BUNDOORA

This project establishes a tailored development container environment to ensure consistent and efficient execution of machine learning projects. By encapsulating all necessary dependencies—including Python, PyTorch, CUDA, and Ubuntu 22.04—within a container, it effectively resolves compatibility issues that often arise in diverse development setups. This approach not only streamlines the research process but also enhances reproducibility and collaboration among researchers.

More about how to use it see document

The environment under maintain

Python	Pytorch	OS	CUDA
3.10	2.5.1	AWS Ubuntu 22.04	12.1

Build the image

To build the image for the default platform (linux/amd64)

make build

To build the image for multiple platforms and load it locally

make build-multi

To build the image for multiple platforms and push it to a registry

make push-multi IMAGE_NAME=yourusername/test

To create and use a new builder instance

make create-builder

Research Projects using this repo

• When Simpler Is Better: Traditional Models Outperform LLMs in ICU Mortality Prediction
• Small Language Model good at specific tasks
• Mimic3 Benchmark
• ChronoScribe: Early fusion with transformer architecture

Citation

@software{Li_Bundoora_2024,
author = {Li, Bowen},
doi = {<>},
month = dec,
title = {{Bundoora}},
url = {https://github.com/SkywardAI/bundoora},
version = {1.0.0},
year = {2024}
}