Contributing To MLRun's Hub

Types of Assets You Can Contribute

• Modules (either a generic module or a model monitoring application)
• Functions (will be converted to MLRun Runtime)

How to Contribute

1. Fork this repository on GitHub and create a new branch for your new asset.
2. Add a new directory for your asset under the appropriate directory (functions/src for functions, modules/src for modules).
3. Populate the directory with your asset files (see the Asset Structure section below).
4. Open a pull request to merge your changes into the main repository, to its development branch.

Asset Structure

Functions

functions
├── src
│   ├── your_function_name
│   │   ├── item.yaml
│   │   ├── function.yaml
│   │   ├── your_function_name.py
│   │   ├── your_function_name.ipynb
│   │   ├── test_your_function_name.py 
│   │   └── requirements.txt

item.yaml

Metadata about the function. Can be generated using the following CLI command:

  python -m cli.cli generate-item-yaml function your_function_name

Then, fill in all the relevant details. For example: kind (either nuclio:serving, serving or job) and categories field (you can browse the MLRun hub UI to see existing categories. You can specify more than one category per function). Important: Be consistent with the module name across the directory name, all relevant item.yaml fields, and the file names.

function.yaml

The MLRun function definition. Can be generated from item.yaml using:

  python -m cli.cli item-to-function --item-path functions/src/your_function_name

your_function_name.py

The main code file for your function. (Notice: keep the code well-documented, the docstrings are used in the hub UI as documentation for the function.)

your_function_name.ipynb

A Jupyter notebook demonstrating the function's usage. (Notice: the notebook must be able to run end-to-end automatically without manual intervention.)

test_your_function_name.py

Unit tests for your function to cover the function functionality as much as possible. (Will run upon each change to your function).

requirements.txt

Any additional Python dependencies required by your function's unit tests. (Notice: The function's own dependencies should be specified in the item.yaml file, not here.)

Modules

modules
├── src
│   ├── your_module_name
│   │   ├── item.yaml
│   │   ├── your_module_name.py
│   │   ├── your_module_name.ipynb
│   │   ├── test_your_module_name.py 
│   │   └── requirements.txt

item.yaml

Metadata about the module. Can be generated using the following CLI command:

  python -m cli.cli generate-item-yaml module your_module_name

Then, fill in all the relevant details. For example: kind (either generic or monitoring_application) and categories (you can browse the MLRun hub UI to see existing categories. You can specify more than one category per module). Important: Be consistent with the module name across the directory name, all relevant item.yaml fields, and the file names.

your_module_name.py

The main code file for your module. (Notice: keep the code well-documented, the docstrings are used in the hub UI as documentation for the module.) For model-monitoring modules, you can see our guidelines for writing model monitoring applications.

your_module_name.ipynb

A Jupyter notebook demonstrating the module's usage.

test_your_module_name.py

Unit tests for your module to cover the module functionality as much as possible. (Will run upon each change to your module).

requirements.txt

Any additional Python dependencies required by your module's unit tests. (Notice: The module's own dependencies should be specified in the item.yaml file, not here.)