{
  "artifacts": [
    {
      "file": "NVIDIA-Nemotron-Labs-Dory1.0-4B_int4.litertlm",
      "sha256": "220daaa90b3b32cdfdec393f71253dcf68467eb16ff67bcc027e2371253a21da",
      "size_mb": 2292.492
    }
  ],
  "benchmarks": [],
  "conversion": {
    "command": "nvidia/NVIDIA-Nemotron-Labs-Dory1.0-4B-BF16 rev b366b87f -> export with the 20 recurrent pairs unrolled twice (weights stored once), per-loop KV caches (56), prefill signatures 128 / 512 / 1024 plus decode over a 2048-token KV budget, the embedding table in its own section; the MLP normalization of 11 of the 36 MLP blocks scaled by 2^-k (k = 1-3) before the sum with the reciprocal square root capped (the same numbers in fp32) -> int4 blockwise-32 with the data-free block scales (absmax / 7, floored at 2^-14) and the codes chosen by GPTQ (group 32, damping 0.01, natural column order, Hessians from 320 calibration samples) -> GPU composites (odml.cache_update + odml.runtime_bmm, 666 in the four graphs; every weight byte for byte) -> metadata (NVIDIA's template with three runtime edits, no start token, stop tokens </s> and <|im_end|> plus 14 stop sequences, the thought channel, max_num_tokens 2048, no activation type declared) (card L130-L137)",
    "quantization": "int4 blockwise-32 on the linear layers, codes chosen by GPTQ, block scales floored at 2^-14; int8 embedding in its own section; GPU composites for the KV-cache writes and attention matmuls; no activation type declared (the GPU runs the runtime's default fp16 activations) (card L32)",
    "tool": "litert-torch 0.9.4 `export_hf` with a model definition of its own (transformers has no Dory class; the formulas follow NVIDIA's vLLM pull request), then hf-to-litertlm's tools/gpu_graph/gpu_graph_retrofit.py loaded unchanged; repro = hf-to-litertlm dory_work/convert_dory_4b.sh (REPRODUCE.md \"NVIDIA-Nemotron-Labs-Dory1.0-4B\") (card L24, L130, L135)",
    "tool_version": "litert-torch 0.9.4, ai-edge-quantizer 0.9.0, litert-converter 0.4.0, torch 2.13.0, transformers 5.14.1; packed with litert-lm-builder 0.18.0 (card L130)"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-24",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (main e93b645; the APK built from 034366c, whose difference to e93b645 is the sample's README only), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1; Run with 'Both' = the one-loop file, then the int4 file, one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = (chunks - 1) / (last - first chunk) as the app prints it ('tok/s' on the screen; tokenizer counts are 1.09-1.60 times the chunks, the sample's README)",
            "int4 final_value 40 (match, GSM8K answer 40), total_ms 26398, first_token_ms 1495.8, chunks 281, tok_s 11.28, load_ms 16496, thermal 0 -> 0; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-24 09:36:17.983 20880 20880 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-54 first_token_ms=1495.8 tok_s=11.28 total_ms=26398.0 chunks=281 thought_chars=0 load_ms=16496 backend=GPU mark=match text=\"**Step-by-step reasoning:**\\n\\n1. **Identify the kittens already in the car:** The prompt states the sisters are driving home with 7 kittens adopted from the local animal shelter. These are 7 kittens currently in the family's possession.\\n\\n2. **Identify the new kittens from the cats:**\\n    *   **Trixie:** The prompt states Trixie has had 12 kittens.\\n    *   **Patchy:** The prompt states Patchy has had \\\"thrice the number of adopted kittens.\\\" This refers to the kittens adopted from the shelter mentioned in the first sentence, which is 7. Therefore, Patchy has had $7 \\\\times 3 = 21$ kittens.\\n\\n3. **Calculate the total number of new kittens:**\\n    *   Trixie's kittens: 12\\n    *   Patchy's kittens: 21\\n    *   Total new kittens: $12 + 21 = 33$\\n\\n4. **Calculate the total number of kittens in the family:**\\n    *   Kittens in the car (adopted): 7\\n    *   New kittens from the cats: 33\\n    *   Total kittens: $7 + 33 = 40$\\n\\nThe Doubtfire family now has 40 kittens.\\n\\n#### 40\"",
            "Hold taken: uptime_s=729901.57 memavail_kb=6129188 thermal=0 battery_dC=267",
            "10-24 09:35:38.807 20880 20947 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-24 09:35:46.955 20880 20947 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-24 09:35:48.000 20880 20947 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-24 09:35:48.955 20880 20947 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode)."
