Pruning recurrent neural network

This is a simple implementation of pruning method from https://arxiv.org/abs/1704.05119

Running instructions

First of all setup the environment by creating virtualenv and installing all the requirements.

# this is done via virtualenvwrapper
mkvirtuaenv pruning
pip install -r requirements.txt
# extra requirement for spacy
python -m spacy download en_core_web_sm

Then run python main.py --help to see options

Train with pruning

In order to get q parameter, run training with --collectq flag, then create your_config.yaml file like this one, and run training again with --prune --config path/to/your/config.yaml flags.

Results

You can see the results here