A Python package for experiment configuration and running using Hydra and OmegaConf.
- Modular experiment runner
- Hydra-based configuration
- Command-line interface:
python -m runexp
For editable install in your development environment, run:
pip install -e .For standard installation, run:
pip install .python -m runexpE.g. to run experiment configured in configs/experiment1.yaml, one should run:
python -m runexp -cn experiment1.yaml
Find examples of how to setup your experiments and configuration files in examples/.
The example configuration automatically logs results into MLFlow. For simplicity, we are using the sqlite backend with this tracking_uri: "sqlite:///mlruns/mlruns.db".
However, for practical purposes, logging is usually faster when MLFlow is run as a separate process, e.g. like this:
mkdir -p mlruns
mlflow server --backend-store-uri=sqlite:///mlruns/mlruns.db --default-artifact-root=file:mlruns --host 0.0.0.0 --port 5000
The tracking_uri would then be "http://localhost:5000".
Wandb logging is also supported.
- Python >= 3.7
- hydra-core
- omegaconf
MIT