Example use cases for the metadata capture, benchmarking, and reproducibility package microbench. Microbench captures runtime metadata from a Python function, such as package versions and timing information, which are saved to a JSON file or a Redis instance.
The first three examples contain a Jupyter Notebook which show how to
analyze the JSON file and perform some basic visualizations. Within
each example is a src directory which contains the original Python
program used to generate the metadata.
The examples are:
- A PySB model called EARM
- A NumPy example showing a (documented) reproducibility difference across versions due to a NumPy API change
- A SLURM example showing how metadata can be caputed for jobs run on a cluster/scheduling system
- A Tellurium example model of the Lorenz attractor