Introduction

This repository contains my solution to the "PetFinder.my - Pawpularity Contest" challenge, hosted in Kaggle.

In this competition, we’ll analyze raw images and metadata to predict the “Pawpularity” of pet photos.

Installation

For reproducibility, we included a Docker image we used to develop and test the application. We defined the Machine Learning pipeline in DVC, a version control system for machine learning projects.

First, we copy our personal kaggle.json file to the project's main directory. This file is used to authenticate to the Kaggle API, and download the competition data from inside the Docker container.

$ cp ~/.kaggle/kaggle.json .

Build the Docker image.

$ make build

Start a Docker container based on the newly built image.

$ make start

Start a bash shell in the container.

$ make attach

Reproduce the DVC pipeline.

$ dvc repro

Contribute

Here is a brief description of what each folder contains:

• data: input and pre-processed data
• nbs: notebooks for exploration analyses
• pipe: Python scripts for each step in the DVC pipeline
• src: source code for companion library
• ckpts: model checkpoints
• outs: model outputs

Other important files are:

• dvc.yaml: list input, output, and parameters used by each DVC step
• params.yaml: parameters used for DVC steps

The companion library (ml) is installed in editable mode. Which means you don't need to rebuild the Docker container every time you make a change to it.

Commit labels

When contributing to this repository, please consider using the following convention to label your commit messages.

• BUG: fixing a bug
• DEV: development environment ― e.g., Docker, TensorBoard, system dependencies
• DOC: documentation
• EDA: exploratory data analysis
• ML: modeling, feature engineering
• MAINT: maintenance ― e.g., refactoring
• OPS: ml ops ― e.g., download/unzip/pre- and post-process data

Tools

• Git
• Docker
• NVIDIA NGC
• DVC
• PyTorch
• PyTorch Lightning
• timm