[Submission] nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8

[Gong-K]

{
  "schema_version": "1.0",
  "suite_id": "suite_F",
  "implementation_id": "nvidia_sglang_c43a8309",
  "chip": {
    "name": "NVIDIA A100-SXM4-40GB",
    "vendor": "NVIDIA",
    "count": 1,
    "memory_gb": 40.0,
    "interconnect_intra_node": null,
    "interconnect_inter_node": null
  },
  "environment": {
    "collected_at": "2026-05-07T10:52:35.700123+00:00",
    "accelerators": [
      {
        "index": 0,
        "name": "NVIDIA A100-SXM4-40GB",
        "vendor": "NVIDIA",
        "memory_gb": 40.0,
        "driver_version": "565.57.01",
        "firmware_version": null,
        "compute_capability": "8.0",
        "supports_bf16": true
      }
    ],
    "accelerator_topology": "\tGPU0\tNIC0\tNIC1\tNIC2\tCPU Affinity\tNUMA Affinity\tGPU NUMA ID\nGPU0\t X \tSYS\tSYS\tSYS\t32-63,96-127\t1\t\tN/A\nNIC0\tSYS\t X \tPIX\tNODE\t\t\t\t\nNIC1\tSYS\tPIX\t X \tNODE\t\t\t\t\nNIC2\tSYS\tNODE\tNODE\t X \t\t\t\t\n\nLegend:\n\n  X    = Self\n  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)\n  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node\n  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)\n  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)\n  PIX  = Connection traversing at most a single PCIe bridge\n  NV#  = Connection traversing a bonded set of # NVLinks\n\nNIC Legend:\n\n  NIC0: mlx5_0\n  NIC1: mlx5_1\n  NIC2: mlx5_2\n\n",
    "intra_node_interconnect": null,
    "cpu": {
      "model": "AMD EPYC 7532 32-Core Processor",
      "physical_cores": 64,
      "logical_cores": 128,
      "numa_nodes": 2
    },
    "system_memory_gb": 1007.7,
    "pcie_generation": "PCIe Gen 4",
    "cpu_accelerator_bandwidth_gbs": null,
    "network_interfaces": [
      {
        "name": "mlx5_0",
        "type": "InfiniBand/RoCE",
        "bandwidth_gbps": null
      },
      {
        "name": "mlx5_1",
        "type": "InfiniBand/RoCE",
        "bandwidth_gbps": null
      },
      {
        "name": "mlx5_2",
        "type": "InfiniBand/RoCE",
        "bandwidth_gbps": null
      }
    ],
    "os": "Ubuntu 22.04.4 LTS",
    "python_version": "3.10.20",
    "kernel_version": "5.15.0-60-generic",
    "runtime_version": "CUDA 12.8",
    "pytorch_version": "2.9.1+cu128"
  },
  "software": {
    "framework": "SGLang",
    "framework_version": "0.5.6",
    "driver_version": "565.57.01",
    "runtime_version": "CUDA 12.8",
    "os": "Ubuntu 22.04.4 LTS",
    "python_version": "3.10.20"
  },
  "model": {
    "model_id": "Qwen/Qwen2.5-0.5B-Instruct",
    "model_revision": "7ae557604adf67be50417f59c2c2f167def9a775",
    "model_name": null,
    "model_note": null,
    "model_source": "local",
    "architecture": "dense",
    "parameter_count_b": 0.5,
    "precision": "BF16",
    "effective_dtype": "bfloat16",
    "quantization_method": null,
    "model_format": "HuggingFace original"
  },
  "task": {
    "scenarios_run": [
      "offline",
      "online",
      "interactive",
      "sustained"
    ],
    "parallelism": {
      "tensor_parallel_size": 1,
      "pipeline_parallel_size": 1,
      "expert_parallel_size": 1,
      "data_parallel_size": 1
    },
    "num_runs": 3,
    "extra_config": null
  },
  "metrics": {
    "derived": {},
    "offline": {
      "results_by_concurrency": [
        {
          "client_concurrency": 4,
          "throughput_tokens_per_sec": 11447.71,
          "throughput_tokens_per_sec_per_chip": 11447.71,
          "elapsed_seconds_median": 3.7,
          "peak_memory_gb": null,
          "power_watts_avg": null,
          "power_watts_peak": null,
          "oom": false,
          "_throughput_note": "output_only",
          "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs."
