{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.02022262477874756, "prefill.throughput.value": 346.14695553054355, "decode.latency.mean": 2.4749503173828122, "decode.throughput.value": 51.314161382560115, "per_token.latency.mean": 0.019561994145510228, "per_token.throughput.value": 51.119532730742336 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.022047264099121094, "prefill.throughput.value": 317.4997119156863, "decode.latency.mean": 2.582226806640625, "decode.throughput.value": 49.182356744728395, "per_token.latency.mean": 0.02041716771634671, "per_token.throughput.value": 48.97838984784185 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "eed9ed679878ada2f6d2eefccdbda368cabc88b1": { "metrics": { "prefill.latency.mean": 0.014052671909332275, "prefill.throughput.value": 498.1259112262738, "decode.latency.mean": 1.5592211303710937, "decode.throughput.value": 81.45092285260019, "per_token.latency.mean": 0.012330162104881798, "per_token.throughput.value": 81.10193454829574 } } } } }