Build your own RAMP challenge¶

If you have a new data set and a predictive problem, you may wish to use RAMP to organise local experiments within your team, log model prototypes and their scores or formalise your problem with reusable workflow building blocks. If you are a researcher you may be interested in setting up a RAMP challenge on RAMP studio to incite data scientists and machine learning researchers to provide machine learning solutions to your predictive problem. In this section we walk you through what you need to do to either use RAMP workflow locally or to launch a RAMP challenge on RAMP studio.

Data¶

First, you will need to prepare your data by cleaning, if necessary, and splitting it into the required public and private, test and training sets. See Preparing your data for more details.

Minimal requirements¶

Next, to setup your predictive problem to use RAMP workflow, the following files/folders are required:

problem.py - this parametrizes the setup and uses building blocks from RAMP workflow. More information about writing this script can be found at The problem.py file.
submissions/ - each solution to be tested should be stored in its own directory within submissions/. The name of this new directory will serve as the ID for the submission. If you wish to launch a RAMP challenge you will need to provide an example solution within submissions/starting_kit/. Even if you are not launching a RAMP challenge on RAMP Studio, it is useful to have an example submission as it shows which files are required, how they need to be named and how each file should be structured.
data files - the data files of the challenge can be stored with your starting kit in a folder named data/. Alternatively, your data may also be downloaded from elsewhere. If this is the case, you will need to provide a download_data.py file. This file should download the data when you open a terminal and run:
```
$ python download_data.py
```

Full starting-kit¶

Once you have the above files, it is quite easy to prepare the additional files required for a full RAMP ‘starting-kit’. These files are not required for RAMP workflow to function locally but are useful for participants and are required to launch a RAMP challenge on RAMP Studio.

starting-kit notebook - this is a jupyter notebook that introduces the predictive problem, provides some background information, exploratory data analysis and data visualisation, explains the workflow and provides a simple example solution. This example solution should generally be the same solution as within the submissions/starting_kit (see above).
requirements.txt - lists the required packages, for participants that wish to use pip.
environment.yml - lists the required packages, for participants that wish to use conda.
README.md - this is the homepage when the challenge is on GitHub and should provide a quick start guide.

The files listed above should be stored in the same RAMP ‘starting-kit’ directory. The base directory of a full RAMP starting-kit should thus look like:

<starting_kit_name>/    # root starting-kit directory
├── README.md
├── download_data.py (optional)
├── problem.py
├── requirements.txt
├── <ramp_kit_name>_starting_kit.ipynb
├── data/
|   ├── train.csv     # any data file format acceptable
|   └── test.csv
└── submissions/
    └── <starting_kit>/      # example solution
        └── <submission_file.py>

If you wish to launch a RAMP challenge on RAMP Studio you will need to upload the full starting-kit to ramp-kits.

Overall directory structure¶

To deploy a RAMP challenge on a RAMP server, you will need a ‘starting-kit’ and a ‘data’ directory. These directories are generally stored with the following directory structure:

├── ramp-kits/
|   ├── <starting_kit_one>   # root starting-kit directories for each challenge
|   └── <starting_kit_two>
└── ramp-data/
    ├── <data_for_kit_one>   # root data directories for each challenge
    └── <data_for_kit_two>

Note in the example above, there are two different RAMP challenges, with corresponding starting-kit and data directories for each.