        },
        {
          "client_concurrency": 16,
          "throughput_tokens_per_sec": 11507.48,
          "throughput_tokens_per_sec_per_chip": 11507.48,
          "elapsed_seconds_median": 3.6,
          "peak_memory_gb": null,
          "power_watts_avg": null,
          "power_watts_peak": null,
          "oom": false,
          "_throughput_note": "output_only",
          "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs."
        },
        {
          "client_concurrency": 64,
          "throughput_tokens_per_sec": 11509.2,
          "throughput_tokens_per_sec_per_chip": 11509.2,
          "elapsed_seconds_median": 3.6,
          "peak_memory_gb": null,
          "power_watts_avg": null,
          "power_watts_peak": null,
          "oom": false,
          "_throughput_note": "output_only",
          "_concurrency_note": "client_concurrency is the number of requests sent simultaneously. The inference engine batches internally; this does not directly set engine parameters like max_num_seqs."
        }
      ]
    },
    "online": {
      "sla_ttft_ms": 500,
      "max_valid_qps": 40,
      "results_by_qps": [
        {
          "target_qps": 10,
          "achieved_qps": 10.0,
          "ttft_ms_p50": 18.63,
          "ttft_ms_p90": 31.07,
          "ttft_ms_p99": 1226.53,
          "tpot_ms_p50": 2.56,
          "tpot_ms_p90": 3.01,
          "tpot_ms_p99": 4.21,
          "elapsed_seconds_median": 31.9,
          "sla_met": false
        },
        {
          "target_qps": 40,
          "achieved_qps": 40.0,
          "ttft_ms_p50": 34.28,
          "ttft_ms_p90": 41.32,
          "ttft_ms_p99": 44.41,
          "tpot_ms_p50": 20.08,
          "tpot_ms_p90": 24.83,
          "tpot_ms_p99": 31.82,
          "elapsed_seconds_median": 10.0,
          "sla_met": true
        }
      ]
    },
    "interactive": {
      "ttft_ms_p50": 16.46,
      "ttft_ms_p90": 17.26,
      "ttft_ms_p99": 18.42,
      "tpot_ms_p50": 1.89,
      "tpot_ms_p90": 1.91,
      "tpot_ms_p99": 1.99,
      "peak_memory_gb": null,
      "elapsed_seconds_median": 56.5
    },
    "sustained": {
      "sustained_concurrency": 32,
      "duration_minutes": 15,
      "warmup_minutes": 1,
      "sample_interval_seconds": 60,
      "samples": [
        {
          "minute": 1.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 6616.4,
          "tokens_out": 397010,
          "tokens_in": 0,
          "requests_completed": 2131,
          "ttft_ms_p50": 19.6,
          "ttft_ms_p99": 4749.5
        },
        {
          "minute": 2.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7181.7,
          "tokens_out": 430976,
          "tokens_in": 0,
          "requests_completed": 2317,
          "ttft_ms_p50": 19.3,
          "ttft_ms_p99": 41.7
        },
        {
          "minute": 3.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7188.9,
          "tokens_out": 431427,
          "tokens_in": 0,
          "requests_completed": 2312,
          "ttft_ms_p50": 19.1,
          "ttft_ms_p99": 41.0
        },
        {
          "minute": 4.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7110.2,
          "tokens_out": 426673,
          "tokens_in": 0,
          "requests_completed": 2292,
          "ttft_ms_p50": 19.4,
          "ttft_ms_p99": 34.0
        },
        {
          "minute": 5.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7106.5,
          "tokens_out": 426157,
          "tokens_in": 0,
          "requests_completed": 2287,
          "ttft_ms_p50": 19.2,
          "ttft_ms_p99": 40.6
        },
        {
          "minute": 6.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7144.4,
          "tokens_out": 428781,
          "tokens_in": 0,
          "requests_completed": 2311,
          "ttft_ms_p50": 19.2,
          "ttft_ms_p99": 33.6
        },
        {
