Nano*
en

Результаты › Отправка

Отправка vlkdhtivryf5y

Машина
unknown
AMD EPYC 7351P · ubuntu 26 · сервер
Модель
gemma-4-e4b-qat df0fd4ee07072c607c29a0a1cb4f98918426cca12f45a2776bdd6ee6d09a4de3
Клиент
NanoLLM b11600-n4-12-gf48f437 · llama.cpp b11600 · ggml haswell · standard/v1
Дата прогона
2026-10-20

Пресеты

ПресетСтатусРежим Загрузка, мсPrefill, ток/сГенерация, ток/с На глубине, ток/сTTFT, мсЧат, ток/с Ускорение MTPПроверкаВывод
ram-94g ok ram-full llamacpp:cpu 5095613322.7 —43622.8 ×1.37 81% в пределах потолка ≤ 84.9 не с чем сравнить
ram-8g пропущен
same plan as ram-94g
ram-full llamacpp:cpu ——— ——— —
ram-4g ok ram-ple-disk llamacpp:cpu 2176511921.8 —47322.9 ×1.30 81% в пределах потолка ≤ 84.9 не с чем сравнить
Отправленный JSON (без ключа установки)
{
  "client": {
    "arch": "amd64",
    "ggml_cpu_variant": "haswell",
    "llama_cpp": {
      "commit": "aad9f95e3ad7d397a4e54770d0e712ae281cd9af",
      "tag": "b11600"
    },
    "nanollm": "b11600-n4-12-gf48f437",
    "os": "linux"
  },
  "consent_version": "2026-10",
  "date": "2026-10-20",
  "duration_s": 203.387028475,
  "machine": {
    "cpu": {
      "cores": 16,
      "isa": [
        "avx",
        "avx2",
        "fma",
        "f16c"
      ],
      "model": "AMD EPYC 7351P",
      "threads": 32,
      "vendor": "AMD"
    },
    "disk": {
      "type": "hdd"
    },
    "extra": [
      {
        "key": "numa_nodes",
        "value": "4"
      },
      {
        "key": "governor",
        "value": "performance"
      },
      {
        "key": "boost",
        "value": "1"
      },
      {
        "key": "thp",
        "value": "madvise"
      },
      {
        "key": "numa_balancing",
        "value": "1"
      },
      {
        "key": "loadavg_1m",
        "value": "1.47"
      },
      {
        "key": "model_disk",
        "value": "md3"
      },
      {
        "key": "model_disk_rotational",
        "value": "1"
      },
      {
        "key": "model_read_ahead_kb",
        "value": "8192"
      },
      {
        "key": "model_max_sectors_kb",
        "value": "4096"
      }
    ],
    "memory": {
      "total_bytes": 135036997632
    },
    "os": {
      "kernel_mm": "7.0",
      "name": "ubuntu",
      "version": "26.04"
    },
    "power": {},
    "system": {
      "chassis": "desktop",
      "form_factor": "server",
      "product": "To Be Filled By O.E.M.",
      "vendor": "To Be Filled By O.E.M."
    }
  },
  "model": {
    "drafter_sha256": "423074e537504b4f9ec5eafed5c639fac82c96631626efccacdd3c4039b20605",
    "id": "gemma-4-e4b-qat",
    "revision": "8c5a9e4fd5482e2be20fe0bf013b4c262a8f4265",
    "sha256": "df0fd4ee07072c607c29a0a1cb4f98918426cca12f45a2776bdd6ee6d09a4de3"
  },
  "runs": [
    {
      "metrics": {
        "buffers": {
          "CPU": 2002947931,
          "CPU_REPACK": 2656053492
        },
        "chat_tok_s": 22.825709699154178,
        "load_ms": 50955.512,
        "mtp": {
          "acceptance": 0.8088235294117647,
          "speedup": 1.367922635624862,
          "tok_s": 31.223804971674955
        },
        "peak_cgroup": 6998732800,
        "peak_page_cache": 50339840,
        "peak_rss": 6959800320,
        "pp_tok_s": 133.46191933961927,
        "stddev": {
          "chat_tok_s": 0.2538995626362098,
          "mtp_tok_s": 1.9801176440363737,
          "pp_tok_s": 13.742013196850209,
          "tg_tok_s.d0": 0.6349941120214484,
          "ttft_ms": 35.93113617741582
        },
        "tg_tok_s": {
          "d0": 22.718494358783815
        },
        "ttft_ms": 436.087