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 11.28,
            "chunks": 281.0,
            "final_value": 40.0,
            "first_token_ms": 1495.8,
            "gsm8k_answer": 40.0,
            "load_ms": 16496.0,
            "screen_recording": 0.0,
            "thermal_status_end": 0.0,
            "thermal_status_start": 0.0,
            "total_ms": 26398.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-54 (GSM8K test question 54, answer 40), thinking off, the one-loop file then the int4 file, screen recording off",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-24/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-24",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (main e93b645; the APK built from 034366c, whose difference to e93b645 is the sample's README only), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1; Run with 'Both' = the one-loop file, then the int4 file, one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = (chunks - 1) / (last - first chunk) as the app prints it ('tok/s' on the screen; tokenizer counts are 1.09-1.60 times the chunks, the sample's README)",
            "int4 final_value 40 (match, GSM8K answer 40), total_ms 30652.2, first_token_ms 1552, chunks 281, tok_s 9.66, load_ms 17814, thermal 0 -> 0; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-24 09:44:34.916 24447 24447 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-54 first_token_ms=1552.0 tok_s=9.66 total_ms=30652.2 chunks=281 thought_chars=0 load_ms=17814 backend=GPU mark=match text=\"**Step-by-step reasoning:**\\n\\n1. **Identify the kittens already in the car:** The prompt states the sisters are driving home with 7 kittens adopted from the local animal shelter. These are 7 kittens currently in the family's possession.\\n\\n2. **Identify the new kittens from the cats:**\\n    *   **Trixie:** The prompt states Trixie has had 12 kittens.\\n    *   **Patchy:** The prompt states Patchy has had \\\"thrice the number of adopted kittens.\\\" This refers to the kittens adopted from the shelter mentioned in the first sentence, which is 7. Therefore, Patchy has had $7 \\\\times 3 = 21$ kittens.\\n\\n3. **Calculate the total number of new kittens:**\\n    *   Trixie's kittens: 12\\n    *   Patchy's kittens: 21\\n    *   Total new kittens: $12 + 21 = 33$\\n\\n4. **Calculate the total number of kittens in the family:**\\n    *   Kittens in the car (adopted): 7\\n    *   New kittens from the cats: 33\\n    *   Total kittens: $7 + 33 = 40$\\n\\nThe Doubtfire family now has 40 kittens.\\n\\n#### 40\"",
            "Hold taken: uptime_s=730395.17 memavail_kb=6840184 thermal=0 battery_dC=302",
            "10-24 09:43:50.309 24447 24484 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-24 09:43:59.046 24447 24484 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-24 09:44:00.185 24447 24484 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-24 09:44:01.259 24447 24484 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "screen recording: the phone's screenrecord at 12 Mbit/s ran through the run (the clip was cut from it)"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 9.66,
            "chunks": 281.0,
            "final_value": 40.0,
            "first_token_ms": 1552.0,
            "gsm8k_answer": 40.0,
            "load_ms": 17814.0,
            "screen_recording": 1.0,
            "thermal_status_end": 0.0,
            "thermal_status_start": 0.0,
            "total_ms": 30652.2
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-54 (GSM8K test question 54, answer 40), thinking off, the one-loop file then the int4 file, screen recording on",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-24/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-26",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (the APK built from b4ab57f, the commit that ships GSM8K test question 75 among the examples; main a582f5c adds only the sample's README and results to it), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1, both files pushed with adb and imported after the SDK's sha256 check; one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = the rate the app prints ('tok/s' on the screen; chunks are not tokens, the sample's README); load_ms = the file's load in that run",
            "int4 gsm8k-75 final_value 60 (match, GSM8K answer 60), total_ms 46424, first_token_ms 1307.2, chunks 320, tok_s 7.09, load_ms 18122, thermal 1 -> 2; files in the run ['int4_1loop', 'int4'], questions 1; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-26 14:21:48.277 23967 23967 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-75 first_token_ms=1307.2 tok_s=7.09 total_ms=46424.0 chunks=320 thought_chars=0 load_ms=18122 backend=GPU mark=match text=\"To find the average square footage of a…",
            "2026-10-26 14:12:16 hold taken: uptime_s=919469.73 memavail_kb=7640420 thermal=2 battery_dC=416",
            "10-26 14:20:47.976 23967 24003 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-26 14:20:56.613 23967 24003 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-26 14:20:57.772 23967 24003 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-26 14:20:58.892 23967 24003 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "hfmodels-android origin/main a582f5c3 samples/loops/results/2026-10-26-s26/t1.json and .log = / .log byte for byte",
            "screen recording: the phone's screenrecord ran through the run (the demo clip was cut from it; the clip's times checked against this file's walls)"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 7.09,
            "chunks": 320.0,
            "file_order_in_run": 2.0,
            "final_value": 60.0,
            "first_token_ms": 1307.2,
            "gsm8k_answer": 60.0,
            "load_ms": 18122.0,
            "question_position": 1.0,
            "screen_recording": 1.0,
            "thermal_status_end": 2.0,
            "thermal_status_start": 1.0,
            "total_ms": 46424.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-75 (GSM8K test question 75, answer 60), thinking off, gsm8k-75 alone, picked from the Examples list with 'Both' (the one-loop file, then the int4 file), run t1, screen recording on",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-26/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-26",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (the APK built from b4ab57f, the commit that ships GSM8K test question 75 among the examples; main a582f5c adds only the sample's README and results to it), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1, both files pushed with adb and imported after the SDK's sha256 check; one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = the rate the app prints ('tok/s' on the screen; chunks are not tokens, the sample's README); load_ms = the file's load in that run",
            "int4 gsm8k-75 final_value 60 (match, GSM8K answer 60), total_ms 42180.8, first_token_ms 1145.8, chunks 320, tok_s 7.8, load_ms 17011, thermal 1 -> 2; files in the run ['int4_1loop', 'int4'], questions 1; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-26 14:12:06.167 19155 19155 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-75 first_token_ms=1145.8 tok_s=7.80 total_ms=42180.8 chunks=320 thought_chars=0 load_ms=17011 backend=GPU mark=match text=\"To find the average square footage of a…",
            "2026-10-26 14:03:44 hold taken: uptime_s=918957.98 memavail_kb=8189648 thermal=1 battery_dC=399",
            "10-26 14:11:10.283 19155 19188 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-26 14:11:18.828 19155 19188 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-26 14:11:20.054 19155 19188 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-26 14:11:21.184 19155 19188 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode)."