          "minute": 7.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7158.3,
          "tokens_out": 429690,
          "tokens_in": 0,
          "requests_completed": 2306,
          "ttft_ms_p50": 19.3,
          "ttft_ms_p99": 40.3
        },
        {
          "minute": 8.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7020.0,
          "tokens_out": 421197,
          "tokens_in": 0,
          "requests_completed": 2265,
          "ttft_ms_p50": 19.3,
          "ttft_ms_p99": 41.4
        },
        {
          "minute": 9.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7160.7,
          "tokens_out": 429349,
          "tokens_in": 0,
          "requests_completed": 2303,
          "ttft_ms_p50": 19.4,
          "ttft_ms_p99": 41.1
        },
        {
          "minute": 10.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7183.9,
          "tokens_out": 431115,
          "tokens_in": 0,
          "requests_completed": 2319,
          "ttft_ms_p50": 19.4,
          "ttft_ms_p99": 38.0
        },
        {
          "minute": 11.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7180.8,
          "tokens_out": 431065,
          "tokens_in": 0,
          "requests_completed": 2308,
          "ttft_ms_p50": 19.2,
          "ttft_ms_p99": 34.1
        },
        {
          "minute": 12.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7127.9,
          "tokens_out": 427694,
          "tokens_in": 0,
          "requests_completed": 2301,
          "ttft_ms_p50": 19.3,
          "ttft_ms_p99": 34.3
        },
        {
          "minute": 13.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7111.1,
          "tokens_out": 426396,
          "tokens_in": 0,
          "requests_completed": 2282,
          "ttft_ms_p50": 19.2,
          "ttft_ms_p99": 41.0
        },
        {
          "minute": 14.0,
          "is_warmup": false,
          "throughput_tokens_per_sec": 7044.6,
          "tokens_out": 422801,
          "tokens_in": 0,
          "requests_completed": 2269,
          "ttft_ms_p50": 19.4,
          "ttft_ms_p99": 41.8
        }
      ],
      "sustained_throughput_tokens_per_sec": 7095.4,
      "throttle_ratio": 0.92,
      "throttle_onset_minute": null,
      "ttft_p99_drift_ms": -4707.7
    }
  },
  "accuracy": {
    "subset_score": 0.41,
    "baseline_delta": 0.03,
    "valid": true,
    "framework": "SGLang",
    "precision": "BF16",
    "notes": "Integrated accuracy check — used same SGLang instance as benchmark."
  },
  "meta": {
    "submitted_by": "Gong-K",
    "submission_type": "individual",
    "date": "2026-05-07",
    "time": "10:56:30",
    "run_id": "435424a8",
    "run_name": "nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8",
    "flagged": null,
    "reproduce_script": "runners/nvidia_sglang_c43a8309/runner.py",
    "env_info_file": "../env_info.json",
    "log_file": "run.log",
    "samples_file": "samples.jsonl",
    "notes": null,
    "benchmark_start_time": "2026-05-07T10:55:44.424768+00:00",
    "benchmark_end_time": "2026-05-07T10:56:30.330070+00:00",
    "benchmark_elapsed_minutes": 20.7,
    "model_load_seconds": 33.4,
    "benchmark_elapsed_minutes_note": "Total across ['offline', 'online', 'interactive', 'sustained'] scenarios.",
    "scenario_dirs": {
      "offline": "results/community/nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8/offline",
      "online": "results/community/nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8/online",
      "interactive": "results/community/nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8/interactive",
      "sustained": "results/community/nvidia_a100_sxm4_40gbx1_suite_F_nvidia_sglang_c43a8309_435424a8/sustained"
    }
  }
}

[github-actions]

Validation passed. PR created: #44 - your result will appear on the leaderboard after the PR is reviewed and merged.