Common Errors
The mistakes people make first, with the messages you will see.
AttributeError: Option 'X' is not valid for this backend
| Symptom | backend.run(qc, time_budget_seconds=1.0) raises AttributeError |
| Cause | MENOSBackend accepts only shots, seed_simulator and want_statevector |
| Solution | Use those three. Time and memory budgets are available through the native API |
ValueError: MENOSBackend.run() does not support batching
| Symptom | backend.run([qc1, qc2]) fails |
| Cause | One circuit per call |
| Solution | Call run once per circuit |
TypeError: MENOSBackend.run() expects a QuantumCircuit
| Symptom | You passed something other than a circuit |
| Solution | Pass a qiskit.QuantumCircuit |
My counts keys have leading zeros
| Symptom | Keys like 001 when you expected 1 |
| Cause | Keys are as wide as the classical register, and unwritten bits are 0 |
| Solution | Declare a register of the width you want, for example QuantumCircuit(3, 1). See Counts |
My counts keys look like 0x3
| Symptom | The native API returns {'0x0': 507, '0x3': 493} |
| Cause | The native API keys counts by hexadecimal string |
| Solution | Use MENOSBackend for Qiskit's binary keys, or convert: format(int(key, 16), "b") |
Results differ between runs
| Symptom | Two runs of the same circuit give different counts |
| Cause | MENOSBackend draws a fresh seed when seed_simulator is None |
| Solution | Pass seed_simulator. See Determinism |
The statevector is very large
| Symptom | Slow or refused jobs with want_statevector=True |
| Cause | A statevector is 16 · 2ⁿ bytes |
| Solution | Ask for counts or an expectation value when that is enough. See Memory Errors |
explain() labels run into values
| Symptom | candidate 0strategy=direct_dense … |
| Cause | Labels are padded to 11 characters, so a label of 11 or more runs into its value. It is the shipped format |
| Solution | Nothing to fix. See Plan Explanation |
InternalError
| Symptom | menos._core.InternalError: … |
| Cause | A MENOS invariant was violated, or a case MENOS does not handle |
| Solution | Check Exceptions for the known case (an all-identity observable) and the CUDA entries in Troubleshooting CUDA. Otherwise it is a bug worth reporting |
ResourceExceededError for a job that used to work
| Symptom | A job now raises ResourceExceededError mentioning an observed runtime |
| Cause | You set a time budget, the run overran it, and no other plan fits |
| Solution | Raise the budget, or remove it. See Replanning |
A replanned job took longer than its budget
| Symptom | Total wall time is a multiple of time_budget_seconds |
| Cause | The budget bounds one attempt; abandoned attempts add to the total |
| Solution | This is by design. See Time Budgets |