As mentioned in the README.md this repository comes with a bunch of dependencies. Dependencies are maintained in conda environments. Environment files are in the environments folder. Currently we support two environments:
environment.yaml: vanilla Tensorflow, no special CPU instruction support.environment-mkl.yaml: Intel MKL support.
Environment can be installed with mamba [1] or conda (not recommended → slow).
To install the packages run
mamba env create --file environment.yaml
mamba activate deepq
The code has also a dependency to a forked version of keras-rl [2]. To install it (in the activated deepq env) run
pip install git+https://github.com/R-Sweke/keras-rl
For convenience install the ipykernel so that jupyter finds it:
python -m ipykernel install --user --name deepq
To remove the kernel you can run
jupyter kernelspec uninstall deepq
That's it, you are ready to go!
Notebooks and cluster scripts make use of the deepq library which can be found under /lib. To install the library in your local environmnent execute following commands:
cd lib
pip3 install -v .