recipes/docker-images/mia-tp4/docs/results/tp4-switchless-ring-20260915.json

{
  "decode_window_idle": {
    "captured_utc": "20260915T194011Z",
    "base_url_redacted": "http://<head>:8888",
    "waves": [
      {
        "concurrency": 1,
        "wall_seconds": 49.321,
        "successes": 1,
        "completion_tokens": 3651,
        "aggregate_tok_s": 74.03,
        "median_window_tps": 67.41,
        "mean_window_tps": 67.41,
        "median_ttft_s": 0.535,
        "finish_reasons": [
          "stop"
        ],
        "idle_before": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "after": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "per_request": [
          {
            "index": 0,
            "seconds": 49.318,
            "ttft_s": 0.535,
            "window_129_641_tps": 67.41,
            "completion_tokens": 3651,
            "finish_reason": "stop"
          }
        ]
      },
      {
        "concurrency": 2,
        "wall_seconds": 61.714,
        "successes": 2,
        "completion_tokens": 6880,
        "aggregate_tok_s": 111.48,
        "median_window_tps": 54.98,
        "mean_window_tps": 54.98,
        "median_ttft_s": 0.738,
        "finish_reasons": [
          "stop",
          "stop"
        ],
        "idle_before": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "after": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "per_request": [
          {
            "index": 0,
            "seconds": 55.228,
            "ttft_s": 0.738,
            "window_129_641_tps": 55.21,
            "completion_tokens": 3190,
            "finish_reason": "stop"
          },
          {
            "index": 1,
            "seconds": 61.713,
            "ttft_s": 0.737,
            "window_129_641_tps": 54.74,
            "completion_tokens": 3690,
            "finish_reason": "stop"
          }
        ]
      },
      {
        "concurrency": 4,
        "wall_seconds": 81.536,
        "successes": 4,
        "completion_tokens": 13050,
        "aggregate_tok_s": 160.05,
        "median_window_tps": 40.84,
        "mean_window_tps": 40.86,
        "median_ttft_s": 0.978,
        "finish_reasons": [
          "stop",
          "stop",
          "stop",
          "stop"
        ],
        "idle_before": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "after": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "per_request": [
          {
            "index": 0,
            "seconds": 80.28,
            "ttft_s": 0.809,
            "window_129_641_tps": 41.35,
            "completion_tokens": 3216,
            "finish_reason": "stop"
          },
          {
            "index": 1,
            "seconds": 76.574,
            "ttft_s": 1.291,
            "window_129_641_tps": 40.42,
            "completion_tokens": 3146,
            "finish_reason": "stop"
          },
          {
            "index": 2,
            "seconds": 81.534,
            "ttft_s": 0.918,
            "window_129_641_tps": 40.91,
            "completion_tokens": 3397,
            "finish_reason": "stop"
          },
          {
            "index": 3,
            "seconds": 80.171,
            "ttft_s": 1.039,
            "window_129_641_tps": 40.77,
            "completion_tokens": 3291,
            "finish_reason": "stop"
          }
        ]
      },
      {
        "concurrency": 8,
        "wall_seconds": 119.922,
        "successes": 8,
        "completion_tokens": 26963,
        "aggregate_tok_s": 224.84,
        "median_window_tps": 29.91,
        "mean_window_tps": 29.76,
        "median_ttft_s": 1.181,
        "finish_reasons": [
          "stop",
          "stop",
          "stop",
          "stop",
          "stop",
          "stop",
          "stop",
          "stop"
        ],
        "idle_before": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "after": {
          "num_running_reqs": 0,
          "num_waiting_reqs": 0,
          "num_used_tokens": 0
        },
        "per_request": [
          {
            "index": 0,
            "seconds": 113.437,
            "ttft_s": 1.181,
            "window_129_641_tps": 32.64,
            "completion_tokens": 3353,
            "finish_reason": "stop"
          },
          {
            "index": 1,
            "seconds": 110.102,
            "ttft_s": 1.181,
            "window_129_641_tps": 29.62,
            "completion_tokens": 3258,
            "finish_reason": "stop"
          },
          {
            "index": 2,
            "seconds": 103.041,
            "ttft_s": 1.222,
            "window_129_641_tps": 30.19,
            "completion_tokens": 3013,
            "finish_reason": "stop"
          },
          {
            "index": 3,
            "seconds": 107.252,
            "ttft_s": 1.181,
            "window_129_641_tps": 30.61,
            "completion_tokens": 3115,
            "finish_reason": "stop"
          },
          {
            "index": 4,
            "seconds": 116.409,
            "ttft_s": 1.181,