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 7.8,
            "chunks": 320.0,
            "file_order_in_run": 2.0,
            "final_value": 60.0,
            "first_token_ms": 1145.8,
            "gsm8k_answer": 60.0,
            "load_ms": 17011.0,
            "question_position": 1.0,
            "screen_recording": 0.0,
            "thermal_status_end": 2.0,
            "thermal_status_start": 1.0,
            "total_ms": 42180.8
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-75 (GSM8K test question 75, answer 60), thinking off, gsm8k-75 alone, picked from the Examples list with 'Both' (the one-loop file, then the int4 file), run u2, screen recording off",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-26/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-26",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (the APK built from b4ab57f, the commit that ships GSM8K test question 75 among the examples; main a582f5c adds only the sample's README and results to it), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1, both files pushed with adb and imported after the SDK's sha256 check; one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = the rate the app prints ('tok/s' on the screen; chunks are not tokens, the sample's README); load_ms = the file's load in that run",
            "int4 gsm8k-75 final_value 60 (match, GSM8K answer 60), total_ms 53717.3, first_token_ms 1392.2, chunks 320, tok_s 6.11, load_ms 12357, thermal 2 -> 2; files in the run ['int4', 'int4_1loop'], questions 8; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-26 13:08:22.225 13427 13427 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-75 first_token_ms=1392.2 tok_s=6.11 total_ms=53717.3 chunks=320 thought_chars=0 load_ms=none backend=GPU mark=match text=\"To find the average square footage of a …",
            "2026-10-26 12:53:33 hold taken: uptime_s=914746.99 memavail_kb=7545512 thermal=3 battery_dC=440",
            "10-26 13:06:46.327 13427 13461 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-26 13:06:54.293 13427 13461 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-26 13:06:54.836 13427 13461 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-26 13:06:55.380 13427 13461 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "hfmodels-android origin/main a582f5c3 samples/loops/results/2026-10-26-s26/b2.json and .log = / .log byte for byte",
            "hfmodels-android origin/main samples/loops/README.md (2026-10-26, each file's first-run pass): | `gsm8k-75` | `#### 7.5` (GSM8K's is 60) | 22.9 | 11.8 | `#### 60` = GSM8K's | 53.7 | 6.1 |"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 6.11,
            "chunks": 320.0,
            "file_order_in_run": 1.0,
            "final_value": 60.0,
            "first_token_ms": 1392.2,
            "gsm8k_answer": 60.0,
            "load_ms": 12357.0,
            "question_position": 6.0,
            "screen_recording": 0.0,
            "thermal_status_end": 2.0,
            "thermal_status_start": 2.0,
            "total_ms": 53717.3
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-75 (GSM8K test question 75, answer 60), thinking off, the sample's eight bundled examples with the int4 file first (this row = gsm8k-75, the sixth), run b2, screen recording off",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-26/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-26",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-loops-app",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/loops (the APK built from b4ab57f, the commit that ships GSM8K test question 75 among the examples; main a582f5c adds only the sample's README and results to it), hfmodels SDK 0.2.1-SNAPSHOT over litertlm-android 0.16.1, both files pushed with adb and imported after the SDK's sha256 check; one model loaded at a time, one conversation per answer, thinking off, the engine's default sampler (greedy); total_ms / first_token_ms = walls from send to the last / first streamed chunk, app_chunks_per_s = the rate the app prints ('tok/s' on the screen; chunks are not tokens, the sample's README); load_ms = the file's load in that run",
            "int4 gsm8k-75 final_value 60 (match, GSM8K answer 60), total_ms 41593.5, first_token_ms 890, chunks 320, tok_s 7.86, load_ms 18402, thermal 2 -> 2; files in the run ['int4_1loop', 'int4'], questions 8; runtime_version 0.16.1, sdk 0.2.1-SNAPSHOT, device SM-S942Q BP4A.251205.006.S942QOPS1AZH9",
            "10-26 13:01:38.625  9812  9812 I loops   : ANSWER loop=2 variant=int4 q=gsm8k-75 first_token_ms=890.0 tok_s=7.86 total_ms=41593.5 chunks=320 thought_chars=0 load_ms=none backend=GPU mark=match text=\"To find the average square footage of a l…",
            "2026-10-26 12:53:33 hold taken: uptime_s=914746.99 memavail_kb=7545512 thermal=3 battery_dC=440",
