2024-09-03/summary.json

{
    "google/gemma-2b": {
        "backend.cache_implementation=null,backend.torch_compile=False": {
            "cff06aac6fad28019930be03f5d467055bf62177": {
                "metrics": {
                    "prefill.latency.mean": 0.02121953582763672,
                    "prefill.throughput.value": 329.8846900733365,
                    "decode.latency.mean": 2.5355677490234374,
                    "decode.throughput.value": 50.08740154898779,
                    "per_token.latency.mean": 0.019963835438405433,
                    "per_token.throughput.value": 50.0905751845785
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=False": {
            "cff06aac6fad28019930be03f5d467055bf62177": {
                "metrics": {
                    "prefill.latency.mean": 0.02198051166534424,
                    "prefill.throughput.value": 318.4639241604466,
                    "decode.latency.mean": 2.5951973876953125,
                    "decode.throughput.value": 48.93654740951456,
                    "per_token.latency.mean": 0.020433629719291146,
                    "per_token.throughput.value": 48.93893124900428
                }
            }
        },
        "backend.cache_implementation=static,backend.torch_compile=True": {
            "cff06aac6fad28019930be03f5d467055bf62177": {
                "metrics": {
                    "prefill.latency.mean": 0.014191952228546143,
                    "prefill.throughput.value": 493.2372859824019,
                    "decode.latency.mean": 1.5656904296874998,
                    "decode.throughput.value": 81.11437458639142,
                    "per_token.latency.mean": 0.012327145779226709,
                    "per_token.throughput.value": 81.12177935667528
                }
            }
        }
    }
}