{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "6aafe5f36d73e730c988ee090233eb757b33821c": { "metrics": { "prefill.latency.mean": 0.0200676155090332, "prefill.throughput.value": 348.8207154880475, "decode.latency.mean": 2.4733016357421875, "decode.throughput.value": 51.34836696207896, "per_token.latency.mean": 0.019552757866297784, "per_token.throughput.value": 51.14368043822889 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "6aafe5f36d73e730c988ee090233eb757b33821c": { "metrics": { "prefill.latency.mean": 0.025525072097778323, "prefill.throughput.value": 274.2401656373489, "decode.latency.mean": 2.5864923095703123, "decode.throughput.value": 49.101247867656795, "per_token.latency.mean": 0.020449494561659017, "per_token.throughput.value": 48.90096412822403 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "6aafe5f36d73e730c988ee090233eb757b33821c": { "metrics": { "prefill.latency.mean": 0.013976719856262207, "prefill.throughput.value": 500.8328185717825, "decode.latency.mean": 1.5733544921874998, "decode.throughput.value": 80.71925343628482, "per_token.latency.mean": 0.012441685001840704, "per_token.throughput.value": 80.37496527617067 } } } } }