3.1. Preparation¶
After downloading and installing Qunomon, start developing an AIT based on the template.
3.1.1. User workflow¶
3.1.2. Fork the AIT template¶
AIT template : https://github.com/aistairc/ait-template
.
3.1.3. Change AIT repository name¶
3.1.3.1. Recommended naming convention of AIT¶
The recommended naming convention of AIT are as follows.
{prefix}_{target}_{task}_{format}_{measure}
name | mandatory | description |
---|---|---|
prefix | Yes | This represents the type of AIT. There are several reserved types. - eval: Quality evaluation - alyz: Analysis - misc: Miscellaneous - generate: Used for AIT that generates data to be input to AIT. |
target | Yes | Choose either "dataset" or "model" depending on the evaluation target of AIT. |
task | No | The problem targeted by AIT. |
format | No | The format of the data processed by AIT. |
measure | No | The quality measured by AIT. |
restriction
The name must be 50 characters or less.
The available characters are half-width alphanumeric characters and underscore
_
.
3.1.4. Clone the AIT repository¶
Follow the URL to clone.
3.1.5. Edit readme.md¶
After create a directory for new AIT, open {YourAITName}/readme.md
and edit title.
Before
# ait-name ...
After
# {YourAITName}
3.1.6. Launch development envionment¶
We are providing Jupyter Lab environment adjusted to the AIT development purpose as AIT development environment. After launching them, they are provided as web application on the localhost. You can use them through your browser.
Launcher
Windows:
{Qunomon_root}/ait_repository/ait/{YourAITName}/tool/launch_devenv.bat
macOS, Linux:
sudo bash {YourAITname}/tool/launch_devenv.sh
3.1.6.1. Screenshots after launch¶
If the launch has succeeded, applications below are displayed automatically.
Jupyter lab
3.1.6.2. Development work flow¶
In this tutorial, we assumes following work flow as default development work flow.
Experiment and development of an AI evaluation program.
Porting AI evaluation program onto the AIT template.
Perform unit testing of an AI evaluation program.
Preparing registration for Qunomon.
Perform integration testing of an AIT with Qunomon testbed.