is a bug. In any case, some feedback would be appricated.

[vanyae-cqc]

Hi @rainij - very sorry for the slow response. I hope I can clear this up.

Regarding the nature of QuantinuumConfig.compiler_options

The QuantinuumConfig.compiler_options is effectively a way that we have used in the past to support experimental extra parameters. It can be used by passing Python kwargs into a class constructor OR as a kind of typed dictionary:

# 1: dictionary form
compiler_options={"experimental_key": "experimental_value"}

OR
# 2: class constructor form
compiler_options = QuantinuumCompilerOptions(experimental_key="experimental_value")

Regarding the max-qubits parameter

Unfortunately, when the max-qubits parameter was added it was realised that it can only be provided in the dictionary form, so if you want to use QuantinuumConfig you will need to specify it like so:

config = qnx.QuantinuumConfig(
    device_name="Helios-1E",
    max_cost=6,
    compiler_options={"max-qubits": 4}
)

For this reason, we are starting to recommend using HeliosConfig for Helios-generation systems

To improve the UX here we have designed a new HeliosConfig class that can be used instead of QuantinuumConfig for systems in the Helios generation.

You can find example usage in this documentation notebook

API docs will be deployed to the qnexus API docs in the next few days. For now you can see the class directly in the source code if you'd like to use it.

Once again very sorry for the confusing error messages and the slow response. Please let me know if you have any issues

[rainij]

Thank you very much @vanyae-cqc. Your reply is very helpful. It is very nice that you took the time to explain it in detail 🙂.