            "10-26 13:00:10.903  9812  9859 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-26 13:00:19.861  9812  9859 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-26 13:00:21.033  9812  9859 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-26 13:00:22.090  9812  9859 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "hfmodels-android origin/main a582f5c3 samples/loops/results/2026-10-26-s26/b1.json and .log = / .log byte for byte",
            "hfmodels-android origin/main samples/loops/README.md (2026-10-26, each file's first-run pass): | `gsm8k-75` | `#### 7.5` (GSM8K's is 60) | 22.9 | 11.8 | `#### 60` = GSM8K's | 53.7 | 6.1 |"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_gsm8k": 1.0,
            "app_chunks_per_s": 7.86,
            "chunks": 320.0,
            "file_order_in_run": 2.0,
            "final_value": 60.0,
            "first_token_ms": 890.0,
            "gsm8k_answer": 60.0,
            "load_ms": 18402.0,
            "question_position": 6.0,
            "screen_recording": 0.0,
            "thermal_status_end": 2.0,
            "thermal_status_start": 2.0,
            "total_ms": 41593.5
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android app (hfmodels-android samples/loops, debug build); gsm8k-75 (GSM8K test question 75, answer 60), thinking off, the sample's eight bundled examples with the one-loop file first (this row = gsm8k-75, the sixth), run b1, screen recording off",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-26/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-22",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-chat-check",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/chat ChatDeviceCheck (androidTest), hfmodels SDK 0.2.1-SNAPSHOT over com.google.ai.edge.litertlm:litertlm-android 0.16.1 (the SDK's pin), debug build; one model load and ONE conversation per run, the descriptor's thinking default (on), the engine's default sampler (the bundle declares none: greedy); load_ms = the wall of fromPretrained (an adb-pushed copy imported after its sha256 check), first_chunk_ms / total_ms = the walls from send to the first and the last streamed chunk; chunks are the streamed pieces, not tokens (the SDK states no token count); hfmodels-android origin/main e93b645a litertlm/results/2026-10-22-RFGL80R6A6H-litertlm0.16.1-dory-chat-check.log: # hfmodels on litertlm-android 0.16.1 (at commit 5e497ba): samples/chat's ChatDeviceCheck with litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B (int4 gpu: 17 + 25, then the model card's question; int8 cpu: 17 + 25), log_thought=true / # serial=RFGL80R6A6H device=SM-S942Q (Galaxy S26) build=BP4A.251205.006.S942QOPS1AZH9 android=16; chat-debug.apk sha256 6178de70…, chat-debug-androidTest.apk sha256 2345a28b…; each run: its start / end line (uptime, MemAvailable, thermal status, free /data), its log lines, and the runtime's delegate lines",
            "litertlm/results/2026-10-22-RFGL80R6A6H-litertlm0.16.1-dory-chat-check.log run x2 (x2 check litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B gpu - May_17 Extract the dates from: The meeting moved from May 3 to May 17.): start 2026-10-22 23:44:07, end 2026-10-22 23:45:26, rc 0",
            "10-22 23:44:08.455  8949  8965 I hfmodels: resolve litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B: commit=61cb6bbe via SAVED_BINDING, descriptor=john-rocky/hfmodels-android@1db6c03d/catalog/entries/litert-community__NVIDIA-Nemotron-Labs-Dory1.0-4B.json sha=ca7a357b",
            "10-22 23:44:25.943  8949  8963 I hfmodels: ready litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B@61cb6bbe variant=int4 profile=gpu language=GPU prepare_ms=17479",
            "10-22 23:44:25.946  8949  8963 I hfmodels-check: RESULT step=load ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B variant=int4 profile=gpu initialized=[language=GPU] thinking=descriptor load_ms=17810 last_event=Ready",
            "10-22 23:45:20.961  8949  8963 I hfmodels-check: RESULT step=answer ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B expect=\"May 17\" chunks=453 thought_chunks=448 first_chunk_ms=1088 total_ms=55012 state=READY thought_chars=1939 reply=\"May 3, May 17\"",
            "10-22 23:45:23.053  8949  8963 I hfmodels-check: RESULT step=stop ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B chunks_at_stop=5 state=INVALID",
            "10-22 23:45:23.873  8949  8963 I hfmodels-check: RESULT step=release ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B close_ms=819",
            "10-22 23:45:23.873  8949  8963 I hfmodels-check: RESULT ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B failed=[] sdk=0.2.1-SNAPSHOT runtime=litert_lm 0.16.1 device=SM-S942Q build=BP4A.251205.006.S942QOPS1AZH9",
            "10-22 23:44:09.715  8949  8965 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-22 23:44:21.472  8949  8965 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-22 23:44:22.281  8949  8965 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-22 23:44:23.095  8949  8965 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "10-22 23:44:25.927  8949  8965 I tflite  : Replacing 1 out of 4 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 2 partitions for subgraph 0 (decode_embedder)."
