Congratulations on making it to this step! You're almost done 🎉
This repository is provided to help with structuring and documenting your solution code before you submit it for review. We recommend cloning this repository as a starting point for your submission.
Your goal is to set up your solution as if it were a finished open-source project. This means providing clear instructions, identifying dependencies and requirements, and structuring code logically with an obvious point of entry. In particular, you should aim for the inference step to be fully reproducible so that it can be run with any new set of data, whether or not it is included in the test set.
.
├── README.md <- You are here!
├── example_documentation_guide.pdf <- Reference for the solution documentation winners are required to submit
├── README_template.md <- Template that you can fill in to document your solution code
├── src <- Folder for your project's source code
├── models <- Folder for your trained models, model predictions, or model summaries
└── Example_submission <- Example of a solution submission
├── README.md <- Example README containing all required information
└── ... <- Codebase for the example submission
The structure of this repo is based on DrivenData's cookiecutter-data-science project template, which we recommend for your submission. Below are a few notes to keep in mind for the purposes of competition solutions.
README: Your solution must include an extremely clear README that explains what code needs to be run to produce your submission starting from a fresh system with no dependencies installed. This includes obvious instructions and a list of all dependencies and requirements. See the provided template for a guide to get started.
Models: Please provide access to all trained model weights necessary to generate predictions from new data samples without needing to retrain your model from scratch. Model weights can be contained in your archive or shared via a cloud storage service.
Data: You do not need to include the raw data provided in the competition. If your solution has additional data that it needs, you can create a data folder and include that in your archive.
Note that the example README and README template are python specific. If you are using another programming language, you may need to adjust the references and tools.
The example submission is based on the benchmark for the Genetic Engineering Attribution Challenge. The goal is to provide an example README that covers all necessary information for a winning submission. Code is provided for reference to help understand the example README.
- Check out our cookiecutter open source data science template as a general resource for sharing data science code effectively.
- The 3rd Place winner from the Deep Chimpact: Depth Estimation for Wildlife Conservation competition provides a good example of a well-written README for more complicated code. A big thank you to user
vecxozfor a well-written and clearly documented solution! - For a full list of what you need to submit, see the
example_documentation_guide.pdf.