The L2L (Learning-to-learn) gradient-free optimization framework contains well documented and tested implementations of various gradient free optimization algorithms. It also defines an API that makes it easy to optimize (hyper-)parameters for any task (optimizee). All the implementations in this package are parallel and can run across different cores and nodes (but equally well on a single core).
NOTE: The L2L framework is currently in BETA
If you are developing a new Optimizee or want to try out a new Optimizee with the Optimizers in the L2L package, install L2L as a python package. See section Installing the L2L Package for details on how to install the package (this automatically installs all requirements).
Documentation is available at https://meta-optimization.github.io/L2L/.
From the Top-Level directory of the directory, run the following command:
python -m pip install -e .
or alternatively (if you do not use a virtual environment)
python -m pip install -e . --user
The --user flag is to be used if you wish to install in the user path as opposed to the root path. However, we recommend to use a virtual enviornment, such as the standard Python env or conda env.
The above will install the package by creating symlinks to the code files in the relevant directory containing python modules. This means that you can change any of the code files and see the changes reflected in the package immediately (i.e. without requiring a reinstall). In order to uninstall one may run the following:
pip uninstall L2L
Note that if the setup was done using sudo access, then the uninstall must also be done using sudo access
Having installed this package, we now have access to the top level l2l module which contains all the relevant modules relevant for using the l2l package.
This should also install the sphinx package which should now enable you to build the documentation as specified below.
Run the following command from the doc directory
make html
And open the documentation with
firefox _build/html/index.html
All further (and extensive) documentation is in the html documentation!
To run the tests go to the folder l2l/tests and execute:
python test_all.py
If the package nosetests is installed run on the top folder (L2L):
nosetests -v --with-coverage --cover-package=l2l/