            "window_129_641_tps": 29.03,
            "completion_tokens": 3470,
            "finish_reason": "stop"
          },
          {
            "index": 5,
            "seconds": 118.41,
            "ttft_s": 1.175,
            "window_129_641_tps": 27.55,
            "completion_tokens": 3585,
            "finish_reason": "stop"
          },
          {
            "index": 6,
            "seconds": 119.919,
            "ttft_s": 1.175,
            "window_129_641_tps": 27.71,
            "completion_tokens": 3610,
            "finish_reason": "stop"
          },
          {
            "index": 7,
            "seconds": 117.113,
            "ttft_s": 1.18,
            "window_129_641_tps": 30.73,
            "completion_tokens": 3559,
            "finish_reason": "stop"
          }
        ]
      }
    ]
  },
  "openai_correctness": {
    "models": {
      "object": "list",
      "data": [
        {
          "id": "deepseek-v4.1-flash",
          "object": "model",
          "created": 1789500134,
          "owned_by": "sglang",
          "root": "deepseek-v4.1-flash",
          "parent": null,
          "max_model_len": 1048576
        }
      ]
    },
    "health_http": 200,
    "arithmetic": {
      "http": 200,
      "seconds": 1.694,
      "finish_reason": "stop",
      "content": "42",
      "reasoning": "",
      "reasoning_tokens": 0,
      "prompt_tokens": 18,
      "completion_tokens": 2,
      "tool_calls": []
    },
    "thinking": {
      "http": 200,
      "seconds": 2.685,
      "finish_reason": "stop",
      "content": "OK99",
      "reasoning": "We need answer exactly OK99 and nothing else. Need comply. Final only.",
      "reasoning_tokens": 17,
      "prompt_tokens": 39,
      "completion_tokens": 20,
      "tool_calls": []
    },
    "tools": {
      "http": 200,
      "seconds": 2.739,
      "finish_reason": "tool_calls",
      "content": "I'll check the weather in Vigo for you.",
      "reasoning": "",
      "reasoning_tokens": 0,
      "prompt_tokens": 270,
      "completion_tokens": 50,
      "tool_calls": [
        {
          "id": "call_9cf65bc578424fff896e7f7c",
          "index": 0,
          "type": "function",
          "function": {
            "name": "get_weather",
            "arguments": "{\"city\": \"Vigo\"}"
          }
        }
      ]
    }
  },
  "openai_short_prose_contaminated": [
    {
      "concurrency": 1,
      "max_tokens": 256,
      "thinking": false,
      "wall_seconds": 16.238,
      "successes": 1,
      "completion_tokens": 256,
      "aggregate_tok_s": 15.77,
      "mean_per_request_tok_s": 15.77,
      "min_per_request_tok_s": 15.77,
      "finish_reasons": [
        "length"
      ],
      "seconds": [
        16.238
      ]
    },
    {
      "concurrency": 2,
      "max_tokens": 256,
      "thinking": false,
      "wall_seconds": 15.378,
      "successes": 2,
      "completion_tokens": 512,
      "aggregate_tok_s": 33.29,
      "mean_per_request_tok_s": 17.47,
      "min_per_request_tok_s": 16.65,
      "finish_reasons": [
        "length",
        "length"
      ],
      "seconds": [
        15.378,
        13.993
      ]
    },
    {
      "concurrency": 4,
      "max_tokens": 256,
      "thinking": false,
      "wall_seconds": 39.182,
      "successes": 4,
      "completion_tokens": 1024,
      "aggregate_tok_s": 26.13,
      "mean_per_request_tok_s": 6.87,
      "min_per_request_tok_s": 6.53,
      "finish_reasons": [
        "length",
        "length",
        "length",
        "length"
      ],
      "seconds": [
        39.181,
        36.689,
        36.691,
        36.694
      ]
    },
    {
      "concurrency": 8,
      "max_tokens": 256,
      "thinking": false,
      "wall_seconds": 33.406,
      "successes": 8,
      "completion_tokens": 2048,
      "aggregate_tok_s": 61.31,
      "mean_per_request_tok_s": 10.2,
      "min_per_request_tok_s": 7.66,
      "finish_reasons": [
        "length",
        "length",
        "length",
        "length",
        "length",
        "length",
        "length",
        "length"
      ],
      "seconds": [
        25.608,
        21.882,
        25.236,
        25.921,
        25.609,
        21.876,
        33.404,
        24.345
      ]
    }
  ],
  "needles": [
    {
      "target_tokens": 32768,
      "concurrency": 1,
      "calibrated_prompt_tokens": [
        26152
      ],
      "wall_seconds": 12.797,
      "correct": 1
    },
    {
      "target_tokens": 32768,
      "concurrency": 4,
      "calibrated_prompt_tokens": [
        26152,
        26150,
        26150,
        27442
      ],
      "wall_seconds": 36.778,
      "correct": 4
    },
    {
      "target_tokens": 131072,
      "concurrency": 1,
      "calibrated_prompt_tokens": [
        107191
      ],
      "wall_seconds": 414.697,
      "correct": 1
    }
  ],
  "long_code_c8": {
    "completion_tokens": 48000,
    "wall_seconds": 222.413,
    "aggregate_tok_s": 215.81,
    "successes": 8,
    "peak_sglang_decode_tok_s": 258.63
  }
}