{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "2505357e4fdea82d1586b8c0862b4eeb46fac457": { "metrics": { "prefill.latency.mean": 0.02032707118988037, "prefill.throughput.value": 344.36835167305753, "decode.latency.mean": 2.403255249023437, "decode.throughput.value": 52.84499016557083, "per_token.latency.mean": 0.0189985674831707, "per_token.throughput.value": 52.63554743723807 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "2505357e4fdea82d1586b8c0862b4eeb46fac457": { "metrics": { "prefill.latency.mean": 0.02256251239776611, "prefill.throughput.value": 310.24913700182884, "decode.latency.mean": 2.5812564697265623, "decode.throughput.value": 49.20084520444935, "per_token.latency.mean": 0.020413658503958362, "per_token.throughput.value": 48.98680948376267 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "2505357e4fdea82d1586b8c0862b4eeb46fac457": { "metrics": { "prefill.latency.mean": 0.01407475233078003, "prefill.throughput.value": 497.3444530666251, "decode.latency.mean": 1.559105651855469, "decode.throughput.value": 81.4569556905006, "per_token.latency.mean": 0.012329652096443025, "per_token.throughput.value": 81.10528927969423 } } } } }