HKUST-KnowComp / R Net
Licence: mit
Tensorflow Implementation of R-Net
Stars: ✭ 582
Programming Languages
python
139335 projects - #7 most used programming language
Labels
Projects that are alternatives of or similar to R Net
Dawn Bench Entries
DAWNBench: An End-to-End Deep Learning Benchmark and Competition
Stars: ✭ 254 (-56.36%)
Mutual labels: squad
co-attention
Pytorch implementation of "Dynamic Coattention Networks For Question Answering"
Stars: ✭ 54 (-90.72%)
Mutual labels: squad
diwako dui
A UI showing unit positions and names of units in your squad
Stars: ✭ 39 (-93.3%)
Mutual labels: squad
extractive rc by runtime mt
Code and datasets of "Multilingual Extractive Reading Comprehension by Runtime Machine Translation"
Stars: ✭ 36 (-93.81%)
Mutual labels: squad
SQUAD2.Q-Augmented-Dataset
Augmented version of SQUAD 2.0 for Questions
Stars: ✭ 31 (-94.67%)
Mutual labels: squad
X3daudio1 7 hrtf
HRTF for Arma 3, Skyrim, and other titles that use XAudio2 + X3DAudio
Stars: ✭ 192 (-67.01%)
Mutual labels: squad
Drqa
A pytorch implementation of Reading Wikipedia to Answer Open-Domain Questions.
Stars: ✭ 378 (-35.05%)
Mutual labels: squad
qa
TensorFlow Models for the Stanford Question Answering Dataset
Stars: ✭ 72 (-87.63%)
Mutual labels: squad
Transformer-QG-on-SQuAD
Implement Question Generator with SOTA pre-trained Language Models (RoBERTa, BERT, GPT, BART, T5, etc.)
Stars: ✭ 28 (-95.19%)
Mutual labels: squad
Squad
Dockerfile for automated build of a Squad gameserver: https://hub.docker.com/r/cm2network/squad/
Stars: ✭ 21 (-96.39%)
Mutual labels: squad
Medi-CoQA
Conversational Question Answering on Clinical Text
Stars: ✭ 22 (-96.22%)
Mutual labels: squad
Learning to retrieve reasoning paths
The official implementation of ICLR 2020, "Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answering".
Stars: ✭ 318 (-45.36%)
Mutual labels: squad
PersianQA
Persian (Farsi) Question Answering Dataset (+ Models)
Stars: ✭ 114 (-80.41%)
Mutual labels: squad
Lambdahack
Haskell game engine library for roguelike dungeon crawlers; please offer feedback, e.g., after trying out the sample game with the web frontend at
Stars: ✭ 439 (-24.57%)
Mutual labels: squad
R Net
A Tensorflow Implementation of R-net: Machine reading comprehension with self matching networks
Stars: ✭ 321 (-44.85%)
Mutual labels: squad
FastFusionNet
A PyTorch Implementation of FastFusionNet on SQuAD 1.1
Stars: ✭ 38 (-93.47%)
Mutual labels: squad
R-Net
- A Tensorflow implementation of R-NET: MACHINE READING COMPREHENSION WITH SELF-MATCHING NETWORKS. This project is specially designed for the SQuAD dataset.
- Should you have any question, please contact Wenxuan Zhou ([email protected]).
Requirements
There have been a lot of known problems caused by using different software versions. Please check your versions before opening issues or emailing me.
General
- Python >= 3.4
- unzip, wget
Python Packages
- tensorflow-gpu >= 1.5.0
- spaCy >= 2.0.0
- tqdm
- ujson
Usage
To download and preprocess the data, run
# download SQuAD and Glove
sh download.sh
# preprocess the data
python config.py --mode prepro
Hyper parameters are stored in config.py. To debug/train/test the model, run
python config.py --mode debug/train/test
To get the official score, run
python evaluate-v1.1.py ~/data/squad/dev-v1.1.json log/answer/answer.json
The default directory for tensorboard log file is log/event
See release for trained model.
Detailed Implementaion
- The original paper uses additive attention, which consumes lots of memory. This project adopts scaled multiplicative attention presented in Attention Is All You Need.
- This project adopts variational dropout presented in A Theoretically Grounded Application of Dropout in Recurrent Neural Networks.
- To solve the degradation problem in stacked RNN, outputs of each layer are concatenated to produce the final output.
- When the loss on dev set increases in a certain period, the learning rate is halved.
- During prediction, the project adopts search method presented in Machine Comprehension Using Match-LSTM and Answer Pointer.
- To address efficiency issue, this implementation uses bucketing method (contributed by xiongyifan) and CudnnGRU. The bucketing method can speedup training, but will lower the F1 score by 0.3%.
Performance
Score
EM | F1 | |
---|---|---|
original paper | 71.1 | 79.5 |
this project | 71.07 | 79.51 |
Training Time (s/it)
Native | Native + Bucket | Cudnn | Cudnn + Bucket | |
---|---|---|---|---|
E5-2640 | 6.21 | 3.56 | - | - |
TITAN X | 2.56 | 1.31 | 0.41 | 0.28 |
Extensions
These settings may increase the score but not used in the model by default. You can turn these settings on in config.py
.
- Pretrained GloVe character embedding. Contributed by yanghanxy.
- Fasttext Embedding. Contributed by xiongyifan. May increase the F1 by 1% (reported by xiongyifan).
Note that the project description data, including the texts, logos, images, and/or trademarks,
for each open source project belongs to its rightful owner.
If you wish to add or remove any projects, please contact us at [email protected].