"""Auto-generated example: 0 Float, 2 Categorical (with repeats)."""
import random
import bencher as bn
class NetworkConfig(bn.ParametrizedSweep):
"""Measures throughput across protocol and region combinations."""
protocol = bn.StringSweep(["http", "grpc"], doc="Network protocol")
region = bn.StringSweep(["us-east", "eu-west", "ap-south"], doc="Deployment region")
throughput = bn.ResultFloat(units="req/s", doc="Request throughput")
def benchmark(self):
proto_factor = {"http": 1.0, "grpc": 1.8}[self.protocol]
region_base = {"us-east": 500, "eu-west": 420, "ap-south": 350}[self.region]
self.throughput = region_base * proto_factor + random.gauss(0, 0.15 * 50)
def example_sweep_0_float_2_cat_with_repeats(run_cfg: bn.BenchRunCfg | None = None) -> bn.Bench:
"""0 Float, 2 Categorical (with repeats)."""
bench = NetworkConfig().to_bench(run_cfg)
bench.plot_sweep(input_vars=['protocol', 'region'], result_vars=['throughput'], description='A 0 float + 2 categorical parameter sweep with multiple repeats per combination. Repeating measurements reveals the noise structure of your benchmark. If your function is deterministic, all repeats will be identical; if it has stochastic components, repeats let you estimate confidence intervals and distinguish signal from noise. The benchmark function must be pure -- if past calls affect future calls through side effects, the statistics will be invalid. Categorical-only sweeps produce bar/swarm plots comparing discrete settings.', post_description='Swarm/violin plots show the distribution of repeated measurements. If repeat has high variance, it suggests either measurement noise or unintended side effects in the benchmark function.')
return bench
if __name__ == "__main__":
bn.run(example_sweep_0_float_2_cat_with_repeats, subsampling_divisions=4, repeats=10)