2024-07-01/summary.json

{
    "google/gemma-2b": {
        "backend.cache_implementation=null,backend.torch_compile=False": {
            "e65502951593a76844e872fee9c56b805598538a": {
                "metrics": {
                    "prefill.latency.mean": 0.02063888072967529,
                    "prefill.throughput.value": 339.1656791705356,
                    "decode.latency.mean": 2.5170850830078124,
                    "decode.throughput.value": 50.4551875728572,
                    "per_token.latency.mean": 0.019902868775982158,
                    "per_token.throughput.value": 50.24401312471862
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=False": {
            "e65502951593a76844e872fee9c56b805598538a": {
                "metrics": {
                    "prefill.latency.mean": 0.021916111946105956,
                    "prefill.throughput.value": 319.39971912964046,
                    "decode.latency.mean": 2.56069580078125,
                    "decode.throughput.value": 49.59589497559733,
                    "per_token.latency.mean": 0.020242197293066696,
                    "per_token.throughput.value": 49.4017514759881
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=True": {
            "e65502951593a76844e872fee9c56b805598538a": {
                "metrics": {
                    "prefill.latency.mean": 0.014443552017211914,
                    "prefill.throughput.value": 484.6453276630517,
                    "decode.latency.mean": 1.573007568359375,
                    "decode.throughput.value": 80.73705591414239,
                    "per_token.latency.mean": 0.012437937103241328,
                    "per_token.throughput.value": 80.3991845029832
                }
            }
        }
    }
}