{ "summary": [ { "input_tokens": 512, "kv_mib": 48.0, "n": 5, "pure_transfer_ms_mean": 17.29, "pure_transfer_ms_p50": 5.16, "pure_transfer_ms_min": 5.13, "pure_transfer_ms_max": 65.8, "throughput_gbps_mean": 7.95, "throughput_gbps_p50": 9.76, "throughput_gbps_max": 9.81 }, { "input_tokens": 1024, "kv_mib": 96.0, "n": 5, "pure_transfer_ms_mean": 10.18, "pure_transfer_ms_p50": 10.15, "pure_transfer_ms_min": 10.06, "pure_transfer_ms_max": 10.37, "throughput_gbps_mean": 9.89, "throughput_gbps_p50": 9.91, "throughput_gbps_max": 10.0 }, { "input_tokens": 2048, "kv_mib": 192.0, "n": 5, "pure_transfer_ms_mean": 20.14, "pure_transfer_ms_p50": 20.01, "pure_transfer_ms_min": 19.99, "pure_transfer_ms_max": 20.51, "throughput_gbps_mean": 10.0, "throughput_gbps_p50": 10.06, "throughput_gbps_max": 10.07 }, { "input_tokens": 4096, "kv_mib": 384.0, "n": 5, "pure_transfer_ms_mean": 40.19, "pure_transfer_ms_p50": 40.12, "pure_transfer_ms_min": 40.07, "pure_transfer_ms_max": 40.5, "throughput_gbps_mean": 10.02, "throughput_gbps_p50": 10.04, "throughput_gbps_max": 10.05 }, { "input_tokens": 8192, "kv_mib": 768.0, "n": 5, "pure_transfer_ms_mean": 81.32, "pure_transfer_ms_p50": 80.86, "pure_transfer_ms_min": 80.71, "pure_transfer_ms_max": 82.5, "throughput_gbps_mean": 9.9, "throughput_gbps_p50": 9.96, "throughput_gbps_max": 9.98 }, { "input_tokens": 16384, "kv_mib": 1536.0, "n": 5, "pure_transfer_ms_mean": 162.76, "pure_transfer_ms_p50": 161.78, "pure_transfer_ms_min": 161.66, "pure_transfer_ms_max": 164.76, "throughput_gbps_mean": 9.9, "throughput_gbps_p50": 9.96, "throughput_gbps_max": 9.96 }, { "input_tokens": 32768, "kv_mib": 3072.0, "n": 5, "pure_transfer_ms_mean": 395.6, "pure_transfer_ms_p50": 309.61, "pure_transfer_ms_min": 307.71, "pure_transfer_ms_max": 526.86, "throughput_gbps_mean": 8.71, "throughput_gbps_p50": 10.4, "throughput_gbps_max": 10.47 }, { "input_tokens": 65536, "kv_mib": 6144.0, "n": 5, "pure_transfer_ms_mean": 1427.92, "pure_transfer_ms_p50": 1733.57, "pure_transfer_ms_min": 653.52, "pure_transfer_ms_max": 1921.19, "throughput_gbps_mean": 5.45, "throughput_gbps_p50": 3.72, "throughput_gbps_max": 9.86 }, { "input_tokens": 131072, "kv_mib": 12288.0, "n": 5, "pure_transfer_ms_mean": 3953.36, "pure_transfer_ms_p50": 2818.43, "pure_transfer_ms_min": 1283.03, "pure_transfer_ms_max": 9158.57, "throughput_gbps_mean": 5.49, "throughput_gbps_p50": 4.57, "throughput_gbps_max": 10.04 } ] }