{ "google/gemma-2b": { "backend.cache_implementation=null,backend.torch_compile=False": { "dcdda5324bcc7a750b5e40e11dd795442204ff27": { "metrics": { "prefill.latency.mean": 0.020058400154113767, "prefill.throughput.value": 348.98097287008073, "decode.latency.mean": 2.4373055419921874, "decode.throughput.value": 52.10672105401839, "per_token.latency.mean": 0.019270506199169538, "per_token.throughput.value": 51.89277280339916 } } }, "backend.cache_implementation=static,backend.torch_compile=False": { "dcdda5324bcc7a750b5e40e11dd795442204ff27": { "metrics": { "prefill.latency.mean": 0.021743040084838867, "prefill.throughput.value": 321.9421006762071, "decode.latency.mean": 2.5406910400390625, "decode.throughput.value": 49.98640054953215, "per_token.latency.mean": 0.02008856086580179, "per_token.throughput.value": 49.779573891844706 } } }, "backend.cache_implementation=static,backend.torch_compile=True": { "dcdda5324bcc7a750b5e40e11dd795442204ff27": { "metrics": { "prefill.latency.mean": 0.014116623878479003, "prefill.throughput.value": 495.86927159486066, "decode.latency.mean": 1.5734918212890625, "decode.throughput.value": 80.71220852991591, "per_token.latency.mean": 0.012442798056621324, "per_token.throughput.value": 80.36777543519312 } } } } }