Function launcher

src/fused_mm_sampling/bench/speed_test.py:172–175

(launch: "nvbench.Launch")

        torch.cuda.synchronize()

        def launcher(launch: "nvbench.Launch"):
            stream = _as_torch_stream(launch.get_stream())
            with torch.cuda.stream(stream):
                sampler.sample(**kwargs)

        state.exec(launcher, batched=False)