forked from embedded-dev-research/OVMobileBench
-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathandroid_example.yaml
More file actions
70 lines (64 loc) · 1.73 KB
/
android_example.yaml
File metadata and controls
70 lines (64 loc) · 1.73 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
project:
name: "ovmobilebench-android"
run_id: "android_benchmark_001"
description: "OpenVINO benchmark on Android device"
build:
enabled: true
openvino_repo: "/path/to/openvino" # UPDATE THIS PATH
openvino_commit: "HEAD"
build_type: "Release"
toolchain:
android_ndk: "/path/to/android-ndk-r26d" # UPDATE THIS PATH
abi: "arm64-v8a"
api_level: 24
cmake: "cmake"
ninja: "ninja"
options:
ENABLE_INTEL_GPU: "OFF"
ENABLE_ONEDNN_FOR_ARM: "OFF"
ENABLE_PYTHON: "OFF"
BUILD_SHARED_LIBS: "ON"
package:
include_symbols: false
extra_files: []
device:
kind: "android"
serials: ["YOUR_DEVICE_SERIAL"] # UPDATE THIS - use 'adb devices' to get serial
push_dir: "/data/local/tmp/ovmobilebench"
use_root: false
models:
directories:
- "/path/to/models" # UPDATE THIS PATH - directory containing model files
- "/path/to/additional/models" # UPDATE THIS PATH - additional model directories
extensions:
- ".xml" # OpenVINO IR format
- ".onnx" # ONNX format
- ".pb" # TensorFlow format
- ".tflite" # TensorFlow Lite format
models: # Optional: explicit models in addition to directory scanning
- name: "custom_model"
path: "/path/to/custom/model.xml" # UPDATE THIS PATH
run:
repeats: 3
matrix:
niter: [200]
api: ["sync"]
nireq: [1, 2]
nstreams: ["1", "2"]
device: ["CPU"]
infer_precision: ["FP16"]
threads: [2, 4]
cooldown_sec: 5
timeout_sec: 120
warmup: true
report:
sinks:
- type: "json"
path: "experiments/results/android_results.json"
- type: "csv"
path: "experiments/results/android_results.csv"
tags:
experiment: "baseline"
version: "v1.0"
aggregate: true
include_raw: false