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_expect": 1.0,
            "chunks": 453.0,
            "close_ms": 819.0,
            "first_chunk_ms": 1088.0,
            "load_ms": 17810.0,
            "memavailable_end_kb": 7369012.0,
            "memavailable_start_kb": 7410812.0,
            "thermal_status_end": 0.0,
            "thermal_status_start": 0.0,
            "thought_chars": 1939.0,
            "thought_chunks": 448.0,
            "total_ms": 55012.0,
            "uptime_end_s": 608259.81,
            "uptime_start_s": 608181.17
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android library (hfmodels-android samples/chat ChatDeviceCheck, debug build); one conversation, 'Extract the dates from: The meeting moved from May 3 to May 17.' (the card's example; expect 'May 17')",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-22/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-22",
          "decode_tokens_per_s": null,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory-chat-check",
            "os_build": "Android 16 (BP4A.251205.006.S942QOPS1AZH9)",
            "runtime": "litert-lm",
            "runtime_version": "0.16.1",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: hfmodels-android samples/chat ChatDeviceCheck (androidTest), hfmodels SDK 0.2.1-SNAPSHOT over com.google.ai.edge.litertlm:litertlm-android 0.16.1 (the SDK's pin), debug build; one model load and ONE conversation per run, the descriptor's thinking default (on), the engine's default sampler (the bundle declares none: greedy); load_ms = the wall of fromPretrained (an adb-pushed copy imported after its sha256 check), first_chunk_ms / total_ms = the walls from send to the first and the last streamed chunk; chunks are the streamed pieces, not tokens (the SDK states no token count); hfmodels-android origin/main e93b645a litertlm/results/2026-10-22-RFGL80R6A6H-litertlm0.16.1-dory-chat-check.log: # hfmodels on litertlm-android 0.16.1 (at commit 5e497ba): samples/chat's ChatDeviceCheck with litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B (int4 gpu: 17 + 25, then the model card's question; int8 cpu: 17 + 25), log_thought=true / # serial=RFGL80R6A6H device=SM-S942Q (Galaxy S26) build=BP4A.251205.006.S942QOPS1AZH9 android=16; chat-debug.apk sha256 6178de70…, chat-debug-androidTest.apk sha256 2345a28b…; each run: its start / end line (uptime, MemAvailable, thermal status, free /data), its log lines, and the runtime's delegate lines",
            "litertlm/results/2026-10-22-RFGL80R6A6H-litertlm0.16.1-dory-chat-check.log run x1 (x1 check litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B gpu): start 2026-10-22 23:37:59, end 2026-10-22 23:44:07, rc 0",
            "10-22 23:38:00.567  8529  8544 I hfmodels: resolve litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B: commit=61cb6bbe via RESOLVED_BRANCH, descriptor=john-rocky/hfmodels-android@1db6c03d/catalog/entries/litert-community__NVIDIA-Nemotron-Labs-Dory1.0-4B.json sha=ca7a357b",
            "10-22 23:43:52.993  8529  8541 I hfmodels: ready litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B@61cb6bbe variant=int4 profile=gpu language=GPU prepare_ms=20939",
            "10-22 23:43:52.994  8529  8541 I hfmodels-check: RESULT step=load ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B variant=int4 profile=gpu initialized=[language=GPU] thinking=descriptor load_ms=353028 last_event=Ready",
            "10-22 23:44:02.789  8529  8541 I hfmodels-check: RESULT step=answer ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B expect=\"42\" chunks=95 thought_chunks=93 first_chunk_ms=1060 total_ms=9792 state=READY thought_chars=369 reply=\"42\"",
            "10-22 23:44:04.501  8529  8541 I hfmodels-check: RESULT step=stop ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B chunks_at_stop=5 state=INVALID",
            "10-22 23:44:05.177  8529  8541 I hfmodels-check: RESULT step=release ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B close_ms=675",
            "10-22 23:44:05.177  8529  8541 I hfmodels-check: RESULT ok=true model=litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B failed=[] sdk=0.2.1-SNAPSHOT runtime=litert_lm 0.16.1 device=SM-S942Q build=BP4A.251205.006.S942QOPS1AZH9",
            "10-22 23:43:33.174  8529  8544 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "10-22 23:43:45.327  8529  8544 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "10-22 23:43:47.126  8529  8544 I tflite  : Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "10-22 23:43:49.005  8529  8544 I tflite  : Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "10-22 23:43:52.973  8529  8544 I tflite  : Replacing 1 out of 4 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 2 partitions for subgraph 0 (decode_embedder)."
