torch.compile takes effect with just a single-line decorator added before a model or function, significantly lowering the barrier to performance optimization compared to the previously required manual use of torch.jit.script or torch.jit.trace to trigger compilation optimization.