Bi-directional Attention Flow (BiDAF) network is a multi-stage hierarchical process that represents context at different levels of granularity and uses a bi-directional attention flow mechanism to achieve a query-aware context representation without early summarization.
APACHE-2.0 License
Bot releases are hidden (Show)
Demo-compatible save files for TensorFlow v1.2. Put these and extract in the source directory.
Published by seominjoon over 7 years ago
Minor bug fixes, improving documentations
CodaLab worksheet link: https://worksheets.codalab.org/worksheets/0x37a9b8c44f6845c28866267ef941c89d/
paper: https://arxiv.org/abs/1611.01603v5
Published by seominjoon over 7 years ago
Bug fixes
Using exponential average for improved accuracy
CodaLab worksheet link: https://worksheets.codalab.org/worksheets/0x37a9b8c44f6845c28866267ef941c89d/
paper: https://arxiv.org/abs/1611.01603v3
Published by seominjoon almost 8 years ago
First released version (for ICLR submission on Fri Nov 4)
Instructions for training on CNN/DailyMail dataset coming soon!
Corresponding paper: 1611.01603v1
Corresponding CodaLab worksheet: link