Work In Progress.
An unofficial implementation of R-net in PyTorch.
Python 3.5/3.6 and PyTorch 0.2
Usage
python main.py --batch_size 64
Current Problems
Char-Level embedding with RNN is too slow and requires much GPU memory, so I removed it from my implementation. I plan to try using CNN to do char-level embedding.
Performance
I would like to recommend https://github.com/HKUST-KnowComp/R-Net (Tensorflow) if you are looking for high performance. Currently my implementation has not fully reached the paper's performance.