Horizon Official Website
Horizon Smart Car Developer Community
version: 3.0.31

English

中文
English

3.0.31

3.9.0
3.8.1
3.7.1
3.7.0
3.5.0
3.2.0
3.0.31

OE Document Introduction

Product Introduction

OpenExplorer Introduction

Toolchain Overview

Environment Deployment

Quick Start

PTQ and QAT Introduction

Algorithm Model PTQ + On-board Deployment Quick Start

Algorithm Model QAT + Deployment Quick Start

Post-training Quantization (PTQ)

PTQ Conversion Principle and Process

PTQ Conversion Tools

PTQ Conversion Tool Overview

The hb_compile Tool

Model Quantized Compilation

Model Modification and HBIR Model Compilation

The quant_config Introduction

The hb_model_info Tool

The hb_verifier Tool

The hb_eval_preprocess Tool

The hb_config_generator Tool

The HBRuntime Inference Library

Accuracy Debug Tool

PTQ Conversion Steps

Prepare Floating-point Model

Prepare Calibration Data

Model Quantization and Compilation

Model Performance Analysis

Model Accuracy Analysis

Model Accuracy Tuning

PTQ Conversion Samples

Release Introduction

Common Algorithm Model Samples

FAQ and Common Failure Resolutions

Post-training Quantization (PTQ) FAQ

Common Failure Resolutions

Appendix

Data Normalization Parameter Configuration

Image Processing Transformer

Quantized Awareness Training (QAT)

Environmental Dependence

Algorithm Model QAT + Deployment Quick Start

Tutorial

Floating-point Model Requirements

QConfig in Detail

Prepare in Detail

Calibration Tutorial

Quantized Awareness Training Guide

Accuracy Tuning Tool Guide

Quantization Accuracy Tuning Guide

Quantization Accuracy Tuning Practice

Advanced Tutorial

FX Quantization Principle

Operator Fusion

API Reference

QAT

quantization.fuse_modules

quantization.fuse_known_modules

fx.fx_helper.wrap

quantization.prepare

Model Export

quantization.hbdk4.export

quantization.hbdk4.get_hbir_input_flattener

quantization.hbdk4.get_hbir_output_unflattener

Horizon Operator

nn.functional.filter

nn.functional.point_pillars_preprocess

nn.PointPillarsScatter

nn.MultiScaleDeformableAttention

bgr2centered_yuv

rgb2centered_yuv

bgr2centered_gray

rgb2centered_gray

FAQ and Common Failure Resolutions

Model Performance Tuning Guide

Model Performance Optimization

Efficient Model Design Advice

Unify Compute Platform (UCP)

