thushv89/attention_keras
A Keras/TensorFlow implementation of Bahdanau attention for RNN-based sequence models used in neural machine translation.
Not currently ranked — collecting fresh signals.
star history
This repository provides a custom Keras layer implementing the Bahdanau attention mechanism, a key component in sequence-to-sequence models. The implementation integrates with TensorFlow 2.x and supports both unidirectional and bidirectional RNN encoders. It includes example code for neural machine translation tasks, complete with training scripts and utilities for visualizing attention weight heatmaps.
Frequently asked
- What is thushv89/attention_keras?
- A Keras/TensorFlow implementation of Bahdanau attention for RNN-based sequence models used in neural machine translation.
- Is attention_keras open source?
- Yes — thushv89/attention_keras is open source, released under the MIT license.
- What language is attention_keras written in?
- thushv89/attention_keras is primarily written in Python.
- How popular is attention_keras?
- thushv89/attention_keras has 444 stars on GitHub.
- Where can I find attention_keras?
- thushv89/attention_keras is on GitHub at https://github.com/thushv89/attention_keras.