{
  "artifacts": [
    {
      "file": "LLaVA-OneVision-0.5B.litertlm",
      "sha256": "7f7cd7ae3d2ae435a1f69651de02a9d51e916be60fdfef5cadc721725310971a",
      "size_mb": 790.205
    }
  ],
  "benchmarks": [],
  "conversion": {
    "command": "TODO (not recorded in the HF card)",
    "quantization": "TODO — the HF card states no quantization recipe; decoder exported with the externalized (tied) embedder (HF card Conversion notes)",
    "tool": "litert-torch (LiteRT-LM fast_vlm bundle: SigLIP VISION_ENCODER [1,384,384,3]->[1,729,1152] + VISION_ADAPTER [1,729,1152]->[1,730,896] (projector + the learned image_newline token) + single-token EMBEDDER + PREFILL_DECODE) (HF card Conversion notes)",
    "tool_version": "TODO (the HF card names the converter but not its version; no export log in the sources)"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-09-05",
          "decode_tokens_per_s": 69.52,
          "delegated_ops": 1337,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-cold-cache-cooled",
            "os_build": "Android 16",
            "runtime": "litert-lm",
            "runtime_version": "0.16.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "VERBOSE: Replacing 961 out of 1063 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 141 partitions for subgraph 0 (prefill_128).",
            "VERBOSE: Replacing 961 out of 1063 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 141 partitions for subgraph 1 (prefill_1024).",
            "VERBOSE: Replacing 869 out of 972 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 146 partitions for subgraph 2 (decode).",
            "VERBOSE: Replacing 8 out of 8 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 1 partitions for subgraph 3 (odml.rms_norm.impl).",
            "results block: prefill=302.52 decode=69.52 tokens/s"
          ],
          "failure_class": null,
          "full_delegation": false,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 69.0,
            "init_s": 0.86042,
            "prefill_tokens": 213.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 302.52,
          "provenance": "measured",
          "runs": true,
          "total_ops": 1392,
          "ttft_ms": 720.0
        },
        "source": "data/device_runs/0.16.0/2026-09-05/llava-onevision-0.5b__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-08-24",
          "decode_tokens_per_s": 86.45,
          "delegated_ops": 1063,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-cl-pinned",
            "os_build": "Android 16",
            "runtime": "litert-lm",
            "runtime_version": "0.16.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "VERBOSE: Replacing 1063 out of 1063 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "VERBOSE: Replacing 1063 out of 1063 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_1024).",
            "VERBOSE: Replacing 972 out of 972 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (decode).",
            "results block: prefill=1376.31 decode=86.45 tokens/s"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 81.0,
            "init_s": 1.34944,
            "prefill_tokens": 203.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 1376.31,
          "provenance": "measured",
          "runs": true,
          "total_ops": 1063,
          "ttft_ms": 160.0
        },
        "source": "data/device_runs/0.16.0/2026-08-24/llava-onevision-0.5b__galaxy-s26.json"
      }
    ]
  },
  "model": {
    "family": "llava-onevision",
    "id": "llava-onevision-0.5b",
    "license": "apache-2.0",
    "source_url": "https://huggingface.co/litert-community/LLaVA-OneVision-0.5B",
    "task": "image-text-to-text"
  },
  "pitfalls": [
    "Best for single-image VQA — one image per conversation: start a new conversation for a different image; single-image VQA produces coherent, image-grounded answers (CPU-verified) (HF card Quality + limitation).",
    "The vision encoder bakes OpenAI-CLIP normalization + the NCHW transpose into the graph; the single base-resolution (no-anyres) path is used so the image always maps to a fixed 730 soft tokens (HF card Conversion notes).",
    "License Apache-2.0 (LLaVA-OneVision + the Qwen2 language component) (HF card License).",
    "LiteRT-LM .litertlm bundle: LiteRT.js cannot run it, so delegation stays null and there is no browser block; device rows come from data/device_runs/ (Pi 5 / S26 / Pixel 8a / Mac / iPhone as measured)."
  ],
  "schema_version": "1.2"
}