oracle cromulon

A neural network for predicting the future

Setup Development Environment

Create conda environment

$ conda create -n cromulon-env python=3.5
$ source activate cromulon-env

Install dependencies

$ pip install -U setuptools --ignore-installed --no-cache-dir
$ pip install -r dev-requirements.txt --src $CONDA_PREFIX

Running the test suite

pytest tests/

Installation on Windows

We need to install scipy and scikit-learn using conda as it requires some external libraries which are not easy to build in Windows.

conda install numpy scipy pandas pytables scikit-learn statsmodels cython

We will need to set the variables for the GSL in the environment first.

set GSL_INCLUDE_DIR=C:\Users\sree\Documents\Software\gsl\include
set GSL_LIBRARY_DIR=C:\Users\sree\Documents\Software\gsl\lib

Now install the requirements using pip.

pip install -r dev-requirements.txt --src %CONDA_PREFIX%

Known Issues

There is an issue with the installation of pytables. The pip installation fails with pip not being able to locate hdf5 library. This can be avoided by setting the variable HDF5_DIR to the directory of anaconda installation. Using the bash this can be done by

export HDF5_DIR=path-to-anaconda

where path-to-anaconda is the full path to anaconda installation (for example export HDF5_DIR=$HOME/anaconda3)

After doing this, it may be necessary to install pytables using conda, which will take care of the necessay hdf5 libraries, and then update it to the correct version using pip:

 conda install pytables
 pip install tables --src $CONDA_PREFIX --ignore-installed --no-cache-dir

Another related issue arises when running the pytest. A runtime error occurs because the lhdf5 is not visible to the executable. This can be solved by setting the DYLD_LIBRARY_PATH to lib directory of anaconda installation. Again, using bash

export DYLD_LIBRARY_PATH=path-to-anaconda/lib