Vision Process

Module Architecture

Performance Indicator

Vision Algorithm

Bilateral Filter

Laplacian Filter

OpticalFlowPyrLK

Warp Perspective

Sample

Basic Image Processing Sample

Image Transformation Sample

Feature Extraction Sample

Optical Flow Sample

Surround View Image Stitching Sample

Video Encoding and Decoding Sample

VP API Introduction

Data Structure

hbVPImageFormat

hbVPFilterKernel

hbVPInterpolationType

Function Interface

hbVPBilateralFilter

hbVPCornerHarris

hbVPEqualizehist

hbVPGaussianBlur

hbVPLaplacianFilter

hbVPOpticalFlowPyrLK

hbVPSepFilter2d

hbVPVideoEncode

hbVPVideoDecode

hbVPWarpPerspective

Environment Variable

Model Inference Development

Model Inference Application Development Guide

Model Inference API Introduction

Model Inference API Overview

Data Type and Data Structure

HB_DNN_TENSOR_MAX_DIMENSIONS

hbDNNPackedHandle_t

hbDNNTensorShape

hbDNNQuantiScale

hbDNNQuantiType

hbDNNTensorProperties

Function Interface

Model Loading & Releasing

Model Inference

Dynamic Input Introduction

Configuration Info

Basic Sample User Guide

AI Benchmark User Guide

Model Inference Tool Introduction

Common Operations

The hrt_model_exec Tool Introduction

High Performance Library

HPL Introduction

Module Architecture

Performance Indicator

High Performance Algorithm

Fast Fourier Transform

Inverse Fast Fourier Transform

Fast Fourier Transform 2D

Inverse Fast Fourier Transform 2D

HPL API Introduction

HPL API Introduction

Data Structure

HB_HPL_MAX_DIMS

hbHPLImaginaryData

Functional Interface

UCP Custom Operator

DSP Development

DSP Development Documentation

DSP Development Process

DSP Sample

Point Cloud Pre-processing Introduction

Quantization and Dequantization Introduction

SLAM Sample Introduction

Demosaicing Sample Introduction

DSP API Instroduction

DSP API Overview

Data Structure

Functional Interface

hb_dsp_env_init

hb_dsp_env_init_ex

hb_dsp_mem_unmap

hb_dsp_register_fn

hb_dsp_unregister_fn

Environment Variables

GPU Development Documentation

Object Detection Full Process Sample

UCP Common API Introduction

Data Structure

Version Information Type

hbUCPTaskHandle_t

hbUCPTaskDoneCb

HB_UCP_ALL_BACKENDS

hbUCPTaskPriority

HB_UCP_INITIALIZE_SCHED_PARAM

hbUCPSchedParam

hbUCPSysMemFlushFlag

Functional Interface

hbUCPGetVersion

hbUCPGetSocName

hbUCPSubmitTask

hbUCPWaitTaskDone

hbUCPSetTaskDoneCb

hbUCPReleaseTask

hbUCPMallocCached

hbUCPGetErrorDesc

Environment Variable

UCP Performance Analysis Tools

UCP Trace Instructions

The hrt_ucp_monitor Tool Introduction

FAQ and Error Code

Model Deployment Practice Guidance

Model Deployment Principle and Process

Prepare the Floating Point Model

Model Conversion and Compilation

Model Modification

Performance Evaluation and Tuning

Board-side Deployment

Model Deployment Practice Guidance Examples

The Practical Guide to Deploying the ResNet18 Model with RGB Input

The Practical Guide to Deploying the ResNet18 Model with Pyramid Input

The Practical Guide to Deploying Multi-Batch ResNet18 Model with Pyramid Input

The Practical Guide to Deploying the ResNet18 Model with Resizer Input

The Practical Guide to Deploying ResNet18 Model with Dequantize Nodes Removed

Advanced Contents

Horizon Torch Samples

Framework

Execution Engine

Tutorials

Quantized Training

Writing Specifications of set_qconfig and Customization of qconfig

How to Turn on AMP

Model Compilation

Registration Mechanism

FCOS-EfficientNetB0 Config Construction

Overriding Config Parameters Using Commands

Start-up Method

Examples

Executing the Script

Multi-machine Instructions

MobileNetV1 Classification Model Training

FCOS Detection Model Training

UNet Segmentation Model Training

PointPillars Detection Model Training

PointPillars Detection Model Training (No config)

PointPillars Detection Model Training (No config and HAT Trainer)

GaNet Lane Line Detection Model Training

FCOS3D Detection Model Training

Motr Multiple Object Track Model Training

CenterPoint Detection Model Training

StereoNet Binocular Depth Estimation Model Training

Bev Multi-task Model Training

Car Keypoint Detection Model Training

lidarMultiTask Model Training

DenseTNT Trajectory Prediction Model Training

Deformable DETR Detection Model Training

MapTROE Training

QCNet Trajectory Prediction Model Training

Occupancy Prediction Model

lidar Fusion MultiTask Model Training

Lidar Fusion Sparse-bev Perception Model

HMCT API Reference

hmct.api.load_model

hmct.api.build_model

hmct.api.ORTExecutor

hmct.api.check_model

HBDK Tool API Reference

Benchmark of Model Performance

Appendix

Toolchain Operator Support Constraint List

Operator Support List

J6P ONNX Operator Support List

J6P Torch Operator Support List

J6E/M ONNX Operator Support List

J6E/M Torch Operator Support List

Operator BPU Constraint List

J6P ONNX Operator BPU Constraint List

J6P Torch Operator BPU Constraint List

J6E/M ONNX Operator BPU Constraint List

J6E/M Torch Operator BPU Constraint List

HBIR Operator Definition

Dataset Download

Common Abbreviations

Community Quality Articles

Common Image Format

OE Development Kit Samples Introduction

Third-party Software Bug Explanation

>Quantized Awareness Training (QAT)>Environmental Dependence

Environmental Dependence

Dependency	GPU	CPU
os	Ubuntu22.04	Ubuntu22.04
cuda	11.8	N/A
python	3.10	3.10
torch	2.3.0+cu118	2.3.0+cpu
torchvision	0.18.0+cu118	0.18.0+cpu

Previous pageTerminology Next pageAlgorithm Model QAT + Deployment Quick Start