      },
      "output": {
        "sha256": "fab7a62f8f866c6db18dbbb1df23b9990c154be774b76c3216b745dd57c83342",
        "tokens": 128
      },
      "plan": {
        "cold": true,
        "device": "llamacpp:cpu",
        "draft_max": 2,
        "draft_p_min": 0.75,
        "enforcement": "strict",
        "fa": true,
        "gpu_layers": 0,
        "kv": "f16",
        "kv_v": "f16",
        "lazy": "off",
        "load_mode": "dio",
        "mode": "ram-full",
        "mtp": true,
        "n_batch": 2048,
        "n_ctx": 32768,
        "n_ubatch": 512,
        "repack": "on",
        "slots": 1,
        "swa_full": true,
        "threads": 16,
        "threads_batch": 16
      },
      "plan_hash": "1f005d502614a4d1",
      "preset": "ram-94g",
      "reps": 3,
      "status": "ok"
    },
    {
      "metrics": {
        "tg_tok_s": {}
      },
      "output": {},
      "plan": {
        "cold": true,
        "device": "llamacpp:cpu",
        "draft_max": 2,
        "draft_p_min": 0.75,
        "fa": true,
        "gpu_layers": 0,
        "kv": "f16",
        "kv_v": "f16",
        "lazy": "off",
        "load_mode": "dio",
        "mode": "ram-full",
        "mtp": true,
        "n_batch": 2048,
        "n_ctx": 32768,
        "n_ubatch": 512,
        "repack": "on",
        "slots": 1,
        "swa_full": true,
        "threads": 16,
        "threads_batch": 16
      },
      "plan_hash": "1f005d502614a4d1",
      "preset": "ram-8g",
      "reason": "same plan as ram-94g",
      "reps": 0,
      "same_as": "ram-94g",
      "status": "skipped"
    },
    {
      "metrics": {
        "buffers": {
          "CPU_Mapped": 2002895427,
          "CPU_REPACK": 2656053492
        },
        "chat_tok_s": 22.872719783422745,
        "load_ms": 21765.265,
        "mtp": {
          "acceptance": 0.8088235294117647,
          "speedup": 1.3031432753635626,
          "tok_s": 29.80643097504247
        },
        "peak_cgroup": 4496293888,
        "peak_locked": 2932736,
        "peak_page_cache": 2198867968,
        "peak_rss": 4507090944,
        "pp_tok_s": 119.00129473338941,
        "stddev": {
          "chat_tok_s": 1.858627438699559,
          "mtp_tok_s": 0.5903183614626974,
          "pp_tok_s": 10.669847774186989,
          "tg_tok_s.d0": 2.373522911877956,
          "ttft_ms": 238.19543614505577
        },
        "tg_tok_s": {
          "d0": 21.786719596761976
        },
        "ttft_ms": 473.415
      },
      "output": {
        "sha256": "fab7a62f8f866c6db18dbbb1df23b9990c154be774b76c3216b745dd57c83342",
        "tokens": 128
      },
      "plan": {
        "cold": true,
        "device": "llamacpp:cpu",
        "draft_max": 2,
        "draft_p_min": 0.75,
        "enforcement": "strict",
        "fa": true,
        "gpu_layers": 0,
        "kv": "f16",
        "kv_v": "f16",
        "lazy": "on",
        "load_mode": "mmap",
        "mode": "ram-ple-disk",
        "mtp": true,
        "n_batch": 2048,
        "n_ctx": 8192,
        "n_ubatch": 512,
        "repack": "on",
        "slots": 1,
        "swa_full": true,
        "threads": 16,
        "threads_batch": 16
      },
      "plan_hash": "c3216e2eb05c6ccb",
      "preset": "ram-4g",
      "reps": 3,
      "status": "ok"
    }
  ],
  "schema": "nanollm.bench/v1",
  "suite": {
    "id": "standard",
    "prompts_sha256": "ad01c12080c9d8d68bf3bd9414940694635d0fdf85c2539ffee4b127c7429754",
    "quick": true,
    "version": "v1"
  }
}

Удалить свою отправку: nanollm bench withdraw vlkdhtivryf5y