How did SGLang HiCache with Mooncake Backend calculate cache hit ratio？

[YummuyWang]

In SGLang, the cache hit ratio calculate formula:

"cache_hit_rate": (
                    0
                    if sum(self.performance_metrics["prompt_len"]) == 0
                    else sum(self.performance_metrics["cached_tokens"])
                    / sum(self.performance_metrics["prompt_len"])
                ),

I think it only calculates cache hit ratio for GPU.
The blog SGLang HiCache with Mooncake Backend Benchmark shows the cache rate change.

But sglang/benchmark/hicache/bench_multirun.py also use get cached_tokens from meta_info.

cached_tokens = (data.get("meta_info") or {}).get(
                                        "cached_tokens", 0
                                    )

Is that means SGLang already support collect cache_hit_rate with data in CPU memory and SSD?