Sync from arm/Optimized-Runtimes via hub-sync

README.md

@@ -24,19 +24,19 @@ variants at square resolutions of 320, 480, 640, 800, 960, and 1280 pixels.
 ## Performance
 
 The following YOLO26n results were measured on Raspberry Pi 5 at 640x640,
-batch 1, against the 128-image, 1,050-instance
-[`coco-128-by-sa`](https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/e7a3ecd065a304caddd2f170b8d41aeda66735b5)
-dataset revision selected for publication. Inference time excludes
+batch 1, against the 128-image, 1,025-instance
+[`coco-128-by-sa`](https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/97d6ea181e9dd0784db880b705e426ad23d424e1)
+published dataset revision. Inference time excludes
 preprocessing and postprocessing.
 
 | Runtime | Inference (ms/image) | mAP50-95 |
 | --- | ---: | ---: |
-| Arm Runtime (this project) | 36.80 | 0.4866 |
-| NCNN | 66.23 | 0.4899 |
-| MNN | 90.38 | 0.4711 |
-| OpenVINO | 102.56 | 0.4787 |
-| ONNX Runtime | 127.29 | 0.4787 |
-| ExecuTorch | 137.90 | 0.4893 |
+| Arm Runtime (this project) | 36.60 | 0.4863 |
+| NCNN | 67.20 | 0.4902 |
+| MNN | 90.90 | 0.4753 |
+| OpenVINO | 102.00 | 0.4812 |
+| ONNX Runtime | 126.90 | 0.4812 |
+| ExecuTorch | 136.90 | 0.4880 |
 
 See the [performance baseline](benchmarks/reporting/baseline.md) for the Arm
 Runtime measurement and host context.

benchmark/accuracy.yaml

@@ -1,7 +1,7 @@
 version: 1.2.0
 report_type: vision-detection
 profile: Arm-Optimized
-created_at: '2026-10-20T17:58:48Z'
+created_at: '2026-10-22T21:59:43Z'
 context:
   target_ref: raspberry-pi-5-8gb
   runtime:
@@ -14,10 +14,10 @@ context:
       - XNNPACK
   dataset:
     name: COCO 128 BY/SA
-    version: e7a3ecd065a304caddd2f170b8d41aeda66735b5
+    version: 97d6ea181e9dd0784db880b705e426ad23d424e1
     slice: all
     sample_count: 128
-    reference_url: https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/e7a3ecd065a304caddd2f170b8d41aeda66735b5
+    reference_url: https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/97d6ea181e9dd0784db880b705e426ad23d424e1
   benchmark:
     batch_size: 1
     input_resolution: 640x640
@@ -25,6 +25,6 @@ context:
     warmup_runs: 0
 accuracy:
   sub_task: coco_map
-  mAP_50_95_pct: 48.6615
-  mAP_50_pct: 65.8740
-  mAP_75_pct: 52.9025
+  mAP_50_95_pct: 48.6264
+  mAP_50_pct: 65.8367
+  mAP_75_pct: 53.0182

benchmarks/reporting/baseline.md

@@ -5,32 +5,31 @@ and `coco-128-by-sa` dataset identified below. Future performance-sensitive
 changes must use the same dataset and validation configuration before comparing
 their results with this measurement.
 
-The measurement uses the unpublished Hugging Face dataset revision identified
-below, which is the candidate intended for publication. Benchmark operators
-with access must download that revision and supply its local validation
-configuration explicitly with `DATA=/path/to/data.yaml`.
+The measurement uses the published Hugging Face dataset revision identified
+below. Benchmark operators must download that revision and supply its local
+validation configuration explicitly with `DATA=/path/to/data.yaml`.
 
