{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "348e2294ac1a3cc744a93f04ab1f886439ee90d3": { "metrics": { "prefill.latency.mean": 0.020395471572875974, "prefill.throughput.value": 343.213442012752, "decode.latency.mean": 2.4265584716796873, "decode.throughput.value": 52.33749834682095, "per_token.latency.mean": 0.019179596934865113, "per_token.throughput.value": 52.13873906714781 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "348e2294ac1a3cc744a93f04ab1f886439ee90d3": { "metrics": { "prefill.latency.mean": 0.02268723201751709, "prefill.throughput.value": 308.54358938962736, "decode.latency.mean": 2.5942015380859376, "decode.throughput.value": 48.955332935969025, "per_token.latency.mean": 0.020515317818864065, "per_token.throughput.value": 48.74406571856707 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "348e2294ac1a3cc744a93f04ab1f886439ee90d3": { "metrics": { "prefill.latency.mean": 0.014029391765594482, "prefill.throughput.value": 498.95249323400594, "decode.latency.mean": 1.5590690307617188, "decode.throughput.value": 81.4588690392697, "per_token.latency.mean": 0.012328717148822287, "per_token.throughput.value": 81.11143989506856 } } } } }