Installation

DeepQuest is written in Python and we highly recommend that you use a Conda in order to keep under control your working environment, without interfering with your system-wide configuration, neither former installation of dependencies.

Assuming you are working in a dedicated Python environment, get DeepQuest as follows:

git clone https://github.com/sheffieldnlp/deepQuest.git
cd deepQuest
conda install theano
pip install -r requirements.txt

Computational Requirements

DeepQuest is GPU compatible and we highly recommend to train your models on GPU with a minimum of 8Go memory available. Of course, if you don’t have access to such resources, you can also use DeepQuest on a CPU server. This will considerably extend the training time, especially for complex architecture, such as the POSTECH model (Kim et al., 2017), on large datasets, but architectures such as BiRNN should work fine and take about 12 hours to be trained (while ~20min on GPU).