 ## Measurement
 
 | Item | Value |
 | --- | --- |
-| Source revision | `c87e781ba97b353c99cd8fac2dd1995e0d2e43b6` |
+| Source revision | `4447eb6ed26b860740ef2077673254e604932c43` |
 | Runtime library SHA-256 | `6b9978243ed1fd325e5b7b6a2bc616280862ed01edc1f653ee55b088658b60b2` |
 | Model manifest SHA-256 | `3723a972dc64782fe86577eabde3d02539e10c778ae5ef77b0c5b9bc9f0d1f54` |
 | Dataset | `coco-128-by-sa` |
 | Dataset repository | [Arm/coco-128-by-sa](https://huggingface.co/datasets/Arm/coco-128-by-sa) |
-| Dataset Hub revision | [`f8af0a317c0b54c3dbd38c70b9757dccb8b02460`](https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/f8af0a317c0b54c3dbd38c70b9757dccb8b02460) |
+| Dataset Hub revision | [`97d6ea181e9dd0784db880b705e426ad23d424e1`](https://huggingface.co/datasets/Arm/coco-128-by-sa/tree/97d6ea181e9dd0784db880b705e426ad23d424e1) |
 | Dataset images | 128 |
-| Dataset instances | 1,050 |
+| Dataset instances | 1,025 |
 | Validation configuration SHA-256 | `accc0ff00b364d39c5f209eeb7e53609db634541551b7d4a392b7aa7e1e5f62b` |
-| Selection metadata SHA-256 | `042d13fd4cfe7d670a09604f4e437829420982effd12a55d91efa7f1d9fb28a4` |
-| License manifest SHA-256 | `38e5daf4bd373f4dc887bb1817ddae0cbe607c788bb46195592a5d9bfacd521c` |
-| COCO annotations SHA-256 | `3c756fbb846ccfcc71218b13b3fa9e274e87764e015346cc143e8a36c08b7388` |
+| Selection metadata SHA-256 | `18205722c822b59635d4b6c911cef4495b5d688b4b23d84fde6d4e4396cec61f` |
+| License manifest SHA-256 | `a56e45ab8e72a21906332dcfcfe8ba4ece9039c2c0cdc2409952042c57675b96` |
+| COCO annotations SHA-256 | `009ef7e4cd3b533774d8755dbf8a3aa788d5af401b8283e87e27b2cad58cda57` |
 | Ultralytics | 8.4.108 |
 | Ultralytics options | `workers=0 plots=False` |
-| Cache state | Cold; host filesystem caches dropped immediately before each run |
-| Evidence directory | `20261020T182235078652000Z-c87e781ba97b` |
-| Recorded | 2026-10-20 18:22 UTC |
+| Cache state | Cold; host filesystem caches dropped immediately before the run |
+| Evidence directory | `20261022T232118602909Z-4447eb6` |
+| Recorded | 2026-10-22 23:21 UTC |
 
 `workers=0` keeps Ultralytics dataset loading in the benchmark process instead
 of starting data-loader subprocesses. It does not change the native runtime's
@@ -41,14 +40,14 @@ writing validation plots; it does not change detector scoring.
 
 | Measurement | Value |
 | --- | ---: |
-| Precision | 0.769 |
-| Recall | 0.577 |
-| mAP50 | 0.6587 |
-| mAP75 | 0.5290 |
-| mAP50-95 | 0.4866 |
-| Preprocess | 2.2 ms/image |
-| Inference | 36.8 ms/image |
-| Postprocess | 3.9 ms/image |
+| Precision | 0.7613 |
+| Recall | 0.5805 |
+| mAP50 | 0.6584 |
+| mAP75 | 0.5302 |
+| mAP50-95 | 0.4863 |
+| Preprocess | 2.3 ms/image |
+| Inference | 36.6 ms/image |
+| Postprocess | 3.7 ms/image |
 
 ## Host context
 
@@ -56,12 +55,12 @@ The measurement ran on a Raspberry Pi 5 Model B Rev 1.1 with Linux
 6.12.75+rpt-rpi-2712. All four cores were available, and the CPU frequency
 policy remained on the `performance` governor at 2.4 GHz.
 
-The host filesystem caches were dropped immediately before each benchmark.
-For the recorded Arm Runtime run, cached and reclaimable data totaled 0.56 GB
-at the initial observation and 0.87 GB afterward. Cold reads caused 1,119
+The host filesystem caches were dropped immediately before the benchmark.
+For the recorded Arm Runtime run, cached and reclaimable data totaled 1.06 GiB
+at the initial observation and 1.37 GiB afterward. Cold reads caused 1,285
 major page faults; swap input and output did not change during the run.
 
-The CPU temperature increased from 50.2 °C to 51.8 °C. The firmware reported
+The CPU temperature increased from 46.9 °C to 54.0 °C. The firmware reported
 no current or historical throttling flags before or after the run. Available
-memory decreased from 7.40 GB to 7.35 GB, and peak benchmark-process-tree RSS
-was 694 MB.
+memory decreased from 6.54 GiB to 6.51 GiB, and peak benchmark-process-tree RSS
+was 445 MiB.