{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.022729344367980957, "prefill.throughput.value": 307.9719276839752, "decode.latency.mean": 2.3878223876953126, "decode.throughput.value": 53.186535420072985, "per_token.latency.mean": 0.018877755681516624, "per_token.throughput.value": 52.97239867232251 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.022289711952209475, "prefill.throughput.value": 314.04622971388926, "decode.latency.mean": 2.5987576904296876, "decode.throughput.value": 48.86950425108752, "per_token.latency.mean": 0.020550745048070616, "per_token.throughput.value": 48.66003629848368 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.014034639835357667, "prefill.throughput.value": 498.76591648364223, "decode.latency.mean": 1.5729868774414062, "decode.throughput.value": 80.73811792160406, "per_token.latency.mean": 0.012438730368029931, "per_token.throughput.value": 80.39405714350104 } } } } }