A proof of concept framework for unsupervised learning of language patterns, models, and vector representations (embeddings).
This project depends on the latest tensorflow installation (Python 3.5) along with other
dependencies listed in the requirements.txt
file provided in this package.
A complete installation would be something as follows:
- Make sure python 3.5 is installed.
- Create a
virtualenv
using virtualenv or virtualenvwrapper
export WORKON_HOME=~/dev/envs
mkdir -p $WORKON_HOME
source /usr/local/bin/virtualenvwrapper.sh
# or
source /usr/bin/virtualenvwrapper.sh
mkvirtualenv --python=/usr/bin/python3 deepsign
Note: To access the environment anytime just run workon deepsign
To install TensorFlow on the newly setup virtualenv
just run the following:
workon deepsign
pip3 install --upgrade https://storage.googleapis.com/tensorflow/linux/cpu/tensorflow-0.10.0rc0-cp35-cp35m-linux_x86_64.whl
This is the CPU only version, if you need other versions see TensorFlow installation. This should take care of the tensorflow dependency.
This is my library that simplifies the development of tensorflow models (attemps to reduce the verbose) when creating typical neural network computation graphs.
Either use some IDE like PyCharm to manage these dependencies separately by downloading the latest version of TensorX from GitHub.
Alternatively, install it in the current environment using:
workon deepsign
pip3 install --upgrade git+https://github.com/davidelnunes/tensorx.git
Note: TensorX also depends on TensorFlow, so this assumes tensorflow is already installed in the environment.
Take the requirements.txt
file provided and run the following
workon deepsign
pip3 install -r /path/to/requirements.txt
workon deepsign
python -m spacy.en.download
To use a GUI backend for matplotlib
sudo apt-get install tcl-dev tk-dev