Skip to content

v2.2.7 pip support and parallelization - cma-es and pickle issue

Compare
Choose a tag to compare
@arturluis arturluis released this 29 Sep 21:49
· 19 commits to master since this release

This release provides multiple fixes for the CMA-ES implementation and the pickle issue caused by multiprocessing in some setups.

==Changes in pip support and parallel MSLS release (v2.2)

This release adds pip support to HyperMapper. HyperMapper can now be installed with
pip install hypermapper

This release also contains the following pip-related changes:

For pip-installations, the HyperMapper main script (formerly called via python scripts/hypermapper.py can now be called by the command hypermapper from the command line.
We also added the following command line hooks for pip installations:
    scripts/compute_pareto.py can now be be called by hm-compute-pareto.
    scripts/plot_pareto.py can now be called by hm-plot-pareto.
    scripts/plot_hvi.py can now be called by hm-plot-hvi.
    The quick start scenario in example_scenarios/quick_start/branin.py can now be called from the command line by hm-quickstart.
    scripts/plot_optimization_results.py can now be called by hm-plot-optimization-results.
The pip installation does not contain the example scenarios, apart from the quick start scenario. However, if you clone the repository to get the examples, make sure to call them as a module (see below). This way, you don’t need to set any environment variables with the pip installation.
The pip installation does not require any environment variables.

We recommend executing the example scenarios by executing them as a module, e.g., by python -m example_scenarios.clients.python.client-server_chakong_haimes.

It is still possible to call scripts/hypermapper.py as well as the other scripts in the scripts folder as before. For this to work, you need to
either

call the scripts from the HyperMapper root directory,
or to set HYPERMAPPER_HOME to your HyperMapper root directory.

We recommend deleting the reference to HYPERMAPPER_HOME/scripts from PYTHONPATH although an existing reference should not break anything with this release (except for the rare case where you have set this reference but not HYPERMAPPER_HOME).

We changed the directory structure. The legacy scripts are in the scripts folder. These import everything from the new hypermapper directory. The scripts directory is not part of the pip installation. For pip installations, the scripts can be executed using the command line hooks (see above).

This release also contains optimization and better parallelism support, which makes HyperMapper significantly faster.

At last, this release also adds support for noiseless GP models, which can be used via the json configuration file.