Related-task warm-starting

Measurement-efficiency compounds when you solve a sequence of related instances — re-optimizing as conditions drift, or a family of similar circuits / devices. Do not restart cold each time; carry the configuration forward.

Reuse the last configuration

copyx = warm_x0 or initial_config
for instance in stream:                # related problems, one after another
    opt = SWCOptimizer(license_key="EVAL-...", n=len(x))
    x = opt.start(x)                   # start from where the last one ended
    for _ in range(rounds):
        x = opt.step(measure(x, instance), score=objective(x, instance))
    opt.end()
    warm_x0 = x                        # carry it to the next instance

When consecutive instances are close, a warm start reaches target in far fewer measurements than a cold one. This is the regime the runtime is built for; the further apart the instances, the smaller the carry-over benefit.