2024-09-13/summary.json

{
    "google/gemma-2b": {
        "backend.cache_implementation=null,backend.torch_compile=False": {
            "a05ce550bf1cda8ae3bffcc62607597770460a42": {
                "metrics": {
                    "prefill.latency.mean": 0.020703951835632325,
                    "prefill.throughput.value": 338.09970461546,
                    "decode.latency.mean": 2.5267764892578124,
                    "decode.throughput.value": 50.26166759898244,
                    "per_token.latency.mean": 0.019894590512974056,
                    "per_token.throughput.value": 50.26491997147969
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=False": {
            "a05ce550bf1cda8ae3bffcc62607597770460a42": {
                "metrics": {
                    "prefill.latency.mean": 0.02243292808532715,
                    "prefill.throughput.value": 312.0412981031458,
                    "decode.latency.mean": 2.6453446044921876,
                    "decode.throughput.value": 48.0088680258652,
                    "per_token.latency.mean": 0.02082844218306654,
                    "per_token.throughput.value": 48.01127185656721
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=True": {
            "a05ce550bf1cda8ae3bffcc62607597770460a42": {
                "metrics": {
                    "prefill.latency.mean": 0.01417632007598877,
                    "prefill.throughput.value": 493.7811761076341,
                    "decode.latency.mean": 1.5648624877929689,
                    "decode.throughput.value": 81.1572908103361,
                    "per_token.latency.mean": 0.012320643023243101,
                    "per_token.throughput.value": 81.16459490900623
                }
            }
        }
    }
}