This package contains all of the python code used for learning. The code is based on Lasagne which is based on Theano.
- sudo apt-get -y install liblapack3 liblapack-dev libblas3 libblas-dev gfortran libspqr1.3.1 libcholmod2.1.2 libmetis5 libmetis-dev libcolamd2.8.0 libccolamd2.8.0 libcamd2.3.1 libamd2.3.1 libx11-dev python-dev
A bunch of math libraries that will be needed - pip install Theano
This is the backbone of the learning framework. - pip install matplotlib
Plotting things is great, especially while things are running - pip install Lasagne==0.1
Makes creating neural networks easier - sudo pip install dill
This will allow for some cool multi-processing later - sudo pip install pathos This library is nice update to the python multiprocessing library allowing for the pickeling of objects as well as other things.
- apt-get -y install swig3.0
Swig is used to generate Python code that wraps the C++ code. - sudo pip install h5py
This library is used to save and load data from files
- Install Anaconda
- Follow the setup instruction for Theano which are
conda install numpy scipy mkl-service libpython m2w64-toolchain <nose> <nose-parameterized> <sphinx> <pydot-ng>
- sudo apt-get install nvidia-cuda-toolkit nvidia-cuda-dev nvidia-modprobe
These libraries are needed to compile code for the GPU as well as to check what GPU devices are available
NOTE: Ran into this issue on Ubuntu 16.04 (Theano/Theano#4425) As a temporary workaround, I use the following hack:
Add cmd.append('-D_FORCE_INLINES') just before p = subprocess.Popen( in the file nvcc_compiler.py
python3 trainModel.py --config settings/particleSim/PPO/PPO.json
To rerun PLAiD see PLAID.md