Multiprocessing

Large HyperStudy or ChangepointStudy runs may require many independent fits over a hyperparameter grid. These fits can be distributed across CPU cores with the joblib backend included in the standard bayesloop installation:

python -m pip install bayesloop

For development checkouts, install the local package in editable mode:

python -m pip install -e .

Once bayesloop is installed, pass n_jobs to fit:

S.fit(silent=True, n_jobs=4)

Use a value that leaves enough memory for one study copy per worker. For small grids the overhead of process startup and serialization can be larger than the speedup, so multiprocessing is most useful for expensive hyperparameter or change-point studies.