English

Registration Mechanism

The registration mechanism is an important part of HAT, which plays an important role in building config.

This section illustrates how to add a new module using the registration mechanism and how to use it in config by using a custom module as an example.

Module Customization

Here we use backbone as an example to demonstrate how to develop a new module (e.g., mobilenet).

Defining a New Backbone (e.g., MobileNet):

Create a new file: hat/models/backbones/mobilenet.py.

import torch.nn as nn
from hat.registry import OBJECT_REGISTRY

__all__ = ["MobileNet"]

@OBJECT_REGISTRY.register
class MobileNet(nn.Module):
    def __init__(self, args1, args2):
        pass
    def forward(self, x):
        pass

Importing Newly Defined Module

You can import the newly defined module by adding the following line in hat/models/backbones/__init__.py:

from .mobilenet import MobileNet

Using the New`Backbone` in`Config`

model = dict(
    ...
    backbone=dict(
        type="MobileNet",
        args1=xxx,
        args2=xxx,
    )
    ...
)

Similarly, You can register and use other registrable module by using this method.

On This Page

The hb_compile Tool

QAT

Model Export

Horizon Operator

Vision Algorithm

Sample

VP API Introduction

Data Structure

Function Interface

Model Inference API Introduction

Data Type and Data Structure

API Interface

Model Inference Tool Introduction

High Performance Algorithm

HPL API Introduction

Data Structure

Functional Interface

Data Structure

Functional Interface

FRAMEWORK

TUTORIALS

EXAMPLES

API REFERENCE

#Registration Mechanism

#Module Customization

#Defining a New Backbone (e.g., MobileNet):

#Importing Newly Defined Module

#Using the NewBackbone inConfig

Registration Mechanism

Module Customization

Defining a New Backbone (e.g., MobileNet):

Importing Newly Defined Module

Using the New`Backbone` in`Config`