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answer_matches_expect": 1.0,
            "chunks": 95.0,
            "close_ms": 675.0,
            "first_chunk_ms": 1060.0,
            "load_ms": 353028.0,
            "memavailable_end_kb": 7419104.0,
            "memavailable_start_kb": 6339116.0,
            "thermal_status_end": 0.0,
            "thermal_status_start": 0.0,
            "thought_chars": 369.0,
            "thought_chunks": 93.0,
            "total_ms": 9792.0,
            "uptime_end_s": 608180.82,
            "uptime_start_s": 607813.08
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "Android library (hfmodels-android samples/chat ChatDeviceCheck, debug build); one conversation, 'What is 17 + 25? Answer briefly.' (expect '42')",
          "total_ops": 6259,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.16.1/2026-10-22/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-22",
          "decode_tokens_per_s": 12.09,
          "delegated_ops": 6259,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-dory",
            "os_build": "Android 16 (BP4A.251205.006)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "binary: LiteRT-LM v0.18.0 litert_lm_advanced_main sha256 cf3636070a8e51b0ea13d9a4a8b26fc9a2f9aa38f9e1c7687fe009ce7550f815 + 7 .so (the edge-llm-bench source build of the v0.18.0 tag, 2026-10-07, android/engine-pins.json v0.18.0), every file sha256-checked on the phone at each hold's start (s0_kit_check.txt: 8 OK)",
            "hold s26-h5 (r13): HOLD_SEEN 2026-10-22 10:08:19: acquired 10:08:17, skip_after 10:33:17, run_end 10:36:17, released by 10:38:17; keeper released the hold at 2026-10-22 10:15:26",
            "(s26-h5, 1792631424 device epoch, ready=clean: clean; rested 124s since 10:08:19; thermal status 0 -> 1): iterations 2-3 prefill 281.64 / 283.35, decode 12.22 / 12.24 tok/s, TTFT 3.72 / 3.7 s, Init Total 13404.89 ms (cold); .adb.txt EXIT=0 WALL=88 START=1792631424 END=1792631512; .meta.json memory 'VmHWM_max_kB 1764780 VmSwap_max_kB 255112 MemAvailable_min_kB 2245584 kgsl_page_alloc_max_B 5430931456'; sampler 40 samples, CPU capped in 12, GPU capped in 3",
            "(s26-h5, 1792631635 device epoch, ready=clean: clean; rested 122s since 10:11:52; thermal status 0 -> 2): iterations 2-3 prefill 288.29 / 282.86, decode 12.05 / 11.85 tok/s, TTFT 3.64 / 3.7 s, Init Total 10616.75 ms (with the caches of the hold); .adb.txt EXIT=0 WALL=86 START=1792631635 END=1792631721; .meta.json memory 'VmHWM_max_kB 1898624 VmSwap_max_kB 0 MemAvailable_min_kB 2193628 kgsl_page_alloc_max_B 5433798656'; sampler 40 samples, CPU capped in 24, GPU capped in 23",
            "command: --backend=gpu --model_path=/data/local/tmp/dory_r13/f4c.litertlm --input_prompt_file=/data/local/tmp/dory_r13/prompts/q1.txt --max_num_tokens=2048 --max_output_tokens=2048 --async=false --benchmark --benchmark_prefill_tokens=0 --benchmark_decode_tokens=0 --use_session=false --use_session=true --benchmark_prefill_tokens=1024 --benchmark_decode_tokens=256 --num_iterations=3",
            "activation_data_type: Not set",
            "Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_128).",
            "Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "Replacing 6259 out of 6259 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_1024).",
            "Replacing 5620 out of 5620 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (decode).",
            "Replacing 1 out of 4 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 2 partitions for subgraph 0 (decode_embedder).",
            "Device: model=SM-S942Q release=16 build_id=BP4A.251205.006 fingerprint=samsung/m1qjpnx/m1q:16/BP4A.251205.006/S942QOPS1AZH9_SJP1AZH9:user/release-keys"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 256.0,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "run1_cold_start": 1.0,
            "run1_cpu_capped_samples": 12.0,
            "run1_decode_tok_s": 12.23,
            "run1_decode_tok_s_it1": 12.08,
            "run1_decode_tok_s_it2": 12.22,
            "run1_decode_tok_s_it3": 12.24,
            "run1_gpu_capped_samples": 3.0,
            "run1_init_total_ms": 13404.89,
            "run1_kgsl_page_alloc_max_b": 5430931456.0,
            "run1_memavail_min_kb": 2245584.0,
            "run1_prefill_tok_s": 282.495,
            "run1_prefill_tok_s_it1": 289.01,
            "run1_prefill_tok_s_it2": 281.64,
            "run1_prefill_tok_s_it3": 283.35,
