Run aws s3 sync s3://project-fruitfly <destination> --endpoint-url=https://s3.us-east-2.wasabisys.com --no-sign-request to pull the data.
A backup copy of our datasets is available at https://zenodo.org/records/13327077 in case of issues with Wasabi.
to run the baselines
bash scripts/get_data.sh
docker compose up -d
bash scripts/run_sweeps.sh
You will need to modify the variables at the top of the run_sweeps.sh script to point to the correct Weights and Biases entity and project.
The results will be available on Weights and Biases. Some scripts for analysis can be found in results_analysis_scripts/
To visualize the data
- First download the data to <repo_path>/datasets/
- install streamlit
pip install streamlit - from the root of the repo run
streamlit run visualization/visualize_data.py