{
  "artifacts": [
    {
      "file": "54b36190cac9da79da0fe4ce3e80b8acecd8b12429c227ac4fa563cbbb7d6d89.tflite",
      "sha256": "1b85a919c19a1832b729f546c3f60ceca179b669258d7ab244fac96890bf630a",
      "size_mb": 44.85
    }
  ],
  "benchmarks": [],
  "browser": {
    "backends": [
      {
        "backend": "wasm_xnnpack",
        "date": "2026-08-20",
        "env": {
          "browser": "chromium",
          "browser_version": "151.0.7922.34",
          "headless": true,
          "jspi": true,
          "litertjs_core_version": "2.5.3",
          "machine_label": "mac-studio-m4-max",
          "os": "macOS",
          "os_version": "27.0.0",
          "webgpu_adapter": {
            "architecture": "metal-3",
            "description": "",
            "device": "",
            "vendor": "apple"
          }
        },
        "full_delegation": null,
        "latency_p50_ms": 2194.563,
        "loads": true,
        "max_rel_diff": null,
        "output_match": null,
        "provenance": "measured",
        "runs": true
      },
      {
        "backend": "webgpu_mldrift",
        "date": "2026-08-20",
        "env": {
          "browser": "chromium",
          "browser_version": "151.0.7922.34",
          "headless": true,
          "jspi": true,
          "litertjs_core_version": "2.5.3",
          "machine_label": "mac-studio-m4-max",
          "os": "macOS",
          "os_version": "27.0.0",
          "webgpu_adapter": {
            "architecture": "metal-3",
            "description": "",
            "device": "",
            "vendor": "apple"
          }
        },
        "full_delegation": true,
        "latency_p50_ms": 24.917,
        "loads": true,
        "max_rel_diff": 0.19101343154907227,
        "output_match": false,
        "provenance": "measured",
        "runs": true
      }
    ],
    "demo_url": null,
    "sweep_source": "data/sweep/2.5.3/2026-08-20/rf-detr-seg-nano__rfdetrseg_grapha_fp16.json"
  },
  "conversion": {
    "command": "python scripts/build_rfdetrseg_split.py",
    "quantization": "fp16",
    "tool": "litert-torch",
    "tool_version": "0.10.0"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu_mldrift",
          "context_length": null,
          "date": "2026-08-26",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-npubench",
            "os_build": "Android 16",
            "runtime": "litert",
            "runtime_version": "2.2.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "npubench sweep row: N=50 median, 1 backend = 1 process, accepted only with thermal NONE->NONE; NPU rows additionally required qnn_partition delegate evidence in logcat (a stock model asked for on the NPU can silently land on XNNPACK and report a CPU number)",
            "mode=jit thermal=NONE->NONE headroom=0.7699999->0.7699999 attempt=0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": 35.283,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "iterations": 50,
            "latency_max_ms": 39.154,
            "latency_min_ms": 34.449,
            "load_ms": 1761.7
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/2.2.0/2026-08-26/rf-detr-seg-nano__rfdetrseg_grapha_fp16__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "npu_qnn",
          "context_length": null,
          "date": "2026-08-26",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-npubench",
            "os_build": "Android 16",
            "runtime": "litert",
            "runtime_version": "2.2.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": "QAIRT (Hexagon, JIT on-device)"
          },
          "error": null,
          "evidence": [
            "npubench sweep row: N=50 median, 1 backend = 1 process, accepted only with thermal NONE->NONE; NPU rows additionally required qnn_partition delegate evidence in logcat (a stock model asked for on the NPU can silently land on XNNPACK and report a CPU number)",
            "mode=jit thermal=NONE->NONE headroom=0.76666665->0.76666665 attempt=0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": 27.833,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "iterations": 50,
            "jit_first_load_ms": 36509.0,
            "latency_max_ms": 28.428,
            "latency_min_ms": 27.424,
            "load_ms": 193.5,
            "qnn_partition_lines": 57
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/2.2.0/2026-08-26/rf-detr-seg-nano__rfdetrseg_grapha_fp16__galaxy-s26.json"
      },
      {
        "device": "raspberry-pi-5",
        "run": {
          "accelerator": "cpu_xnnpack",
          "context_length": null,
          "date": "2026-08-31",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Raspberry Pi 5 Model B Rev 1.1",
            "machine_label": "raspberry-pi-5",
            "os_build": "Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41",
            "runtime": "litert",
            "runtime_version": "2.2.0.dev20260804",
            "soc": null,
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "pi5 sweep row: LiteRT benchmark_model, CPU/XNNPACK at --num_threads=4, 3 invocations per file of 10 warm-up + 50 timed runs (the tool caps a phase at 150 s, so very slow graphs run fewer); latency = median of the three per-invocation medians over the timed phase; a row counts as measured only when every invocation exited 0 with XNNPACK engaged and vcgencmd get_throttled 0x0 before and after",
            "versions: ai-edge-litert-nightly=2.2.0.dev20260804, benchmark_model_sha256=babc9275addd8612caf0842294cb5b3f301c3cce46e6c5979cd2803f215f6bee, cpu=Raspberry Pi 5 Model B Rev 1.1, litert-cli-nightly=0.2.0.dev20260805, platform=Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41, python=3.13.5",
            "invocation 0: exit=0 wall_s=41.8 temp 48.8->63.7C throttled=0x0 xnnpack=True median_us=693061.0 runs=50 footprint_peak_mb=182.08",
            "invocation 1: exit=0 wall_s=41.9 temp 49.9->63.7C throttled=0x0 xnnpack=True median_us=695348.0 runs=50 footprint_peak_mb=181.58",
            "invocation 2: exit=0 wall_s=42.0 temp 51.0->64.2C throttled=0x0 xnnpack=True median_us=696368.0 runs=50 footprint_peak_mb=181.59"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": 695.348,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "init_ms": 91.901,
            "invocations": 3,
            "iterations": 150,
            "latency_max_ms": 701.769,
            "latency_min_ms": 691.301,
            "threads": 4,
            "warmup_runs_per_invocation": 10
          },
          "output_match": null,
          "peak_mem_mb": 182.08,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/2.2.0.dev20260804/2026-08-31/rf-detr-seg-nano__rfdetrseg_grapha_fp16__raspberry-pi-5.json"
      }
    ]
  },
  "model": {
    "family": "rf-detr",
    "id": "rf-detr-seg-nano__rfdetrseg_grapha_fp16",
    "license": "apache-2.0",
    "source_url": "https://huggingface.co/litert-community/RF-DETR-Seg-Nano-LiteRT",
    "task": "image-segmentation"
  },
  "pitfalls": [
    "The cls+pos embedding and patch pos-embed are HOST-FED inputs (clspos.bin, pospatch.bin published beside the model): the GPU delegate mis-executes compute chains that consume large baked-constant tensors (fp32 identical wrong numbers) — HF card 'baked-constant execution bug' section.",
    "memory is emitted as memory*2: a [1,N,C] tensor that is both consumed in-graph and a graph output comes back zeroed on the delegate; the host halves it — HF card.",
    "The two-stage query selection (top-100 + gather + reparam with refpoint_embed.bin) runs on the host between graph A and graph B — HF card pipeline diagram."
  ],
  "schema_version": "1.2"
}