            "run1_sampler_samples": 40.0,
            "run1_t_s": 24.557,
            "run1_thermal_status_end": 1.0,
            "run1_thermal_status_start": 0.0,
            "run1_ttft_s": 3.71,
            "run1_vmhwm_kb": 1764780.0,
            "run1_vmswap_max_kb": 255112.0,
            "run2_cold_start": 0.0,
            "run2_cpu_capped_samples": 24.0,
            "run2_decode_tok_s": 11.95,
            "run2_decode_tok_s_it1": 12.18,
            "run2_decode_tok_s_it2": 12.05,
            "run2_decode_tok_s_it3": 11.85,
            "run2_gpu_capped_samples": 23.0,
            "run2_init_total_ms": 10616.75,
            "run2_kgsl_page_alloc_max_b": 5433798656.0,
            "run2_memavail_min_kb": 2193628.0,
            "run2_prefill_tok_s": 285.575,
            "run2_prefill_tok_s_it1": 287.74,
            "run2_prefill_tok_s_it2": 288.29,
            "run2_prefill_tok_s_it3": 282.86,
            "run2_sampler_samples": 40.0,
            "run2_t_s": 25.0083,
            "run2_thermal_status_end": 2.0,
            "run2_thermal_status_start": 0.0,
            "run2_ttft_s": 3.67,
            "run2_vmhwm_kb": 1898624.0,
            "run2_vmswap_max_kb": 0.0,
            "runs": 2.0,
            "t_1024_256_s": 24.7797
          },
          "output_match": null,
          "peak_mem_mb": 1875.663,
          "prefill_tokens_per_s": 284.035,
          "provenance": "measured",
          "runs": true,
          "total_ops": 6259,
          "ttft_ms": 3690.0
        },
        "source": "data/device_runs/0.18.0/2026-10-22/nemotron-labs-dory1.0-4b-int4__galaxy-s26.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-22",
          "decode_tokens_per_s": 57.705,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "Settle: quiet after 0 s; warm-up + 3 runs; prefill 1033.45 / decode 57.77 tok/s, TTFT 1.0082 s, init 8.0203 s (the CLI prints the mean of the 3 runs); maximum resident set size 3151593472 B, peak memory footprint 4003204264 B; 0 'Validation error' lines",
            "2026-10-22 03:31:26 [w1] START c2_gpu_b1 cap 320s (disk 351 GB free) / 2026-10-22 03:32:06 [w1] DONE c2_gpu_b1 rc=0 40s",
            "Settle: quiet after 0 s; warm-up + 3 runs; prefill 1034.34 / decode 57.64 tok/s, TTFT 1.0074 s, init 7.9788 s (the CLI prints the mean of the 3 runs); maximum resident set size 3157360640 B, peak memory footprint 4016082088 B; 0 'Validation error' lines",
            "2026-10-22 03:32:06 [w1] START c2_gpu_b2 cap 320s (disk 351 GB free) / 2026-10-22 03:32:44 [w1] DONE c2_gpu_b2 rc=0 38s",
            "2026-10-22 03:33:27 [w1] JUDGE R12_C c2_gpu: A 1012.54 / 45.265, B (F4c) 1033.895 / 57.705 tok/s, decode B/A 1.2748 -> faster (> 1.03)"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 256.0,
            "init_s": 7.9996,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "run1_decode_tok_s": 57.77,
            "run1_init_s": 8.0203,
            "run1_max_rss_mb": 3151.593,
            "run1_peak_footprint_mb": 4003.204,
            "run1_prefill_tok_s": 1033.45,
            "run1_t_s": 5.4222,
            "run1_ttft_s": 1.0082,
            "run2_decode_tok_s": 57.64,
            "run2_init_s": 7.9788,
            "run2_max_rss_mb": 3157.361,
            "run2_peak_footprint_mb": 4016.082,
            "run2_prefill_tok_s": 1034.34,
            "run2_t_s": 5.4314,
            "run2_ttft_s": 1.0074,
            "runs": 2.0,
            "t_1024_256_s": 5.4268,
            "warmup_runs": 1.0
          },
          "output_match": null,
          "peak_mem_mb": 4009.643,
          "prefill_tokens_per_s": 1033.895,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": 1007.8
        },
        "source": "data/device_runs/0.18.0/2026-10-22/nemotron-labs-dory1.0-4b-int4__mac-studio-m4-max.json"
      }
    ]
  },
  "model": {
    "family": "dory",
    "id": "nemotron-labs-dory1.0-4b-int4",
    "license": "other (openmdw-1.1)",
    "source_url": "https://huggingface.co/litert-community/NVIDIA-Nemotron-Labs-Dory1.0-4B",
    "task": "text-generation"
  },
  "pitfalls": [
    "GPU file: use the int4 file on a GPU and the int8 file on a CPU. An earlier int4 build of this model prefilled at a third of the int8 rate on the Galaxy S26 CPU (29.0 against 92.1 tok/s), and the int4 weights' greedy answers change from one CPU to another (card L36, L50).",
