Datasets:

Modalities:
Image
Text
Formats:
text
Libraries:
Datasets
License:
hilaus-arm commited on
Commit
e7e6003
·
verified ·
1 Parent(s): f8af0a3

Remove model benchmark metadata from dataset

Browse files

Dataset-card metadata is maintained in README.md frontmatter. The standalone metadata.yaml contains model and benchmark fields, is not a Hugging Face dataset-card metadata document, and is outside the AI Portal model schema contract.

Files changed (1) hide show
  1. metadata.yaml +0 -43
metadata.yaml DELETED
@@ -1,43 +0,0 @@
1
- # Metadata generated by ARM Model Optimization Pipeline
2
- version: 1.0.0
3
- task: vision-detection
4
- created_at: '2026-09-12T01:01:35Z'
5
- context:
6
- model:
7
- id: yolo26n
8
- base_model_id: yolo26n
9
- profile: Arm-Optimized
10
- weight_dtype: fp32
11
- model_size_mb: 11.287
12
- parameter_count: 2572280
13
- format: custom-bundle
14
- target:
15
- name: Raspberry Pi 5
16
- hardware_class: Edge Linux
17
- cpu_architecture: arm64
18
- cpu_model: Cortex-A76
19
- cpu_core_count: 4
20
- cpu_clock_ghz: 2.4
21
- system_memory_gb: 8
22
- os: linux
23
- os_version: Raspberry Pi OS Bookworm
24
- runtime:
25
- name: Arm Runtime for YOLO26 on Raspberry Pi 5
26
- execution_backend: cpu
27
- config:
28
- optimisations:
29
- - XNNPACK
30
- dataset:
31
- name: COCO 2017
32
- sample_count: 128
33
- reference_url: https://cocodataset.org/
34
- slice: val2017
35
- benchmark:
36
- batch_size: 1
37
- input_resolution: 640x640
38
- num_runs: 50
39
- warmup_runs: 10
40
- accuracy:
41
- mAP_50_95_pct: 46.16
42
- mAP_50_pct: 49.47
43
- mAP_75_pct: 42.84