Benchmark models¶

Larq Compute Engine provides prebuilt binaries to benchmark Larq converted models on Android phones or 64-bit ARM based systems like the Raspberry Pi.

Raspberry Pi (64-bit)

Download the prebuilt benchmarking binary from the latest release:

wget https://github.com/larq/compute-engine/releases/download/v0.4.3/lce_benchmark_model_aarch64 -O lce_benchmark_model

Make the binary executable:
```
chmod +x lce_benchmark_model
```
Benchmark the converted .tflite model:
```
./lce_benchmark_model --graph=quicknet.tflite --num_threads=4
```
Add --help to the command for a detailed description of the available benchmarking options.

Android phone

Install the Android Debug Bridge (adb) on your host machine. E.g. on macOS you can install it via brew:
```
brew cask install android-platform-tools
```
Follow the instructions here to enable "USB debugging" on your Android phone.
Connect your phone and run the following command to confirm that your host computer recognises your phone:
```
adb devices
```

Download the prebuilt Android benchmarking binary from the latest release:

wget https://github.com/larq/compute-engine/releases/download/v0.4.3/lce_benchmark_model_android_arm64 -O lce_benchmark_model

Make the binary executable:
```
chmod +x lce_benchmark_model
```
Transfer the LCE inference binary to your phone:
```
adb push lce_benchmark_model /data/local/tmp
```
Transfer the converted .tflite model file to your phone:
```
adb push quicknet.tflite /data/local/tmp
```

Benchmark the model:

adb shell /data/local/tmp/lce_benchmark_model \
    --graph=/data/local/tmp/quicknet.tflite \
    --num_threads=4

Add --help to the command for a detailed description of the available benchmarking options.