{
 "schema": "local-ai-bench/result/v1",
 "run_id": "20261008T105246Z-glm53-offload-v4.1-nvme-90g-rtx-3090-24gb-gpu2-kit_sweep",
 "ticket": "T42",
 "kind": "kit_sweep",
 "model": "glm-5.3-flash",
 "config_id": "glm53-offload-v4.1-nvme-90g-rtx-3090-24gb-gpu2",
 "outcome": "pass",
 "failed_stage": null,
 "exit_status": 0,
 "started_at": "2026-10-08T10:52:46.487157+00:00",
 "finished_at": "2026-10-08T12:23:51.679983+00:00",
 "launch": {
  "file": "launches/glm53-offload-v4.1-nvme-90g-rtx-3090-24gb.json",
  "sha256": "6952b4d74789c9d18463812a52173d024806e5cfe0ec985fdb8b37db8d68c29c",
  "image": "ghcr.io/sybil-solutions/glm53-flash-offload@sha256:99b8926e01480b1acc35251a60f61cba6a2ad6309491a6ab2bb6b5688f843c16",
  "argv": [],
  "env": {
   "PORT": "30000",
   "GLM53_MODE": "nvme",
   "GLM53_EC_MAX_SLOTS": "1376",
   "GLM53_MAIN_CPUS": "1",
   "GLM53_NV_CTL_CPU": "16",
   "GLM53_NV_CPU_CPUS": "3-15",
   "GLM53_NV_READER_CPUS": "19-31",
   "HF_HUB_OFFLINE": "1"
  },
  "engine": "exllamav3",
  "engine_source_commit": "https://github.com/0xSero/glm53-flash-offload/tree/7f1ee89",
  "mtp": null,
  "entrypoint": null,
  "port": 30000,
  "shm": "1g",
  "flags": [
   "--memory",
   "90g",
   "--memory-swap",
   "90g",
   "--ulimit",
   "memlock=-1",
   "--cpuset-cpus",
   "1,3-16,19-31"
  ],
  "weights_at": [
   "/models"
  ]
 },
 "weights": [
  {
   "repo": "turboderp/GLM-5.3-Flash-exl3",
   "revision": "332ab457b709b7ba30dd9a448be5de03b80a7ac9",
   "manifest": "models/turboderp-GLM-5.3-Flash-exl3-3.05bpw.json"
  }
 ],
 "host": {
  "hostname": "omarchy-gpu",
  "cpu": "AMD Ryzen 9 5950X 16-Core Processor",
  "ram_total_gib": 125.7,
  "ram_layout": "DIMM_A1 32 GiB DDR4 3200 MT/s, DIMM_A2 32 GiB DDR4 3200 MT/s, DIMM_B1 32 GiB DDR4 3200 MT/s, DIMM_B2 32 GiB DDR4 3200 MT/s (AM4: 2 channels)",
  "kernel": "7.2.5-3-omarchy",
  "driver": "610.57.04",
  "cuda": "13.3",
  "cpus_online": "0-1,3-17,19-31",
  "cpus_offline": "2,18",
  "cpu_boost": false,
  "cpu_max_mhz": 3401,
  "machine_checks_this_boot": 6
 },
 "gpus": [
  {
   "index": 0,
   "uuid": "GPU-816b43e4-8d65-dfd7-2234-8517b3dfbf2d",
   "name": "NVIDIA GeForce RTX 3090",
   "bus_id": "00000000:04:00.0",
   "pcie_gen_max": 4,
   "pcie_width": 8,
   "display_active": false,
   "used_by_run": false
  },
  {
   "index": 1,
   "uuid": "GPU-c67ac872-3371-a88c-aa92-d969daf6405e",
   "name": "NVIDIA GeForce RTX 3090",
   "bus_id": "00000000:0B:00.0",
   "pcie_gen_max": 4,
   "pcie_width": 8,
   "display_active": true,
   "used_by_run": false
  },
  {
   "index": 2,
   "uuid": "GPU-76d3c6af-7f18-69d2-3145-899103de1722",
   "name": "NVIDIA GeForce RTX 3090",
   "bus_id": "00000000:0C:00.0",
   "pcie_gen_max": 4,
   "pcie_width": 8,
   "display_active": false,
   "used_by_run": true
  }
 ],
 "layout": {
  "name": "current",
  "cards": 1,
  "display": "on"
 },
 "context": {
  "configured": 131072,
  "occupied_max": {
   "state": "unavailable",
   "reason": "no occupancy measured in this run"
  },
  "headroom_min": {
   "state": "unavailable",
   "reason": "no occupancy measured in this run"
  }
 },
 "cache_state": {
  "cold": false,
  "reset": []
 },
 "provenance": {
  "bench_commit": "6d451232712d567aa4886576429949cd7a6729bf",
  "bench_dirty": true,
  "pins_sha256": "858564b6e07125ac48c2ef6c5178d0a53596f21eb5f0ebda43906472b01b4d78",
  "registry_commit": "d21258dd744e7c78be28177c90060c6af8e10b7e",
  "kit_commit": "ef883d269e50ecbea290f919f095e2f3ca633b42",
  "harness_commit": "04d809ceab9df28f9adaed044884180159172930"
 },
 "metrics": {
  "prefill_8k_tps": {
   "value": 379.3,
   "unit": "tok/s",
   "detail": {
    "min": 360.7,
    "max": 381.4,
    "n": 3
   }
  },
  "prefill_16k_tps": {
   "value": 519.8,
   "unit": "tok/s",
   "detail": {
    "min": 513.0,
    "max": 520.5,
    "n": 3
   }
  },
  "prefill_32k_tps": {
   "value": 620.8,
   "unit": "tok/s",
   "detail": {
    "min": 545.1,
    "max": 633.6,
    "n": 3
   }
  },
  "prefill_64k_tps": {
   "value": 675.5,
   "unit": "tok/s",
   "detail": {
    "min": 667.4,
    "max": 677.6,
    "n": 3
   }
  },
  "decode_c1_tps": {
   "value": 10.95,
   "unit": "tok/s",
   "detail": {
    "min": 10.5,
    "max": 11.39,
    "power_w": null
   }
  },
  "decode_c2_tps": {
   "value": 10.99,
   "unit": "tok/s",
   "detail": {
    "min": 10.57,
    "max": 11.41,
    "power_w": null
   }
  },
  "decode_c3_tps": {
   "value": 10.9,
   "unit": "tok/s",
   "detail": {
    "min": 10.53,
    "max": 11.28,
    "power_w": null
   }
  },
  "decode_c4_tps": {
   "value": 10.53,
   "unit": "tok/s",
   "detail": {
    "min": 10.32,
    "max": 10.74,
    "power_w": null
   }
  },
  "decode_c1_32k_tps": {
   "value": 10.48,
   "unit": "tok/s",
   "detail": {
    "min": 9.6,
    "max": 11.36,
    "power_w": null
   }
  },
  "mtp_accepted_tps": {
   "state": "skipped",
   "reason": "speculative decoding off in this launch"
  },
  "mtp_acceptance_rate": {
   "state": "skipped",
   "reason": "speculative decoding off in this launch"
  }
 },
 "eligibility": "none",
 "artifacts": [
  "artifacts/runs/20261008T105246Z-lab-glm53-offload-v4.1-nvme-90g-rtx-3090-24gb-gpu2-kit_sweep"
 ],
 "notes": "sweep status: DONE"
}
