CLI tool for machine learning model training.

Install from pip via:

pip install ravenml

Create the conda environment from the environment.yml file using:

conda env create -f environment.yml

Activate the conda environment with:

conda activate ravenml      # may require source activate ravenml depending on system setup

Install ravenML from the root of this repository using:

pip install --editable .

ravenML must be configured with the name of the S3 buckets you wish to pull Jigsaw-created datasets from and upload trained models to.

After installation, set this configuration by running:

ravenml config update

You can check your configuration anytime by running ravenml config show, and update it anytime with ravenml config update.

ravenML provides core functionality while unique model training pipelines are implemented via plugins dynamically loaded at runtime. A default set of plugins is located at ravenML-plugins. See the README there for more information about how plugins work and how to make your own.

To install all default plugins for use with ravenML you just need to clone the repository and run a script.

Clone the respository with:

git clone https://github.com/autognc/ravenML-plugins

Install default plugins by navigating to the downloaded ravenML directory and using:

./install_all.sh

To test your installation run ravenml train list and verify that the training plugin names appear on your console.

We will use commitizen for all commit messages. The repository is set up to use commitizen via the .czrc file. If you have commitizen already installed globally, you can use it to commit for this repository.

If you do not have commitizen installed, follow the instructions on their GitHub.

If you do not have npm installed, you will need to do that before installing commitizen. npm is distributed with Node.js. Install Node.js here.