Hailo Command Line Interface

Table of Contents

1. Introduction
2. Hailo CLI command groups
2.1. Model translation related commands
2.2. Benchmark command
2.3. Quantization analysis command
2.4. FW Related commands
2.5. Hailo general commands
2.6. Hailo tutorial
3. Model Translation Related Commands
3.1. TB – Convert *.hn or *.ckpt to Tensorboard
3.2. Parser – Translate network to HN
3.3. Visualizer – HN Visualization tool
3.4. Profiler – Hailo models profiler
3.5. Quantize – Quantize network params
3.6. Compiler – Compile network to HEF Binary files
3.7. Infer – Run inference on target device

Home Training and support Hailo Command Line Interface

1. Introduction

This document will go over the Hailo CLI (Pythonic CLI).

To use the Hailo CLI you will have to source the SDK virtual environment.

Once inside the SDK virtual environment you will be able to call the

Hailo command in order to get all the supported commands.

At any stage when not sure what are the needed parameters you will be able to add the — help flag to the command and get information about different options and flags for the command you are using.

2. Hailo CLI command groups

The CLI commands can be divided into several groups.

Model translation related commands
Benchmark command
Quantization tool commands
FW related commands
General commands
Tutorial

Model translation related commands
Visualize *.hn or *.ckpt in Tensorboard
```
$ hailo tb HN
```
HN visualization tool
```
$ hailo visualizer HN
```
Hailo models Profiler
```
$ hailo profiler HN
```
Quantize network params
```
$ hailo quantize HAR
```
Compile a network to HEF binary files
```
$ hailo compiler HAR
```
Run inference on a target device
```
$ hailo infer HEF
```

2.2. Benchmark command

New Benchmarks tool
```
$ hailo benchmark
```

2.3. Quantization analysis command

Analyze network quantization noise
```
$ hailo analyze HAR
```

Firmware update tool
```
$ hailo fw-update
```
Firmware configuration tool
```
$ hailo fw-config
```
Useful firmware control operations
```
$ hailo fw-control
```

2.5. Hailo general commands

Scans for devices (Ethernet or PCIE)
```
$ hailo scan
```
Scans for devices on a given interface
```
$ hailo broadcast
```
UDP rate limitation tool, relevant only for a devuce connected on the Ethernet link
```
$ hailo udp-limiter
```
Alias to udp-limiter
```
$ hailo udp
```

2.6. Hailo tutorial

Runs the tutorials in jupyter notebook
```
$ hailo tutorial
```

3.1. TB – Convert .hn or .ckpt to Tensorboard

The Hailo tb (TensorBoard) is a shortcut to run TensorFlow TensorBoard tool

This tool is used when trying to find a model’s start and end nodes for the parser tool.

When pressing the link above a TensorBoard session will be started.

3.2. Parser – Translate network to HN

The Hailo parser – run the parsing CLI in order to parse a model in the supported frameworks (TF, TF2, ONNX) to a Hailo archive (HAR) including the HN file and original weights in NPZ format

Dependent on the original model format more parameters will be required.

For example the TF2 additional needed parameters:

On a successful parsing the output will look as follows:

3.3. Visualizer – HN Visualization tool

The Hailo visualier– will take the HN as the input and shows a visualization of the model

And the output will look as follows:

1. Introduction

2. Hailo CLI command groups

2.1. Model translation related commands

2.2. Benchmark command

2.3. Quantization analysis command

2.4. FW Related commands

2.5. Hailo general commands

2.6. Hailo tutorial

3. Model Translation Related Commands

3.1. TB – Convert *.hn or *.ckpt to Tensorboard

3.2. Parser – Translate network to HN

3.3. Visualizer – HN Visualization tool

3.4. Profiler – Hailo models profiler

3.5. Quantize – Quantize network params

3.6. Compiler – Compile network to HEF Binary files

3.7. Infer – Run inference on target device

3.1. TB – Convert .hn or .ckpt to Tensorboard