If you are looking for a simple challenge configuration that you can replicate to create a challenge on EvalAI, then you are at the right place. Follow the instructions given below to get started.
.
├── README.md
├── annotations # Contains the annotations for Dataset splits
│ ├── test_annotations_devsplit.json # Annotations of dev split
│ └── test_annotations_testsplit.json # Annotations for test split
├── challenge_data # Contains scripts to test the evalautaion script locally
│ ├── challenge_1 # Contains evaluation script for the challenge
| ├── __init__.py # Imports the main.py file for evaluation
| └── main.py # Challenge evaluation script
│ └── __init__.py # Imports the modules which involve evaluation script loading
├── challenge_config.yaml # Configuration file to define challenge setup
├── evaluation_script # Contains the evaluation script
│ ├── __init__.py # Imports the modules that involve annotations loading etc
│ └── main.py # Contains the main `evaluate()` method
├── logo.jpg # Logo image of the challenge
├── submission.json # Sample submission file
└── templates # Contains challenge related HTML templates
├── challenge_phase_1_description.html # Challenge Phase 1 description template
├── challenge_phase_2_description.html # Challenge Phase 2 description template
├── description.html # Challenge description template
├── evaluation_details.html # Contains description about how submissions will be evalauted for each challenge phase
├── submission_guidelines.html # Contains information about how to make submissions to the challenge
└── terms_and_conditions.html # Contains terms and conditions related to the challenge
├── worker # Contains the scripts to test evaluation script locally
│ ├── __init__.py # Imports the module that ionvolves loading evaluation script
│ └── run.py # Contains the code to run the evaluation locally
Use this repository as template.
Generate your github personal acccess token and copy it in clipboard.
Add the github personal access token in the forked repository’s secrets with the name AUTH_TOKEN.
evalai_user_auth_token - Go to profile page after logging in and click on Get your Auth Token to copy your auth token.host_team_pk - Go to host team page and copy the ID for the team you want to use for challenge creation.evalai_host_url - Use https://eval.ai for production server and https://staging.eval.ai for staging server.Create a branch with name challenge in the forked repository from the master branch.
Note: Only changes in challenge branch will be synchronized with challenge on EvalAI.
Add evalai_user_auth_token and host_team_pk in github/host_config.json.
Read EvalAI challenge creation documentation to know more about how you want to structure your challenge. Once you are ready, start making changes in the yaml file, HTML templates, evaluation script according to your need.
Commit the changes and push the challenge branch in the repository and wait for the build to complete. View the logs of your build.
If challenge config contains errors then a issue will be opened automatically in the repository with the errors otherwise the challenge will be created on EvalAI.
Go to Hosted Challenges to view your challenge. The challenge will be publicly available once EvalAI admin approves the challenge.
challenge branch and wait for the build to complete.To add custom dependency packages in the evaluation script, refer to this guide.
In order to test the evaluation script locally before uploading it to EvalAI server, please follow the below instructions -
Copy the evaluation script i.e __init__.py , main.py and other relevant files from evaluation_script/ directory to challenge_data/challenge_1/ directory.
Now, edit challenge_phase name, annotation file name and submission file name in the worker/run.py file to the challenge phase codename (which you want to test for), annotation file name in the annotations/ folder (for specific phase) and corresponding submission file respectively.
Run the command python -m worker.run from the directory where annotations/ challenge_data/ and worker/ directories are present. If the command runs successfully, then the evaluation script works locally and will work on the server as well.
Use this when you want to test everything against a local EvalAI server before pushing to the real site.
Spin up EvalAI locally
cd <path-to-EvalAI>
docker-compose up --build # --build only the first time or after code changes
Backend API → http://localhost:8000, Frontend → http://localhost:8888.
Register a self-hosted runner
Point host_config.json to localhost
{
"token": "<your_local_evalai_auth_token>",
"team_pk": "<your_local_team_pk>",
"evalai_host_url": "http://host.docker.internal:8000"
}
host.docker.internal : Docker’s built-in hostname that points to the Docker host (your machine). 8000 : Port where Backend API for the EvalAI server used to create challenges runs.
Create (or switch to) the challenge branch locally
Commit your config / template / script changes here , as you would when creating a challenge using Github.
Verify the result Go to Hosted Challenges on your local server and confirm your challenge appears and renders correctly.
host_config.json file includes default placeholders like:
<evalai_user_auth_token><host_team_pk><evalai_host_url>Please replace them with real values before pushing changes to avoid build errors.
Please feel free to open issues on our GitHub Repository or contact us at team@cloudcv.org if you have issues.