    "fp16 activations: the MLP normalization sums 7,168 squared values; in fp16 that sum overflows (680 of 361,872 rows in an fp16 emulation over 25 prompts) and the GPTQ weights scored 83 on GSM8K; the file scales those values by 2^-k in 11 of the 36 MLP blocks and caps the reciprocal square root (the same numbers in fp32), which keeps the sum inside fp16 (card L134).",
    "GPU composites: on the Mac GPU decode went from 45.27 to 57.71 tok/s (x 1.275) against the same graph without them; on the Galaxy S26 GPU decode x 1.17 over one run in each order, with 0.7 GB more GPU memory (5.43 against 4.71 GB); on the Mac CPU they do not help (decode x 0.967); with fp16 activations GSM8K is 90 against 91 (card L135).",
    "Accuracy: GSM8K questions 0-99, thinking off, greedy, at most 2,048 tokens of prompt and reply: this file 90 / 100 on the Mac GPU with fp16 activations, the BF16 reference (llama.cpp, two loops) 95 and 86 with one loop; with thinking on the 2048-token budget is short for GSM8K (the two-loop reference left 29 of the first 32 questions with the thought still open) (card L36, L58-L68).",
    "Phone heat: Galaxy S26 GPU rates follow the temperature; from a warmer start (thermal status 1) the same file fell to 9.2-9.5 tok/s by the third iteration against 12.09 from a cool start (card L122).",
    "Pixel 8a GPU: none of the GPU builds run fitted the 8 GB phone under the harness's 400,000 kB MemAvailable rule (earlier builds without the GPU composites, Mali memory 3.3-4.1 GB); this file was not run there; use the CPU (card L124).",
    "NPU: a static-range build (int16 activations, int8 weights) on the Galaxy S26: the Qualcomm plugin took the whole graph, but within 9 s of the on-phone compile free memory fell from 6.93 to 0.92 GB and the harness stopped it; an ahead-of-time build was not tried; there is no NPU file (card L125).",
    "Thinking and template: thinking is on by default (the generation prompt ends with <think> and a newline, the bundle declares the thought channel); rendered with enable_thinking false the prompt ends with <think></think>; NVIDIA recommends temperature 1.0 and top-p 0.95, the bundle sets no sampler; model type generic_model (tool calls not parsed; tool calling not tested) (card L97, L99).",
    "Final mark on LiteRT-LM 0.16.1 and earlier: of the 14 stop sequences, 13 put one of 13 characters (12 punctuation marks and a space) in front of <|im_end|> and a newline; an answer whose last mark is a token of its own, such as a ? or . after a word, arrives without that mark on LiteRT-LM 0.16.1 (the runtime of the Android app under Loops) and on 0.11.0 to 0.16.0 (source read): the runtime holds the mark back, the answer ends on <|im_end|> alone and the held mark is never written; a mark the tokenizer merges with the text before it (**., ).) is kept; 0.17.0 and later write the held mark (card L97).",
    "Second turn with thinking off: from the second turn of a conversation on, LiteRT-LM puts the previous answer's stop token into the KV cache with a zero embedding, because the file holds the embedding table in its own section and the decoder takes embeddings (read in the source of 0.16.0 to 0.18.1; measured with 0.18.1 on WeVisOne-VL-2B and Qwen2.5-Coder-1.5B-Instruct, not on these files), so later answers can differ from the answer to the same messages in one prompt; with thinking on the runtime rebuilds the cache at each turn, and the source shows a zero embedding at another token from the third turn on (not measured); one question per conversation, or a new conversation that starts with the earlier messages and the new question, avoids both (card L101).",
    "Conversation preface with tools and no messages: the template reads messages[0] without checking that the list is non-empty, so LiteRT-LM's render fails with 'Failed to apply template: undefined value' when the preface is prefilled at creation (prefillPrefaceOnInit, default false) and in the preface render (renderPrefaceIntoString and its C++ and C API forms); the same with extra context (such as enable_thinking) and no messages; read in the source of 0.17.0, 0.18.1 and main, measured with the 0.18.1 PyPI package through its C API on these files' template text; the default path, where the first message is rendered with the preface, is not affected; a system message or any initial message in the conversation config avoids it (card L141).",
    "Cache: without --cache no, the CLI can write a compiled cache file next to the model on its first run; without --backend, `run` uses the CPU (card L95)."
  ],
  "schema_version": "1.2"
}
