General purpose experimentation framework using Tensorflow and Keras
The code base works is compatible with >= Python 3.6 but it is highly proposed to stick on Python 3.6 and something bigger. It is expected that interpreter is installed before continuing with this guide.
It is proposed to use python virtual virtual environments to handle different projects. To create and activate a virtual environment for python 3 you need to run:
$ python3.6 -m venv venv
$ source venv/bin/activate
Make sure you have updated your setuptools and pip versions using:
$ pip install --upgrade setuptools pip
Next you have to install package dependencies, cli tool using the pip command. If you work on development environment
it is prefered to use -e option:
$ pip install -e .
DeepResearch framework provides different cli entry points in order to be able to use different modules from the
framework. These entry points are {resources, trainer, hpt, utils} with the prefix of deep command which is the
entry point for the whole framework.
Some functionality related with text processing depends on NLTK's external resources to work properly. After installation of requirements you need to manually download these resources:
$ python -m nltk.downloader punkt
In order to be able to use this framework on DeepLab's servers with GPUJ support you have to add these lines:
export LD_LIBRARY_PATH=/usr/local/cuda-9.0/lib64
export PATH=/usr/local/cuda-9.0/bin:$PATH
in the end of your .bashrc and then run source ~/.bashrc.