Skip to content

CLI Installation

The Rockfish CLI is a tool to help you run our use case tutorials quickly.

Installing the Rockfish CLI

We recommend using Python >=3.9 to run the Rockfish CLI. This section assumes you are using pip, venv, and macOS/Linux.

Create a fresh virtual environment and activate it:

$ python -m venv rf-venv
$ source rf-venv/bin/activate

Note: You can replace rf-venv here with the virtual environment name of your choice.

Ensure you have the latest version of pip:

$ python -m pip install --upgrade pip

Install the latest version of the Rockfish CLI: 

$ pip install -U rockfish-tutorial -f https://docs142.rockfish.ai/packages/index.html

This will also install dependencies (which includes Rockfish SDK) to your virtual environment.

Now, you can view and run the available tutorials!

Example: Run the Central Global Model Accuracy use case tutorial

Let's see how to use the CLI to run an existing tutorial. In this example, we will run the Central Global Model Accuracy tutorial.

First, initialize the tutorial. This step will download the dataset files required to run the rest of the tutorial. 

$ rockfish-tutorial init central_network

Then, run the onboarding step. This step recommends a Rockfish workflow based on the dataset downloaded in the previous step. It also runs this workflow to validate that the synthetic data matches the real data according to some fidelity and privacy requirements. 

$ rockfish-tutorial exec central_network run_small onboard

Next, run the train step. This step runs the recommended workflow from the onboarding step on the full dataset, and adds the trained Rockfish models to the model store. 

$ rockfish-tutorial exec central_network run_small train

Finally, run the generate step. This step uses the recommended workflow and the trained models to generate a synthetic version of the full dataset. The central_network tutorial, in particular, will also show the results of using Rockfish synthetic data for improving a downstream central model.

$ rockfish-tutorial exec central_network run_small generate

Note: The run_small argument runs the tutorial on 20 hours of the full dataset. You can use the run_entire argument to use all 215 hours of the dataset.

If you want to only view the final results of this tutorial, use the view_results argument:

$ rockfish-tutorial init central_network
$ rockfish-tutorial exec central_network view_results onboard
$ rockfish-tutorial exec central_network view_results train
$ rockfish-tutorial exec central_network view_results generate

Refer to the CLI commands section below for more information on the available options.

CLI Commands

List all available tutorials

$ rockfish-tutorial list

Initialize dataset files for a specific tutorial

$ rockfish-tutorial init <tutorial_name> <optional-args>

Optional arguments:

  1. --save-path <PATH>: Specify the directory path where the dataset files are stored. If unspecified, a default path will be used.

Run a specific tutorial

rockfish-tutorial exec <tutorial_name> <size> <step> <optional_args>

Positional arguments:

  1. <size>: Specify one of view_results, run_small, run_entire to run a tutorial step or view the results.
  2. <step>: Specify one of onboard, train, generate to run the corresponding part of the tutorial.

Optional arguments:

  1. --logging <TRUE/FALSE>: Enable or disable displaying logs from the Rockfish platform. Disabled by default.
  2. --data-path <PATH>: Specify the directory path where the dataset files are stored. Uses the path shown during onboarding by default.

Help Commands

Each CLI command has a --help flag that will display more information (e.g. available arguments) about the command.

To view more information about the main CLI commands: 

$ rockfish-tutorial -h

To view more information about init CLI command: 

$ rockfish-tutorial init -h

To view more information about exec CLI command for a particular tutorial: 

$ rockfish-tutorial exec <tutorial_name> -h